Upload folder using huggingface_hub

checkpoint-1054/config.json

@@ -0,0 +1,27 @@
+{
+  "_name_or_path": "/mnt/parscratch/users/acp23ay/private/models/OLMo-2-1124-7B-mni_Beng-coli-cw2v/",
+  "architectures": [
+    "Olmo2ForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "eos_token_id": 100257,
+  "hidden_act": "silu",
+  "hidden_size": 4096,
+  "initializer_range": 0.02,
+  "intermediate_size": 11008,
+  "max_position_embeddings": 4096,
+  "model_type": "olmo2",
+  "num_attention_heads": 32,
+  "num_hidden_layers": 32,
+  "num_key_value_heads": 32,
+  "pad_token_id": 100277,
+  "rms_norm_eps": 1e-06,
+  "rope_scaling": null,
+  "rope_theta": 500000,
+  "tie_word_embeddings": false,
+  "torch_dtype": "float32",
+  "transformers_version": "4.48.0.dev0",
+  "use_cache": true,
+  "vocab_size": 100384
+}

checkpoint-1054/generation_config.json

@@ -0,0 +1,6 @@
+{
+  "_from_model_config": true,
+  "eos_token_id": 100257,
+  "pad_token_id": 100277,
+  "transformers_version": "4.48.0.dev0"
+}

checkpoint-1054/model-00001-of-00006.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7aec87eb603416187ba146f5478b1d3b2db3699ebbcfdd27b16819425dc12ea7
+size 4950070472

checkpoint-1054/model-00002-of-00006.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:de0b6a357d03676c4986588aaf4dc96afc3f44a0498514961d2b2cfa1940655e
+size 4991622816

checkpoint-1054/model-00003-of-00006.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c1038a7d346a681a8fb6949e054c2246f3f3c80e826855c1cec6af1155cc607d
+size 4924546896

checkpoint-1054/model-00004-of-00006.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:33c0a21146ace02d36928787ae7f34be868f7c81b894dea98ec7b0e5a3ea01b2
+size 4857404928

checkpoint-1054/model-00005-of-00006.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e7f82a3014ccfc0f67208f63785ae4bc71f09aba86db48bf6139d18178abbfed
+size 4857404928

checkpoint-1054/model-00006-of-00006.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1fc931f1d954432e62cc1d24d0f192d768b5fb3a281942af9348f5869ddef0fe
+size 4614509080

checkpoint-1054/model.safetensors.index.json

@@ -0,0 +1,362 @@
+{
+  "metadata": {
+    "total_size": 29195517952
+  },
+  "weight_map": {
+    "lm_head.weight": "model-00006-of-00006.safetensors",
+    "model.embed_tokens.weight": "model-00001-of-00006.safetensors",
+    "model.layers.0.mlp.down_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.0.mlp.gate_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.0.mlp.up_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.0.post_attention_layernorm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.0.post_feedforward_layernorm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.0.self_attn.k_norm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.0.self_attn.k_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.0.self_attn.o_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.0.self_attn.q_norm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.0.self_attn.q_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.0.self_attn.v_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.1.mlp.down_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.1.mlp.gate_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.1.mlp.up_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.1.post_attention_layernorm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.1.post_feedforward_layernorm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.1.self_attn.k_norm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.1.self_attn.k_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.1.self_attn.o_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.1.self_attn.q_norm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.1.self_attn.q_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.1.self_attn.v_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.10.mlp.down_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.10.mlp.gate_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.10.mlp.up_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.10.post_attention_layernorm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.10.post_feedforward_layernorm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.10.self_attn.k_norm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.10.self_attn.k_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.10.self_attn.o_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.10.self_attn.q_norm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.10.self_attn.q_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.10.self_attn.v_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.11.mlp.down_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.11.mlp.gate_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.11.mlp.up_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.11.post_attention_layernorm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.11.post_feedforward_layernorm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.11.self_attn.k_norm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.11.self_attn.k_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.11.self_attn.o_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.11.self_attn.q_norm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.11.self_attn.q_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.11.self_attn.v_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.12.mlp.down_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.12.mlp.gate_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.12.mlp.up_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.12.post_attention_layernorm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.12.post_feedforward_layernorm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.12.self_attn.k_norm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.12.self_attn.k_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.12.self_attn.o_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.12.self_attn.q_norm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.12.self_attn.q_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.12.self_attn.v_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.13.mlp.down_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.13.mlp.gate_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.13.mlp.up_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.13.post_attention_layernorm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.13.post_feedforward_layernorm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.13.self_attn.k_norm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.13.self_attn.k_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.13.self_attn.o_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.13.self_attn.q_norm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.13.self_attn.q_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.13.self_attn.v_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.14.mlp.down_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.14.mlp.gate_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.14.mlp.up_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.14.post_attention_layernorm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.14.post_feedforward_layernorm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.14.self_attn.k_norm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.14.self_attn.k_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.14.self_attn.o_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.14.self_attn.q_norm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.14.self_attn.q_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.14.self_attn.v_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.15.mlp.down_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.15.mlp.gate_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.15.mlp.up_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.15.post_attention_layernorm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.15.post_feedforward_layernorm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.15.self_attn.k_norm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.15.self_attn.k_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.15.self_attn.o_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.15.self_attn.q_norm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.15.self_attn.q_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.15.self_attn.v_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.16.mlp.down_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.16.mlp.gate_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.16.mlp.up_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.16.post_attention_layernorm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.16.post_feedforward_layernorm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.16.self_attn.k_norm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.16.self_attn.k_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.16.self_attn.o_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.16.self_attn.q_norm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.16.self_attn.q_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.16.self_attn.v_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.17.mlp.down_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.17.mlp.gate_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.17.mlp.up_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.17.post_attention_layernorm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.17.post_feedforward_layernorm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.17.self_attn.k_norm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.17.self_attn.k_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.17.self_attn.o_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.17.self_attn.q_norm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.17.self_attn.q_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.17.self_attn.v_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.18.mlp.down_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.18.mlp.gate_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.18.mlp.up_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.18.post_attention_layernorm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.18.post_feedforward_layernorm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.18.self_attn.k_norm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.18.self_attn.k_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.18.self_attn.o_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.18.self_attn.q_norm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.18.self_attn.q_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.18.self_attn.v_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.19.mlp.down_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.19.mlp.gate_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.19.mlp.up_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.19.post_attention_layernorm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.19.post_feedforward_layernorm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.19.self_attn.k_norm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.19.self_attn.k_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.19.self_attn.o_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.19.self_attn.q_norm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.19.self_attn.q_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.19.self_attn.v_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.2.mlp.down_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.2.mlp.gate_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.2.mlp.up_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.2.post_attention_layernorm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.2.post_feedforward_layernorm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.2.self_attn.k_norm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.2.self_attn.k_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.2.self_attn.o_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.2.self_attn.q_norm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.2.self_attn.q_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.2.self_attn.v_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.20.mlp.down_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.20.mlp.gate_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.20.mlp.up_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.20.post_attention_layernorm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.20.post_feedforward_layernorm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.20.self_attn.k_norm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.20.self_attn.k_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.20.self_attn.o_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.20.self_attn.q_norm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.20.self_attn.q_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.20.self_attn.v_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.21.mlp.down_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.21.mlp.gate_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.21.mlp.up_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.21.post_attention_layernorm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.21.post_feedforward_layernorm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.21.self_attn.k_norm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.21.self_attn.k_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.21.self_attn.o_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.21.self_attn.q_norm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.21.self_attn.q_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.21.self_attn.v_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.22.mlp.down_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.22.mlp.gate_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.22.mlp.up_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.22.post_attention_layernorm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.22.post_feedforward_layernorm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.22.self_attn.k_norm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.22.self_attn.k_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.22.self_attn.o_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.22.self_attn.q_norm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.22.self_attn.q_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.22.self_attn.v_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.23.mlp.down_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.23.mlp.gate_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.23.mlp.up_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.23.post_attention_layernorm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.23.post_feedforward_layernorm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.23.self_attn.k_norm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.23.self_attn.k_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.23.self_attn.o_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.23.self_attn.q_norm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.23.self_attn.q_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.23.self_attn.v_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.24.mlp.down_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.24.mlp.gate_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.24.mlp.up_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.24.post_attention_layernorm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.24.post_feedforward_layernorm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.24.self_attn.k_norm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.24.self_attn.k_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.24.self_attn.o_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.24.self_attn.q_norm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.24.self_attn.q_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.24.self_attn.v_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.25.mlp.down_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.25.mlp.gate_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.25.mlp.up_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.25.post_attention_layernorm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.25.post_feedforward_layernorm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.25.self_attn.k_norm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.25.self_attn.k_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.25.self_attn.o_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.25.self_attn.q_norm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.25.self_attn.q_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.25.self_attn.v_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.26.mlp.down_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.26.mlp.gate_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.26.mlp.up_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.26.post_attention_layernorm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.26.post_feedforward_layernorm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.26.self_attn.k_norm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.26.self_attn.k_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.26.self_attn.o_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.26.self_attn.q_norm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.26.self_attn.q_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.26.self_attn.v_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.27.mlp.down_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.27.mlp.gate_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.27.mlp.up_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.27.post_attention_layernorm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.27.post_feedforward_layernorm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.27.self_attn.k_norm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.27.self_attn.k_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.27.self_attn.o_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.27.self_attn.q_norm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.27.self_attn.q_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.27.self_attn.v_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.28.mlp.down_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.28.mlp.gate_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.28.mlp.up_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.28.post_attention_layernorm.weight": "model-00006-of-00006.safetensors",
+    "model.layers.28.post_feedforward_layernorm.weight": "model-00006-of-00006.safetensors",
+    "model.layers.28.self_attn.k_norm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.28.self_attn.k_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.28.self_attn.o_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.28.self_attn.q_norm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.28.self_attn.q_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.28.self_attn.v_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.29.mlp.down_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.29.mlp.gate_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.29.mlp.up_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.29.post_attention_layernorm.weight": "model-00006-of-00006.safetensors",
+    "model.layers.29.post_feedforward_layernorm.weight": "model-00006-of-00006.safetensors",
+    "model.layers.29.self_attn.k_norm.weight": "model-00006-of-00006.safetensors",
+    "model.layers.29.self_attn.k_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.29.self_attn.o_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.29.self_attn.q_norm.weight": "model-00006-of-00006.safetensors",
+    "model.layers.29.self_attn.q_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.29.self_attn.v_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.3.mlp.down_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.3.mlp.gate_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.3.mlp.up_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.3.post_attention_layernorm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.3.post_feedforward_layernorm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.3.self_attn.k_norm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.3.self_attn.k_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.3.self_attn.o_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.3.self_attn.q_norm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.3.self_attn.q_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.3.self_attn.v_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.30.mlp.down_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.30.mlp.gate_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.30.mlp.up_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.30.post_attention_layernorm.weight": "model-00006-of-00006.safetensors",
+    "model.layers.30.post_feedforward_layernorm.weight": "model-00006-of-00006.safetensors",
+    "model.layers.30.self_attn.k_norm.weight": "model-00006-of-00006.safetensors",
+    "model.layers.30.self_attn.k_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.30.self_attn.o_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.30.self_attn.q_norm.weight": "model-00006-of-00006.safetensors",
+    "model.layers.30.self_attn.q_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.30.self_attn.v_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.31.mlp.down_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.31.mlp.gate_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.31.mlp.up_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.31.post_attention_layernorm.weight": "model-00006-of-00006.safetensors",
+    "model.layers.31.post_feedforward_layernorm.weight": "model-00006-of-00006.safetensors",
+    "model.layers.31.self_attn.k_norm.weight": "model-00006-of-00006.safetensors",
+    "model.layers.31.self_attn.k_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.31.self_attn.o_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.31.self_attn.q_norm.weight": "model-00006-of-00006.safetensors",
+    "model.layers.31.self_attn.q_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.31.self_attn.v_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.4.mlp.down_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.4.mlp.gate_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.4.mlp.up_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.4.post_attention_layernorm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.4.post_feedforward_layernorm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.4.self_attn.k_norm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.4.self_attn.k_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.4.self_attn.o_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.4.self_attn.q_norm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.4.self_attn.q_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.4.self_attn.v_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.5.mlp.down_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.5.mlp.gate_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.5.mlp.up_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.5.post_attention_layernorm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.5.post_feedforward_layernorm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.5.self_attn.k_norm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.5.self_attn.k_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.5.self_attn.o_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.5.self_attn.q_norm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.5.self_attn.q_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.5.self_attn.v_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.6.mlp.down_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.6.mlp.gate_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.6.mlp.up_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.6.post_attention_layernorm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.6.post_feedforward_layernorm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.6.self_attn.k_norm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.6.self_attn.k_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.6.self_attn.o_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.6.self_attn.q_norm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.6.self_attn.q_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.6.self_attn.v_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.7.mlp.down_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.7.mlp.gate_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.7.mlp.up_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.7.post_attention_layernorm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.7.post_feedforward_layernorm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.7.self_attn.k_norm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.7.self_attn.k_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.7.self_attn.o_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.7.self_attn.q_norm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.7.self_attn.q_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.7.self_attn.v_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.8.mlp.down_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.8.mlp.gate_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.8.mlp.up_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.8.post_attention_layernorm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.8.post_feedforward_layernorm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.8.self_attn.k_norm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.8.self_attn.k_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.8.self_attn.o_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.8.self_attn.q_norm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.8.self_attn.q_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.8.self_attn.v_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.9.mlp.down_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.9.mlp.gate_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.9.mlp.up_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.9.post_attention_layernorm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.9.post_feedforward_layernorm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.9.self_attn.k_norm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.9.self_attn.k_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.9.self_attn.o_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.9.self_attn.q_norm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.9.self_attn.q_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.9.self_attn.v_proj.weight": "model-00002-of-00006.safetensors",
+    "model.norm.weight": "model-00006-of-00006.safetensors"
+  }
+}

checkpoint-1054/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:164b9264729d640dca8a3b1e40ebac7b896d19e27dd1cade81702e9fbdad0af6
+size 13055369358

checkpoint-1054/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1ff264f99d31b522cc7e2a4eac9d38606d0c58a34c0adc74d71e0ca8b371dc36
+size 14244

checkpoint-1054/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fc23b7ea1627a8ce887ca6d1916045e948746b712746e84e85c5585c1ef54898
+size 1064

checkpoint-1054/trainer_state.json

@@ -0,0 +1,3722 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.0,
+  "eval_steps": 500,
+  "global_step": 1054,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.003795066413662239,
+      "grad_norm": 58.99103546142578,
+      "learning_rate": 3.7735849056603773e-06,
+      "loss": 11.9758,
+      "step": 2
+    },
+    {
+      "epoch": 0.007590132827324478,
+      "grad_norm": 60.11953353881836,
+      "learning_rate": 7.547169811320755e-06,
+      "loss": 10.2142,
+      "step": 4
+    },
+    {
+      "epoch": 0.011385199240986717,
+      "grad_norm": 15.779712677001953,
+      "learning_rate": 1.1320754716981132e-05,
+      "loss": 6.4991,
+      "step": 6
+    },
+    {
+      "epoch": 0.015180265654648957,
+      "grad_norm": 19.793678283691406,
+      "learning_rate": 1.509433962264151e-05,
+      "loss": 5.2038,
+      "step": 8
+    },
+    {
+      "epoch": 0.018975332068311195,
+      "grad_norm": 23.680116653442383,
+      "learning_rate": 1.8867924528301888e-05,
+      "loss": 4.8628,
+      "step": 10
+    },
+    {
+      "epoch": 0.022770398481973434,
+      "grad_norm": 12.569281578063965,
+      "learning_rate": 2.2641509433962265e-05,
+      "loss": 4.6113,
+      "step": 12
+    },
+    {
+      "epoch": 0.026565464895635674,
+      "grad_norm": 16.667444229125977,
+      "learning_rate": 2.641509433962264e-05,
+      "loss": 4.5817,
+      "step": 14
+    },
+    {
+      "epoch": 0.030360531309297913,
+      "grad_norm": 9.480690956115723,
+      "learning_rate": 3.018867924528302e-05,
+      "loss": 4.4784,
+      "step": 16
+    },
+    {
+      "epoch": 0.03415559772296015,
+      "grad_norm": 10.323396682739258,
+      "learning_rate": 3.39622641509434e-05,
+      "loss": 4.5034,
+      "step": 18
+    },
+    {
+      "epoch": 0.03795066413662239,
+      "grad_norm": 6.275529384613037,
+      "learning_rate": 3.7735849056603776e-05,
+      "loss": 4.3612,
+      "step": 20
+    },
+    {
+      "epoch": 0.04174573055028463,
+      "grad_norm": 18.33019256591797,
+      "learning_rate": 4.150943396226415e-05,
+      "loss": 4.3324,
+      "step": 22
+    },
+    {
+      "epoch": 0.04554079696394687,
+      "grad_norm": 5.673777103424072,
+      "learning_rate": 4.528301886792453e-05,
+      "loss": 4.2084,
+      "step": 24
+    },
+    {
+      "epoch": 0.04933586337760911,
+      "grad_norm": 5.893975734710693,
+      "learning_rate": 4.9056603773584906e-05,
+      "loss": 4.2536,
+      "step": 26
+    },
+    {
+      "epoch": 0.05313092979127135,
+      "grad_norm": 7.583183765411377,
+      "learning_rate": 5.283018867924528e-05,
+      "loss": 4.2715,
+      "step": 28
+    },
+    {
+      "epoch": 0.056925996204933584,
+      "grad_norm": 29.12498664855957,
+      "learning_rate": 5.660377358490566e-05,
+      "loss": 4.2941,
+      "step": 30
+    },
+    {
+      "epoch": 0.06072106261859583,
+      "grad_norm": 13.786356925964355,
+      "learning_rate": 6.037735849056604e-05,
+      "loss": 4.1344,
+      "step": 32
+    },
+    {
+      "epoch": 0.06451612903225806,
+      "grad_norm": 9.63448715209961,
+      "learning_rate": 6.415094339622641e-05,
+      "loss": 4.016,
+      "step": 34
+    },
+    {
+      "epoch": 0.0683111954459203,
+      "grad_norm": 24.18341636657715,
+      "learning_rate": 6.79245283018868e-05,
+      "loss": 4.2109,
+      "step": 36
+    },
+    {
+      "epoch": 0.07210626185958255,
+      "grad_norm": 10.357343673706055,
+      "learning_rate": 7.169811320754717e-05,
+      "loss": 4.0338,
+      "step": 38
+    },
+    {
+      "epoch": 0.07590132827324478,
+      "grad_norm": 6.5294060707092285,
+      "learning_rate": 7.547169811320755e-05,
+      "loss": 4.1091,
+      "step": 40
+    },
+    {
+      "epoch": 0.07969639468690702,
+      "grad_norm": 24.691822052001953,
+      "learning_rate": 7.924528301886794e-05,
+      "loss": 3.8535,
+      "step": 42
+    },
+    {
+      "epoch": 0.08349146110056926,
+      "grad_norm": 5.220446586608887,
+      "learning_rate": 8.30188679245283e-05,
+      "loss": 3.9548,
+      "step": 44
+    },
+    {
+      "epoch": 0.0872865275142315,
+      "grad_norm": 10.426956176757812,
+      "learning_rate": 8.679245283018869e-05,
+      "loss": 4.0897,
+      "step": 46
+    },
+    {
+      "epoch": 0.09108159392789374,
+      "grad_norm": 7.928496360778809,
+      "learning_rate": 9.056603773584906e-05,
+      "loss": 3.8346,
+      "step": 48
+    },
+    {
+      "epoch": 0.09487666034155598,
+      "grad_norm": 4.195712566375732,
+      "learning_rate": 9.433962264150944e-05,
+      "loss": 3.9885,
+      "step": 50
+    },
+    {
+      "epoch": 0.09867172675521822,
+      "grad_norm": 8.46008014678955,
+      "learning_rate": 9.811320754716981e-05,
+      "loss": 3.9167,
+      "step": 52
+    },
+    {
+      "epoch": 0.10246679316888045,
+      "grad_norm": 11.128891944885254,
+      "learning_rate": 9.999975375283309e-05,
+      "loss": 3.8611,
+      "step": 54
+    },
+    {
+      "epoch": 0.1062618595825427,
+      "grad_norm": 7.3787031173706055,
+      "learning_rate": 9.999778379005078e-05,
+      "loss": 3.9224,
+      "step": 56
+    },
+    {
+      "epoch": 0.11005692599620494,
+      "grad_norm": 9.250240325927734,
+      "learning_rate": 9.99938439421016e-05,
+      "loss": 3.9686,
+      "step": 58
+    },
+    {
+      "epoch": 0.11385199240986717,
+      "grad_norm": 3.8329930305480957,
+      "learning_rate": 9.99879343642134e-05,
+      "loss": 3.8928,
+      "step": 60
+    },
+    {
+      "epoch": 0.11764705882352941,
+      "grad_norm": 4.931491374969482,
+      "learning_rate": 9.99800552892203e-05,
+      "loss": 3.9112,
+      "step": 62
+    },
+    {
+      "epoch": 0.12144212523719165,
+      "grad_norm": 4.240759372711182,
+      "learning_rate": 9.997020702755353e-05,
+      "loss": 3.7947,
+      "step": 64
+    },
+    {
+      "epoch": 0.1252371916508539,
+      "grad_norm": 2.899198532104492,
+      "learning_rate": 9.995838996722914e-05,
+      "loss": 3.6796,
+      "step": 66
+    },
+    {
+      "epoch": 0.12903225806451613,
+      "grad_norm": 3.4333019256591797,
+      "learning_rate": 9.994460457383284e-05,
+      "loss": 3.9,
+      "step": 68
+    },
+    {
+      "epoch": 0.13282732447817835,
+      "grad_norm": 3.4416115283966064,
+      "learning_rate": 9.992885139050154e-05,
+      "loss": 3.6634,
+      "step": 70
+    },
+    {
+      "epoch": 0.1366223908918406,
+      "grad_norm": 5.570509433746338,
+      "learning_rate": 9.991113103790198e-05,
+      "loss": 3.6589,
+      "step": 72
+    },
+    {
+      "epoch": 0.14041745730550284,
+      "grad_norm": 4.77843713760376,
+      "learning_rate": 9.98914442142063e-05,
+      "loss": 3.5929,
+      "step": 74
+    },
+    {
+      "epoch": 0.1442125237191651,
+      "grad_norm": 3.334336996078491,
+      "learning_rate": 9.986979169506453e-05,
+      "loss": 3.6783,
+      "step": 76
+    },
+    {
+      "epoch": 0.14800759013282733,
+      "grad_norm": 6.237847805023193,
+      "learning_rate": 9.9846174333574e-05,
+      "loss": 3.8635,
+      "step": 78
+    },
+    {
+      "epoch": 0.15180265654648956,
+      "grad_norm": 5.3761887550354,
+      "learning_rate": 9.982059306024577e-05,
+      "loss": 3.7552,
+      "step": 80
+    },
+    {
+      "epoch": 0.1555977229601518,
+      "grad_norm": 3.021200656890869,
+      "learning_rate": 9.979304888296792e-05,
+      "loss": 3.7418,
+      "step": 82
+    },
+    {
+      "epoch": 0.15939278937381404,
+      "grad_norm": 4.499217510223389,
+      "learning_rate": 9.976354288696588e-05,
+      "loss": 3.5939,
+      "step": 84
+    },
+    {
+      "epoch": 0.16318785578747627,
+      "grad_norm": 2.8771920204162598,
+      "learning_rate": 9.973207623475965e-05,
+      "loss": 3.7072,
+      "step": 86
+    },
+    {
+      "epoch": 0.16698292220113853,
+      "grad_norm": 5.786422252655029,
+      "learning_rate": 9.9698650166118e-05,
+      "loss": 3.6476,
+      "step": 88
+    },
+    {
+      "epoch": 0.17077798861480076,
+      "grad_norm": 4.563845157623291,
+      "learning_rate": 9.966326599800967e-05,
+      "loss": 3.6845,
+      "step": 90
+    },
+    {
+      "epoch": 0.174573055028463,
+      "grad_norm": 7.240761756896973,
+      "learning_rate": 9.962592512455138e-05,
+      "loss": 3.4461,
+      "step": 92
+    },
+    {
+      "epoch": 0.17836812144212524,
+      "grad_norm": 7.457889556884766,
+      "learning_rate": 9.958662901695303e-05,
+      "loss": 3.6345,
+      "step": 94
+    },
+    {
+      "epoch": 0.18216318785578747,
+      "grad_norm": 6.383795261383057,
+      "learning_rate": 9.954537922345961e-05,
+      "loss": 3.4164,
+      "step": 96
+    },
+    {
+      "epoch": 0.1859582542694497,
+      "grad_norm": 4.423782825469971,
+      "learning_rate": 9.950217736929029e-05,
+      "loss": 3.4743,
+      "step": 98
+    },
+    {
+      "epoch": 0.18975332068311196,
+      "grad_norm": 4.3913469314575195,
+      "learning_rate": 9.945702515657434e-05,
+      "loss": 3.4078,
+      "step": 100
+    },
+    {
+      "epoch": 0.1935483870967742,
+      "grad_norm": 4.806511878967285,
+      "learning_rate": 9.940992436428409e-05,
+      "loss": 3.49,
+      "step": 102
+    },
+    {
+      "epoch": 0.19734345351043645,
+      "grad_norm": 4.089972496032715,
+      "learning_rate": 9.936087684816486e-05,
+      "loss": 3.6553,
+      "step": 104
+    },
+    {
+      "epoch": 0.20113851992409867,
+      "grad_norm": 4.213963985443115,
+      "learning_rate": 9.930988454066177e-05,
+      "loss": 3.3105,
+      "step": 106
+    },
+    {
+      "epoch": 0.2049335863377609,
+      "grad_norm": 5.072576999664307,
+      "learning_rate": 9.92569494508437e-05,
+      "loss": 3.3807,
+      "step": 108
+    },
+    {
+      "epoch": 0.20872865275142316,
+      "grad_norm": 3.9879724979400635,
+      "learning_rate": 9.920207366432402e-05,
+      "loss": 3.3832,
+      "step": 110
+    },
+    {
+      "epoch": 0.2125237191650854,
+      "grad_norm": 3.9012928009033203,
+      "learning_rate": 9.914525934317855e-05,
+      "loss": 3.4129,
+      "step": 112
+    },
+    {
+      "epoch": 0.21631878557874762,
+      "grad_norm": 3.9006423950195312,
+      "learning_rate": 9.908650872586029e-05,
+      "loss": 3.3788,
+      "step": 114
+    },
+    {
+      "epoch": 0.22011385199240988,
+      "grad_norm": 4.585041046142578,
+      "learning_rate": 9.90258241271112e-05,
+      "loss": 3.2844,
+      "step": 116
+    },
+    {
+      "epoch": 0.2239089184060721,
+      "grad_norm": 3.21873140335083,
+      "learning_rate": 9.896320793787106e-05,
+      "loss": 3.2972,
+      "step": 118
+    },
+    {
+      "epoch": 0.22770398481973433,
+      "grad_norm": 4.861288547515869,
+      "learning_rate": 9.889866262518331e-05,
+      "loss": 3.3681,
+      "step": 120
+    },
+    {
+      "epoch": 0.2314990512333966,
+      "grad_norm": 3.7863895893096924,
+      "learning_rate": 9.883219073209772e-05,
+      "loss": 3.2602,
+      "step": 122
+    },
+    {
+      "epoch": 0.23529411764705882,
+      "grad_norm": 5.968108177185059,
+      "learning_rate": 9.876379487757034e-05,
+      "loss": 3.2957,
+      "step": 124
+    },
+    {
+      "epoch": 0.23908918406072105,
+      "grad_norm": 4.3137993812561035,
+      "learning_rate": 9.869347775636015e-05,
+      "loss": 3.211,
+      "step": 126
+    },
+    {
+      "epoch": 0.2428842504743833,
+      "grad_norm": 6.66726016998291,
+      "learning_rate": 9.862124213892304e-05,
+      "loss": 3.253,
+      "step": 128
+    },
+    {
+      "epoch": 0.24667931688804554,
+      "grad_norm": 3.351065158843994,
+      "learning_rate": 9.85470908713026e-05,
+      "loss": 3.1391,
+      "step": 130
+    },
+    {
+      "epoch": 0.2504743833017078,
+      "grad_norm": 5.007400035858154,
+      "learning_rate": 9.847102687501797e-05,
+      "loss": 3.2324,
+      "step": 132
+    },
+    {
+      "epoch": 0.25426944971537,
+      "grad_norm": 3.5997180938720703,
+      "learning_rate": 9.839305314694873e-05,
+      "loss": 3.1649,
+      "step": 134
+    },
+    {
+      "epoch": 0.25806451612903225,
+      "grad_norm": 5.135311126708984,
+      "learning_rate": 9.831317275921685e-05,
+      "loss": 3.1186,
+      "step": 136
+    },
+    {
+      "epoch": 0.2618595825426945,
+      "grad_norm": 4.529127597808838,
+      "learning_rate": 9.823138885906566e-05,
+      "loss": 3.0933,
+      "step": 138
+    },
+    {
+      "epoch": 0.2656546489563567,
+      "grad_norm": 7.555715084075928,
+      "learning_rate": 9.814770466873585e-05,
+      "loss": 3.1694,
+      "step": 140
+    },
+    {
+      "epoch": 0.269449715370019,
+      "grad_norm": 6.438807487487793,
+      "learning_rate": 9.806212348533838e-05,
+      "loss": 3.0059,
+      "step": 142
+    },
+    {
+      "epoch": 0.2732447817836812,
+      "grad_norm": 3.579599380493164,
+      "learning_rate": 9.797464868072488e-05,
+      "loss": 3.0959,
+      "step": 144
+    },
+    {
+      "epoch": 0.27703984819734345,
+      "grad_norm": 2.4115488529205322,
+      "learning_rate": 9.788528370135443e-05,
+      "loss": 3.0124,
+      "step": 146
+    },
+    {
+      "epoch": 0.2808349146110057,
+      "grad_norm": 4.941457748413086,
+      "learning_rate": 9.77940320681581e-05,
+      "loss": 3.0013,
+      "step": 148
+    },
+    {
+      "epoch": 0.2846299810246679,
+      "grad_norm": 3.0970795154571533,
+      "learning_rate": 9.77008973764e-05,
+      "loss": 2.9156,
+      "step": 150
+    },
+    {
+      "epoch": 0.2884250474383302,
+      "grad_norm": 3.7954447269439697,
+      "learning_rate": 9.760588329553571e-05,
+      "loss": 2.9065,
+      "step": 152
+    },
+    {
+      "epoch": 0.2922201138519924,
+      "grad_norm": 6.178630352020264,
+      "learning_rate": 9.750899356906775e-05,
+      "loss": 2.9992,
+      "step": 154
+    },
+    {
+      "epoch": 0.29601518026565465,
+      "grad_norm": 4.201640605926514,
+      "learning_rate": 9.741023201439803e-05,
+      "loss": 2.9465,
+      "step": 156
+    },
+    {
+      "epoch": 0.2998102466793169,
+      "grad_norm": 6.396300792694092,
+      "learning_rate": 9.730960252267743e-05,
+      "loss": 2.9635,
+      "step": 158
+    },
+    {
+      "epoch": 0.3036053130929791,
+      "grad_norm": 3.5059707164764404,
+      "learning_rate": 9.720710905865256e-05,
+      "loss": 2.9565,
+      "step": 160
+    },
+    {
+      "epoch": 0.30740037950664134,
+      "grad_norm": 3.6058707237243652,
+      "learning_rate": 9.710275566050951e-05,
+      "loss": 2.9546,
+      "step": 162
+    },
+    {
+      "epoch": 0.3111954459203036,
+      "grad_norm": 3.369931697845459,
+      "learning_rate": 9.699654643971472e-05,
+      "loss": 2.9414,
+      "step": 164
+    },
+    {
+      "epoch": 0.31499051233396586,
+      "grad_norm": 3.4260523319244385,
+      "learning_rate": 9.688848558085306e-05,
+      "loss": 2.8259,
+      "step": 166
+    },
+    {
+      "epoch": 0.3187855787476281,
+      "grad_norm": 2.820384979248047,
+      "learning_rate": 9.677857734146289e-05,
+      "loss": 2.8689,
+      "step": 168
+    },
+    {
+      "epoch": 0.3225806451612903,
+      "grad_norm": 6.160562992095947,
+      "learning_rate": 9.666682605186835e-05,
+      "loss": 2.8823,
+      "step": 170
+    },
+    {
+      "epoch": 0.32637571157495254,
+      "grad_norm": 8.138015747070312,
+      "learning_rate": 9.655323611500875e-05,
+      "loss": 2.9995,
+      "step": 172
+    },
+    {
+      "epoch": 0.3301707779886148,
+      "grad_norm": 8.836729049682617,
+      "learning_rate": 9.643781200626511e-05,
+      "loss": 2.7727,
+      "step": 174
+    },
+    {
+      "epoch": 0.33396584440227706,
+      "grad_norm": 5.794134616851807,
+      "learning_rate": 9.632055827328382e-05,
+      "loss": 2.8787,
+      "step": 176
+    },
+    {
+      "epoch": 0.3377609108159393,
+      "grad_norm": 8.271780967712402,
+      "learning_rate": 9.620147953579737e-05,
+      "loss": 2.9325,
+      "step": 178
+    },
+    {
+      "epoch": 0.3415559772296015,
+      "grad_norm": 7.904753684997559,
+      "learning_rate": 9.608058048544251e-05,
+      "loss": 2.9486,
+      "step": 180
+    },
+    {
+      "epoch": 0.34535104364326374,
+      "grad_norm": 5.133846759796143,
+      "learning_rate": 9.595786588557532e-05,
+      "loss": 2.8692,
+      "step": 182
+    },
+    {
+      "epoch": 0.349146110056926,
+      "grad_norm": 4.022572040557861,
+      "learning_rate": 9.583334057108346e-05,
+      "loss": 2.7638,
+      "step": 184
+    },
+    {
+      "epoch": 0.35294117647058826,
+      "grad_norm": 6.008692264556885,
+      "learning_rate": 9.570700944819584e-05,
+      "loss": 2.9448,
+      "step": 186
+    },
+    {
+      "epoch": 0.3567362428842505,
+      "grad_norm": 4.208110809326172,
+      "learning_rate": 9.557887749428913e-05,
+      "loss": 2.7578,
+      "step": 188
+    },
+    {
+      "epoch": 0.3605313092979127,
+      "grad_norm": 3.0022525787353516,
+      "learning_rate": 9.544894975769186e-05,
+      "loss": 2.7091,
+      "step": 190
+    },
+    {
+      "epoch": 0.36432637571157495,
+      "grad_norm": 2.1430158615112305,
+      "learning_rate": 9.531723135748529e-05,
+      "loss": 2.6014,
+      "step": 192
+    },
+    {
+      "epoch": 0.3681214421252372,
+      "grad_norm": 3.903315782546997,
+      "learning_rate": 9.518372748330194e-05,
+      "loss": 2.7102,
+      "step": 194
+    },
+    {
+      "epoch": 0.3719165085388994,
+      "grad_norm": 4.798587322235107,
+      "learning_rate": 9.504844339512095e-05,
+      "loss": 2.5969,
+      "step": 196
+    },
+    {
+      "epoch": 0.3757115749525617,
+      "grad_norm": 2.9411423206329346,
+      "learning_rate": 9.4911384423061e-05,
+      "loss": 2.7626,
+      "step": 198
+    },
+    {
+      "epoch": 0.3795066413662239,
+      "grad_norm": 2.5336225032806396,
+      "learning_rate": 9.477255596717012e-05,
+      "loss": 2.6405,
+      "step": 200
+    },
+    {
+      "epoch": 0.38330170777988615,
+      "grad_norm": 2.6455540657043457,
+      "learning_rate": 9.463196349721308e-05,
+      "loss": 2.699,
+      "step": 202
+    },
+    {
+      "epoch": 0.3870967741935484,
+      "grad_norm": 3.346395969390869,
+      "learning_rate": 9.448961255245584e-05,
+      "loss": 2.4129,
+      "step": 204
+    },
+    {
+      "epoch": 0.3908918406072106,
+      "grad_norm": 4.412266254425049,
+      "learning_rate": 9.434550874144728e-05,
+      "loss": 2.5733,
+      "step": 206
+    },
+    {
+      "epoch": 0.3946869070208729,
+      "grad_norm": 7.3123579025268555,
+      "learning_rate": 9.419965774179824e-05,
+      "loss": 2.6668,
+      "step": 208
+    },
+    {
+      "epoch": 0.3984819734345351,
+      "grad_norm": 7.745405197143555,
+      "learning_rate": 9.405206529995785e-05,
+      "loss": 2.8233,
+      "step": 210
+    },
+    {
+      "epoch": 0.40227703984819735,
+      "grad_norm": 6.307023048400879,
+      "learning_rate": 9.39027372309871e-05,
+      "loss": 2.649,
+      "step": 212
+    },
+    {
+      "epoch": 0.4060721062618596,
+      "grad_norm": 3.971792697906494,
+      "learning_rate": 9.375167941832973e-05,
+      "loss": 2.7105,
+      "step": 214
+    },
+    {
+      "epoch": 0.4098671726755218,
+      "grad_norm": 2.6765499114990234,
+      "learning_rate": 9.359889781358042e-05,
+      "loss": 2.5993,
+      "step": 216
+    },
+    {
+      "epoch": 0.41366223908918404,
+      "grad_norm": 2.9960570335388184,
+      "learning_rate": 9.344439843625034e-05,
+      "loss": 2.6444,
+      "step": 218
+    },
+    {
+      "epoch": 0.4174573055028463,
+      "grad_norm": 4.103157997131348,
+      "learning_rate": 9.32881873735299e-05,
+      "loss": 2.6523,
+      "step": 220
+    },
+    {
+      "epoch": 0.42125237191650855,
+      "grad_norm": 3.6402695178985596,
+      "learning_rate": 9.313027078004903e-05,
+      "loss": 2.5287,
+      "step": 222
+    },
+    {
+      "epoch": 0.4250474383301708,
+      "grad_norm": 5.530549049377441,
+      "learning_rate": 9.297065487763462e-05,
+      "loss": 2.6326,
+      "step": 224
+    },
+    {
+      "epoch": 0.428842504743833,
+      "grad_norm": 4.8087310791015625,
+      "learning_rate": 9.280934595506536e-05,
+      "loss": 2.4784,
+      "step": 226
+    },
+    {
+      "epoch": 0.43263757115749524,
+      "grad_norm": 6.805646896362305,
+      "learning_rate": 9.264635036782405e-05,
+      "loss": 2.5274,
+      "step": 228
+    },
+    {
+      "epoch": 0.4364326375711575,
+      "grad_norm": 4.533606052398682,
+      "learning_rate": 9.248167453784711e-05,
+      "loss": 2.5572,
+      "step": 230
+    },
+    {
+      "epoch": 0.44022770398481975,
+      "grad_norm": 4.55278205871582,
+      "learning_rate": 9.231532495327165e-05,
+      "loss": 2.5945,
+      "step": 232
+    },
+    {
+      "epoch": 0.444022770398482,
+      "grad_norm": 2.897782802581787,
+      "learning_rate": 9.21473081681797e-05,
+      "loss": 2.6323,
+      "step": 234
+    },
+    {
+      "epoch": 0.4478178368121442,
+      "grad_norm": 3.3823184967041016,
+      "learning_rate": 9.197763080234019e-05,
+      "loss": 2.4394,
+      "step": 236
+    },
+    {
+      "epoch": 0.45161290322580644,
+      "grad_norm": 4.916838645935059,
+      "learning_rate": 9.180629954094792e-05,
+      "loss": 2.514,
+      "step": 238
+    },
+    {
+      "epoch": 0.45540796963946867,
+      "grad_norm": 4.822211265563965,
+      "learning_rate": 9.163332113436032e-05,
+      "loss": 2.439,
+      "step": 240
+    },
+    {
+      "epoch": 0.45920303605313095,
+      "grad_norm": 4.708934783935547,
+      "learning_rate": 9.145870239783142e-05,
+      "loss": 2.5608,
+      "step": 242
+    },
+    {
+      "epoch": 0.4629981024667932,
+      "grad_norm": 2.962392568588257,
+      "learning_rate": 9.128245021124334e-05,
+      "loss": 2.5684,
+      "step": 244
+    },
+    {
+      "epoch": 0.4667931688804554,
+      "grad_norm": 2.4564027786254883,
+      "learning_rate": 9.110457151883523e-05,
+      "loss": 2.4745,
+      "step": 246
+    },
+    {
+      "epoch": 0.47058823529411764,
+      "grad_norm": 2.94209623336792,
+      "learning_rate": 9.092507332892968e-05,
+      "loss": 2.4497,
+      "step": 248
+    },
+    {
+      "epoch": 0.47438330170777987,
+      "grad_norm": 2.5642683506011963,
+      "learning_rate": 9.07439627136566e-05,
+      "loss": 2.4282,
+      "step": 250
+    },
+    {
+      "epoch": 0.4781783681214421,
+      "grad_norm": 2.714101791381836,
+      "learning_rate": 9.056124680867457e-05,
+      "loss": 2.3797,
+      "step": 252
+    },
+    {
+      "epoch": 0.4819734345351044,
+      "grad_norm": 2.010376453399658,
+      "learning_rate": 9.037693281288969e-05,
+      "loss": 2.4963,
+      "step": 254
+    },
+    {
+      "epoch": 0.4857685009487666,
+      "grad_norm": 2.4395999908447266,
+      "learning_rate": 9.019102798817197e-05,
+      "loss": 2.4386,
+      "step": 256
+    },
+    {
+      "epoch": 0.48956356736242884,
+      "grad_norm": 2.391928195953369,
+      "learning_rate": 9.000353965906917e-05,
+      "loss": 2.5203,
+      "step": 258
+    },
+    {
+      "epoch": 0.49335863377609107,
+      "grad_norm": 2.7171072959899902,
+      "learning_rate": 8.981447521251831e-05,
+      "loss": 2.4105,
+      "step": 260
+    },
+    {
+      "epoch": 0.4971537001897533,
+      "grad_norm": 3.4616568088531494,
+      "learning_rate": 8.962384209755452e-05,
+      "loss": 2.2693,
+      "step": 262
+    },
+    {
+      "epoch": 0.5009487666034156,
+      "grad_norm": 3.0641255378723145,
+      "learning_rate": 8.943164782501765e-05,
+      "loss": 2.4696,
+      "step": 264
+    },
+    {
+      "epoch": 0.5047438330170778,
+      "grad_norm": 2.5761353969573975,
+      "learning_rate": 8.923789996725624e-05,
+      "loss": 2.3747,
+      "step": 266
+    },
+    {
+      "epoch": 0.50853889943074,
+      "grad_norm": 3.255844831466675,
+      "learning_rate": 8.904260615782927e-05,
+      "loss": 2.3592,
+      "step": 268
+    },
+    {
+      "epoch": 0.5123339658444023,
+      "grad_norm": 3.119816303253174,
+      "learning_rate": 8.884577409120535e-05,
+      "loss": 2.404,
+      "step": 270
+    },
+    {
+      "epoch": 0.5161290322580645,
+      "grad_norm": 4.314058780670166,
+      "learning_rate": 8.864741152245963e-05,
+      "loss": 2.3595,
+      "step": 272
+    },
+    {
+      "epoch": 0.5199240986717267,
+      "grad_norm": 3.1923677921295166,
+      "learning_rate": 8.84475262669681e-05,
+      "loss": 2.4839,
+      "step": 274
+    },
+    {
+      "epoch": 0.523719165085389,
+      "grad_norm": 5.698940753936768,
+      "learning_rate": 8.824612620009987e-05,
+      "loss": 2.5254,
+      "step": 276
+    },
+    {
+      "epoch": 0.5275142314990512,
+      "grad_norm": 4.198439598083496,
+      "learning_rate": 8.804321925690672e-05,
+      "loss": 2.4509,
+      "step": 278
+    },
+    {
+      "epoch": 0.5313092979127134,
+      "grad_norm": 3.021700620651245,
+      "learning_rate": 8.783881343181055e-05,
+      "loss": 2.3725,
+      "step": 280
+    },
+    {
+      "epoch": 0.5351043643263758,
+      "grad_norm": 3.3499674797058105,
+      "learning_rate": 8.763291677828838e-05,
+      "loss": 2.2391,
+      "step": 282
+    },
+    {
+      "epoch": 0.538899430740038,
+      "grad_norm": 3.2401561737060547,
+      "learning_rate": 8.742553740855506e-05,
+      "loss": 2.4657,
+      "step": 284
+    },
+    {
+      "epoch": 0.5426944971537002,
+      "grad_norm": 2.584604263305664,
+      "learning_rate": 8.721668349324364e-05,
+      "loss": 2.472,
+      "step": 286
+    },
+    {
+      "epoch": 0.5464895635673624,
+      "grad_norm": 2.86765456199646,
+      "learning_rate": 8.700636326108342e-05,
+      "loss": 2.5246,
+      "step": 288
+    },
+    {
+      "epoch": 0.5502846299810247,
+      "grad_norm": 3.742485284805298,
+      "learning_rate": 8.679458499857582e-05,
+      "loss": 2.3875,
+      "step": 290
+    },
+    {
+      "epoch": 0.5540796963946869,
+      "grad_norm": 2.3582913875579834,
+      "learning_rate": 8.658135704966786e-05,
+      "loss": 2.3662,
+      "step": 292
+    },
+    {
+      "epoch": 0.5578747628083491,
+      "grad_norm": 2.188462495803833,
+      "learning_rate": 8.636668781542336e-05,
+      "loss": 2.2918,
+      "step": 294
+    },
+    {
+      "epoch": 0.5616698292220114,
+      "grad_norm": 3.121817111968994,
+      "learning_rate": 8.615058575369202e-05,
+      "loss": 2.3115,
+      "step": 296
+    },
+    {
+      "epoch": 0.5654648956356736,
+      "grad_norm": 3.2461938858032227,
+      "learning_rate": 8.593305937877614e-05,
+      "loss": 2.3089,
+      "step": 298
+    },
+    {
+      "epoch": 0.5692599620493358,
+      "grad_norm": 2.2377259731292725,
+      "learning_rate": 8.571411726109519e-05,
+      "loss": 2.4239,
+      "step": 300
+    },
+    {
+      "epoch": 0.573055028462998,
+      "grad_norm": 2.2074265480041504,
+      "learning_rate": 8.549376802684812e-05,
+      "loss": 2.4559,
+      "step": 302
+    },
+    {
+      "epoch": 0.5768500948766604,
+      "grad_norm": 2.276872396469116,
+      "learning_rate": 8.527202035767349e-05,
+      "loss": 2.4151,
+      "step": 304
+    },
+    {
+      "epoch": 0.5806451612903226,
+      "grad_norm": 2.8724167346954346,
+      "learning_rate": 8.504888299030747e-05,
+      "loss": 2.3258,
+      "step": 306
+    },
+    {
+      "epoch": 0.5844402277039848,
+      "grad_norm": 3.028040647506714,
+      "learning_rate": 8.482436471623951e-05,
+      "loss": 2.3436,
+      "step": 308
+    },
+    {
+      "epoch": 0.5882352941176471,
+      "grad_norm": 2.2716915607452393,
+      "learning_rate": 8.459847438136605e-05,
+      "loss": 2.207,
+      "step": 310
+    },
+    {
+      "epoch": 0.5920303605313093,
+      "grad_norm": 2.3537518978118896,
+      "learning_rate": 8.437122088564198e-05,
+      "loss": 2.2379,
+      "step": 312
+    },
+    {
+      "epoch": 0.5958254269449715,
+      "grad_norm": 2.7615747451782227,
+      "learning_rate": 8.414261318272996e-05,
+      "loss": 2.353,
+      "step": 314
+    },
+    {
+      "epoch": 0.5996204933586338,
+      "grad_norm": 2.191373586654663,
+      "learning_rate": 8.391266027964771e-05,
+      "loss": 2.406,
+      "step": 316
+    },
+    {
+      "epoch": 0.603415559772296,
+      "grad_norm": 2.1172335147857666,
+      "learning_rate": 8.368137123641302e-05,
+      "loss": 2.287,
+      "step": 318
+    },
+    {
+      "epoch": 0.6072106261859582,
+      "grad_norm": 2.387768030166626,
+      "learning_rate": 8.344875516568695e-05,
+      "loss": 2.3713,
+      "step": 320
+    },
+    {
+      "epoch": 0.6110056925996205,
+      "grad_norm": 2.8590333461761475,
+      "learning_rate": 8.321482123241464e-05,
+      "loss": 2.2492,
+      "step": 322
+    },
+    {
+      "epoch": 0.6148007590132827,
+      "grad_norm": 2.2510037422180176,
+      "learning_rate": 8.297957865346437e-05,
+      "loss": 2.3075,
+      "step": 324
+    },
+    {
+      "epoch": 0.618595825426945,
+      "grad_norm": 2.747525930404663,
+      "learning_rate": 8.274303669726426e-05,
+      "loss": 2.089,
+      "step": 326
+    },
+    {
+      "epoch": 0.6223908918406073,
+      "grad_norm": 2.4523816108703613,
+      "learning_rate": 8.250520468343722e-05,
+      "loss": 2.2978,
+      "step": 328
+    },
+    {
+      "epoch": 0.6261859582542695,
+      "grad_norm": 2.1454994678497314,
+      "learning_rate": 8.226609198243372e-05,
+      "loss": 2.3123,
+      "step": 330
+    },
+    {
+      "epoch": 0.6299810246679317,
+      "grad_norm": 2.733635663986206,
+      "learning_rate": 8.202570801516256e-05,
+      "loss": 2.2186,
+      "step": 332
+    },
+    {
+      "epoch": 0.6337760910815939,
+      "grad_norm": 2.4969663619995117,
+      "learning_rate": 8.178406225261981e-05,
+      "loss": 2.2239,
+      "step": 334
+    },
+    {
+      "epoch": 0.6375711574952562,
+      "grad_norm": 3.0612945556640625,
+      "learning_rate": 8.15411642155155e-05,
+      "loss": 2.2033,
+      "step": 336
+    },
+    {
+      "epoch": 0.6413662239089184,
+      "grad_norm": 3.4063143730163574,
+      "learning_rate": 8.129702347389865e-05,
+      "loss": 2.3211,
+      "step": 338
+    },
+    {
+      "epoch": 0.6451612903225806,
+      "grad_norm": 2.2077441215515137,
+      "learning_rate": 8.105164964678009e-05,
+      "loss": 2.5048,
+      "step": 340
+    },
+    {
+      "epoch": 0.6489563567362429,
+      "grad_norm": 2.7867558002471924,
+      "learning_rate": 8.080505240175363e-05,
+      "loss": 2.3034,
+      "step": 342
+    },
+    {
+      "epoch": 0.6527514231499051,
+      "grad_norm": 2.510138511657715,
+      "learning_rate": 8.055724145461495e-05,
+      "loss": 2.1329,
+      "step": 344
+    },
+    {
+      "epoch": 0.6565464895635673,
+      "grad_norm": 2.863159418106079,
+      "learning_rate": 8.030822656897902e-05,
+      "loss": 2.2401,
+      "step": 346
+    },
+    {
+      "epoch": 0.6603415559772297,
+      "grad_norm": 2.664177179336548,
+      "learning_rate": 8.005801755589532e-05,
+      "loss": 2.1028,
+      "step": 348
+    },
+    {
+      "epoch": 0.6641366223908919,
+      "grad_norm": 2.095933675765991,
+      "learning_rate": 7.980662427346127e-05,
+      "loss": 1.9901,
+      "step": 350
+    },
+    {
+      "epoch": 0.6679316888045541,
+      "grad_norm": 2.5369749069213867,
+      "learning_rate": 7.955405662643384e-05,
+      "loss": 2.0314,
+      "step": 352
+    },
+    {
+      "epoch": 0.6717267552182163,
+      "grad_norm": 2.400635242462158,
+      "learning_rate": 7.930032456583931e-05,
+      "loss": 2.2097,
+      "step": 354
+    },
+    {
+      "epoch": 0.6755218216318786,
+      "grad_norm": 2.1800007820129395,
+      "learning_rate": 7.904543808858127e-05,
+      "loss": 2.1188,
+      "step": 356
+    },
+    {
+      "epoch": 0.6793168880455408,
+      "grad_norm": 2.3743057250976562,
+      "learning_rate": 7.878940723704664e-05,
+      "loss": 2.1496,
+      "step": 358
+    },
+    {
+      "epoch": 0.683111954459203,
+      "grad_norm": 2.9855995178222656,
+      "learning_rate": 7.853224209871007e-05,
+      "loss": 2.2357,
+      "step": 360
+    },
+    {
+      "epoch": 0.6869070208728653,
+      "grad_norm": 2.130502700805664,
+      "learning_rate": 7.82739528057365e-05,
+      "loss": 2.1077,
+      "step": 362
+    },
+    {
+      "epoch": 0.6907020872865275,
+      "grad_norm": 2.2894301414489746,
+      "learning_rate": 7.801454953458193e-05,
+      "loss": 2.1231,
+      "step": 364
+    },
+    {
+      "epoch": 0.6944971537001897,
+      "grad_norm": 2.3042802810668945,
+      "learning_rate": 7.775404250559249e-05,
+      "loss": 2.3711,
+      "step": 366
+    },
+    {
+      "epoch": 0.698292220113852,
+      "grad_norm": 2.517493724822998,
+      "learning_rate": 7.749244198260175e-05,
+      "loss": 2.2039,
+      "step": 368
+    },
+    {
+      "epoch": 0.7020872865275142,
+      "grad_norm": 2.50030255317688,
+      "learning_rate": 7.722975827252638e-05,
+      "loss": 2.3868,
+      "step": 370
+    },
+    {
+      "epoch": 0.7058823529411765,
+      "grad_norm": 2.252354621887207,
+      "learning_rate": 7.696600172495997e-05,
+      "loss": 2.2204,
+      "step": 372
+    },
+    {
+      "epoch": 0.7096774193548387,
+      "grad_norm": 2.182471990585327,
+      "learning_rate": 7.670118273176534e-05,
+      "loss": 2.2011,
+      "step": 374
+    },
+    {
+      "epoch": 0.713472485768501,
+      "grad_norm": 2.4250662326812744,
+      "learning_rate": 7.643531172666513e-05,
+      "loss": 2.2251,
+      "step": 376
+    },
+    {
+      "epoch": 0.7172675521821632,
+      "grad_norm": 2.91322922706604,
+      "learning_rate": 7.616839918483061e-05,
+      "loss": 2.2859,
+      "step": 378
+    },
+    {
+      "epoch": 0.7210626185958254,
+      "grad_norm": 2.963437080383301,
+      "learning_rate": 7.590045562246902e-05,
+      "loss": 2.1708,
+      "step": 380
+    },
+    {
+      "epoch": 0.7248576850094877,
+      "grad_norm": 3.132415771484375,
+      "learning_rate": 7.563149159640929e-05,
+      "loss": 2.1548,
+      "step": 382
+    },
+    {
+      "epoch": 0.7286527514231499,
+      "grad_norm": 2.5823440551757812,
+      "learning_rate": 7.5361517703686e-05,
+      "loss": 2.1451,
+      "step": 384
+    },
+    {
+      "epoch": 0.7324478178368121,
+      "grad_norm": 2.6822919845581055,
+      "learning_rate": 7.509054458112202e-05,
+      "loss": 2.3556,
+      "step": 386
+    },
+    {
+      "epoch": 0.7362428842504743,
+      "grad_norm": 2.1513512134552,
+      "learning_rate": 7.481858290490917e-05,
+      "loss": 2.1135,
+      "step": 388
+    },
+    {
+      "epoch": 0.7400379506641366,
+      "grad_norm": 2.6822762489318848,
+      "learning_rate": 7.45456433901879e-05,
+      "loss": 2.2004,
+      "step": 390
+    },
+    {
+      "epoch": 0.7438330170777988,
+      "grad_norm": 2.8601059913635254,
+      "learning_rate": 7.427173679062484e-05,
+      "loss": 2.187,
+      "step": 392
+    },
+    {
+      "epoch": 0.7476280834914611,
+      "grad_norm": 2.706089496612549,
+      "learning_rate": 7.399687389798933e-05,
+      "loss": 2.2509,
+      "step": 394
+    },
+    {
+      "epoch": 0.7514231499051234,
+      "grad_norm": 3.9849417209625244,
+      "learning_rate": 7.372106554172802e-05,
+      "loss": 2.1557,
+      "step": 396
+    },
+    {
+      "epoch": 0.7552182163187856,
+      "grad_norm": 2.7597997188568115,
+      "learning_rate": 7.344432258853841e-05,
+      "loss": 2.1147,
+      "step": 398
+    },
+    {
+      "epoch": 0.7590132827324478,
+      "grad_norm": 2.172595500946045,
+      "learning_rate": 7.316665594194053e-05,
+      "loss": 2.2463,
+      "step": 400
+    },
+    {
+      "epoch": 0.7628083491461101,
+      "grad_norm": 1.9581433534622192,
+      "learning_rate": 7.288807654184747e-05,
+      "loss": 2.1798,
+      "step": 402
+    },
+    {
+      "epoch": 0.7666034155597723,
+      "grad_norm": 2.4689390659332275,
+      "learning_rate": 7.260859536413429e-05,
+      "loss": 2.1782,
+      "step": 404
+    },
+    {
+      "epoch": 0.7703984819734345,
+      "grad_norm": 3.047780752182007,
+      "learning_rate": 7.232822342020557e-05,
+      "loss": 2.0098,
+      "step": 406
+    },
+    {
+      "epoch": 0.7741935483870968,
+      "grad_norm": 3.0638253688812256,
+      "learning_rate": 7.204697175656165e-05,
+      "loss": 2.2135,
+      "step": 408
+    },
+    {
+      "epoch": 0.777988614800759,
+      "grad_norm": 2.14133882522583,
+      "learning_rate": 7.176485145436325e-05,
+      "loss": 2.1974,
+      "step": 410
+    },
+    {
+      "epoch": 0.7817836812144212,
+      "grad_norm": 3.42946720123291,
+      "learning_rate": 7.148187362899505e-05,
+      "loss": 2.2132,
+      "step": 412
+    },
+    {
+      "epoch": 0.7855787476280834,
+      "grad_norm": 2.9244472980499268,
+      "learning_rate": 7.119804942962762e-05,
+      "loss": 1.9828,
+      "step": 414
+    },
+    {
+      "epoch": 0.7893738140417458,
+      "grad_norm": 2.4446115493774414,
+      "learning_rate": 7.091339003877826e-05,
+      "loss": 2.2205,
+      "step": 416
+    },
+    {
+      "epoch": 0.793168880455408,
+      "grad_norm": 2.8077917098999023,
+      "learning_rate": 7.062790667187029e-05,
+      "loss": 2.0921,
+      "step": 418
+    },
+    {
+      "epoch": 0.7969639468690702,
+      "grad_norm": 2.168386697769165,
+      "learning_rate": 7.034161057679127e-05,
+      "loss": 2.1198,
+      "step": 420
+    },
+    {
+      "epoch": 0.8007590132827325,
+      "grad_norm": 2.11982798576355,
+      "learning_rate": 7.005451303344979e-05,
+      "loss": 2.0766,
+      "step": 422
+    },
+    {
+      "epoch": 0.8045540796963947,
+      "grad_norm": 2.1477482318878174,
+      "learning_rate": 6.976662535333107e-05,
+      "loss": 2.0088,
+      "step": 424
+    },
+    {
+      "epoch": 0.8083491461100569,
+      "grad_norm": 2.2626473903656006,
+      "learning_rate": 6.947795887905127e-05,
+      "loss": 1.974,
+      "step": 426
+    },
+    {
+      "epoch": 0.8121442125237192,
+      "grad_norm": 2.2598705291748047,
+      "learning_rate": 6.918852498391063e-05,
+      "loss": 2.0463,
+      "step": 428
+    },
+    {
+      "epoch": 0.8159392789373814,
+      "grad_norm": 2.132086992263794,
+      "learning_rate": 6.889833507144532e-05,
+      "loss": 2.2507,
+      "step": 430
+    },
+    {
+      "epoch": 0.8197343453510436,
+      "grad_norm": 2.307889223098755,
+      "learning_rate": 6.860740057497823e-05,
+      "loss": 2.1272,
+      "step": 432
+    },
+    {
+      "epoch": 0.8235294117647058,
+      "grad_norm": 2.638432264328003,
+      "learning_rate": 6.831573295716837e-05,
+      "loss": 2.1772,
+      "step": 434
+    },
+    {
+      "epoch": 0.8273244781783681,
+      "grad_norm": 2.552748680114746,
+      "learning_rate": 6.802334370955941e-05,
+      "loss": 2.0331,
+      "step": 436
+    },
+    {
+      "epoch": 0.8311195445920304,
+      "grad_norm": 2.7252237796783447,
+      "learning_rate": 6.773024435212678e-05,
+      "loss": 2.1394,
+      "step": 438
+    },
+    {
+      "epoch": 0.8349146110056926,
+      "grad_norm": 2.4146969318389893,
+      "learning_rate": 6.743644643282388e-05,
+      "loss": 2.2107,
+      "step": 440
+    },
+    {
+      "epoch": 0.8387096774193549,
+      "grad_norm": 2.2211036682128906,
+      "learning_rate": 6.714196152712704e-05,
+      "loss": 2.2715,
+      "step": 442
+    },
+    {
+      "epoch": 0.8425047438330171,
+      "grad_norm": 2.484332323074341,
+      "learning_rate": 6.684680123757949e-05,
+      "loss": 2.0996,
+      "step": 444
+    },
+    {
+      "epoch": 0.8462998102466793,
+      "grad_norm": 2.478273868560791,
+      "learning_rate": 6.65509771933342e-05,
+      "loss": 1.9356,
+      "step": 446
+    },
+    {
+      "epoch": 0.8500948766603416,
+      "grad_norm": 2.3699731826782227,
+      "learning_rate": 6.625450104969572e-05,
+      "loss": 2.1316,
+      "step": 448
+    },
+    {
+      "epoch": 0.8538899430740038,
+      "grad_norm": 2.419955253601074,
+      "learning_rate": 6.595738448766095e-05,
+      "loss": 2.1218,
+      "step": 450
+    },
+    {
+      "epoch": 0.857685009487666,
+      "grad_norm": 2.293102979660034,
+      "learning_rate": 6.565963921345895e-05,
+      "loss": 2.0795,
+      "step": 452
+    },
+    {
+      "epoch": 0.8614800759013282,
+      "grad_norm": 2.706804037094116,
+      "learning_rate": 6.536127695808964e-05,
+      "loss": 2.0764,
+      "step": 454
+    },
+    {
+      "epoch": 0.8652751423149905,
+      "grad_norm": 2.3462438583374023,
+      "learning_rate": 6.506230947686172e-05,
+      "loss": 2.0065,
+      "step": 456
+    },
+    {
+      "epoch": 0.8690702087286527,
+      "grad_norm": 2.636672258377075,
+      "learning_rate": 6.47627485489294e-05,
+      "loss": 2.1865,
+      "step": 458
+    },
+    {
+      "epoch": 0.872865275142315,
+      "grad_norm": 2.1746952533721924,
+      "learning_rate": 6.446260597682839e-05,
+      "loss": 1.9698,
+      "step": 460
+    },
+    {
+      "epoch": 0.8766603415559773,
+      "grad_norm": 2.1193501949310303,
+      "learning_rate": 6.416189358601088e-05,
+      "loss": 1.9559,
+      "step": 462
+    },
+    {
+      "epoch": 0.8804554079696395,
+      "grad_norm": 2.4427428245544434,
+      "learning_rate": 6.386062322437954e-05,
+      "loss": 1.9442,
+      "step": 464
+    },
+    {
+      "epoch": 0.8842504743833017,
+      "grad_norm": 2.3021249771118164,
+      "learning_rate": 6.355880676182086e-05,
+      "loss": 2.1573,
+      "step": 466
+    },
+    {
+      "epoch": 0.888045540796964,
+      "grad_norm": 2.224827527999878,
+      "learning_rate": 6.325645608973735e-05,
+      "loss": 2.0547,
+      "step": 468
+    },
+    {
+      "epoch": 0.8918406072106262,
+      "grad_norm": 2.451848030090332,
+      "learning_rate": 6.295358312057914e-05,
+      "loss": 2.0589,
+      "step": 470
+    },
+    {
+      "epoch": 0.8956356736242884,
+      "grad_norm": 2.2723612785339355,
+      "learning_rate": 6.26501997873745e-05,
+      "loss": 1.9392,
+      "step": 472
+    },
+    {
+      "epoch": 0.8994307400379506,
+      "grad_norm": 2.340785503387451,
+      "learning_rate": 6.234631804325981e-05,
+      "loss": 2.1117,
+      "step": 474
+    },
+    {
+      "epoch": 0.9032258064516129,
+      "grad_norm": 2.963451385498047,
+      "learning_rate": 6.204194986100857e-05,
+      "loss": 2.0476,
+      "step": 476
+    },
+    {
+      "epoch": 0.9070208728652751,
+      "grad_norm": 2.2216341495513916,
+      "learning_rate": 6.173710723255966e-05,
+      "loss": 2.154,
+      "step": 478
+    },
+    {
+      "epoch": 0.9108159392789373,
+      "grad_norm": 1.8229964971542358,
+      "learning_rate": 6.143180216854487e-05,
+      "loss": 1.8867,
+      "step": 480
+    },
+    {
+      "epoch": 0.9146110056925996,
+      "grad_norm": 2.198394775390625,
+      "learning_rate": 6.112604669781572e-05,
+      "loss": 2.0622,
+      "step": 482
+    },
+    {
+      "epoch": 0.9184060721062619,
+      "grad_norm": 1.7389253377914429,
+      "learning_rate": 6.081985286696949e-05,
+      "loss": 2.1547,
+      "step": 484
+    },
+    {
+      "epoch": 0.9222011385199241,
+      "grad_norm": 2.4822840690612793,
+      "learning_rate": 6.051323273987463e-05,
+      "loss": 1.9973,
+      "step": 486
+    },
+    {
+      "epoch": 0.9259962049335864,
+      "grad_norm": 2.384826183319092,
+      "learning_rate": 6.020619839719538e-05,
+      "loss": 2.0659,
+      "step": 488
+    },
+    {
+      "epoch": 0.9297912713472486,
+      "grad_norm": 1.8725968599319458,
+      "learning_rate": 5.989876193591589e-05,
+      "loss": 2.082,
+      "step": 490
+    },
+    {
+      "epoch": 0.9335863377609108,
+      "grad_norm": 2.7569420337677,
+      "learning_rate": 5.959093546886356e-05,
+      "loss": 2.0654,
+      "step": 492
+    },
+    {
+      "epoch": 0.937381404174573,
+      "grad_norm": 1.96099853515625,
+      "learning_rate": 5.928273112423177e-05,
+      "loss": 1.9152,
+      "step": 494
+    },
+    {
+      "epoch": 0.9411764705882353,
+      "grad_norm": 2.6206164360046387,
+      "learning_rate": 5.897416104510211e-05,
+      "loss": 2.0493,
+      "step": 496
+    },
+    {
+      "epoch": 0.9449715370018975,
+      "grad_norm": 1.8731147050857544,
+      "learning_rate": 5.866523738896587e-05,
+      "loss": 2.0,
+      "step": 498
+    },
+    {
+      "epoch": 0.9487666034155597,
+      "grad_norm": 2.0109100341796875,
+      "learning_rate": 5.835597232724511e-05,
+      "loss": 2.0603,
+      "step": 500
+    },
+    {
+      "epoch": 0.952561669829222,
+      "grad_norm": 2.412135362625122,
+      "learning_rate": 5.804637804481306e-05,
+      "loss": 2.0034,
+      "step": 502
+    },
+    {
+      "epoch": 0.9563567362428842,
+      "grad_norm": 1.894960880279541,
+      "learning_rate": 5.773646673951406e-05,
+      "loss": 2.0708,
+      "step": 504
+    },
+    {
+      "epoch": 0.9601518026565465,
+      "grad_norm": 3.9883475303649902,
+      "learning_rate": 5.742625062168303e-05,
+      "loss": 2.1407,
+      "step": 506
+    },
+    {
+      "epoch": 0.9639468690702088,
+      "grad_norm": 2.3427770137786865,
+      "learning_rate": 5.7115741913664264e-05,
+      "loss": 1.9919,
+      "step": 508
+    },
+    {
+      "epoch": 0.967741935483871,
+      "grad_norm": 2.0752058029174805,
+      "learning_rate": 5.680495284933e-05,
+      "loss": 2.0542,
+      "step": 510
+    },
+    {
+      "epoch": 0.9715370018975332,
+      "grad_norm": 1.9299733638763428,
+      "learning_rate": 5.649389567359831e-05,
+      "loss": 2.0434,
+      "step": 512
+    },
+    {
+      "epoch": 0.9753320683111955,
+      "grad_norm": 2.1606550216674805,
+      "learning_rate": 5.6182582641950764e-05,
+      "loss": 1.9045,
+      "step": 514
+    },
+    {
+      "epoch": 0.9791271347248577,
+      "grad_norm": 1.9549524784088135,
+      "learning_rate": 5.58710260199495e-05,
+      "loss": 1.9951,
+      "step": 516
+    },
+    {
+      "epoch": 0.9829222011385199,
+      "grad_norm": 1.8001773357391357,
+      "learning_rate": 5.555923808275395e-05,
+      "loss": 1.8519,
+      "step": 518
+    },
+    {
+      "epoch": 0.9867172675521821,
+      "grad_norm": 2.168572187423706,
+      "learning_rate": 5.5247231114637256e-05,
+      "loss": 1.8828,
+      "step": 520
+    },
+    {
+      "epoch": 0.9905123339658444,
+      "grad_norm": 2.0529568195343018,
+      "learning_rate": 5.4935017408502274e-05,
+      "loss": 2.0863,
+      "step": 522
+    },
+    {
+      "epoch": 0.9943074003795066,
+      "grad_norm": 2.078700304031372,
+      "learning_rate": 5.462260926539722e-05,
+      "loss": 2.0463,
+      "step": 524
+    },
+    {
+      "epoch": 0.9981024667931688,
+      "grad_norm": 2.1271212100982666,
+      "learning_rate": 5.431001899403098e-05,
+      "loss": 2.0645,
+      "step": 526
+    },
+    {
+      "epoch": 1.0018975332068312,
+      "grad_norm": 1.7772611379623413,
+      "learning_rate": 5.399725891028824e-05,
+      "loss": 1.7716,
+      "step": 528
+    },
+    {
+      "epoch": 1.0056925996204933,
+      "grad_norm": 1.8182088136672974,
+      "learning_rate": 5.368434133674417e-05,
+      "loss": 1.7655,
+      "step": 530
+    },
+    {
+      "epoch": 1.0094876660341556,
+      "grad_norm": 1.876814842224121,
+      "learning_rate": 5.3371278602179e-05,
+      "loss": 1.8835,
+      "step": 532
+    },
+    {
+      "epoch": 1.0132827324478177,
+      "grad_norm": 2.333099603652954,
+      "learning_rate": 5.305808304109214e-05,
+      "loss": 1.8472,
+      "step": 534
+    },
+    {
+      "epoch": 1.01707779886148,
+      "grad_norm": 2.7806167602539062,
+      "learning_rate": 5.274476699321638e-05,
+      "loss": 1.813,
+      "step": 536
+    },
+    {
+      "epoch": 1.0208728652751424,
+      "grad_norm": 1.8221896886825562,
+      "learning_rate": 5.243134280303156e-05,
+      "loss": 1.7012,
+      "step": 538
+    },
+    {
+      "epoch": 1.0246679316888045,
+      "grad_norm": 2.117246150970459,
+      "learning_rate": 5.211782281927832e-05,
+      "loss": 1.9813,
+      "step": 540
+    },
+    {
+      "epoch": 1.0284629981024669,
+      "grad_norm": 2.018366813659668,
+      "learning_rate": 5.1804219394471456e-05,
+      "loss": 1.8759,
+      "step": 542
+    },
+    {
+      "epoch": 1.032258064516129,
+      "grad_norm": 1.881903886795044,
+      "learning_rate": 5.149054488441333e-05,
+      "loss": 1.9816,
+      "step": 544
+    },
+    {
+      "epoch": 1.0360531309297913,
+      "grad_norm": 1.7794609069824219,
+      "learning_rate": 5.117681164770704e-05,
+      "loss": 1.7131,
+      "step": 546
+    },
+    {
+      "epoch": 1.0398481973434535,
+      "grad_norm": 2.2241082191467285,
+      "learning_rate": 5.086303204526943e-05,
+      "loss": 1.7683,
+      "step": 548
+    },
+    {
+      "epoch": 1.0436432637571158,
+      "grad_norm": 2.8065714836120605,
+      "learning_rate": 5.054921843984418e-05,
+      "loss": 1.7091,
+      "step": 550
+    },
+    {
+      "epoch": 1.047438330170778,
+      "grad_norm": 2.32865571975708,
+      "learning_rate": 5.023538319551465e-05,
+      "loss": 1.7517,
+      "step": 552
+    },
+    {
+      "epoch": 1.0512333965844403,
+      "grad_norm": 1.928222894668579,
+      "learning_rate": 4.9921538677216736e-05,
+      "loss": 1.8336,
+      "step": 554
+    },
+    {
+      "epoch": 1.0550284629981024,
+      "grad_norm": 2.499162435531616,
+      "learning_rate": 4.9607697250251786e-05,
+      "loss": 1.833,
+      "step": 556
+    },
+    {
+      "epoch": 1.0588235294117647,
+      "grad_norm": 2.649643898010254,
+      "learning_rate": 4.929387127979927e-05,
+      "loss": 1.8476,
+      "step": 558
+    },
+    {
+      "epoch": 1.0626185958254268,
+      "grad_norm": 1.7383689880371094,
+      "learning_rate": 4.898007313042975e-05,
+      "loss": 1.8564,
+      "step": 560
+    },
+    {
+      "epoch": 1.0664136622390892,
+      "grad_norm": 2.4569990634918213,
+      "learning_rate": 4.866631516561759e-05,
+      "loss": 1.7267,
+      "step": 562
+    },
+    {
+      "epoch": 1.0702087286527515,
+      "grad_norm": 2.2858664989471436,
+      "learning_rate": 4.835260974725397e-05,
+      "loss": 1.9215,
+      "step": 564
+    },
+    {
+      "epoch": 1.0740037950664136,
+      "grad_norm": 1.9371929168701172,
+      "learning_rate": 4.803896923515974e-05,
+      "loss": 1.776,
+      "step": 566
+    },
+    {
+      "epoch": 1.077798861480076,
+      "grad_norm": 1.8845841884613037,
+      "learning_rate": 4.772540598659844e-05,
+      "loss": 1.9276,
+      "step": 568
+    },
+    {
+      "epoch": 1.081593927893738,
+      "grad_norm": 2.1376278400421143,
+      "learning_rate": 4.741193235578952e-05,
+      "loss": 1.7701,
+      "step": 570
+    },
+    {
+      "epoch": 1.0853889943074004,
+      "grad_norm": 1.7806552648544312,
+      "learning_rate": 4.7098560693421565e-05,
+      "loss": 1.7934,
+      "step": 572
+    },
+    {
+      "epoch": 1.0891840607210626,
+      "grad_norm": 1.7085773944854736,
+      "learning_rate": 4.678530334616557e-05,
+      "loss": 1.8279,
+      "step": 574
+    },
+    {
+      "epoch": 1.092979127134725,
+      "grad_norm": 2.3024978637695312,
+      "learning_rate": 4.647217265618866e-05,
+      "loss": 1.7992,
+      "step": 576
+    },
+    {
+      "epoch": 1.096774193548387,
+      "grad_norm": 2.357330322265625,
+      "learning_rate": 4.615918096066766e-05,
+      "loss": 1.7864,
+      "step": 578
+    },
+    {
+      "epoch": 1.1005692599620494,
+      "grad_norm": 1.9522347450256348,
+      "learning_rate": 4.584634059130317e-05,
+      "loss": 1.7091,
+      "step": 580
+    },
+    {
+      "epoch": 1.1043643263757117,
+      "grad_norm": 1.706804633140564,
+      "learning_rate": 4.553366387383357e-05,
+      "loss": 1.7207,
+      "step": 582
+    },
+    {
+      "epoch": 1.1081593927893738,
+      "grad_norm": 2.086026668548584,
+      "learning_rate": 4.522116312754944e-05,
+      "loss": 1.905,
+      "step": 584
+    },
+    {
+      "epoch": 1.1119544592030361,
+      "grad_norm": 1.659766435623169,
+      "learning_rate": 4.4908850664808245e-05,
+      "loss": 1.8103,
+      "step": 586
+    },
+    {
+      "epoch": 1.1157495256166983,
+      "grad_norm": 1.8237513303756714,
+      "learning_rate": 4.4596738790549114e-05,
+      "loss": 1.897,
+      "step": 588
+    },
+    {
+      "epoch": 1.1195445920303606,
+      "grad_norm": 2.0152337551116943,
+      "learning_rate": 4.428483980180814e-05,
+      "loss": 1.7799,
+      "step": 590
+    },
+    {
+      "epoch": 1.1233396584440227,
+      "grad_norm": 2.138798952102661,
+      "learning_rate": 4.397316598723385e-05,
+      "loss": 1.8348,
+      "step": 592
+    },
+    {
+      "epoch": 1.127134724857685,
+      "grad_norm": 1.8744384050369263,
+      "learning_rate": 4.366172962660301e-05,
+      "loss": 1.8539,
+      "step": 594
+    },
+    {
+      "epoch": 1.1309297912713472,
+      "grad_norm": 1.8008323907852173,
+      "learning_rate": 4.335054299033686e-05,
+      "loss": 1.8384,
+      "step": 596
+    },
+    {
+      "epoch": 1.1347248576850095,
+      "grad_norm": 1.9751280546188354,
+      "learning_rate": 4.3039618339017595e-05,
+      "loss": 1.7037,
+      "step": 598
+    },
+    {
+      "epoch": 1.1385199240986716,
+      "grad_norm": 2.2664952278137207,
+      "learning_rate": 4.272896792290537e-05,
+      "loss": 1.8883,
+      "step": 600
+    },
+    {
+      "epoch": 1.142314990512334,
+      "grad_norm": 1.7477809190750122,
+      "learning_rate": 4.241860398145565e-05,
+      "loss": 1.7767,
+      "step": 602
+    },
+    {
+      "epoch": 1.146110056925996,
+      "grad_norm": 1.8783535957336426,
+      "learning_rate": 4.2108538742836906e-05,
+      "loss": 1.8748,
+      "step": 604
+    },
+    {
+      "epoch": 1.1499051233396584,
+      "grad_norm": 2.1925675868988037,
+      "learning_rate": 4.179878442344892e-05,
+      "loss": 1.9109,
+      "step": 606
+    },
+    {
+      "epoch": 1.1537001897533208,
+      "grad_norm": 1.8848633766174316,
+      "learning_rate": 4.14893532274414e-05,
+      "loss": 1.7008,
+      "step": 608
+    },
+    {
+      "epoch": 1.157495256166983,
+      "grad_norm": 1.8942062854766846,
+      "learning_rate": 4.1180257346233186e-05,
+      "loss": 1.6562,
+      "step": 610
+    },
+    {
+      "epoch": 1.1612903225806452,
+      "grad_norm": 2.463782787322998,
+      "learning_rate": 4.0871508958031924e-05,
+      "loss": 1.8924,
+      "step": 612
+    },
+    {
+      "epoch": 1.1650853889943074,
+      "grad_norm": 2.1543052196502686,
+      "learning_rate": 4.056312022735417e-05,
+      "loss": 1.7368,
+      "step": 614
+    },
+    {
+      "epoch": 1.1688804554079697,
+      "grad_norm": 1.9905352592468262,
+      "learning_rate": 4.025510330454621e-05,
+      "loss": 1.7367,
+      "step": 616
+    },
+    {
+      "epoch": 1.1726755218216318,
+      "grad_norm": 2.0409960746765137,
+      "learning_rate": 3.994747032530532e-05,
+      "loss": 2.0239,
+      "step": 618
+    },
+    {
+      "epoch": 1.1764705882352942,
+      "grad_norm": 2.049666404724121,
+      "learning_rate": 3.9640233410201553e-05,
+      "loss": 1.6878,
+      "step": 620
+    },
+    {
+      "epoch": 1.1802656546489563,
+      "grad_norm": 1.962066411972046,
+      "learning_rate": 3.933340466420032e-05,
+      "loss": 1.7774,
+      "step": 622
+    },
+    {
+      "epoch": 1.1840607210626186,
+      "grad_norm": 1.9107617139816284,
+      "learning_rate": 3.902699617618531e-05,
+      "loss": 1.7755,
+      "step": 624
+    },
+    {
+      "epoch": 1.187855787476281,
+      "grad_norm": 1.8392717838287354,
+      "learning_rate": 3.872102001848238e-05,
+      "loss": 1.7025,
+      "step": 626
+    },
+    {
+      "epoch": 1.191650853889943,
+      "grad_norm": 1.8730952739715576,
+      "learning_rate": 3.841548824638376e-05,
+      "loss": 1.8045,
+      "step": 628
+    },
+    {
+      "epoch": 1.1954459203036052,
+      "grad_norm": 1.7919378280639648,
+      "learning_rate": 3.8110412897673096e-05,
+      "loss": 1.7641,
+      "step": 630
+    },
+    {
+      "epoch": 1.1992409867172675,
+      "grad_norm": 2.541914939880371,
+      "learning_rate": 3.7805805992151284e-05,
+      "loss": 1.723,
+      "step": 632
+    },
+    {
+      "epoch": 1.2030360531309299,
+      "grad_norm": 1.7635172605514526,
+      "learning_rate": 3.750167953116272e-05,
+      "loss": 1.8093,
+      "step": 634
+    },
+    {
+      "epoch": 1.206831119544592,
+      "grad_norm": 1.9383831024169922,
+      "learning_rate": 3.719804549712265e-05,
+      "loss": 1.7736,
+      "step": 636
+    },
+    {
+      "epoch": 1.2106261859582543,
+      "grad_norm": 2.1055490970611572,
+      "learning_rate": 3.689491585304491e-05,
+      "loss": 1.6286,
+      "step": 638
+    },
+    {
+      "epoch": 1.2144212523719164,
+      "grad_norm": 2.0924720764160156,
+      "learning_rate": 3.6592302542070624e-05,
+      "loss": 1.7897,
+      "step": 640
+    },
+    {
+      "epoch": 1.2182163187855788,
+      "grad_norm": 2.072434425354004,
+      "learning_rate": 3.629021748699777e-05,
+      "loss": 1.8087,
+      "step": 642
+    },
+    {
+      "epoch": 1.222011385199241,
+      "grad_norm": 1.8711074590682983,
+      "learning_rate": 3.598867258981122e-05,
+      "loss": 1.8447,
+      "step": 644
+    },
+    {
+      "epoch": 1.2258064516129032,
+      "grad_norm": 1.9876536130905151,
+      "learning_rate": 3.5687679731214016e-05,
+      "loss": 1.7634,
+      "step": 646
+    },
+    {
+      "epoch": 1.2296015180265654,
+      "grad_norm": 1.8552366495132446,
+      "learning_rate": 3.538725077015915e-05,
+      "loss": 1.7555,
+      "step": 648
+    },
+    {
+      "epoch": 1.2333965844402277,
+      "grad_norm": 2.034203290939331,
+      "learning_rate": 3.5087397543382326e-05,
+      "loss": 1.8566,
+      "step": 650
+    },
+    {
+      "epoch": 1.23719165085389,
+      "grad_norm": 2.132357358932495,
+      "learning_rate": 3.478813186493569e-05,
+      "loss": 1.7685,
+      "step": 652
+    },
+    {
+      "epoch": 1.2409867172675522,
+      "grad_norm": 1.622307538986206,
+      "learning_rate": 3.448946552572229e-05,
+      "loss": 1.8548,
+      "step": 654
+    },
+    {
+      "epoch": 1.2447817836812145,
+      "grad_norm": 1.8897305727005005,
+      "learning_rate": 3.419141029303149e-05,
+      "loss": 1.5911,
+      "step": 656
+    },
+    {
+      "epoch": 1.2485768500948766,
+      "grad_norm": 1.7892314195632935,
+      "learning_rate": 3.389397791007548e-05,
+      "loss": 1.7389,
+      "step": 658
+    },
+    {
+      "epoch": 1.252371916508539,
+      "grad_norm": 1.8299717903137207,
+      "learning_rate": 3.3597180095526434e-05,
+      "loss": 1.7719,
+      "step": 660
+    },
+    {
+      "epoch": 1.256166982922201,
+      "grad_norm": 1.807536244392395,
+      "learning_rate": 3.3301028543054935e-05,
+      "loss": 1.8034,
+      "step": 662
+    },
+    {
+      "epoch": 1.2599620493358634,
+      "grad_norm": 1.920238733291626,
+      "learning_rate": 3.3005534920869176e-05,
+      "loss": 1.8372,
+      "step": 664
+    },
+    {
+      "epoch": 1.2637571157495255,
+      "grad_norm": 2.0501198768615723,
+      "learning_rate": 3.271071087125522e-05,
+      "loss": 1.6749,
+      "step": 666
+    },
+    {
+      "epoch": 1.2675521821631879,
+      "grad_norm": 1.8483867645263672,
+      "learning_rate": 3.2416568010118435e-05,
+      "loss": 1.8974,
+      "step": 668
+    },
+    {
+      "epoch": 1.2713472485768502,
+      "grad_norm": 1.8510785102844238,
+      "learning_rate": 3.212311792652564e-05,
+      "loss": 1.7007,
+      "step": 670
+    },
+    {
+      "epoch": 1.2751423149905123,
+      "grad_norm": 1.8836129903793335,
+      "learning_rate": 3.183037218224862e-05,
+      "loss": 1.9495,
+      "step": 672
+    },
+    {
+      "epoch": 1.2789373814041745,
+      "grad_norm": 1.859937071800232,
+      "learning_rate": 3.153834231130866e-05,
+      "loss": 1.8093,
+      "step": 674
+    },
+    {
+      "epoch": 1.2827324478178368,
+      "grad_norm": 1.8940603733062744,
+      "learning_rate": 3.124703981952191e-05,
+      "loss": 1.7062,
+      "step": 676
+    },
+    {
+      "epoch": 1.2865275142314991,
+      "grad_norm": 1.9347773790359497,
+      "learning_rate": 3.0956476184046275e-05,
+      "loss": 1.6123,
+      "step": 678
+    },
+    {
+      "epoch": 1.2903225806451613,
+      "grad_norm": 1.9208083152770996,
+      "learning_rate": 3.066666285292906e-05,
+      "loss": 1.7506,
+      "step": 680
+    },
+    {
+      "epoch": 1.2941176470588236,
+      "grad_norm": 1.7094935178756714,
+      "learning_rate": 3.037761124465604e-05,
+      "loss": 1.8333,
+      "step": 682
+    },
+    {
+      "epoch": 1.2979127134724857,
+      "grad_norm": 1.676564335823059,
+      "learning_rate": 3.0089332747701527e-05,
+      "loss": 1.6712,
+      "step": 684
+    },
+    {
+      "epoch": 1.301707779886148,
+      "grad_norm": 2.0832326412200928,
+      "learning_rate": 2.9801838720079633e-05,
+      "loss": 1.8002,
+      "step": 686
+    },
+    {
+      "epoch": 1.3055028462998102,
+      "grad_norm": 1.8611640930175781,
+      "learning_rate": 2.9515140488896847e-05,
+      "loss": 1.7021,
+      "step": 688
+    },
+    {
+      "epoch": 1.3092979127134725,
+      "grad_norm": 1.9767136573791504,
+      "learning_rate": 2.9229249349905684e-05,
+      "loss": 1.7101,
+      "step": 690
+    },
+    {
+      "epoch": 1.3130929791271346,
+      "grad_norm": 1.8870338201522827,
+      "learning_rate": 2.8944176567059706e-05,
+      "loss": 1.7298,
+      "step": 692
+    },
+    {
+      "epoch": 1.316888045540797,
+      "grad_norm": 1.785990834236145,
+      "learning_rate": 2.86599333720697e-05,
+      "loss": 1.8476,
+      "step": 694
+    },
+    {
+      "epoch": 1.3206831119544593,
+      "grad_norm": 1.8696433305740356,
+      "learning_rate": 2.837653096396108e-05,
+      "loss": 1.6997,
+      "step": 696
+    },
+    {
+      "epoch": 1.3244781783681214,
+      "grad_norm": 1.8706103563308716,
+      "learning_rate": 2.809398050863279e-05,
+      "loss": 1.8175,
+      "step": 698
+    },
+    {
+      "epoch": 1.3282732447817835,
+      "grad_norm": 1.8281570672988892,
+      "learning_rate": 2.7812293138417312e-05,
+      "loss": 1.6426,
+      "step": 700
+    },
+    {
+      "epoch": 1.3320683111954459,
+      "grad_norm": 1.893720030784607,
+      "learning_rate": 2.7531479951641924e-05,
+      "loss": 1.7798,
+      "step": 702
+    },
+    {
+      "epoch": 1.3358633776091082,
+      "grad_norm": 1.8282567262649536,
+      "learning_rate": 2.7251552012191762e-05,
+      "loss": 1.863,
+      "step": 704
+    },
+    {
+      "epoch": 1.3396584440227703,
+      "grad_norm": 1.6841830015182495,
+      "learning_rate": 2.697252034907351e-05,
+      "loss": 1.7531,
+      "step": 706
+    },
+    {
+      "epoch": 1.3434535104364327,
+      "grad_norm": 1.837632179260254,
+      "learning_rate": 2.6694395955981143e-05,
+      "loss": 1.7933,
+      "step": 708
+    },
+    {
+      "epoch": 1.3472485768500948,
+      "grad_norm": 2.0122556686401367,
+      "learning_rate": 2.641718979086277e-05,
+      "loss": 1.8014,
+      "step": 710
+    },
+    {
+      "epoch": 1.3510436432637571,
+      "grad_norm": 1.693137288093567,
+      "learning_rate": 2.614091277548864e-05,
+      "loss": 1.7138,
+      "step": 712
+    },
+    {
+      "epoch": 1.3548387096774195,
+      "grad_norm": 2.4528486728668213,
+      "learning_rate": 2.5865575795021218e-05,
+      "loss": 1.8396,
+      "step": 714
+    },
+    {
+      "epoch": 1.3586337760910816,
+      "grad_norm": 1.9499150514602661,
+      "learning_rate": 2.559118969758595e-05,
+      "loss": 1.6793,
+      "step": 716
+    },
+    {
+      "epoch": 1.3624288425047437,
+      "grad_norm": 1.8878819942474365,
+      "learning_rate": 2.531776529384407e-05,
+      "loss": 1.7692,
+      "step": 718
+    },
+    {
+      "epoch": 1.366223908918406,
+      "grad_norm": 1.8431209325790405,
+      "learning_rate": 2.504531335656668e-05,
+      "loss": 1.7763,
+      "step": 720
+    },
+    {
+      "epoch": 1.3700189753320684,
+      "grad_norm": 1.5216484069824219,
+      "learning_rate": 2.4773844620210118e-05,
+      "loss": 1.608,
+      "step": 722
+    },
+    {
+      "epoch": 1.3738140417457305,
+      "grad_norm": 1.7268508672714233,
+      "learning_rate": 2.450336978049322e-05,
+      "loss": 1.752,
+      "step": 724
+    },
+    {
+      "epoch": 1.3776091081593929,
+      "grad_norm": 1.669382095336914,
+      "learning_rate": 2.423389949397582e-05,
+      "loss": 1.6645,
+      "step": 726
+    },
+    {
+      "epoch": 1.381404174573055,
+      "grad_norm": 1.9102509021759033,
+      "learning_rate": 2.3965444377638906e-05,
+      "loss": 1.7733,
+      "step": 728
+    },
+    {
+      "epoch": 1.3851992409867173,
+      "grad_norm": 1.9738460779190063,
+      "learning_rate": 2.3698015008466372e-05,
+      "loss": 1.6467,
+      "step": 730
+    },
+    {
+      "epoch": 1.3889943074003794,
+      "grad_norm": 1.833115816116333,
+      "learning_rate": 2.3431621923028145e-05,
+      "loss": 1.7282,
+      "step": 732
+    },
+    {
+      "epoch": 1.3927893738140418,
+      "grad_norm": 1.8798232078552246,
+      "learning_rate": 2.3166275617065213e-05,
+      "loss": 1.8703,
+      "step": 734
+    },
+    {
+      "epoch": 1.396584440227704,
+      "grad_norm": 1.78863525390625,
+      "learning_rate": 2.2901986545076e-05,
+      "loss": 1.8465,
+      "step": 736
+    },
+    {
+      "epoch": 1.4003795066413662,
+      "grad_norm": 1.7265403270721436,
+      "learning_rate": 2.26387651199045e-05,
+      "loss": 1.8277,
+      "step": 738
+    },
+    {
+      "epoch": 1.4041745730550286,
+      "grad_norm": 1.874920129776001,
+      "learning_rate": 2.2376621712330015e-05,
+      "loss": 1.7515,
+      "step": 740
+    },
+    {
+      "epoch": 1.4079696394686907,
+      "grad_norm": 1.729573130607605,
+      "learning_rate": 2.2115566650658536e-05,
+      "loss": 1.6828,
+      "step": 742
+    },
+    {
+      "epoch": 1.4117647058823528,
+      "grad_norm": 1.8121289014816284,
+      "learning_rate": 2.185561022031582e-05,
+      "loss": 1.7124,
+      "step": 744
+    },
+    {
+      "epoch": 1.4155597722960152,
+      "grad_norm": 1.6570336818695068,
+      "learning_rate": 2.1596762663442218e-05,
+      "loss": 1.7264,
+      "step": 746
+    },
+    {
+      "epoch": 1.4193548387096775,
+      "grad_norm": 3.0217766761779785,
+      "learning_rate": 2.1339034178488964e-05,
+      "loss": 1.66,
+      "step": 748
+    },
+    {
+      "epoch": 1.4231499051233396,
+      "grad_norm": 1.637597680091858,
+      "learning_rate": 2.1082434919816557e-05,
+      "loss": 1.6517,
+      "step": 750
+    },
+    {
+      "epoch": 1.426944971537002,
+      "grad_norm": 1.7600562572479248,
+      "learning_rate": 2.0826974997294574e-05,
+      "loss": 1.7542,
+      "step": 752
+    },
+    {
+      "epoch": 1.430740037950664,
+      "grad_norm": 1.7863155603408813,
+      "learning_rate": 2.0572664475903386e-05,
+      "loss": 1.5719,
+      "step": 754
+    },
+    {
+      "epoch": 1.4345351043643264,
+      "grad_norm": 1.7714210748672485,
+      "learning_rate": 2.0319513375337578e-05,
+      "loss": 1.6499,
+      "step": 756
+    },
+    {
+      "epoch": 1.4383301707779887,
+      "grad_norm": 2.219492197036743,
+      "learning_rate": 2.0067531669611206e-05,
+      "loss": 1.7686,
+      "step": 758
+    },
+    {
+      "epoch": 1.4421252371916509,
+      "grad_norm": 1.8768947124481201,
+      "learning_rate": 1.9816729286664798e-05,
+      "loss": 1.7189,
+      "step": 760
+    },
+    {
+      "epoch": 1.445920303605313,
+      "grad_norm": 2.026461601257324,
+      "learning_rate": 1.9567116107974232e-05,
+      "loss": 1.7267,
+      "step": 762
+    },
+    {
+      "epoch": 1.4497153700189753,
+      "grad_norm": 1.8134167194366455,
+      "learning_rate": 1.931870196816138e-05,
+      "loss": 1.6333,
+      "step": 764
+    },
+    {
+      "epoch": 1.4535104364326377,
+      "grad_norm": 1.9153342247009277,
+      "learning_rate": 1.907149665460664e-05,
+      "loss": 1.645,
+      "step": 766
+    },
+    {
+      "epoch": 1.4573055028462998,
+      "grad_norm": 1.8665708303451538,
+      "learning_rate": 1.8825509907063327e-05,
+      "loss": 1.7551,
+      "step": 768
+    },
+    {
+      "epoch": 1.4611005692599621,
+      "grad_norm": 2.1076242923736572,
+      "learning_rate": 1.8580751417273928e-05,
+      "loss": 1.7156,
+      "step": 770
+    },
+    {
+      "epoch": 1.4648956356736242,
+      "grad_norm": 1.7849063873291016,
+      "learning_rate": 1.833723082858825e-05,
+      "loss": 1.6047,
+      "step": 772
+    },
+    {
+      "epoch": 1.4686907020872866,
+      "grad_norm": 3.0725789070129395,
+      "learning_rate": 1.8094957735583463e-05,
+      "loss": 1.7424,
+      "step": 774
+    },
+    {
+      "epoch": 1.4724857685009487,
+      "grad_norm": 1.8258955478668213,
+      "learning_rate": 1.7853941683686114e-05,
+      "loss": 1.656,
+      "step": 776
+    },
+    {
+      "epoch": 1.476280834914611,
+      "grad_norm": 1.6560546159744263,
+      "learning_rate": 1.761419216879601e-05,
+      "loss": 1.8144,
+      "step": 778
+    },
+    {
+      "epoch": 1.4800759013282732,
+      "grad_norm": 1.7182631492614746,
+      "learning_rate": 1.7375718636912103e-05,
+      "loss": 1.5824,
+      "step": 780
+    },
+    {
+      "epoch": 1.4838709677419355,
+      "grad_norm": 1.7388415336608887,
+      "learning_rate": 1.7138530483760314e-05,
+      "loss": 1.7111,
+      "step": 782
+    },
+    {
+      "epoch": 1.4876660341555978,
+      "grad_norm": 1.8411345481872559,
+      "learning_rate": 1.690263705442334e-05,
+      "loss": 1.6691,
+      "step": 784
+    },
+    {
+      "epoch": 1.49146110056926,
+      "grad_norm": 1.7131088972091675,
+      "learning_rate": 1.6668047642972494e-05,
+      "loss": 1.7838,
+      "step": 786
+    },
+    {
+      "epoch": 1.495256166982922,
+      "grad_norm": 1.9504882097244263,
+      "learning_rate": 1.6434771492101485e-05,
+      "loss": 1.747,
+      "step": 788
+    },
+    {
+      "epoch": 1.4990512333965844,
+      "grad_norm": 1.534936547279358,
+      "learning_rate": 1.6202817792762282e-05,
+      "loss": 1.8543,
+      "step": 790
+    },
+    {
+      "epoch": 1.5028462998102468,
+      "grad_norm": 1.6542677879333496,
+      "learning_rate": 1.5972195683802992e-05,
+      "loss": 1.9799,
+      "step": 792
+    },
+    {
+      "epoch": 1.5066413662239089,
+      "grad_norm": 1.5913175344467163,
+      "learning_rate": 1.5742914251607793e-05,
+      "loss": 1.6424,
+      "step": 794
+    },
+    {
+      "epoch": 1.510436432637571,
+      "grad_norm": 1.8262730836868286,
+      "learning_rate": 1.5514982529738946e-05,
+      "loss": 1.788,
+      "step": 796
+    },
+    {
+      "epoch": 1.5142314990512333,
+      "grad_norm": 1.6641391515731812,
+      "learning_rate": 1.5288409498580824e-05,
+      "loss": 1.854,
+      "step": 798
+    },
+    {
+      "epoch": 1.5180265654648957,
+      "grad_norm": 1.7485277652740479,
+      "learning_rate": 1.506320408498621e-05,
+      "loss": 1.7398,
+      "step": 800
+    },
+    {
+      "epoch": 1.521821631878558,
+      "grad_norm": 1.7265832424163818,
+      "learning_rate": 1.4839375161924446e-05,
+      "loss": 1.7167,
+      "step": 802
+    },
+    {
+      "epoch": 1.5256166982922201,
+      "grad_norm": 1.7332998514175415,
+      "learning_rate": 1.4616931548131929e-05,
+      "loss": 1.6337,
+      "step": 804
+    },
+    {
+      "epoch": 1.5294117647058822,
+      "grad_norm": 1.5418668985366821,
+      "learning_rate": 1.4395882007764644e-05,
+      "loss": 1.5121,
+      "step": 806
+    },
+    {
+      "epoch": 1.5332068311195446,
+      "grad_norm": 1.712519884109497,
+      "learning_rate": 1.4176235250052788e-05,
+      "loss": 1.8379,
+      "step": 808
+    },
+    {
+      "epoch": 1.537001897533207,
+      "grad_norm": 1.718359351158142,
+      "learning_rate": 1.3957999928957787e-05,
+      "loss": 1.7405,
+      "step": 810
+    },
+    {
+      "epoch": 1.540796963946869,
+      "grad_norm": 1.679713249206543,
+      "learning_rate": 1.3741184642831189e-05,
+      "loss": 1.6968,
+      "step": 812
+    },
+    {
+      "epoch": 1.5445920303605312,
+      "grad_norm": 1.5460269451141357,
+      "learning_rate": 1.3525797934075912e-05,
+      "loss": 1.6123,
+      "step": 814
+    },
+    {
+      "epoch": 1.5483870967741935,
+      "grad_norm": 1.6191962957382202,
+      "learning_rate": 1.3311848288809813e-05,
+      "loss": 1.7576,
+      "step": 816
+    },
+    {
+      "epoch": 1.5521821631878558,
+      "grad_norm": 1.6308246850967407,
+      "learning_rate": 1.3099344136531105e-05,
+      "loss": 1.7816,
+      "step": 818
+    },
+    {
+      "epoch": 1.5559772296015182,
+      "grad_norm": 1.646103024482727,
+      "learning_rate": 1.2888293849786503e-05,
+      "loss": 1.6321,
+      "step": 820
+    },
+    {
+      "epoch": 1.5597722960151803,
+      "grad_norm": 1.7340853214263916,
+      "learning_rate": 1.2678705743841152e-05,
+      "loss": 1.6944,
+      "step": 822
+    },
+    {
+      "epoch": 1.5635673624288424,
+      "grad_norm": 1.631166696548462,
+      "learning_rate": 1.2470588076351036e-05,
+      "loss": 1.5587,
+      "step": 824
+    },
+    {
+      "epoch": 1.5673624288425048,
+      "grad_norm": 1.665603756904602,
+      "learning_rate": 1.2263949047037786e-05,
+      "loss": 1.5882,
+      "step": 826
+    },
+    {
+      "epoch": 1.571157495256167,
+      "grad_norm": 1.6381057500839233,
+      "learning_rate": 1.2058796797365423e-05,
+      "loss": 1.5904,
+      "step": 828
+    },
+    {
+      "epoch": 1.5749525616698292,
+      "grad_norm": 1.5420763492584229,
+      "learning_rate": 1.1855139410219657e-05,
+      "loss": 1.6369,
+      "step": 830
+    },
+    {
+      "epoch": 1.5787476280834913,
+      "grad_norm": 1.6090302467346191,
+      "learning_rate": 1.1652984909589515e-05,
+      "loss": 1.6626,
+      "step": 832
+    },
+    {
+      "epoch": 1.5825426944971537,
+      "grad_norm": 1.7544772624969482,
+      "learning_rate": 1.145234126025102e-05,
+      "loss": 1.654,
+      "step": 834
+    },
+    {
+      "epoch": 1.586337760910816,
+      "grad_norm": 1.609275221824646,
+      "learning_rate": 1.1253216367453578e-05,
+      "loss": 1.6235,
+      "step": 836
+    },
+    {
+      "epoch": 1.5901328273244781,
+      "grad_norm": 1.7230455875396729,
+      "learning_rate": 1.1055618076608381e-05,
+      "loss": 1.6404,
+      "step": 838
+    },
+    {
+      "epoch": 1.5939278937381403,
+      "grad_norm": 1.7173190116882324,
+      "learning_rate": 1.085955417297932e-05,
+      "loss": 1.629,
+      "step": 840
+    },
+    {
+      "epoch": 1.5977229601518026,
+      "grad_norm": 1.5825388431549072,
+      "learning_rate": 1.0665032381376338e-05,
+      "loss": 1.6566,
+      "step": 842
+    },
+    {
+      "epoch": 1.601518026565465,
+      "grad_norm": 1.640567421913147,
+      "learning_rate": 1.047206036585095e-05,
+      "loss": 1.6087,
+      "step": 844
+    },
+    {
+      "epoch": 1.6053130929791273,
+      "grad_norm": 1.9455435276031494,
+      "learning_rate": 1.0280645729394366e-05,
+      "loss": 1.7938,
+      "step": 846
+    },
+    {
+      "epoch": 1.6091081593927894,
+      "grad_norm": 1.6935428380966187,
+      "learning_rate": 1.0090796013637965e-05,
+      "loss": 1.7387,
+      "step": 848
+    },
+    {
+      "epoch": 1.6129032258064515,
+      "grad_norm": 1.7314420938491821,
+      "learning_rate": 9.902518698556018e-06,
+      "loss": 1.6033,
+      "step": 850
+    },
+    {
+      "epoch": 1.6166982922201139,
+      "grad_norm": 1.6285747289657593,
+      "learning_rate": 9.715821202171178e-06,
+      "loss": 1.7201,
+      "step": 852
+    },
+    {
+      "epoch": 1.6204933586337762,
+      "grad_norm": 1.6489043235778809,
+      "learning_rate": 9.530710880262023e-06,
+      "loss": 1.6788,
+      "step": 854
+    },
+    {
+      "epoch": 1.6242884250474383,
+      "grad_norm": 1.6721750497817993,
+      "learning_rate": 9.347195026073369e-06,
+      "loss": 1.6642,
+      "step": 856
+    },
+    {
+      "epoch": 1.6280834914611004,
+      "grad_norm": 1.8302478790283203,
+      "learning_rate": 9.16528087002892e-06,
+      "loss": 1.79,
+      "step": 858
+    },
+    {
+      "epoch": 1.6318785578747628,
+      "grad_norm": 1.5789343118667603,
+      "learning_rate": 8.98497557944627e-06,
+      "loss": 1.6721,
+      "step": 860
+    },
+    {
+      "epoch": 1.635673624288425,
+      "grad_norm": 1.4637460708618164,
+      "learning_rate": 8.806286258254664e-06,
+      "loss": 1.5916,
+      "step": 862
+    },
+    {
+      "epoch": 1.6394686907020875,
+      "grad_norm": 1.703425407409668,
+      "learning_rate": 8.62921994671501e-06,
+      "loss": 1.8469,
+      "step": 864
+    },
+    {
+      "epoch": 1.6432637571157496,
+      "grad_norm": 1.585274338722229,
+      "learning_rate": 8.453783621142542e-06,
+      "loss": 1.6635,
+      "step": 866
+    },
+    {
+      "epoch": 1.6470588235294117,
+      "grad_norm": 1.7186620235443115,
+      "learning_rate": 8.279984193631967e-06,
+      "loss": 1.6559,
+      "step": 868
+    },
+    {
+      "epoch": 1.650853889943074,
+      "grad_norm": 1.6305322647094727,
+      "learning_rate": 8.10782851178506e-06,
+      "loss": 1.5428,
+      "step": 870
+    },
+    {
+      "epoch": 1.6546489563567364,
+      "grad_norm": 1.6739540100097656,
+      "learning_rate": 7.937323358440935e-06,
+      "loss": 1.6924,
+      "step": 872
+    },
+    {
+      "epoch": 1.6584440227703985,
+      "grad_norm": 1.6253348588943481,
+      "learning_rate": 7.768475451408847e-06,
+      "loss": 1.6914,
+      "step": 874
+    },
+    {
+      "epoch": 1.6622390891840606,
+      "grad_norm": 1.6650753021240234,
+      "learning_rate": 7.601291443203373e-06,
+      "loss": 1.7853,
+      "step": 876
+    },
+    {
+      "epoch": 1.666034155597723,
+      "grad_norm": 1.617315411567688,
+      "learning_rate": 7.435777920782444e-06,
+      "loss": 1.6255,
+      "step": 878
+    },
+    {
+      "epoch": 1.6698292220113853,
+      "grad_norm": 1.6971100568771362,
+      "learning_rate": 7.271941405287763e-06,
+      "loss": 1.5836,
+      "step": 880
+    },
+    {
+      "epoch": 1.6736242884250474,
+      "grad_norm": 1.6293379068374634,
+      "learning_rate": 7.109788351787866e-06,
+      "loss": 1.7387,
+      "step": 882
+    },
+    {
+      "epoch": 1.6774193548387095,
+      "grad_norm": 1.6460471153259277,
+      "learning_rate": 6.949325149023861e-06,
+      "loss": 1.6863,
+      "step": 884
+    },
+    {
+      "epoch": 1.6812144212523719,
+      "grad_norm": 1.654508352279663,
+      "learning_rate": 6.790558119157597e-06,
+      "loss": 1.6287,
+      "step": 886
+    },
+    {
+      "epoch": 1.6850094876660342,
+      "grad_norm": 1.5705487728118896,
+      "learning_rate": 6.633493517522687e-06,
+      "loss": 1.6695,
+      "step": 888
+    },
+    {
+      "epoch": 1.6888045540796965,
+      "grad_norm": 1.7113432884216309,
+      "learning_rate": 6.478137532378004e-06,
+      "loss": 1.5901,
+      "step": 890
+    },
+    {
+      "epoch": 1.6925996204933587,
+      "grad_norm": 1.6679534912109375,
+      "learning_rate": 6.324496284663867e-06,
+      "loss": 1.578,
+      "step": 892
+    },
+    {
+      "epoch": 1.6963946869070208,
+      "grad_norm": 1.9186716079711914,
+      "learning_rate": 6.172575827760885e-06,
+      "loss": 1.7412,
+      "step": 894
+    },
+    {
+      "epoch": 1.7001897533206831,
+      "grad_norm": 1.471213698387146,
+      "learning_rate": 6.022382147251454e-06,
+      "loss": 1.6411,
+      "step": 896
+    },
+    {
+      "epoch": 1.7039848197343455,
+      "grad_norm": 1.6535768508911133,
+      "learning_rate": 5.873921160683943e-06,
+      "loss": 1.633,
+      "step": 898
+    },
+    {
+      "epoch": 1.7077798861480076,
+      "grad_norm": 1.5547441244125366,
+      "learning_rate": 5.727198717339511e-06,
+      "loss": 1.7347,
+      "step": 900
+    },
+    {
+      "epoch": 1.7115749525616697,
+      "grad_norm": 1.5353294610977173,
+      "learning_rate": 5.582220598001681e-06,
+      "loss": 1.7455,
+      "step": 902
+    },
+    {
+      "epoch": 1.715370018975332,
+      "grad_norm": 1.5471186637878418,
+      "learning_rate": 5.438992514728586e-06,
+      "loss": 1.6082,
+      "step": 904
+    },
+    {
+      "epoch": 1.7191650853889944,
+      "grad_norm": 1.568590521812439,
+      "learning_rate": 5.297520110627868e-06,
+      "loss": 1.7422,
+      "step": 906
+    },
+    {
+      "epoch": 1.7229601518026565,
+      "grad_norm": 1.7015557289123535,
+      "learning_rate": 5.157808959634408e-06,
+      "loss": 1.6793,
+      "step": 908
+    },
+    {
+      "epoch": 1.7267552182163188,
+      "grad_norm": 2.917369842529297,
+      "learning_rate": 5.0198645662906666e-06,
+      "loss": 1.7334,
+      "step": 910
+    },
+    {
+      "epoch": 1.730550284629981,
+      "grad_norm": 1.633530855178833,
+      "learning_rate": 4.88369236552983e-06,
+      "loss": 1.6879,
+      "step": 912
+    },
+    {
+      "epoch": 1.7343453510436433,
+      "grad_norm": 1.6307895183563232,
+      "learning_rate": 4.74929772246166e-06,
+      "loss": 1.7291,
+      "step": 914
+    },
+    {
+      "epoch": 1.7381404174573056,
+      "grad_norm": 1.5775015354156494,
+      "learning_rate": 4.616685932161152e-06,
+      "loss": 1.701,
+      "step": 916
+    },
+    {
+      "epoch": 1.7419354838709677,
+      "grad_norm": 1.5769219398498535,
+      "learning_rate": 4.4858622194598525e-06,
+      "loss": 1.6994,
+      "step": 918
+    },
+    {
+      "epoch": 1.7457305502846299,
+      "grad_norm": 1.5686330795288086,
+      "learning_rate": 4.356831738740053e-06,
+      "loss": 1.7386,
+      "step": 920
+    },
+    {
+      "epoch": 1.7495256166982922,
+      "grad_norm": 1.6501996517181396,
+      "learning_rate": 4.229599573731685e-06,
+      "loss": 1.6726,
+      "step": 922
+    },
+    {
+      "epoch": 1.7533206831119545,
+      "grad_norm": 1.6141939163208008,
+      "learning_rate": 4.1041707373120356e-06,
+      "loss": 1.654,
+      "step": 924
+    },
+    {
+      "epoch": 1.7571157495256167,
+      "grad_norm": 1.5882351398468018,
+      "learning_rate": 3.980550171308228e-06,
+      "loss": 1.6573,
+      "step": 926
+    },
+    {
+      "epoch": 1.7609108159392788,
+      "grad_norm": 1.517470359802246,
+      "learning_rate": 3.858742746302535e-06,
+      "loss": 1.6852,
+      "step": 928
+    },
+    {
+      "epoch": 1.7647058823529411,
+      "grad_norm": 1.597163438796997,
+      "learning_rate": 3.738753261440475e-06,
+      "loss": 1.607,
+      "step": 930
+    },
+    {
+      "epoch": 1.7685009487666035,
+      "grad_norm": 1.59674870967865,
+      "learning_rate": 3.6205864442417136e-06,
+      "loss": 1.7626,
+      "step": 932
+    },
+    {
+      "epoch": 1.7722960151802658,
+      "grad_norm": 1.625994086265564,
+      "learning_rate": 3.5042469504138363e-06,
+      "loss": 1.6589,
+      "step": 934
+    },
+    {
+      "epoch": 1.776091081593928,
+      "grad_norm": 1.5371334552764893,
+      "learning_rate": 3.3897393636688368e-06,
+      "loss": 1.6479,
+      "step": 936
+    },
+    {
+      "epoch": 1.77988614800759,
+      "grad_norm": 1.4965029954910278,
+      "learning_rate": 3.277068195542654e-06,
+      "loss": 1.577,
+      "step": 938
+    },
+    {
+      "epoch": 1.7836812144212524,
+      "grad_norm": 1.6149548292160034,
+      "learning_rate": 3.166237885217299e-06,
+      "loss": 1.8349,
+      "step": 940
+    },
+    {
+      "epoch": 1.7874762808349147,
+      "grad_norm": 1.5945125818252563,
+      "learning_rate": 3.0572527993460053e-06,
+      "loss": 1.6663,
+      "step": 942
+    },
+    {
+      "epoch": 1.7912713472485768,
+      "grad_norm": 1.5761655569076538,
+      "learning_rate": 2.950117231881183e-06,
+      "loss": 1.6158,
+      "step": 944
+    },
+    {
+      "epoch": 1.795066413662239,
+      "grad_norm": 1.7141841650009155,
+      "learning_rate": 2.8448354039052216e-06,
+      "loss": 1.6652,
+      "step": 946
+    },
+    {
+      "epoch": 1.7988614800759013,
+      "grad_norm": 1.666662335395813,
+      "learning_rate": 2.741411463464211e-06,
+      "loss": 1.7688,
+      "step": 948
+    },
+    {
+      "epoch": 1.8026565464895636,
+      "grad_norm": 1.5551278591156006,
+      "learning_rate": 2.639849485404505e-06,
+      "loss": 1.5906,
+      "step": 950
+    },
+    {
+      "epoch": 1.8064516129032258,
+      "grad_norm": 1.698128342628479,
+      "learning_rate": 2.5401534712121146e-06,
+      "loss": 1.7194,
+      "step": 952
+    },
+    {
+      "epoch": 1.810246679316888,
+      "grad_norm": 1.75882887840271,
+      "learning_rate": 2.442327348855161e-06,
+      "loss": 1.6539,
+      "step": 954
+    },
+    {
+      "epoch": 1.8140417457305502,
+      "grad_norm": 1.8165512084960938,
+      "learning_rate": 2.3463749726290286e-06,
+      "loss": 1.6697,
+      "step": 956
+    },
+    {
+      "epoch": 1.8178368121442126,
+      "grad_norm": 1.616357684135437,
+      "learning_rate": 2.252300123004525e-06,
+      "loss": 1.7466,
+      "step": 958
+    },
+    {
+      "epoch": 1.821631878557875,
+      "grad_norm": 1.6569586992263794,
+      "learning_rate": 2.1601065064789704e-06,
+      "loss": 1.5822,
+      "step": 960
+    },
+    {
+      "epoch": 1.825426944971537,
+      "grad_norm": 1.5467514991760254,
+      "learning_rate": 2.069797755430097e-06,
+      "loss": 1.5533,
+      "step": 962
+    },
+    {
+      "epoch": 1.8292220113851991,
+      "grad_norm": 1.6702818870544434,
+      "learning_rate": 1.981377427973019e-06,
+      "loss": 1.7528,
+      "step": 964
+    },
+    {
+      "epoch": 1.8330170777988615,
+      "grad_norm": 1.6937930583953857,
+      "learning_rate": 1.8948490078199764e-06,
+      "loss": 1.7521,
+      "step": 966
+    },
+    {
+      "epoch": 1.8368121442125238,
+      "grad_norm": 3.543177604675293,
+      "learning_rate": 1.8102159041430922e-06,
+      "loss": 1.7019,
+      "step": 968
+    },
+    {
+      "epoch": 1.840607210626186,
+      "grad_norm": 1.5496528148651123,
+      "learning_rate": 1.7274814514400994e-06,
+      "loss": 1.6022,
+      "step": 970
+    },
+    {
+      "epoch": 1.844402277039848,
+      "grad_norm": 1.6045300960540771,
+      "learning_rate": 1.6466489094028759e-06,
+      "loss": 1.6172,
+      "step": 972
+    },
+    {
+      "epoch": 1.8481973434535104,
+      "grad_norm": 1.582979440689087,
+      "learning_rate": 1.5677214627891179e-06,
+      "loss": 1.7938,
+      "step": 974
+    },
+    {
+      "epoch": 1.8519924098671727,
+      "grad_norm": 1.6310341358184814,
+      "learning_rate": 1.4907022212967803e-06,
+      "loss": 1.7066,
+      "step": 976
+    },
+    {
+      "epoch": 1.855787476280835,
+      "grad_norm": 1.6005839109420776,
+      "learning_rate": 1.415594219441585e-06,
+      "loss": 1.5591,
+      "step": 978
+    },
+    {
+      "epoch": 1.8595825426944972,
+      "grad_norm": 1.6014922857284546,
+      "learning_rate": 1.3424004164374837e-06,
+      "loss": 1.6571,
+      "step": 980
+    },
+    {
+      "epoch": 1.8633776091081593,
+      "grad_norm": 1.5527675151824951,
+      "learning_rate": 1.2711236960800287e-06,
+      "loss": 1.5875,
+      "step": 982
+    },
+    {
+      "epoch": 1.8671726755218216,
+      "grad_norm": 1.5006603002548218,
+      "learning_rate": 1.2017668666327753e-06,
+      "loss": 1.5288,
+      "step": 984
+    },
+    {
+      "epoch": 1.870967741935484,
+      "grad_norm": 1.4573849439620972,
+      "learning_rate": 1.1343326607166527e-06,
+      "loss": 1.6196,
+      "step": 986
+    },
+    {
+      "epoch": 1.874762808349146,
+      "grad_norm": 1.4461921453475952,
+      "learning_rate": 1.0688237352022345e-06,
+      "loss": 1.7204,
+      "step": 988
+    },
+    {
+      "epoch": 1.8785578747628082,
+      "grad_norm": 1.6633641719818115,
+      "learning_rate": 1.0052426711051666e-06,
+      "loss": 1.5931,
+      "step": 990
+    },
+    {
+      "epoch": 1.8823529411764706,
+      "grad_norm": 1.6192902326583862,
+      "learning_rate": 9.435919734843645e-07,
+      "loss": 1.5492,
+      "step": 992
+    },
+    {
+      "epoch": 1.886148007590133,
+      "grad_norm": 1.8350443840026855,
+      "learning_rate": 8.83874071343388e-07,
+      "loss": 1.7133,
+      "step": 994
+    },
+    {
+      "epoch": 1.889943074003795,
+      "grad_norm": 1.5247050523757935,
+      "learning_rate": 8.260913175347273e-07,
+      "loss": 1.53,
+      "step": 996
+    },
+    {
+      "epoch": 1.8937381404174574,
+      "grad_norm": 1.5474799871444702,
+      "learning_rate": 7.702459886670788e-07,
+      "loss": 1.7518,
+      "step": 998
+    },
+    {
+      "epoch": 1.8975332068311195,
+      "grad_norm": 1.5862666368484497,
+      "learning_rate": 7.163402850156653e-07,
+      "loss": 1.7613,
+      "step": 1000
+    },
+    {
+      "epoch": 1.9013282732447818,
+      "grad_norm": 1.5844755172729492,
+      "learning_rate": 6.643763304355566e-07,
+      "loss": 1.6749,
+      "step": 1002
+    },
+    {
+      "epoch": 1.9051233396584442,
+      "grad_norm": 1.8030189275741577,
+      "learning_rate": 6.143561722779523e-07,
+      "loss": 1.727,
+      "step": 1004
+    },
+    {
+      "epoch": 1.9089184060721063,
+      "grad_norm": 1.7325502634048462,
+      "learning_rate": 5.662817813095633e-07,
+      "loss": 1.7329,
+      "step": 1006
+    },
+    {
+      "epoch": 1.9127134724857684,
+      "grad_norm": 1.6039164066314697,
+      "learning_rate": 5.201550516349407e-07,
+      "loss": 1.7462,
+      "step": 1008
+    },
+    {
+      "epoch": 1.9165085388994307,
+      "grad_norm": 1.4843109846115112,
+      "learning_rate": 4.7597780062184073e-07,
+      "loss": 1.6243,
+      "step": 1010
+    },
+    {
+      "epoch": 1.920303605313093,
+      "grad_norm": 1.6233309507369995,
+      "learning_rate": 4.337517688296544e-07,
+      "loss": 1.622,
+      "step": 1012
+    },
+    {
+      "epoch": 1.9240986717267552,
+      "grad_norm": 1.6170657873153687,
+      "learning_rate": 3.9347861994078474e-07,
+      "loss": 1.8024,
+      "step": 1014
+    },
+    {
+      "epoch": 1.9278937381404173,
+      "grad_norm": 1.656578779220581,
+      "learning_rate": 3.551599406951434e-07,
+      "loss": 1.8554,
+      "step": 1016
+    },
+    {
+      "epoch": 1.9316888045540797,
+      "grad_norm": 1.888298511505127,
+      "learning_rate": 3.1879724082760076e-07,
+      "loss": 1.7063,
+      "step": 1018
+    },
+    {
+      "epoch": 1.935483870967742,
+      "grad_norm": 1.6982399225234985,
+      "learning_rate": 2.843919530085226e-07,
+      "loss": 1.7374,
+      "step": 1020
+    },
+    {
+      "epoch": 1.9392789373814043,
+      "grad_norm": 1.4739940166473389,
+      "learning_rate": 2.51945432787315e-07,
+      "loss": 1.6844,
+      "step": 1022
+    },
+    {
+      "epoch": 1.9430740037950665,
+      "grad_norm": 1.4712330102920532,
+      "learning_rate": 2.21458958539017e-07,
+      "loss": 1.6772,
+      "step": 1024
+    },
+    {
+      "epoch": 1.9468690702087286,
+      "grad_norm": 1.6379010677337646,
+      "learning_rate": 1.9293373141394122e-07,
+      "loss": 1.807,
+      "step": 1026
+    },
+    {
+      "epoch": 1.950664136622391,
+      "grad_norm": 1.4777188301086426,
+      "learning_rate": 1.6637087529033923e-07,
+      "loss": 1.7502,
+      "step": 1028
+    },
+    {
+      "epoch": 1.9544592030360532,
+      "grad_norm": 1.5368890762329102,
+      "learning_rate": 1.4177143673011484e-07,
+      "loss": 1.7178,
+      "step": 1030
+    },
+    {
+      "epoch": 1.9582542694497154,
+      "grad_norm": 1.5123056173324585,
+      "learning_rate": 1.191363849376237e-07,
+      "loss": 1.6945,
+      "step": 1032
+    },
+    {
+      "epoch": 1.9620493358633775,
+      "grad_norm": 1.4532966613769531,
+      "learning_rate": 9.846661172143723e-08,
+      "loss": 1.6243,
+      "step": 1034
+    },
+    {
+      "epoch": 1.9658444022770398,
+      "grad_norm": 1.576069951057434,
+      "learning_rate": 7.976293145924852e-08,
+      "loss": 1.797,
+      "step": 1036
+    },
+    {
+      "epoch": 1.9696394686907022,
+      "grad_norm": 1.595721960067749,
+      "learning_rate": 6.302608106577568e-08,
+      "loss": 1.6501,
+      "step": 1038
+    },
+    {
+      "epoch": 1.9734345351043643,
+      "grad_norm": 1.6888916492462158,
+      "learning_rate": 4.825671996370185e-08,
+      "loss": 1.6579,
+      "step": 1040
+    },
+    {
+      "epoch": 1.9772296015180264,
+      "grad_norm": 1.7020574808120728,
+      "learning_rate": 3.545543005773477e-08,
+      "loss": 1.7715,
+      "step": 1042
+    },
+    {
+      "epoch": 1.9810246679316887,
+      "grad_norm": 1.5102561712265015,
+      "learning_rate": 2.4622715711647426e-08,
+      "loss": 1.6051,
+      "step": 1044
+    },
+    {
+      "epoch": 1.984819734345351,
+      "grad_norm": 1.3985369205474854,
+      "learning_rate": 1.5759003728427202e-08,
+      "loss": 1.7225,
+      "step": 1046
+    },
+    {
+      "epoch": 1.9886148007590134,
+      "grad_norm": 1.5577125549316406,
+      "learning_rate": 8.864643333450495e-09,
+      "loss": 1.6524,
+      "step": 1048
+    },
+    {
+      "epoch": 1.9924098671726755,
+      "grad_norm": 1.5508955717086792,
+      "learning_rate": 3.9399061607103825e-09,
+      "loss": 1.6419,
+      "step": 1050
+    },
+    {
+      "epoch": 1.9962049335863377,
+      "grad_norm": 1.5975019931793213,
+      "learning_rate": 9.84986242158481e-10,
+      "loss": 1.7808,
+      "step": 1052
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 2.207545280456543,
+      "learning_rate": 0.0,
+      "loss": 1.6642,
+      "step": 1054
+    }
+  ],
+  "logging_steps": 2,
+  "max_steps": 1054,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 2,
+  "save_steps": 527,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.7824370872496947e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoint-1054/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:721e80e1778cdc5974559fb76036f3bd489ad9c204cc63b34e1b4276ae1c1d8e
+size 5432

checkpoint-527/config.json

@@ -0,0 +1,27 @@
+{
+  "_name_or_path": "/mnt/parscratch/users/acp23ay/private/models/OLMo-2-1124-7B-mni_Beng-coli-cw2v/",
+  "architectures": [
+    "Olmo2ForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "eos_token_id": 100257,
+  "hidden_act": "silu",
+  "hidden_size": 4096,
+  "initializer_range": 0.02,
+  "intermediate_size": 11008,
+  "max_position_embeddings": 4096,
+  "model_type": "olmo2",
+  "num_attention_heads": 32,
+  "num_hidden_layers": 32,
+  "num_key_value_heads": 32,
+  "pad_token_id": 100277,
+  "rms_norm_eps": 1e-06,
+  "rope_scaling": null,
+  "rope_theta": 500000,
+  "tie_word_embeddings": false,
+  "torch_dtype": "float32",
+  "transformers_version": "4.48.0.dev0",
+  "use_cache": true,
+  "vocab_size": 100384
+}

checkpoint-527/generation_config.json

@@ -0,0 +1,6 @@
+{
+  "_from_model_config": true,
+  "eos_token_id": 100257,
+  "pad_token_id": 100277,
+  "transformers_version": "4.48.0.dev0"
+}

checkpoint-527/model-00001-of-00006.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4491793e3d7f5d1c9810b9790443465f38ce32985f291173a5b0e8660f8e6b12
+size 4950070472

checkpoint-527/model-00002-of-00006.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:de0b6a357d03676c4986588aaf4dc96afc3f44a0498514961d2b2cfa1940655e
+size 4991622816

checkpoint-527/model-00003-of-00006.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c1038a7d346a681a8fb6949e054c2246f3f3c80e826855c1cec6af1155cc607d
+size 4924546896

checkpoint-527/model-00004-of-00006.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:33c0a21146ace02d36928787ae7f34be868f7c81b894dea98ec7b0e5a3ea01b2
+size 4857404928

checkpoint-527/model-00005-of-00006.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e7f82a3014ccfc0f67208f63785ae4bc71f09aba86db48bf6139d18178abbfed
+size 4857404928

checkpoint-527/model-00006-of-00006.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bc1b19cac6846f3864e657ce812c1a97bf07e9145ad7766476a515e7f4574852
+size 4614509080

checkpoint-527/model.safetensors.index.json

@@ -0,0 +1,362 @@
+{
+  "metadata": {
+    "total_size": 29195517952
+  },
+  "weight_map": {
+    "lm_head.weight": "model-00006-of-00006.safetensors",
+    "model.embed_tokens.weight": "model-00001-of-00006.safetensors",
+    "model.layers.0.mlp.down_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.0.mlp.gate_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.0.mlp.up_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.0.post_attention_layernorm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.0.post_feedforward_layernorm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.0.self_attn.k_norm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.0.self_attn.k_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.0.self_attn.o_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.0.self_attn.q_norm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.0.self_attn.q_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.0.self_attn.v_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.1.mlp.down_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.1.mlp.gate_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.1.mlp.up_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.1.post_attention_layernorm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.1.post_feedforward_layernorm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.1.self_attn.k_norm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.1.self_attn.k_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.1.self_attn.o_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.1.self_attn.q_norm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.1.self_attn.q_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.1.self_attn.v_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.10.mlp.down_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.10.mlp.gate_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.10.mlp.up_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.10.post_attention_layernorm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.10.post_feedforward_layernorm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.10.self_attn.k_norm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.10.self_attn.k_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.10.self_attn.o_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.10.self_attn.q_norm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.10.self_attn.q_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.10.self_attn.v_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.11.mlp.down_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.11.mlp.gate_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.11.mlp.up_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.11.post_attention_layernorm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.11.post_feedforward_layernorm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.11.self_attn.k_norm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.11.self_attn.k_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.11.self_attn.o_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.11.self_attn.q_norm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.11.self_attn.q_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.11.self_attn.v_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.12.mlp.down_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.12.mlp.gate_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.12.mlp.up_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.12.post_attention_layernorm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.12.post_feedforward_layernorm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.12.self_attn.k_norm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.12.self_attn.k_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.12.self_attn.o_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.12.self_attn.q_norm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.12.self_attn.q_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.12.self_attn.v_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.13.mlp.down_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.13.mlp.gate_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.13.mlp.up_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.13.post_attention_layernorm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.13.post_feedforward_layernorm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.13.self_attn.k_norm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.13.self_attn.k_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.13.self_attn.o_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.13.self_attn.q_norm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.13.self_attn.q_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.13.self_attn.v_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.14.mlp.down_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.14.mlp.gate_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.14.mlp.up_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.14.post_attention_layernorm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.14.post_feedforward_layernorm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.14.self_attn.k_norm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.14.self_attn.k_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.14.self_attn.o_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.14.self_attn.q_norm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.14.self_attn.q_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.14.self_attn.v_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.15.mlp.down_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.15.mlp.gate_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.15.mlp.up_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.15.post_attention_layernorm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.15.post_feedforward_layernorm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.15.self_attn.k_norm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.15.self_attn.k_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.15.self_attn.o_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.15.self_attn.q_norm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.15.self_attn.q_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.15.self_attn.v_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.16.mlp.down_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.16.mlp.gate_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.16.mlp.up_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.16.post_attention_layernorm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.16.post_feedforward_layernorm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.16.self_attn.k_norm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.16.self_attn.k_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.16.self_attn.o_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.16.self_attn.q_norm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.16.self_attn.q_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.16.self_attn.v_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.17.mlp.down_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.17.mlp.gate_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.17.mlp.up_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.17.post_attention_layernorm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.17.post_feedforward_layernorm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.17.self_attn.k_norm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.17.self_attn.k_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.17.self_attn.o_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.17.self_attn.q_norm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.17.self_attn.q_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.17.self_attn.v_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.18.mlp.down_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.18.mlp.gate_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.18.mlp.up_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.18.post_attention_layernorm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.18.post_feedforward_layernorm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.18.self_attn.k_norm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.18.self_attn.k_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.18.self_attn.o_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.18.self_attn.q_norm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.18.self_attn.q_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.18.self_attn.v_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.19.mlp.down_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.19.mlp.gate_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.19.mlp.up_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.19.post_attention_layernorm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.19.post_feedforward_layernorm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.19.self_attn.k_norm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.19.self_attn.k_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.19.self_attn.o_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.19.self_attn.q_norm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.19.self_attn.q_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.19.self_attn.v_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.2.mlp.down_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.2.mlp.gate_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.2.mlp.up_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.2.post_attention_layernorm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.2.post_feedforward_layernorm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.2.self_attn.k_norm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.2.self_attn.k_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.2.self_attn.o_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.2.self_attn.q_norm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.2.self_attn.q_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.2.self_attn.v_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.20.mlp.down_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.20.mlp.gate_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.20.mlp.up_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.20.post_attention_layernorm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.20.post_feedforward_layernorm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.20.self_attn.k_norm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.20.self_attn.k_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.20.self_attn.o_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.20.self_attn.q_norm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.20.self_attn.q_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.20.self_attn.v_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.21.mlp.down_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.21.mlp.gate_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.21.mlp.up_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.21.post_attention_layernorm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.21.post_feedforward_layernorm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.21.self_attn.k_norm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.21.self_attn.k_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.21.self_attn.o_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.21.self_attn.q_norm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.21.self_attn.q_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.21.self_attn.v_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.22.mlp.down_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.22.mlp.gate_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.22.mlp.up_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.22.post_attention_layernorm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.22.post_feedforward_layernorm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.22.self_attn.k_norm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.22.self_attn.k_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.22.self_attn.o_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.22.self_attn.q_norm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.22.self_attn.q_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.22.self_attn.v_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.23.mlp.down_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.23.mlp.gate_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.23.mlp.up_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.23.post_attention_layernorm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.23.post_feedforward_layernorm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.23.self_attn.k_norm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.23.self_attn.k_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.23.self_attn.o_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.23.self_attn.q_norm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.23.self_attn.q_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.23.self_attn.v_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.24.mlp.down_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.24.mlp.gate_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.24.mlp.up_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.24.post_attention_layernorm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.24.post_feedforward_layernorm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.24.self_attn.k_norm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.24.self_attn.k_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.24.self_attn.o_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.24.self_attn.q_norm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.24.self_attn.q_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.24.self_attn.v_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.25.mlp.down_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.25.mlp.gate_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.25.mlp.up_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.25.post_attention_layernorm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.25.post_feedforward_layernorm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.25.self_attn.k_norm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.25.self_attn.k_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.25.self_attn.o_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.25.self_attn.q_norm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.25.self_attn.q_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.25.self_attn.v_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.26.mlp.down_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.26.mlp.gate_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.26.mlp.up_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.26.post_attention_layernorm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.26.post_feedforward_layernorm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.26.self_attn.k_norm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.26.self_attn.k_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.26.self_attn.o_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.26.self_attn.q_norm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.26.self_attn.q_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.26.self_attn.v_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.27.mlp.down_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.27.mlp.gate_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.27.mlp.up_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.27.post_attention_layernorm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.27.post_feedforward_layernorm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.27.self_attn.k_norm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.27.self_attn.k_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.27.self_attn.o_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.27.self_attn.q_norm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.27.self_attn.q_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.27.self_attn.v_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.28.mlp.down_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.28.mlp.gate_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.28.mlp.up_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.28.post_attention_layernorm.weight": "model-00006-of-00006.safetensors",
+    "model.layers.28.post_feedforward_layernorm.weight": "model-00006-of-00006.safetensors",
+    "model.layers.28.self_attn.k_norm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.28.self_attn.k_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.28.self_attn.o_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.28.self_attn.q_norm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.28.self_attn.q_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.28.self_attn.v_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.29.mlp.down_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.29.mlp.gate_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.29.mlp.up_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.29.post_attention_layernorm.weight": "model-00006-of-00006.safetensors",
+    "model.layers.29.post_feedforward_layernorm.weight": "model-00006-of-00006.safetensors",
+    "model.layers.29.self_attn.k_norm.weight": "model-00006-of-00006.safetensors",
+    "model.layers.29.self_attn.k_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.29.self_attn.o_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.29.self_attn.q_norm.weight": "model-00006-of-00006.safetensors",
+    "model.layers.29.self_attn.q_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.29.self_attn.v_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.3.mlp.down_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.3.mlp.gate_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.3.mlp.up_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.3.post_attention_layernorm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.3.post_feedforward_layernorm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.3.self_attn.k_norm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.3.self_attn.k_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.3.self_attn.o_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.3.self_attn.q_norm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.3.self_attn.q_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.3.self_attn.v_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.30.mlp.down_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.30.mlp.gate_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.30.mlp.up_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.30.post_attention_layernorm.weight": "model-00006-of-00006.safetensors",
+    "model.layers.30.post_feedforward_layernorm.weight": "model-00006-of-00006.safetensors",
+    "model.layers.30.self_attn.k_norm.weight": "model-00006-of-00006.safetensors",
+    "model.layers.30.self_attn.k_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.30.self_attn.o_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.30.self_attn.q_norm.weight": "model-00006-of-00006.safetensors",
+    "model.layers.30.self_attn.q_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.30.self_attn.v_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.31.mlp.down_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.31.mlp.gate_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.31.mlp.up_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.31.post_attention_layernorm.weight": "model-00006-of-00006.safetensors",
+    "model.layers.31.post_feedforward_layernorm.weight": "model-00006-of-00006.safetensors",
+    "model.layers.31.self_attn.k_norm.weight": "model-00006-of-00006.safetensors",
+    "model.layers.31.self_attn.k_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.31.self_attn.o_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.31.self_attn.q_norm.weight": "model-00006-of-00006.safetensors",
+    "model.layers.31.self_attn.q_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.31.self_attn.v_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.4.mlp.down_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.4.mlp.gate_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.4.mlp.up_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.4.post_attention_layernorm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.4.post_feedforward_layernorm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.4.self_attn.k_norm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.4.self_attn.k_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.4.self_attn.o_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.4.self_attn.q_norm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.4.self_attn.q_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.4.self_attn.v_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.5.mlp.down_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.5.mlp.gate_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.5.mlp.up_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.5.post_attention_layernorm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.5.post_feedforward_layernorm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.5.self_attn.k_norm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.5.self_attn.k_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.5.self_attn.o_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.5.self_attn.q_norm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.5.self_attn.q_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.5.self_attn.v_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.6.mlp.down_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.6.mlp.gate_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.6.mlp.up_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.6.post_attention_layernorm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.6.post_feedforward_layernorm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.6.self_attn.k_norm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.6.self_attn.k_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.6.self_attn.o_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.6.self_attn.q_norm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.6.self_attn.q_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.6.self_attn.v_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.7.mlp.down_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.7.mlp.gate_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.7.mlp.up_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.7.post_attention_layernorm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.7.post_feedforward_layernorm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.7.self_attn.k_norm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.7.self_attn.k_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.7.self_attn.o_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.7.self_attn.q_norm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.7.self_attn.q_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.7.self_attn.v_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.8.mlp.down_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.8.mlp.gate_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.8.mlp.up_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.8.post_attention_layernorm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.8.post_feedforward_layernorm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.8.self_attn.k_norm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.8.self_attn.k_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.8.self_attn.o_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.8.self_attn.q_norm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.8.self_attn.q_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.8.self_attn.v_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.9.mlp.down_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.9.mlp.gate_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.9.mlp.up_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.9.post_attention_layernorm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.9.post_feedforward_layernorm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.9.self_attn.k_norm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.9.self_attn.k_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.9.self_attn.o_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.9.self_attn.q_norm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.9.self_attn.q_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.9.self_attn.v_proj.weight": "model-00002-of-00006.safetensors",
+    "model.norm.weight": "model-00006-of-00006.safetensors"
+  }
+}

checkpoint-527/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:238257fca90155a265f016fba4bba5b1ba83e2fa52bcd491e0613efb1abea996
+size 13055369358

checkpoint-527/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:386fcc8cc1089aade9450d86fb239ea3483f455fd2d78d8378645feecfec9d69
+size 14244

checkpoint-527/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4988f45d958973fdae7bab9a472bd46eb48f56f526b60df9b90e4e353c03eae1
+size 1064

checkpoint-527/trainer_state.json

@@ -0,0 +1,1874 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "eval_steps": 500,
+  "global_step": 527,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.003795066413662239,
+      "grad_norm": 58.99103546142578,
+      "learning_rate": 3.7735849056603773e-06,
+      "loss": 11.9758,
+      "step": 2
+    },
+    {
+      "epoch": 0.007590132827324478,
+      "grad_norm": 60.11953353881836,
+      "learning_rate": 7.547169811320755e-06,
+      "loss": 10.2142,
+      "step": 4
+    },
+    {
+      "epoch": 0.011385199240986717,
+      "grad_norm": 15.779712677001953,
+      "learning_rate": 1.1320754716981132e-05,
+      "loss": 6.4991,
+      "step": 6
+    },
+    {
+      "epoch": 0.015180265654648957,
+      "grad_norm": 19.793678283691406,
+      "learning_rate": 1.509433962264151e-05,
+      "loss": 5.2038,
+      "step": 8
+    },
+    {
+      "epoch": 0.018975332068311195,
+      "grad_norm": 23.680116653442383,
+      "learning_rate": 1.8867924528301888e-05,
+      "loss": 4.8628,
+      "step": 10
+    },
+    {
+      "epoch": 0.022770398481973434,
+      "grad_norm": 12.569281578063965,
+      "learning_rate": 2.2641509433962265e-05,
+      "loss": 4.6113,
+      "step": 12
+    },
+    {
+      "epoch": 0.026565464895635674,
+      "grad_norm": 16.667444229125977,
+      "learning_rate": 2.641509433962264e-05,
+      "loss": 4.5817,
+      "step": 14
+    },
+    {
+      "epoch": 0.030360531309297913,
+      "grad_norm": 9.480690956115723,
+      "learning_rate": 3.018867924528302e-05,
+      "loss": 4.4784,
+      "step": 16
+    },
+    {
+      "epoch": 0.03415559772296015,
+      "grad_norm": 10.323396682739258,
+      "learning_rate": 3.39622641509434e-05,
+      "loss": 4.5034,
+      "step": 18
+    },
+    {
+      "epoch": 0.03795066413662239,
+      "grad_norm": 6.275529384613037,
+      "learning_rate": 3.7735849056603776e-05,
+      "loss": 4.3612,
+      "step": 20
+    },
+    {
+      "epoch": 0.04174573055028463,
+      "grad_norm": 18.33019256591797,
+      "learning_rate": 4.150943396226415e-05,
+      "loss": 4.3324,
+      "step": 22
+    },
+    {
+      "epoch": 0.04554079696394687,
+      "grad_norm": 5.673777103424072,
+      "learning_rate": 4.528301886792453e-05,
+      "loss": 4.2084,
+      "step": 24
+    },
+    {
+      "epoch": 0.04933586337760911,
+      "grad_norm": 5.893975734710693,
+      "learning_rate": 4.9056603773584906e-05,
+      "loss": 4.2536,
+      "step": 26
+    },
+    {
+      "epoch": 0.05313092979127135,
+      "grad_norm": 7.583183765411377,
+      "learning_rate": 5.283018867924528e-05,
+      "loss": 4.2715,
+      "step": 28
+    },
+    {
+      "epoch": 0.056925996204933584,
+      "grad_norm": 29.12498664855957,
+      "learning_rate": 5.660377358490566e-05,
+      "loss": 4.2941,
+      "step": 30
+    },
+    {
+      "epoch": 0.06072106261859583,
+      "grad_norm": 13.786356925964355,
+      "learning_rate": 6.037735849056604e-05,
+      "loss": 4.1344,
+      "step": 32
+    },
+    {
+      "epoch": 0.06451612903225806,
+      "grad_norm": 9.63448715209961,
+      "learning_rate": 6.415094339622641e-05,
+      "loss": 4.016,
+      "step": 34
+    },
+    {
+      "epoch": 0.0683111954459203,
+      "grad_norm": 24.18341636657715,
+      "learning_rate": 6.79245283018868e-05,
+      "loss": 4.2109,
+      "step": 36
+    },
+    {
+      "epoch": 0.07210626185958255,
+      "grad_norm": 10.357343673706055,
+      "learning_rate": 7.169811320754717e-05,
+      "loss": 4.0338,
+      "step": 38
+    },
+    {
+      "epoch": 0.07590132827324478,
+      "grad_norm": 6.5294060707092285,
+      "learning_rate": 7.547169811320755e-05,
+      "loss": 4.1091,
+      "step": 40
+    },
+    {
+      "epoch": 0.07969639468690702,
+      "grad_norm": 24.691822052001953,
+      "learning_rate": 7.924528301886794e-05,
+      "loss": 3.8535,
+      "step": 42
+    },
+    {
+      "epoch": 0.08349146110056926,
+      "grad_norm": 5.220446586608887,
+      "learning_rate": 8.30188679245283e-05,
+      "loss": 3.9548,
+      "step": 44
+    },
+    {
+      "epoch": 0.0872865275142315,
+      "grad_norm": 10.426956176757812,
+      "learning_rate": 8.679245283018869e-05,
+      "loss": 4.0897,
+      "step": 46
+    },
+    {
+      "epoch": 0.09108159392789374,
+      "grad_norm": 7.928496360778809,
+      "learning_rate": 9.056603773584906e-05,
+      "loss": 3.8346,
+      "step": 48
+    },
+    {
+      "epoch": 0.09487666034155598,
+      "grad_norm": 4.195712566375732,
+      "learning_rate": 9.433962264150944e-05,
+      "loss": 3.9885,
+      "step": 50
+    },
+    {
+      "epoch": 0.09867172675521822,
+      "grad_norm": 8.46008014678955,
+      "learning_rate": 9.811320754716981e-05,
+      "loss": 3.9167,
+      "step": 52
+    },
+    {
+      "epoch": 0.10246679316888045,
+      "grad_norm": 11.128891944885254,
+      "learning_rate": 9.999975375283309e-05,
+      "loss": 3.8611,
+      "step": 54
+    },
+    {
+      "epoch": 0.1062618595825427,
+      "grad_norm": 7.3787031173706055,
+      "learning_rate": 9.999778379005078e-05,
+      "loss": 3.9224,
+      "step": 56
+    },
+    {
+      "epoch": 0.11005692599620494,
+      "grad_norm": 9.250240325927734,
+      "learning_rate": 9.99938439421016e-05,
+      "loss": 3.9686,
+      "step": 58
+    },
+    {
+      "epoch": 0.11385199240986717,
+      "grad_norm": 3.8329930305480957,
+      "learning_rate": 9.99879343642134e-05,
+      "loss": 3.8928,
+      "step": 60
+    },
+    {
+      "epoch": 0.11764705882352941,
+      "grad_norm": 4.931491374969482,
+      "learning_rate": 9.99800552892203e-05,
+      "loss": 3.9112,
+      "step": 62
+    },
+    {
+      "epoch": 0.12144212523719165,
+      "grad_norm": 4.240759372711182,
+      "learning_rate": 9.997020702755353e-05,
+      "loss": 3.7947,
+      "step": 64
+    },
+    {
+      "epoch": 0.1252371916508539,
+      "grad_norm": 2.899198532104492,
+      "learning_rate": 9.995838996722914e-05,
+      "loss": 3.6796,
+      "step": 66
+    },
+    {
+      "epoch": 0.12903225806451613,
+      "grad_norm": 3.4333019256591797,
+      "learning_rate": 9.994460457383284e-05,
+      "loss": 3.9,
+      "step": 68
+    },
+    {
+      "epoch": 0.13282732447817835,
+      "grad_norm": 3.4416115283966064,
+      "learning_rate": 9.992885139050154e-05,
+      "loss": 3.6634,
+      "step": 70
+    },
+    {
+      "epoch": 0.1366223908918406,
+      "grad_norm": 5.570509433746338,
+      "learning_rate": 9.991113103790198e-05,
+      "loss": 3.6589,
+      "step": 72
+    },
+    {
+      "epoch": 0.14041745730550284,
+      "grad_norm": 4.77843713760376,
+      "learning_rate": 9.98914442142063e-05,
+      "loss": 3.5929,
+      "step": 74
+    },
+    {
+      "epoch": 0.1442125237191651,
+      "grad_norm": 3.334336996078491,
+      "learning_rate": 9.986979169506453e-05,
+      "loss": 3.6783,
+      "step": 76
+    },
+    {
+      "epoch": 0.14800759013282733,
+      "grad_norm": 6.237847805023193,
+      "learning_rate": 9.9846174333574e-05,
+      "loss": 3.8635,
+      "step": 78
+    },
+    {
+      "epoch": 0.15180265654648956,
+      "grad_norm": 5.3761887550354,
+      "learning_rate": 9.982059306024577e-05,
+      "loss": 3.7552,
+      "step": 80
+    },
+    {
+      "epoch": 0.1555977229601518,
+      "grad_norm": 3.021200656890869,
+      "learning_rate": 9.979304888296792e-05,
+      "loss": 3.7418,
+      "step": 82
+    },
+    {
+      "epoch": 0.15939278937381404,
+      "grad_norm": 4.499217510223389,
+      "learning_rate": 9.976354288696588e-05,
+      "loss": 3.5939,
+      "step": 84
+    },
+    {
+      "epoch": 0.16318785578747627,
+      "grad_norm": 2.8771920204162598,
+      "learning_rate": 9.973207623475965e-05,
+      "loss": 3.7072,
+      "step": 86
+    },
+    {
+      "epoch": 0.16698292220113853,
+      "grad_norm": 5.786422252655029,
+      "learning_rate": 9.9698650166118e-05,
+      "loss": 3.6476,
+      "step": 88
+    },
+    {
+      "epoch": 0.17077798861480076,
+      "grad_norm": 4.563845157623291,
+      "learning_rate": 9.966326599800967e-05,
+      "loss": 3.6845,
+      "step": 90
+    },
+    {
+      "epoch": 0.174573055028463,
+      "grad_norm": 7.240761756896973,
+      "learning_rate": 9.962592512455138e-05,
+      "loss": 3.4461,
+      "step": 92
+    },
+    {
+      "epoch": 0.17836812144212524,
+      "grad_norm": 7.457889556884766,
+      "learning_rate": 9.958662901695303e-05,
+      "loss": 3.6345,
+      "step": 94
+    },
+    {
+      "epoch": 0.18216318785578747,
+      "grad_norm": 6.383795261383057,
+      "learning_rate": 9.954537922345961e-05,
+      "loss": 3.4164,
+      "step": 96
+    },
+    {
+      "epoch": 0.1859582542694497,
+      "grad_norm": 4.423782825469971,
+      "learning_rate": 9.950217736929029e-05,
+      "loss": 3.4743,
+      "step": 98
+    },
+    {
+      "epoch": 0.18975332068311196,
+      "grad_norm": 4.3913469314575195,
+      "learning_rate": 9.945702515657434e-05,
+      "loss": 3.4078,
+      "step": 100
+    },
+    {
+      "epoch": 0.1935483870967742,
+      "grad_norm": 4.806511878967285,
+      "learning_rate": 9.940992436428409e-05,
+      "loss": 3.49,
+      "step": 102
+    },
+    {
+      "epoch": 0.19734345351043645,
+      "grad_norm": 4.089972496032715,
+      "learning_rate": 9.936087684816486e-05,
+      "loss": 3.6553,
+      "step": 104
+    },
+    {
+      "epoch": 0.20113851992409867,
+      "grad_norm": 4.213963985443115,
+      "learning_rate": 9.930988454066177e-05,
+      "loss": 3.3105,
+      "step": 106
+    },
+    {
+      "epoch": 0.2049335863377609,
+      "grad_norm": 5.072576999664307,
+      "learning_rate": 9.92569494508437e-05,
+      "loss": 3.3807,
+      "step": 108
+    },
+    {
+      "epoch": 0.20872865275142316,
+      "grad_norm": 3.9879724979400635,
+      "learning_rate": 9.920207366432402e-05,
+      "loss": 3.3832,
+      "step": 110
+    },
+    {
+      "epoch": 0.2125237191650854,
+      "grad_norm": 3.9012928009033203,
+      "learning_rate": 9.914525934317855e-05,
+      "loss": 3.4129,
+      "step": 112
+    },
+    {
+      "epoch": 0.21631878557874762,
+      "grad_norm": 3.9006423950195312,
+      "learning_rate": 9.908650872586029e-05,
+      "loss": 3.3788,
+      "step": 114
+    },
+    {
+      "epoch": 0.22011385199240988,
+      "grad_norm": 4.585041046142578,
+      "learning_rate": 9.90258241271112e-05,
+      "loss": 3.2844,
+      "step": 116
+    },
+    {
+      "epoch": 0.2239089184060721,
+      "grad_norm": 3.21873140335083,
+      "learning_rate": 9.896320793787106e-05,
+      "loss": 3.2972,
+      "step": 118
+    },
+    {
+      "epoch": 0.22770398481973433,
+      "grad_norm": 4.861288547515869,
+      "learning_rate": 9.889866262518331e-05,
+      "loss": 3.3681,
+      "step": 120
+    },
+    {
+      "epoch": 0.2314990512333966,
+      "grad_norm": 3.7863895893096924,
+      "learning_rate": 9.883219073209772e-05,
+      "loss": 3.2602,
+      "step": 122
+    },
+    {
+      "epoch": 0.23529411764705882,
+      "grad_norm": 5.968108177185059,
+      "learning_rate": 9.876379487757034e-05,
+      "loss": 3.2957,
+      "step": 124
+    },
+    {
+      "epoch": 0.23908918406072105,
+      "grad_norm": 4.3137993812561035,
+      "learning_rate": 9.869347775636015e-05,
+      "loss": 3.211,
+      "step": 126
+    },
+    {
+      "epoch": 0.2428842504743833,
+      "grad_norm": 6.66726016998291,
+      "learning_rate": 9.862124213892304e-05,
+      "loss": 3.253,
+      "step": 128
+    },
+    {
+      "epoch": 0.24667931688804554,
+      "grad_norm": 3.351065158843994,
+      "learning_rate": 9.85470908713026e-05,
+      "loss": 3.1391,
+      "step": 130
+    },
+    {
+      "epoch": 0.2504743833017078,
+      "grad_norm": 5.007400035858154,
+      "learning_rate": 9.847102687501797e-05,
+      "loss": 3.2324,
+      "step": 132
+    },
+    {
+      "epoch": 0.25426944971537,
+      "grad_norm": 3.5997180938720703,
+      "learning_rate": 9.839305314694873e-05,
+      "loss": 3.1649,
+      "step": 134
+    },
+    {
+      "epoch": 0.25806451612903225,
+      "grad_norm": 5.135311126708984,
+      "learning_rate": 9.831317275921685e-05,
+      "loss": 3.1186,
+      "step": 136
+    },
+    {
+      "epoch": 0.2618595825426945,
+      "grad_norm": 4.529127597808838,
+      "learning_rate": 9.823138885906566e-05,
+      "loss": 3.0933,
+      "step": 138
+    },
+    {
+      "epoch": 0.2656546489563567,
+      "grad_norm": 7.555715084075928,
+      "learning_rate": 9.814770466873585e-05,
+      "loss": 3.1694,
+      "step": 140
+    },
+    {
+      "epoch": 0.269449715370019,
+      "grad_norm": 6.438807487487793,
+      "learning_rate": 9.806212348533838e-05,
+      "loss": 3.0059,
+      "step": 142
+    },
+    {
+      "epoch": 0.2732447817836812,
+      "grad_norm": 3.579599380493164,
+      "learning_rate": 9.797464868072488e-05,
+      "loss": 3.0959,
+      "step": 144
+    },
+    {
+      "epoch": 0.27703984819734345,
+      "grad_norm": 2.4115488529205322,
+      "learning_rate": 9.788528370135443e-05,
+      "loss": 3.0124,
+      "step": 146
+    },
+    {
+      "epoch": 0.2808349146110057,
+      "grad_norm": 4.941457748413086,
+      "learning_rate": 9.77940320681581e-05,
+      "loss": 3.0013,
+      "step": 148
+    },
+    {
+      "epoch": 0.2846299810246679,
+      "grad_norm": 3.0970795154571533,
+      "learning_rate": 9.77008973764e-05,
+      "loss": 2.9156,
+      "step": 150
+    },
+    {
+      "epoch": 0.2884250474383302,
+      "grad_norm": 3.7954447269439697,
+      "learning_rate": 9.760588329553571e-05,
+      "loss": 2.9065,
+      "step": 152
+    },
+    {
+      "epoch": 0.2922201138519924,
+      "grad_norm": 6.178630352020264,
+      "learning_rate": 9.750899356906775e-05,
+      "loss": 2.9992,
+      "step": 154
+    },
+    {
+      "epoch": 0.29601518026565465,
+      "grad_norm": 4.201640605926514,
+      "learning_rate": 9.741023201439803e-05,
+      "loss": 2.9465,
+      "step": 156
+    },
+    {
+      "epoch": 0.2998102466793169,
+      "grad_norm": 6.396300792694092,
+      "learning_rate": 9.730960252267743e-05,
+      "loss": 2.9635,
+      "step": 158
+    },
+    {
+      "epoch": 0.3036053130929791,
+      "grad_norm": 3.5059707164764404,
+      "learning_rate": 9.720710905865256e-05,
+      "loss": 2.9565,
+      "step": 160
+    },
+    {
+      "epoch": 0.30740037950664134,
+      "grad_norm": 3.6058707237243652,
+      "learning_rate": 9.710275566050951e-05,
+      "loss": 2.9546,
+      "step": 162
+    },
+    {
+      "epoch": 0.3111954459203036,
+      "grad_norm": 3.369931697845459,
+      "learning_rate": 9.699654643971472e-05,
+      "loss": 2.9414,
+      "step": 164
+    },
+    {
+      "epoch": 0.31499051233396586,
+      "grad_norm": 3.4260523319244385,
+      "learning_rate": 9.688848558085306e-05,
+      "loss": 2.8259,
+      "step": 166
+    },
+    {
+      "epoch": 0.3187855787476281,
+      "grad_norm": 2.820384979248047,
+      "learning_rate": 9.677857734146289e-05,
+      "loss": 2.8689,
+      "step": 168
+    },
+    {
+      "epoch": 0.3225806451612903,
+      "grad_norm": 6.160562992095947,
+      "learning_rate": 9.666682605186835e-05,
+      "loss": 2.8823,
+      "step": 170
+    },
+    {
+      "epoch": 0.32637571157495254,
+      "grad_norm": 8.138015747070312,
+      "learning_rate": 9.655323611500875e-05,
+      "loss": 2.9995,
+      "step": 172
+    },
+    {
+      "epoch": 0.3301707779886148,
+      "grad_norm": 8.836729049682617,
+      "learning_rate": 9.643781200626511e-05,
+      "loss": 2.7727,
+      "step": 174
+    },
+    {
+      "epoch": 0.33396584440227706,
+      "grad_norm": 5.794134616851807,
+      "learning_rate": 9.632055827328382e-05,
+      "loss": 2.8787,
+      "step": 176
+    },
+    {
+      "epoch": 0.3377609108159393,
+      "grad_norm": 8.271780967712402,
+      "learning_rate": 9.620147953579737e-05,
+      "loss": 2.9325,
+      "step": 178
+    },
+    {
+      "epoch": 0.3415559772296015,
+      "grad_norm": 7.904753684997559,
+      "learning_rate": 9.608058048544251e-05,
+      "loss": 2.9486,
+      "step": 180
+    },
+    {
+      "epoch": 0.34535104364326374,
+      "grad_norm": 5.133846759796143,
+      "learning_rate": 9.595786588557532e-05,
+      "loss": 2.8692,
+      "step": 182
+    },
+    {
+      "epoch": 0.349146110056926,
+      "grad_norm": 4.022572040557861,
+      "learning_rate": 9.583334057108346e-05,
+      "loss": 2.7638,
+      "step": 184
+    },
+    {
+      "epoch": 0.35294117647058826,
+      "grad_norm": 6.008692264556885,
+      "learning_rate": 9.570700944819584e-05,
+      "loss": 2.9448,
+      "step": 186
+    },
+    {
+      "epoch": 0.3567362428842505,
+      "grad_norm": 4.208110809326172,
+      "learning_rate": 9.557887749428913e-05,
+      "loss": 2.7578,
+      "step": 188
+    },
+    {
+      "epoch": 0.3605313092979127,
+      "grad_norm": 3.0022525787353516,
+      "learning_rate": 9.544894975769186e-05,
+      "loss": 2.7091,
+      "step": 190
+    },
+    {
+      "epoch": 0.36432637571157495,
+      "grad_norm": 2.1430158615112305,
+      "learning_rate": 9.531723135748529e-05,
+      "loss": 2.6014,
+      "step": 192
+    },
+    {
+      "epoch": 0.3681214421252372,
+      "grad_norm": 3.903315782546997,
+      "learning_rate": 9.518372748330194e-05,
+      "loss": 2.7102,
+      "step": 194
+    },
+    {
+      "epoch": 0.3719165085388994,
+      "grad_norm": 4.798587322235107,
+      "learning_rate": 9.504844339512095e-05,
+      "loss": 2.5969,
+      "step": 196
+    },
+    {
+      "epoch": 0.3757115749525617,
+      "grad_norm": 2.9411423206329346,
+      "learning_rate": 9.4911384423061e-05,
+      "loss": 2.7626,
+      "step": 198
+    },
+    {
+      "epoch": 0.3795066413662239,
+      "grad_norm": 2.5336225032806396,
+      "learning_rate": 9.477255596717012e-05,
+      "loss": 2.6405,
+      "step": 200
+    },
+    {
+      "epoch": 0.38330170777988615,
+      "grad_norm": 2.6455540657043457,
+      "learning_rate": 9.463196349721308e-05,
+      "loss": 2.699,
+      "step": 202
+    },
+    {
+      "epoch": 0.3870967741935484,
+      "grad_norm": 3.346395969390869,
+      "learning_rate": 9.448961255245584e-05,
+      "loss": 2.4129,
+      "step": 204
+    },
+    {
+      "epoch": 0.3908918406072106,
+      "grad_norm": 4.412266254425049,
+      "learning_rate": 9.434550874144728e-05,
+      "loss": 2.5733,
+      "step": 206
+    },
+    {
+      "epoch": 0.3946869070208729,
+      "grad_norm": 7.3123579025268555,
+      "learning_rate": 9.419965774179824e-05,
+      "loss": 2.6668,
+      "step": 208
+    },
+    {
+      "epoch": 0.3984819734345351,
+      "grad_norm": 7.745405197143555,
+      "learning_rate": 9.405206529995785e-05,
+      "loss": 2.8233,
+      "step": 210
+    },
+    {
+      "epoch": 0.40227703984819735,
+      "grad_norm": 6.307023048400879,
+      "learning_rate": 9.39027372309871e-05,
+      "loss": 2.649,
+      "step": 212
+    },
+    {
+      "epoch": 0.4060721062618596,
+      "grad_norm": 3.971792697906494,
+      "learning_rate": 9.375167941832973e-05,
+      "loss": 2.7105,
+      "step": 214
+    },
+    {
+      "epoch": 0.4098671726755218,
+      "grad_norm": 2.6765499114990234,
+      "learning_rate": 9.359889781358042e-05,
+      "loss": 2.5993,
+      "step": 216
+    },
+    {
+      "epoch": 0.41366223908918404,
+      "grad_norm": 2.9960570335388184,
+      "learning_rate": 9.344439843625034e-05,
+      "loss": 2.6444,
+      "step": 218
+    },
+    {
+      "epoch": 0.4174573055028463,
+      "grad_norm": 4.103157997131348,
+      "learning_rate": 9.32881873735299e-05,
+      "loss": 2.6523,
+      "step": 220
+    },
+    {
+      "epoch": 0.42125237191650855,
+      "grad_norm": 3.6402695178985596,
+      "learning_rate": 9.313027078004903e-05,
+      "loss": 2.5287,
+      "step": 222
+    },
+    {
+      "epoch": 0.4250474383301708,
+      "grad_norm": 5.530549049377441,
+      "learning_rate": 9.297065487763462e-05,
+      "loss": 2.6326,
+      "step": 224
+    },
+    {
+      "epoch": 0.428842504743833,
+      "grad_norm": 4.8087310791015625,
+      "learning_rate": 9.280934595506536e-05,
+      "loss": 2.4784,
+      "step": 226
+    },
+    {
+      "epoch": 0.43263757115749524,
+      "grad_norm": 6.805646896362305,
+      "learning_rate": 9.264635036782405e-05,
+      "loss": 2.5274,
+      "step": 228
+    },
+    {
+      "epoch": 0.4364326375711575,
+      "grad_norm": 4.533606052398682,
+      "learning_rate": 9.248167453784711e-05,
+      "loss": 2.5572,
+      "step": 230
+    },
+    {
+      "epoch": 0.44022770398481975,
+      "grad_norm": 4.55278205871582,
+      "learning_rate": 9.231532495327165e-05,
+      "loss": 2.5945,
+      "step": 232
+    },
+    {
+      "epoch": 0.444022770398482,
+      "grad_norm": 2.897782802581787,
+      "learning_rate": 9.21473081681797e-05,
+      "loss": 2.6323,
+      "step": 234
+    },
+    {
+      "epoch": 0.4478178368121442,
+      "grad_norm": 3.3823184967041016,
+      "learning_rate": 9.197763080234019e-05,
+      "loss": 2.4394,
+      "step": 236
+    },
+    {
+      "epoch": 0.45161290322580644,
+      "grad_norm": 4.916838645935059,
+      "learning_rate": 9.180629954094792e-05,
+      "loss": 2.514,
+      "step": 238
+    },
+    {
+      "epoch": 0.45540796963946867,
+      "grad_norm": 4.822211265563965,
+      "learning_rate": 9.163332113436032e-05,
+      "loss": 2.439,
+      "step": 240
+    },
+    {
+      "epoch": 0.45920303605313095,
+      "grad_norm": 4.708934783935547,
+      "learning_rate": 9.145870239783142e-05,
+      "loss": 2.5608,
+      "step": 242
+    },
+    {
+      "epoch": 0.4629981024667932,
+      "grad_norm": 2.962392568588257,
+      "learning_rate": 9.128245021124334e-05,
+      "loss": 2.5684,
+      "step": 244
+    },
+    {
+      "epoch": 0.4667931688804554,
+      "grad_norm": 2.4564027786254883,
+      "learning_rate": 9.110457151883523e-05,
+      "loss": 2.4745,
+      "step": 246
+    },
+    {
+      "epoch": 0.47058823529411764,
+      "grad_norm": 2.94209623336792,
+      "learning_rate": 9.092507332892968e-05,
+      "loss": 2.4497,
+      "step": 248
+    },
+    {
+      "epoch": 0.47438330170777987,
+      "grad_norm": 2.5642683506011963,
+      "learning_rate": 9.07439627136566e-05,
+      "loss": 2.4282,
+      "step": 250
+    },
+    {
+      "epoch": 0.4781783681214421,
+      "grad_norm": 2.714101791381836,
+      "learning_rate": 9.056124680867457e-05,
+      "loss": 2.3797,
+      "step": 252
+    },
+    {
+      "epoch": 0.4819734345351044,
+      "grad_norm": 2.010376453399658,
+      "learning_rate": 9.037693281288969e-05,
+      "loss": 2.4963,
+      "step": 254
+    },
+    {
+      "epoch": 0.4857685009487666,
+      "grad_norm": 2.4395999908447266,
+      "learning_rate": 9.019102798817197e-05,
+      "loss": 2.4386,
+      "step": 256
+    },
+    {
+      "epoch": 0.48956356736242884,
+      "grad_norm": 2.391928195953369,
+      "learning_rate": 9.000353965906917e-05,
+      "loss": 2.5203,
+      "step": 258
+    },
+    {
+      "epoch": 0.49335863377609107,
+      "grad_norm": 2.7171072959899902,
+      "learning_rate": 8.981447521251831e-05,
+      "loss": 2.4105,
+      "step": 260
+    },
+    {
+      "epoch": 0.4971537001897533,
+      "grad_norm": 3.4616568088531494,
+      "learning_rate": 8.962384209755452e-05,
+      "loss": 2.2693,
+      "step": 262
+    },
+    {
+      "epoch": 0.5009487666034156,
+      "grad_norm": 3.0641255378723145,
+      "learning_rate": 8.943164782501765e-05,
+      "loss": 2.4696,
+      "step": 264
+    },
+    {
+      "epoch": 0.5047438330170778,
+      "grad_norm": 2.5761353969573975,
+      "learning_rate": 8.923789996725624e-05,
+      "loss": 2.3747,
+      "step": 266
+    },
+    {
+      "epoch": 0.50853889943074,
+      "grad_norm": 3.255844831466675,
+      "learning_rate": 8.904260615782927e-05,
+      "loss": 2.3592,
+      "step": 268
+    },
+    {
+      "epoch": 0.5123339658444023,
+      "grad_norm": 3.119816303253174,
+      "learning_rate": 8.884577409120535e-05,
+      "loss": 2.404,
+      "step": 270
+    },
+    {
+      "epoch": 0.5161290322580645,
+      "grad_norm": 4.314058780670166,
+      "learning_rate": 8.864741152245963e-05,
+      "loss": 2.3595,
+      "step": 272
+    },
+    {
+      "epoch": 0.5199240986717267,
+      "grad_norm": 3.1923677921295166,
+      "learning_rate": 8.84475262669681e-05,
+      "loss": 2.4839,
+      "step": 274
+    },
+    {
+      "epoch": 0.523719165085389,
+      "grad_norm": 5.698940753936768,
+      "learning_rate": 8.824612620009987e-05,
+      "loss": 2.5254,
+      "step": 276
+    },
+    {
+      "epoch": 0.5275142314990512,
+      "grad_norm": 4.198439598083496,
+      "learning_rate": 8.804321925690672e-05,
+      "loss": 2.4509,
+      "step": 278
+    },
+    {
+      "epoch": 0.5313092979127134,
+      "grad_norm": 3.021700620651245,
+      "learning_rate": 8.783881343181055e-05,
+      "loss": 2.3725,
+      "step": 280
+    },
+    {
+      "epoch": 0.5351043643263758,
+      "grad_norm": 3.3499674797058105,
+      "learning_rate": 8.763291677828838e-05,
+      "loss": 2.2391,
+      "step": 282
+    },
+    {
+      "epoch": 0.538899430740038,
+      "grad_norm": 3.2401561737060547,
+      "learning_rate": 8.742553740855506e-05,
+      "loss": 2.4657,
+      "step": 284
+    },
+    {
+      "epoch": 0.5426944971537002,
+      "grad_norm": 2.584604263305664,
+      "learning_rate": 8.721668349324364e-05,
+      "loss": 2.472,
+      "step": 286
+    },
+    {
+      "epoch": 0.5464895635673624,
+      "grad_norm": 2.86765456199646,
+      "learning_rate": 8.700636326108342e-05,
+      "loss": 2.5246,
+      "step": 288
+    },
+    {
+      "epoch": 0.5502846299810247,
+      "grad_norm": 3.742485284805298,
+      "learning_rate": 8.679458499857582e-05,
+      "loss": 2.3875,
+      "step": 290
+    },
+    {
+      "epoch": 0.5540796963946869,
+      "grad_norm": 2.3582913875579834,
+      "learning_rate": 8.658135704966786e-05,
+      "loss": 2.3662,
+      "step": 292
+    },
+    {
+      "epoch": 0.5578747628083491,
+      "grad_norm": 2.188462495803833,
+      "learning_rate": 8.636668781542336e-05,
+      "loss": 2.2918,
+      "step": 294
+    },
+    {
+      "epoch": 0.5616698292220114,
+      "grad_norm": 3.121817111968994,
+      "learning_rate": 8.615058575369202e-05,
+      "loss": 2.3115,
+      "step": 296
+    },
+    {
+      "epoch": 0.5654648956356736,
+      "grad_norm": 3.2461938858032227,
+      "learning_rate": 8.593305937877614e-05,
+      "loss": 2.3089,
+      "step": 298
+    },
+    {
+      "epoch": 0.5692599620493358,
+      "grad_norm": 2.2377259731292725,
+      "learning_rate": 8.571411726109519e-05,
+      "loss": 2.4239,
+      "step": 300
+    },
+    {
+      "epoch": 0.573055028462998,
+      "grad_norm": 2.2074265480041504,
+      "learning_rate": 8.549376802684812e-05,
+      "loss": 2.4559,
+      "step": 302
+    },
+    {
+      "epoch": 0.5768500948766604,
+      "grad_norm": 2.276872396469116,
+      "learning_rate": 8.527202035767349e-05,
+      "loss": 2.4151,
+      "step": 304
+    },
+    {
+      "epoch": 0.5806451612903226,
+      "grad_norm": 2.8724167346954346,
+      "learning_rate": 8.504888299030747e-05,
+      "loss": 2.3258,
+      "step": 306
+    },
+    {
+      "epoch": 0.5844402277039848,
+      "grad_norm": 3.028040647506714,
+      "learning_rate": 8.482436471623951e-05,
+      "loss": 2.3436,
+      "step": 308
+    },
+    {
+      "epoch": 0.5882352941176471,
+      "grad_norm": 2.2716915607452393,
+      "learning_rate": 8.459847438136605e-05,
+      "loss": 2.207,
+      "step": 310
+    },
+    {
+      "epoch": 0.5920303605313093,
+      "grad_norm": 2.3537518978118896,
+      "learning_rate": 8.437122088564198e-05,
+      "loss": 2.2379,
+      "step": 312
+    },
+    {
+      "epoch": 0.5958254269449715,
+      "grad_norm": 2.7615747451782227,
+      "learning_rate": 8.414261318272996e-05,
+      "loss": 2.353,
+      "step": 314
+    },
+    {
+      "epoch": 0.5996204933586338,
+      "grad_norm": 2.191373586654663,
+      "learning_rate": 8.391266027964771e-05,
+      "loss": 2.406,
+      "step": 316
+    },
+    {
+      "epoch": 0.603415559772296,
+      "grad_norm": 2.1172335147857666,
+      "learning_rate": 8.368137123641302e-05,
+      "loss": 2.287,
+      "step": 318
+    },
+    {
+      "epoch": 0.6072106261859582,
+      "grad_norm": 2.387768030166626,
+      "learning_rate": 8.344875516568695e-05,
+      "loss": 2.3713,
+      "step": 320
+    },
+    {
+      "epoch": 0.6110056925996205,
+      "grad_norm": 2.8590333461761475,
+      "learning_rate": 8.321482123241464e-05,
+      "loss": 2.2492,
+      "step": 322
+    },
+    {
+      "epoch": 0.6148007590132827,
+      "grad_norm": 2.2510037422180176,
+      "learning_rate": 8.297957865346437e-05,
+      "loss": 2.3075,
+      "step": 324
+    },
+    {
+      "epoch": 0.618595825426945,
+      "grad_norm": 2.747525930404663,
+      "learning_rate": 8.274303669726426e-05,
+      "loss": 2.089,
+      "step": 326
+    },
+    {
+      "epoch": 0.6223908918406073,
+      "grad_norm": 2.4523816108703613,
+      "learning_rate": 8.250520468343722e-05,
+      "loss": 2.2978,
+      "step": 328
+    },
+    {
+      "epoch": 0.6261859582542695,
+      "grad_norm": 2.1454994678497314,
+      "learning_rate": 8.226609198243372e-05,
+      "loss": 2.3123,
+      "step": 330
+    },
+    {
+      "epoch": 0.6299810246679317,
+      "grad_norm": 2.733635663986206,
+      "learning_rate": 8.202570801516256e-05,
+      "loss": 2.2186,
+      "step": 332
+    },
+    {
+      "epoch": 0.6337760910815939,
+      "grad_norm": 2.4969663619995117,
+      "learning_rate": 8.178406225261981e-05,
+      "loss": 2.2239,
+      "step": 334
+    },
+    {
+      "epoch": 0.6375711574952562,
+      "grad_norm": 3.0612945556640625,
+      "learning_rate": 8.15411642155155e-05,
+      "loss": 2.2033,
+      "step": 336
+    },
+    {
+      "epoch": 0.6413662239089184,
+      "grad_norm": 3.4063143730163574,
+      "learning_rate": 8.129702347389865e-05,
+      "loss": 2.3211,
+      "step": 338
+    },
+    {
+      "epoch": 0.6451612903225806,
+      "grad_norm": 2.2077441215515137,
+      "learning_rate": 8.105164964678009e-05,
+      "loss": 2.5048,
+      "step": 340
+    },
+    {
+      "epoch": 0.6489563567362429,
+      "grad_norm": 2.7867558002471924,
+      "learning_rate": 8.080505240175363e-05,
+      "loss": 2.3034,
+      "step": 342
+    },
+    {
+      "epoch": 0.6527514231499051,
+      "grad_norm": 2.510138511657715,
+      "learning_rate": 8.055724145461495e-05,
+      "loss": 2.1329,
+      "step": 344
+    },
+    {
+      "epoch": 0.6565464895635673,
+      "grad_norm": 2.863159418106079,
+      "learning_rate": 8.030822656897902e-05,
+      "loss": 2.2401,
+      "step": 346
+    },
+    {
+      "epoch": 0.6603415559772297,
+      "grad_norm": 2.664177179336548,
+      "learning_rate": 8.005801755589532e-05,
+      "loss": 2.1028,
+      "step": 348
+    },
+    {
+      "epoch": 0.6641366223908919,
+      "grad_norm": 2.095933675765991,
+      "learning_rate": 7.980662427346127e-05,
+      "loss": 1.9901,
+      "step": 350
+    },
+    {
+      "epoch": 0.6679316888045541,
+      "grad_norm": 2.5369749069213867,
+      "learning_rate": 7.955405662643384e-05,
+      "loss": 2.0314,
+      "step": 352
+    },
+    {
+      "epoch": 0.6717267552182163,
+      "grad_norm": 2.400635242462158,
+      "learning_rate": 7.930032456583931e-05,
+      "loss": 2.2097,
+      "step": 354
+    },
+    {
+      "epoch": 0.6755218216318786,
+      "grad_norm": 2.1800007820129395,
+      "learning_rate": 7.904543808858127e-05,
+      "loss": 2.1188,
+      "step": 356
+    },
+    {
+      "epoch": 0.6793168880455408,
+      "grad_norm": 2.3743057250976562,
+      "learning_rate": 7.878940723704664e-05,
+      "loss": 2.1496,
+      "step": 358
+    },
+    {
+      "epoch": 0.683111954459203,
+      "grad_norm": 2.9855995178222656,
+      "learning_rate": 7.853224209871007e-05,
+      "loss": 2.2357,
+      "step": 360
+    },
+    {
+      "epoch": 0.6869070208728653,
+      "grad_norm": 2.130502700805664,
+      "learning_rate": 7.82739528057365e-05,
+      "loss": 2.1077,
+      "step": 362
+    },
+    {
+      "epoch": 0.6907020872865275,
+      "grad_norm": 2.2894301414489746,
+      "learning_rate": 7.801454953458193e-05,
+      "loss": 2.1231,
+      "step": 364
+    },
+    {
+      "epoch": 0.6944971537001897,
+      "grad_norm": 2.3042802810668945,
+      "learning_rate": 7.775404250559249e-05,
+      "loss": 2.3711,
+      "step": 366
+    },
+    {
+      "epoch": 0.698292220113852,
+      "grad_norm": 2.517493724822998,
+      "learning_rate": 7.749244198260175e-05,
+      "loss": 2.2039,
+      "step": 368
+    },
+    {
+      "epoch": 0.7020872865275142,
+      "grad_norm": 2.50030255317688,
+      "learning_rate": 7.722975827252638e-05,
+      "loss": 2.3868,
+      "step": 370
+    },
+    {
+      "epoch": 0.7058823529411765,
+      "grad_norm": 2.252354621887207,
+      "learning_rate": 7.696600172495997e-05,
+      "loss": 2.2204,
+      "step": 372
+    },
+    {
+      "epoch": 0.7096774193548387,
+      "grad_norm": 2.182471990585327,
+      "learning_rate": 7.670118273176534e-05,
+      "loss": 2.2011,
+      "step": 374
+    },
+    {
+      "epoch": 0.713472485768501,
+      "grad_norm": 2.4250662326812744,
+      "learning_rate": 7.643531172666513e-05,
+      "loss": 2.2251,
+      "step": 376
+    },
+    {
+      "epoch": 0.7172675521821632,
+      "grad_norm": 2.91322922706604,
+      "learning_rate": 7.616839918483061e-05,
+      "loss": 2.2859,
+      "step": 378
+    },
+    {
+      "epoch": 0.7210626185958254,
+      "grad_norm": 2.963437080383301,
+      "learning_rate": 7.590045562246902e-05,
+      "loss": 2.1708,
+      "step": 380
+    },
+    {
+      "epoch": 0.7248576850094877,
+      "grad_norm": 3.132415771484375,
+      "learning_rate": 7.563149159640929e-05,
+      "loss": 2.1548,
+      "step": 382
+    },
+    {
+      "epoch": 0.7286527514231499,
+      "grad_norm": 2.5823440551757812,
+      "learning_rate": 7.5361517703686e-05,
+      "loss": 2.1451,
+      "step": 384
+    },
+    {
+      "epoch": 0.7324478178368121,
+      "grad_norm": 2.6822919845581055,
+      "learning_rate": 7.509054458112202e-05,
+      "loss": 2.3556,
+      "step": 386
+    },
+    {
+      "epoch": 0.7362428842504743,
+      "grad_norm": 2.1513512134552,
+      "learning_rate": 7.481858290490917e-05,
+      "loss": 2.1135,
+      "step": 388
+    },
+    {
+      "epoch": 0.7400379506641366,
+      "grad_norm": 2.6822762489318848,
+      "learning_rate": 7.45456433901879e-05,
+      "loss": 2.2004,
+      "step": 390
+    },
+    {
+      "epoch": 0.7438330170777988,
+      "grad_norm": 2.8601059913635254,
+      "learning_rate": 7.427173679062484e-05,
+      "loss": 2.187,
+      "step": 392
+    },
+    {
+      "epoch": 0.7476280834914611,
+      "grad_norm": 2.706089496612549,
+      "learning_rate": 7.399687389798933e-05,
+      "loss": 2.2509,
+      "step": 394
+    },
+    {
+      "epoch": 0.7514231499051234,
+      "grad_norm": 3.9849417209625244,
+      "learning_rate": 7.372106554172802e-05,
+      "loss": 2.1557,
+      "step": 396
+    },
+    {
+      "epoch": 0.7552182163187856,
+      "grad_norm": 2.7597997188568115,
+      "learning_rate": 7.344432258853841e-05,
+      "loss": 2.1147,
+      "step": 398
+    },
+    {
+      "epoch": 0.7590132827324478,
+      "grad_norm": 2.172595500946045,
+      "learning_rate": 7.316665594194053e-05,
+      "loss": 2.2463,
+      "step": 400
+    },
+    {
+      "epoch": 0.7628083491461101,
+      "grad_norm": 1.9581433534622192,
+      "learning_rate": 7.288807654184747e-05,
+      "loss": 2.1798,
+      "step": 402
+    },
+    {
+      "epoch": 0.7666034155597723,
+      "grad_norm": 2.4689390659332275,
+      "learning_rate": 7.260859536413429e-05,
+      "loss": 2.1782,
+      "step": 404
+    },
+    {
+      "epoch": 0.7703984819734345,
+      "grad_norm": 3.047780752182007,
+      "learning_rate": 7.232822342020557e-05,
+      "loss": 2.0098,
+      "step": 406
+    },
+    {
+      "epoch": 0.7741935483870968,
+      "grad_norm": 3.0638253688812256,
+      "learning_rate": 7.204697175656165e-05,
+      "loss": 2.2135,
+      "step": 408
+    },
+    {
+      "epoch": 0.777988614800759,
+      "grad_norm": 2.14133882522583,
+      "learning_rate": 7.176485145436325e-05,
+      "loss": 2.1974,
+      "step": 410
+    },
+    {
+      "epoch": 0.7817836812144212,
+      "grad_norm": 3.42946720123291,
+      "learning_rate": 7.148187362899505e-05,
+      "loss": 2.2132,
+      "step": 412
+    },
+    {
+      "epoch": 0.7855787476280834,
+      "grad_norm": 2.9244472980499268,
+      "learning_rate": 7.119804942962762e-05,
+      "loss": 1.9828,
+      "step": 414
+    },
+    {
+      "epoch": 0.7893738140417458,
+      "grad_norm": 2.4446115493774414,
+      "learning_rate": 7.091339003877826e-05,
+      "loss": 2.2205,
+      "step": 416
+    },
+    {
+      "epoch": 0.793168880455408,
+      "grad_norm": 2.8077917098999023,
+      "learning_rate": 7.062790667187029e-05,
+      "loss": 2.0921,
+      "step": 418
+    },
+    {
+      "epoch": 0.7969639468690702,
+      "grad_norm": 2.168386697769165,
+      "learning_rate": 7.034161057679127e-05,
+      "loss": 2.1198,
+      "step": 420
+    },
+    {
+      "epoch": 0.8007590132827325,
+      "grad_norm": 2.11982798576355,
+      "learning_rate": 7.005451303344979e-05,
+      "loss": 2.0766,
+      "step": 422
+    },
+    {
+      "epoch": 0.8045540796963947,
+      "grad_norm": 2.1477482318878174,
+      "learning_rate": 6.976662535333107e-05,
+      "loss": 2.0088,
+      "step": 424
+    },
+    {
+      "epoch": 0.8083491461100569,
+      "grad_norm": 2.2626473903656006,
+      "learning_rate": 6.947795887905127e-05,
+      "loss": 1.974,
+      "step": 426
+    },
+    {
+      "epoch": 0.8121442125237192,
+      "grad_norm": 2.2598705291748047,
+      "learning_rate": 6.918852498391063e-05,
+      "loss": 2.0463,
+      "step": 428
+    },
+    {
+      "epoch": 0.8159392789373814,
+      "grad_norm": 2.132086992263794,
+      "learning_rate": 6.889833507144532e-05,
+      "loss": 2.2507,
+      "step": 430
+    },
+    {
+      "epoch": 0.8197343453510436,
+      "grad_norm": 2.307889223098755,
+      "learning_rate": 6.860740057497823e-05,
+      "loss": 2.1272,
+      "step": 432
+    },
+    {
+      "epoch": 0.8235294117647058,
+      "grad_norm": 2.638432264328003,
+      "learning_rate": 6.831573295716837e-05,
+      "loss": 2.1772,
+      "step": 434
+    },
+    {
+      "epoch": 0.8273244781783681,
+      "grad_norm": 2.552748680114746,
+      "learning_rate": 6.802334370955941e-05,
+      "loss": 2.0331,
+      "step": 436
+    },
+    {
+      "epoch": 0.8311195445920304,
+      "grad_norm": 2.7252237796783447,
+      "learning_rate": 6.773024435212678e-05,
+      "loss": 2.1394,
+      "step": 438
+    },
+    {
+      "epoch": 0.8349146110056926,
+      "grad_norm": 2.4146969318389893,
+      "learning_rate": 6.743644643282388e-05,
+      "loss": 2.2107,
+      "step": 440
+    },
+    {
+      "epoch": 0.8387096774193549,
+      "grad_norm": 2.2211036682128906,
+      "learning_rate": 6.714196152712704e-05,
+      "loss": 2.2715,
+      "step": 442
+    },
+    {
+      "epoch": 0.8425047438330171,
+      "grad_norm": 2.484332323074341,
+      "learning_rate": 6.684680123757949e-05,
+      "loss": 2.0996,
+      "step": 444
+    },
+    {
+      "epoch": 0.8462998102466793,
+      "grad_norm": 2.478273868560791,
+      "learning_rate": 6.65509771933342e-05,
+      "loss": 1.9356,
+      "step": 446
+    },
+    {
+      "epoch": 0.8500948766603416,
+      "grad_norm": 2.3699731826782227,
+      "learning_rate": 6.625450104969572e-05,
+      "loss": 2.1316,
+      "step": 448
+    },
+    {
+      "epoch": 0.8538899430740038,
+      "grad_norm": 2.419955253601074,
+      "learning_rate": 6.595738448766095e-05,
+      "loss": 2.1218,
+      "step": 450
+    },
+    {
+      "epoch": 0.857685009487666,
+      "grad_norm": 2.293102979660034,
+      "learning_rate": 6.565963921345895e-05,
+      "loss": 2.0795,
+      "step": 452
+    },
+    {
+      "epoch": 0.8614800759013282,
+      "grad_norm": 2.706804037094116,
+      "learning_rate": 6.536127695808964e-05,
+      "loss": 2.0764,
+      "step": 454
+    },
+    {
+      "epoch": 0.8652751423149905,
+      "grad_norm": 2.3462438583374023,
+      "learning_rate": 6.506230947686172e-05,
+      "loss": 2.0065,
+      "step": 456
+    },
+    {
+      "epoch": 0.8690702087286527,
+      "grad_norm": 2.636672258377075,
+      "learning_rate": 6.47627485489294e-05,
+      "loss": 2.1865,
+      "step": 458
+    },
+    {
+      "epoch": 0.872865275142315,
+      "grad_norm": 2.1746952533721924,
+      "learning_rate": 6.446260597682839e-05,
+      "loss": 1.9698,
+      "step": 460
+    },
+    {
+      "epoch": 0.8766603415559773,
+      "grad_norm": 2.1193501949310303,
+      "learning_rate": 6.416189358601088e-05,
+      "loss": 1.9559,
+      "step": 462
+    },
+    {
+      "epoch": 0.8804554079696395,
+      "grad_norm": 2.4427428245544434,
+      "learning_rate": 6.386062322437954e-05,
+      "loss": 1.9442,
+      "step": 464
+    },
+    {
+      "epoch": 0.8842504743833017,
+      "grad_norm": 2.3021249771118164,
+      "learning_rate": 6.355880676182086e-05,
+      "loss": 2.1573,
+      "step": 466
+    },
+    {
+      "epoch": 0.888045540796964,
+      "grad_norm": 2.224827527999878,
+      "learning_rate": 6.325645608973735e-05,
+      "loss": 2.0547,
+      "step": 468
+    },
+    {
+      "epoch": 0.8918406072106262,
+      "grad_norm": 2.451848030090332,
+      "learning_rate": 6.295358312057914e-05,
+      "loss": 2.0589,
+      "step": 470
+    },
+    {
+      "epoch": 0.8956356736242884,
+      "grad_norm": 2.2723612785339355,
+      "learning_rate": 6.26501997873745e-05,
+      "loss": 1.9392,
+      "step": 472
+    },
+    {
+      "epoch": 0.8994307400379506,
+      "grad_norm": 2.340785503387451,
+      "learning_rate": 6.234631804325981e-05,
+      "loss": 2.1117,
+      "step": 474
+    },
+    {
+      "epoch": 0.9032258064516129,
+      "grad_norm": 2.963451385498047,
+      "learning_rate": 6.204194986100857e-05,
+      "loss": 2.0476,
+      "step": 476
+    },
+    {
+      "epoch": 0.9070208728652751,
+      "grad_norm": 2.2216341495513916,
+      "learning_rate": 6.173710723255966e-05,
+      "loss": 2.154,
+      "step": 478
+    },
+    {
+      "epoch": 0.9108159392789373,
+      "grad_norm": 1.8229964971542358,
+      "learning_rate": 6.143180216854487e-05,
+      "loss": 1.8867,
+      "step": 480
+    },
+    {
+      "epoch": 0.9146110056925996,
+      "grad_norm": 2.198394775390625,
+      "learning_rate": 6.112604669781572e-05,
+      "loss": 2.0622,
+      "step": 482
+    },
+    {
+      "epoch": 0.9184060721062619,
+      "grad_norm": 1.7389253377914429,
+      "learning_rate": 6.081985286696949e-05,
+      "loss": 2.1547,
+      "step": 484
+    },
+    {
+      "epoch": 0.9222011385199241,
+      "grad_norm": 2.4822840690612793,
+      "learning_rate": 6.051323273987463e-05,
+      "loss": 1.9973,
+      "step": 486
+    },
+    {
+      "epoch": 0.9259962049335864,
+      "grad_norm": 2.384826183319092,
+      "learning_rate": 6.020619839719538e-05,
+      "loss": 2.0659,
+      "step": 488
+    },
+    {
+      "epoch": 0.9297912713472486,
+      "grad_norm": 1.8725968599319458,
+      "learning_rate": 5.989876193591589e-05,
+      "loss": 2.082,
+      "step": 490
+    },
+    {
+      "epoch": 0.9335863377609108,
+      "grad_norm": 2.7569420337677,
+      "learning_rate": 5.959093546886356e-05,
+      "loss": 2.0654,
+      "step": 492
+    },
+    {
+      "epoch": 0.937381404174573,
+      "grad_norm": 1.96099853515625,
+      "learning_rate": 5.928273112423177e-05,
+      "loss": 1.9152,
+      "step": 494
+    },
+    {
+      "epoch": 0.9411764705882353,
+      "grad_norm": 2.6206164360046387,
+      "learning_rate": 5.897416104510211e-05,
+      "loss": 2.0493,
+      "step": 496
+    },
+    {
+      "epoch": 0.9449715370018975,
+      "grad_norm": 1.8731147050857544,
+      "learning_rate": 5.866523738896587e-05,
+      "loss": 2.0,
+      "step": 498
+    },
+    {
+      "epoch": 0.9487666034155597,
+      "grad_norm": 2.0109100341796875,
+      "learning_rate": 5.835597232724511e-05,
+      "loss": 2.0603,
+      "step": 500
+    },
+    {
+      "epoch": 0.952561669829222,
+      "grad_norm": 2.412135362625122,
+      "learning_rate": 5.804637804481306e-05,
+      "loss": 2.0034,
+      "step": 502
+    },
+    {
+      "epoch": 0.9563567362428842,
+      "grad_norm": 1.894960880279541,
+      "learning_rate": 5.773646673951406e-05,
+      "loss": 2.0708,
+      "step": 504
+    },
+    {
+      "epoch": 0.9601518026565465,
+      "grad_norm": 3.9883475303649902,
+      "learning_rate": 5.742625062168303e-05,
+      "loss": 2.1407,
+      "step": 506
+    },
+    {
+      "epoch": 0.9639468690702088,
+      "grad_norm": 2.3427770137786865,
+      "learning_rate": 5.7115741913664264e-05,
+      "loss": 1.9919,
+      "step": 508
+    },
+    {
+      "epoch": 0.967741935483871,
+      "grad_norm": 2.0752058029174805,
+      "learning_rate": 5.680495284933e-05,
+      "loss": 2.0542,
+      "step": 510
+    },
+    {
+      "epoch": 0.9715370018975332,
+      "grad_norm": 1.9299733638763428,
+      "learning_rate": 5.649389567359831e-05,
+      "loss": 2.0434,
+      "step": 512
+    },
+    {
+      "epoch": 0.9753320683111955,
+      "grad_norm": 2.1606550216674805,
+      "learning_rate": 5.6182582641950764e-05,
+      "loss": 1.9045,
+      "step": 514
+    },
+    {
+      "epoch": 0.9791271347248577,
+      "grad_norm": 1.9549524784088135,
+      "learning_rate": 5.58710260199495e-05,
+      "loss": 1.9951,
+      "step": 516
+    },
+    {
+      "epoch": 0.9829222011385199,
+      "grad_norm": 1.8001773357391357,
+      "learning_rate": 5.555923808275395e-05,
+      "loss": 1.8519,
+      "step": 518
+    },
+    {
+      "epoch": 0.9867172675521821,
+      "grad_norm": 2.168572187423706,
+      "learning_rate": 5.5247231114637256e-05,
+      "loss": 1.8828,
+      "step": 520
+    },
+    {
+      "epoch": 0.9905123339658444,
+      "grad_norm": 2.0529568195343018,
+      "learning_rate": 5.4935017408502274e-05,
+      "loss": 2.0863,
+      "step": 522
+    },
+    {
+      "epoch": 0.9943074003795066,
+      "grad_norm": 2.078700304031372,
+      "learning_rate": 5.462260926539722e-05,
+      "loss": 2.0463,
+      "step": 524
+    },
+    {
+      "epoch": 0.9981024667931688,
+      "grad_norm": 2.1271212100982666,
+      "learning_rate": 5.431001899403098e-05,
+      "loss": 2.0645,
+      "step": 526
+    }
+  ],
+  "logging_steps": 2,
+  "max_steps": 1054,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 2,
+  "save_steps": 527,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 8.912185436248474e+16,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoint-527/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:721e80e1778cdc5974559fb76036f3bd489ad9c204cc63b34e1b4276ae1c1d8e
+size 5432

config.json

@@ -0,0 +1,27 @@
+{
+  "_name_or_path": "/mnt/parscratch/users/acp23ay/private/models/OLMo-2-1124-7B-mni_Beng-coli-cw2v/",
+  "architectures": [
+    "Olmo2ForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "eos_token_id": 100257,
+  "hidden_act": "silu",
+  "hidden_size": 4096,
+  "initializer_range": 0.02,
+  "intermediate_size": 11008,
+  "max_position_embeddings": 4096,
+  "model_type": "olmo2",
+  "num_attention_heads": 32,
+  "num_hidden_layers": 32,
+  "num_key_value_heads": 32,
+  "pad_token_id": 100277,
+  "rms_norm_eps": 1e-06,
+  "rope_scaling": null,
+  "rope_theta": 500000,
+  "tie_word_embeddings": false,
+  "torch_dtype": "float32",
+  "transformers_version": "4.48.0.dev0",
+  "use_cache": true,
+  "vocab_size": 100384
+}

generation_config.json

@@ -0,0 +1,6 @@
+{
+  "_from_model_config": true,
+  "eos_token_id": 100257,
+  "pad_token_id": 100277,
+  "transformers_version": "4.48.0.dev0"
+}

model-00001-of-00006.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7aec87eb603416187ba146f5478b1d3b2db3699ebbcfdd27b16819425dc12ea7
+size 4950070472

model-00002-of-00006.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:de0b6a357d03676c4986588aaf4dc96afc3f44a0498514961d2b2cfa1940655e
+size 4991622816

model-00003-of-00006.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c1038a7d346a681a8fb6949e054c2246f3f3c80e826855c1cec6af1155cc607d
+size 4924546896

model-00004-of-00006.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:33c0a21146ace02d36928787ae7f34be868f7c81b894dea98ec7b0e5a3ea01b2
+size 4857404928

model-00005-of-00006.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e7f82a3014ccfc0f67208f63785ae4bc71f09aba86db48bf6139d18178abbfed
+size 4857404928

model-00006-of-00006.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1fc931f1d954432e62cc1d24d0f192d768b5fb3a281942af9348f5869ddef0fe
+size 4614509080

model.safetensors.index.json

@@ -0,0 +1,362 @@
+{
+  "metadata": {
+    "total_size": 29195517952
+  },
+  "weight_map": {
+    "lm_head.weight": "model-00006-of-00006.safetensors",
+    "model.embed_tokens.weight": "model-00001-of-00006.safetensors",
+    "model.layers.0.mlp.down_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.0.mlp.gate_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.0.mlp.up_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.0.post_attention_layernorm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.0.post_feedforward_layernorm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.0.self_attn.k_norm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.0.self_attn.k_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.0.self_attn.o_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.0.self_attn.q_norm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.0.self_attn.q_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.0.self_attn.v_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.1.mlp.down_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.1.mlp.gate_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.1.mlp.up_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.1.post_attention_layernorm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.1.post_feedforward_layernorm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.1.self_attn.k_norm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.1.self_attn.k_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.1.self_attn.o_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.1.self_attn.q_norm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.1.self_attn.q_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.1.self_attn.v_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.10.mlp.down_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.10.mlp.gate_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.10.mlp.up_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.10.post_attention_layernorm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.10.post_feedforward_layernorm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.10.self_attn.k_norm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.10.self_attn.k_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.10.self_attn.o_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.10.self_attn.q_norm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.10.self_attn.q_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.10.self_attn.v_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.11.mlp.down_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.11.mlp.gate_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.11.mlp.up_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.11.post_attention_layernorm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.11.post_feedforward_layernorm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.11.self_attn.k_norm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.11.self_attn.k_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.11.self_attn.o_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.11.self_attn.q_norm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.11.self_attn.q_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.11.self_attn.v_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.12.mlp.down_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.12.mlp.gate_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.12.mlp.up_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.12.post_attention_layernorm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.12.post_feedforward_layernorm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.12.self_attn.k_norm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.12.self_attn.k_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.12.self_attn.o_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.12.self_attn.q_norm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.12.self_attn.q_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.12.self_attn.v_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.13.mlp.down_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.13.mlp.gate_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.13.mlp.up_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.13.post_attention_layernorm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.13.post_feedforward_layernorm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.13.self_attn.k_norm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.13.self_attn.k_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.13.self_attn.o_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.13.self_attn.q_norm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.13.self_attn.q_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.13.self_attn.v_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.14.mlp.down_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.14.mlp.gate_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.14.mlp.up_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.14.post_attention_layernorm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.14.post_feedforward_layernorm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.14.self_attn.k_norm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.14.self_attn.k_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.14.self_attn.o_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.14.self_attn.q_norm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.14.self_attn.q_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.14.self_attn.v_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.15.mlp.down_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.15.mlp.gate_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.15.mlp.up_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.15.post_attention_layernorm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.15.post_feedforward_layernorm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.15.self_attn.k_norm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.15.self_attn.k_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.15.self_attn.o_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.15.self_attn.q_norm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.15.self_attn.q_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.15.self_attn.v_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.16.mlp.down_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.16.mlp.gate_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.16.mlp.up_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.16.post_attention_layernorm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.16.post_feedforward_layernorm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.16.self_attn.k_norm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.16.self_attn.k_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.16.self_attn.o_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.16.self_attn.q_norm.weight": "model-00003-of-00006.safetensors",
+    "model.layers.16.self_attn.q_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.16.self_attn.v_proj.weight": "model-00003-of-00006.safetensors",
+    "model.layers.17.mlp.down_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.17.mlp.gate_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.17.mlp.up_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.17.post_attention_layernorm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.17.post_feedforward_layernorm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.17.self_attn.k_norm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.17.self_attn.k_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.17.self_attn.o_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.17.self_attn.q_norm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.17.self_attn.q_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.17.self_attn.v_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.18.mlp.down_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.18.mlp.gate_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.18.mlp.up_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.18.post_attention_layernorm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.18.post_feedforward_layernorm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.18.self_attn.k_norm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.18.self_attn.k_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.18.self_attn.o_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.18.self_attn.q_norm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.18.self_attn.q_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.18.self_attn.v_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.19.mlp.down_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.19.mlp.gate_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.19.mlp.up_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.19.post_attention_layernorm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.19.post_feedforward_layernorm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.19.self_attn.k_norm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.19.self_attn.k_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.19.self_attn.o_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.19.self_attn.q_norm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.19.self_attn.q_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.19.self_attn.v_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.2.mlp.down_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.2.mlp.gate_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.2.mlp.up_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.2.post_attention_layernorm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.2.post_feedforward_layernorm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.2.self_attn.k_norm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.2.self_attn.k_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.2.self_attn.o_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.2.self_attn.q_norm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.2.self_attn.q_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.2.self_attn.v_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.20.mlp.down_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.20.mlp.gate_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.20.mlp.up_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.20.post_attention_layernorm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.20.post_feedforward_layernorm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.20.self_attn.k_norm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.20.self_attn.k_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.20.self_attn.o_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.20.self_attn.q_norm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.20.self_attn.q_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.20.self_attn.v_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.21.mlp.down_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.21.mlp.gate_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.21.mlp.up_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.21.post_attention_layernorm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.21.post_feedforward_layernorm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.21.self_attn.k_norm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.21.self_attn.k_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.21.self_attn.o_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.21.self_attn.q_norm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.21.self_attn.q_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.21.self_attn.v_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.22.mlp.down_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.22.mlp.gate_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.22.mlp.up_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.22.post_attention_layernorm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.22.post_feedforward_layernorm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.22.self_attn.k_norm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.22.self_attn.k_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.22.self_attn.o_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.22.self_attn.q_norm.weight": "model-00004-of-00006.safetensors",
+    "model.layers.22.self_attn.q_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.22.self_attn.v_proj.weight": "model-00004-of-00006.safetensors",
+    "model.layers.23.mlp.down_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.23.mlp.gate_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.23.mlp.up_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.23.post_attention_layernorm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.23.post_feedforward_layernorm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.23.self_attn.k_norm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.23.self_attn.k_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.23.self_attn.o_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.23.self_attn.q_norm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.23.self_attn.q_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.23.self_attn.v_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.24.mlp.down_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.24.mlp.gate_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.24.mlp.up_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.24.post_attention_layernorm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.24.post_feedforward_layernorm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.24.self_attn.k_norm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.24.self_attn.k_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.24.self_attn.o_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.24.self_attn.q_norm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.24.self_attn.q_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.24.self_attn.v_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.25.mlp.down_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.25.mlp.gate_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.25.mlp.up_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.25.post_attention_layernorm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.25.post_feedforward_layernorm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.25.self_attn.k_norm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.25.self_attn.k_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.25.self_attn.o_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.25.self_attn.q_norm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.25.self_attn.q_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.25.self_attn.v_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.26.mlp.down_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.26.mlp.gate_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.26.mlp.up_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.26.post_attention_layernorm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.26.post_feedforward_layernorm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.26.self_attn.k_norm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.26.self_attn.k_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.26.self_attn.o_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.26.self_attn.q_norm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.26.self_attn.q_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.26.self_attn.v_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.27.mlp.down_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.27.mlp.gate_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.27.mlp.up_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.27.post_attention_layernorm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.27.post_feedforward_layernorm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.27.self_attn.k_norm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.27.self_attn.k_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.27.self_attn.o_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.27.self_attn.q_norm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.27.self_attn.q_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.27.self_attn.v_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.28.mlp.down_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.28.mlp.gate_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.28.mlp.up_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.28.post_attention_layernorm.weight": "model-00006-of-00006.safetensors",
+    "model.layers.28.post_feedforward_layernorm.weight": "model-00006-of-00006.safetensors",
+    "model.layers.28.self_attn.k_norm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.28.self_attn.k_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.28.self_attn.o_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.28.self_attn.q_norm.weight": "model-00005-of-00006.safetensors",
+    "model.layers.28.self_attn.q_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.28.self_attn.v_proj.weight": "model-00005-of-00006.safetensors",
+    "model.layers.29.mlp.down_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.29.mlp.gate_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.29.mlp.up_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.29.post_attention_layernorm.weight": "model-00006-of-00006.safetensors",
+    "model.layers.29.post_feedforward_layernorm.weight": "model-00006-of-00006.safetensors",
+    "model.layers.29.self_attn.k_norm.weight": "model-00006-of-00006.safetensors",
+    "model.layers.29.self_attn.k_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.29.self_attn.o_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.29.self_attn.q_norm.weight": "model-00006-of-00006.safetensors",
+    "model.layers.29.self_attn.q_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.29.self_attn.v_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.3.mlp.down_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.3.mlp.gate_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.3.mlp.up_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.3.post_attention_layernorm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.3.post_feedforward_layernorm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.3.self_attn.k_norm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.3.self_attn.k_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.3.self_attn.o_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.3.self_attn.q_norm.weight": "model-00001-of-00006.safetensors",
+    "model.layers.3.self_attn.q_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.3.self_attn.v_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.30.mlp.down_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.30.mlp.gate_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.30.mlp.up_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.30.post_attention_layernorm.weight": "model-00006-of-00006.safetensors",
+    "model.layers.30.post_feedforward_layernorm.weight": "model-00006-of-00006.safetensors",
+    "model.layers.30.self_attn.k_norm.weight": "model-00006-of-00006.safetensors",
+    "model.layers.30.self_attn.k_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.30.self_attn.o_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.30.self_attn.q_norm.weight": "model-00006-of-00006.safetensors",
+    "model.layers.30.self_attn.q_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.30.self_attn.v_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.31.mlp.down_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.31.mlp.gate_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.31.mlp.up_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.31.post_attention_layernorm.weight": "model-00006-of-00006.safetensors",
+    "model.layers.31.post_feedforward_layernorm.weight": "model-00006-of-00006.safetensors",
+    "model.layers.31.self_attn.k_norm.weight": "model-00006-of-00006.safetensors",
+    "model.layers.31.self_attn.k_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.31.self_attn.o_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.31.self_attn.q_norm.weight": "model-00006-of-00006.safetensors",
+    "model.layers.31.self_attn.q_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.31.self_attn.v_proj.weight": "model-00006-of-00006.safetensors",
+    "model.layers.4.mlp.down_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.4.mlp.gate_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.4.mlp.up_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.4.post_attention_layernorm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.4.post_feedforward_layernorm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.4.self_attn.k_norm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.4.self_attn.k_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.4.self_attn.o_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.4.self_attn.q_norm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.4.self_attn.q_proj.weight": "model-00001-of-00006.safetensors",
+    "model.layers.4.self_attn.v_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.5.mlp.down_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.5.mlp.gate_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.5.mlp.up_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.5.post_attention_layernorm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.5.post_feedforward_layernorm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.5.self_attn.k_norm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.5.self_attn.k_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.5.self_attn.o_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.5.self_attn.q_norm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.5.self_attn.q_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.5.self_attn.v_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.6.mlp.down_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.6.mlp.gate_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.6.mlp.up_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.6.post_attention_layernorm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.6.post_feedforward_layernorm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.6.self_attn.k_norm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.6.self_attn.k_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.6.self_attn.o_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.6.self_attn.q_norm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.6.self_attn.q_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.6.self_attn.v_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.7.mlp.down_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.7.mlp.gate_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.7.mlp.up_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.7.post_attention_layernorm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.7.post_feedforward_layernorm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.7.self_attn.k_norm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.7.self_attn.k_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.7.self_attn.o_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.7.self_attn.q_norm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.7.self_attn.q_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.7.self_attn.v_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.8.mlp.down_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.8.mlp.gate_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.8.mlp.up_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.8.post_attention_layernorm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.8.post_feedforward_layernorm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.8.self_attn.k_norm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.8.self_attn.k_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.8.self_attn.o_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.8.self_attn.q_norm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.8.self_attn.q_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.8.self_attn.v_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.9.mlp.down_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.9.mlp.gate_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.9.mlp.up_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.9.post_attention_layernorm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.9.post_feedforward_layernorm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.9.self_attn.k_norm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.9.self_attn.k_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.9.self_attn.o_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.9.self_attn.q_norm.weight": "model-00002-of-00006.safetensors",
+    "model.layers.9.self_attn.q_proj.weight": "model-00002-of-00006.safetensors",
+    "model.layers.9.self_attn.v_proj.weight": "model-00002-of-00006.safetensors",
+    "model.norm.weight": "model-00006-of-00006.safetensors"
+  }
+}

special_tokens_map.json

@@ -0,0 +1,30 @@
+{
+  "bos_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|pad|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

tokenizer_config.json

@@ -0,0 +1,189 @@
+{
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "100256": {
+      "content": "<|extra_id_0|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "100257": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "100258": {
+      "content": "<|fim_prefix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "100259": {
+      "content": "<|fim_middle|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "100260": {
+      "content": "<|fim_suffix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "100261": {
+      "content": "|||PHONE_NUMBER|||",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "100262": {
+      "content": "|||EMAIL_ADDRESS|||",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "100263": {
+      "content": "|||IP_ADDRESS|||",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "100264": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "100265": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "100266": {
+      "content": "<|extra_id_1|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "100267": {
+      "content": "<|extra_id_2|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "100268": {
+      "content": "<|extra_id_3|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "100269": {
+      "content": "<|extra_id_4|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "100270": {
+      "content": "<|extra_id_5|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "100271": {
+      "content": "<|extra_id_6|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "100272": {
+      "content": "<|extra_id_7|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "100273": {
+      "content": "<|extra_id_8|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "100274": {
+      "content": "<|extra_id_9|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "100275": {
+      "content": "<|extra_id_10|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "100276": {
+      "content": "<|endofprompt|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "100277": {
+      "content": "<|pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "bos_token": "<|endoftext|>",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "extra_special_tokens": {},
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_token": "<|pad|>",
+  "tokenizer_class": "GPT2Tokenizer",
+  "unk_token": "<|endoftext|>"
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:721e80e1778cdc5974559fb76036f3bd489ad9c204cc63b34e1b4276ae1c1d8e
+size 5432