Upload folder using huggingface_hub

checkpoint-1042/config.json

@@ -0,0 +1,27 @@
+{
+  "_name_or_path": "/mnt/parscratch/users/acp23ay/private/models/OLMo-2-1124-7B-ary_Arab-coli-align/",
+  "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-1042/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-1042/model-00001-of-00006.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3d4bce1a45701e55a95ce02f30455c3413855ce255ec0102192770b2d8d34e30
+size 4950070472

checkpoint-1042/model-00002-of-00006.safetensors

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

checkpoint-1042/model-00003-of-00006.safetensors

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

checkpoint-1042/model-00004-of-00006.safetensors

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

checkpoint-1042/model-00005-of-00006.safetensors

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

checkpoint-1042/model-00006-of-00006.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:16c198b2a120e6355187683d79784abf4b91d01c842ff312d327a8fc4b1d694b
+size 4614509080

checkpoint-1042/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-1042/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:de0caffbbfdc347ebb669918fcaa9412e9591282fb486e8732cf7b6c7a4106b8
+size 13055369358

checkpoint-1042/rng_state.pth

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

checkpoint-1042/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d7958675868298b58d93c029d088813924b8fdcb24ce82dad99471ebd0b58806
+size 1064

checkpoint-1042/trainer_state.json

@@ -0,0 +1,3680 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.0,
+  "eval_steps": 500,
+  "global_step": 1042,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.003838771593090211,
+      "grad_norm": 39.1374626159668,
+      "learning_rate": 3.7735849056603773e-06,
+      "loss": 5.6737,
+      "step": 2
+    },
+    {
+      "epoch": 0.007677543186180422,
+      "grad_norm": 24.925661087036133,
+      "learning_rate": 7.547169811320755e-06,
+      "loss": 5.612,
+      "step": 4
+    },
+    {
+      "epoch": 0.011516314779270634,
+      "grad_norm": 12.562352180480957,
+      "learning_rate": 1.1320754716981132e-05,
+      "loss": 5.0919,
+      "step": 6
+    },
+    {
+      "epoch": 0.015355086372360844,
+      "grad_norm": 7.011171817779541,
+      "learning_rate": 1.509433962264151e-05,
+      "loss": 4.6621,
+      "step": 8
+    },
+    {
+      "epoch": 0.019193857965451054,
+      "grad_norm": 5.962992191314697,
+      "learning_rate": 1.8867924528301888e-05,
+      "loss": 4.6095,
+      "step": 10
+    },
+    {
+      "epoch": 0.023032629558541268,
+      "grad_norm": 6.030173301696777,
+      "learning_rate": 2.2641509433962265e-05,
+      "loss": 4.6055,
+      "step": 12
+    },
+    {
+      "epoch": 0.026871401151631478,
+      "grad_norm": 5.021393775939941,
+      "learning_rate": 2.641509433962264e-05,
+      "loss": 4.2814,
+      "step": 14
+    },
+    {
+      "epoch": 0.030710172744721688,
+      "grad_norm": 4.441908359527588,
+      "learning_rate": 3.018867924528302e-05,
+      "loss": 4.2399,
+      "step": 16
+    },
+    {
+      "epoch": 0.0345489443378119,
+      "grad_norm": 3.59828782081604,
+      "learning_rate": 3.39622641509434e-05,
+      "loss": 4.1171,
+      "step": 18
+    },
+    {
+      "epoch": 0.03838771593090211,
+      "grad_norm": 3.9370954036712646,
+      "learning_rate": 3.7735849056603776e-05,
+      "loss": 4.1206,
+      "step": 20
+    },
+    {
+      "epoch": 0.04222648752399232,
+      "grad_norm": 2.9339046478271484,
+      "learning_rate": 4.150943396226415e-05,
+      "loss": 3.9446,
+      "step": 22
+    },
+    {
+      "epoch": 0.046065259117082535,
+      "grad_norm": 3.2044146060943604,
+      "learning_rate": 4.528301886792453e-05,
+      "loss": 3.7939,
+      "step": 24
+    },
+    {
+      "epoch": 0.04990403071017274,
+      "grad_norm": 3.1344337463378906,
+      "learning_rate": 4.9056603773584906e-05,
+      "loss": 3.7661,
+      "step": 26
+    },
+    {
+      "epoch": 0.053742802303262956,
+      "grad_norm": 2.632190704345703,
+      "learning_rate": 5.283018867924528e-05,
+      "loss": 3.7593,
+      "step": 28
+    },
+    {
+      "epoch": 0.05758157389635317,
+      "grad_norm": 2.875645399093628,
+      "learning_rate": 5.660377358490566e-05,
+      "loss": 3.458,
+      "step": 30
+    },
+    {
+      "epoch": 0.061420345489443376,
+      "grad_norm": 2.202479600906372,
+      "learning_rate": 6.037735849056604e-05,
+      "loss": 3.7317,
+      "step": 32
+    },
+    {
+      "epoch": 0.06525911708253358,
+      "grad_norm": 3.269009590148926,
+      "learning_rate": 6.415094339622641e-05,
+      "loss": 3.6346,
+      "step": 34
+    },
+    {
+      "epoch": 0.0690978886756238,
+      "grad_norm": 2.55442476272583,
+      "learning_rate": 6.79245283018868e-05,
+      "loss": 3.5382,
+      "step": 36
+    },
+    {
+      "epoch": 0.07293666026871401,
+      "grad_norm": 2.7956719398498535,
+      "learning_rate": 7.169811320754717e-05,
+      "loss": 3.6172,
+      "step": 38
+    },
+    {
+      "epoch": 0.07677543186180422,
+      "grad_norm": 2.430014133453369,
+      "learning_rate": 7.547169811320755e-05,
+      "loss": 3.4135,
+      "step": 40
+    },
+    {
+      "epoch": 0.08061420345489444,
+      "grad_norm": 2.5124785900115967,
+      "learning_rate": 7.924528301886794e-05,
+      "loss": 3.565,
+      "step": 42
+    },
+    {
+      "epoch": 0.08445297504798464,
+      "grad_norm": 2.733428955078125,
+      "learning_rate": 8.30188679245283e-05,
+      "loss": 3.4902,
+      "step": 44
+    },
+    {
+      "epoch": 0.08829174664107485,
+      "grad_norm": 2.421130657196045,
+      "learning_rate": 8.679245283018869e-05,
+      "loss": 3.4303,
+      "step": 46
+    },
+    {
+      "epoch": 0.09213051823416507,
+      "grad_norm": 2.4804866313934326,
+      "learning_rate": 9.056603773584906e-05,
+      "loss": 3.2669,
+      "step": 48
+    },
+    {
+      "epoch": 0.09596928982725528,
+      "grad_norm": 2.927652359008789,
+      "learning_rate": 9.433962264150944e-05,
+      "loss": 3.4791,
+      "step": 50
+    },
+    {
+      "epoch": 0.09980806142034548,
+      "grad_norm": 2.6607537269592285,
+      "learning_rate": 9.811320754716981e-05,
+      "loss": 3.5444,
+      "step": 52
+    },
+    {
+      "epoch": 0.1036468330134357,
+      "grad_norm": 2.635265827178955,
+      "learning_rate": 9.999974774092106e-05,
+      "loss": 3.2915,
+      "step": 54
+    },
+    {
+      "epoch": 0.10748560460652591,
+      "grad_norm": 2.650139093399048,
+      "learning_rate": 9.999772968356182e-05,
+      "loss": 3.3209,
+      "step": 56
+    },
+    {
+      "epoch": 0.11132437619961612,
+      "grad_norm": 3.0350704193115234,
+      "learning_rate": 9.999369365029487e-05,
+      "loss": 3.3741,
+      "step": 58
+    },
+    {
+      "epoch": 0.11516314779270634,
+      "grad_norm": 2.821930170059204,
+      "learning_rate": 9.998763980401998e-05,
+      "loss": 3.3345,
+      "step": 60
+    },
+    {
+      "epoch": 0.11900191938579655,
+      "grad_norm": 2.58774471282959,
+      "learning_rate": 9.997956838907852e-05,
+      "loss": 3.3121,
+      "step": 62
+    },
+    {
+      "epoch": 0.12284069097888675,
+      "grad_norm": 2.615741491317749,
+      "learning_rate": 9.996947973124372e-05,
+      "loss": 3.4007,
+      "step": 64
+    },
+    {
+      "epoch": 0.12667946257197696,
+      "grad_norm": 2.759786605834961,
+      "learning_rate": 9.995737423770745e-05,
+      "loss": 3.1918,
+      "step": 66
+    },
+    {
+      "epoch": 0.13051823416506717,
+      "grad_norm": 2.426225423812866,
+      "learning_rate": 9.994325239706377e-05,
+      "loss": 3.252,
+      "step": 68
+    },
+    {
+      "epoch": 0.1343570057581574,
+      "grad_norm": 2.713718891143799,
+      "learning_rate": 9.992711477928926e-05,
+      "loss": 3.125,
+      "step": 70
+    },
+    {
+      "epoch": 0.1381957773512476,
+      "grad_norm": 2.7484843730926514,
+      "learning_rate": 9.990896203571995e-05,
+      "loss": 3.361,
+      "step": 72
+    },
+    {
+      "epoch": 0.1420345489443378,
+      "grad_norm": 2.6632022857666016,
+      "learning_rate": 9.988879489902509e-05,
+      "loss": 3.2453,
+      "step": 74
+    },
+    {
+      "epoch": 0.14587332053742802,
+      "grad_norm": 2.7256522178649902,
+      "learning_rate": 9.986661418317759e-05,
+      "loss": 3.2655,
+      "step": 76
+    },
+    {
+      "epoch": 0.14971209213051823,
+      "grad_norm": 2.542839288711548,
+      "learning_rate": 9.984242078342107e-05,
+      "loss": 3.317,
+      "step": 78
+    },
+    {
+      "epoch": 0.15355086372360843,
+      "grad_norm": 2.6361520290374756,
+      "learning_rate": 9.981621567623385e-05,
+      "loss": 3.2138,
+      "step": 80
+    },
+    {
+      "epoch": 0.15738963531669867,
+      "grad_norm": 2.4558825492858887,
+      "learning_rate": 9.978799991928944e-05,
+      "loss": 3.2643,
+      "step": 82
+    },
+    {
+      "epoch": 0.16122840690978887,
+      "grad_norm": 2.361485719680786,
+      "learning_rate": 9.97577746514139e-05,
+      "loss": 3.2489,
+      "step": 84
+    },
+    {
+      "epoch": 0.16506717850287908,
+      "grad_norm": 2.5965819358825684,
+      "learning_rate": 9.972554109253989e-05,
+      "loss": 3.2573,
+      "step": 86
+    },
+    {
+      "epoch": 0.1689059500959693,
+      "grad_norm": 2.2212257385253906,
+      "learning_rate": 9.969130054365736e-05,
+      "loss": 3.1332,
+      "step": 88
+    },
+    {
+      "epoch": 0.1727447216890595,
+      "grad_norm": 2.350938320159912,
+      "learning_rate": 9.965505438676115e-05,
+      "loss": 3.1953,
+      "step": 90
+    },
+    {
+      "epoch": 0.1765834932821497,
+      "grad_norm": 2.1297547817230225,
+      "learning_rate": 9.961680408479507e-05,
+      "loss": 3.0385,
+      "step": 92
+    },
+    {
+      "epoch": 0.18042226487523993,
+      "grad_norm": 2.6407082080841064,
+      "learning_rate": 9.957655118159304e-05,
+      "loss": 3.1181,
+      "step": 94
+    },
+    {
+      "epoch": 0.18426103646833014,
+      "grad_norm": 2.049705982208252,
+      "learning_rate": 9.953429730181653e-05,
+      "loss": 3.1154,
+      "step": 96
+    },
+    {
+      "epoch": 0.18809980806142035,
+      "grad_norm": 2.397480010986328,
+      "learning_rate": 9.949004415088927e-05,
+      "loss": 3.1141,
+      "step": 98
+    },
+    {
+      "epoch": 0.19193857965451055,
+      "grad_norm": 2.1459343433380127,
+      "learning_rate": 9.944379351492818e-05,
+      "loss": 3.0847,
+      "step": 100
+    },
+    {
+      "epoch": 0.19577735124760076,
+      "grad_norm": 2.409905433654785,
+      "learning_rate": 9.939554726067141e-05,
+      "loss": 3.1939,
+      "step": 102
+    },
+    {
+      "epoch": 0.19961612284069097,
+      "grad_norm": 2.6216790676116943,
+      "learning_rate": 9.934530733540294e-05,
+      "loss": 3.1399,
+      "step": 104
+    },
+    {
+      "epoch": 0.2034548944337812,
+      "grad_norm": 2.3735368251800537,
+      "learning_rate": 9.929307576687403e-05,
+      "loss": 3.1325,
+      "step": 106
+    },
+    {
+      "epoch": 0.2072936660268714,
+      "grad_norm": 2.61091685295105,
+      "learning_rate": 9.923885466322135e-05,
+      "loss": 3.2417,
+      "step": 108
+    },
+    {
+      "epoch": 0.21113243761996162,
+      "grad_norm": 2.302191734313965,
+      "learning_rate": 9.918264621288187e-05,
+      "loss": 3.192,
+      "step": 110
+    },
+    {
+      "epoch": 0.21497120921305182,
+      "grad_norm": 2.6899170875549316,
+      "learning_rate": 9.912445268450458e-05,
+      "loss": 3.1271,
+      "step": 112
+    },
+    {
+      "epoch": 0.21880998080614203,
+      "grad_norm": 2.581143856048584,
+      "learning_rate": 9.906427642685889e-05,
+      "loss": 3.1607,
+      "step": 114
+    },
+    {
+      "epoch": 0.22264875239923224,
+      "grad_norm": 2.510422706604004,
+      "learning_rate": 9.900211986873985e-05,
+      "loss": 3.159,
+      "step": 116
+    },
+    {
+      "epoch": 0.22648752399232247,
+      "grad_norm": 2.3196523189544678,
+      "learning_rate": 9.89379855188701e-05,
+      "loss": 3.0774,
+      "step": 118
+    },
+    {
+      "epoch": 0.23032629558541268,
+      "grad_norm": 2.4071953296661377,
+      "learning_rate": 9.887187596579866e-05,
+      "loss": 3.0611,
+      "step": 120
+    },
+    {
+      "epoch": 0.23416506717850288,
+      "grad_norm": 2.733366012573242,
+      "learning_rate": 9.880379387779637e-05,
+      "loss": 3.0284,
+      "step": 122
+    },
+    {
+      "epoch": 0.2380038387715931,
+      "grad_norm": 2.3916118144989014,
+      "learning_rate": 9.873374200274826e-05,
+      "loss": 3.0748,
+      "step": 124
+    },
+    {
+      "epoch": 0.2418426103646833,
+      "grad_norm": 13.829148292541504,
+      "learning_rate": 9.866172316804266e-05,
+      "loss": 3.1429,
+      "step": 126
+    },
+    {
+      "epoch": 0.2456813819577735,
+      "grad_norm": 2.5056142807006836,
+      "learning_rate": 9.8587740280457e-05,
+      "loss": 2.9744,
+      "step": 128
+    },
+    {
+      "epoch": 0.2495201535508637,
+      "grad_norm": 2.326772689819336,
+      "learning_rate": 9.851179632604057e-05,
+      "loss": 3.136,
+      "step": 130
+    },
+    {
+      "epoch": 0.2533589251439539,
+      "grad_norm": 2.530068874359131,
+      "learning_rate": 9.843389436999396e-05,
+      "loss": 3.0951,
+      "step": 132
+    },
+    {
+      "epoch": 0.2571976967370441,
+      "grad_norm": 2.2792017459869385,
+      "learning_rate": 9.835403755654536e-05,
+      "loss": 3.1231,
+      "step": 134
+    },
+    {
+      "epoch": 0.26103646833013433,
+      "grad_norm": 2.3225743770599365,
+      "learning_rate": 9.827222910882358e-05,
+      "loss": 3.1742,
+      "step": 136
+    },
+    {
+      "epoch": 0.2648752399232246,
+      "grad_norm": 2.5904088020324707,
+      "learning_rate": 9.818847232872815e-05,
+      "loss": 3.0991,
+      "step": 138
+    },
+    {
+      "epoch": 0.2687140115163148,
+      "grad_norm": 2.1772799491882324,
+      "learning_rate": 9.81027705967958e-05,
+      "loss": 3.055,
+      "step": 140
+    },
+    {
+      "epoch": 0.272552783109405,
+      "grad_norm": 2.2711241245269775,
+      "learning_rate": 9.801512737206422e-05,
+      "loss": 3.1628,
+      "step": 142
+    },
+    {
+      "epoch": 0.2763915547024952,
+      "grad_norm": 2.514880895614624,
+      "learning_rate": 9.792554619193235e-05,
+      "loss": 2.9655,
+      "step": 144
+    },
+    {
+      "epoch": 0.2802303262955854,
+      "grad_norm": 2.381726026535034,
+      "learning_rate": 9.783403067201762e-05,
+      "loss": 2.9727,
+      "step": 146
+    },
+    {
+      "epoch": 0.2840690978886756,
+      "grad_norm": 2.410158395767212,
+      "learning_rate": 9.774058450601002e-05,
+      "loss": 3.114,
+      "step": 148
+    },
+    {
+      "epoch": 0.28790786948176583,
+      "grad_norm": 2.2826764583587646,
+      "learning_rate": 9.764521146552309e-05,
+      "loss": 3.0518,
+      "step": 150
+    },
+    {
+      "epoch": 0.29174664107485604,
+      "grad_norm": 2.6153564453125,
+      "learning_rate": 9.754791539994152e-05,
+      "loss": 3.0053,
+      "step": 152
+    },
+    {
+      "epoch": 0.29558541266794625,
+      "grad_norm": 2.96071195602417,
+      "learning_rate": 9.744870023626599e-05,
+      "loss": 2.9478,
+      "step": 154
+    },
+    {
+      "epoch": 0.29942418426103645,
+      "grad_norm": 2.3832027912139893,
+      "learning_rate": 9.734756997895448e-05,
+      "loss": 3.029,
+      "step": 156
+    },
+    {
+      "epoch": 0.30326295585412666,
+      "grad_norm": 2.4208216667175293,
+      "learning_rate": 9.724452870976085e-05,
+      "loss": 3.1009,
+      "step": 158
+    },
+    {
+      "epoch": 0.30710172744721687,
+      "grad_norm": 2.337291717529297,
+      "learning_rate": 9.713958058756985e-05,
+      "loss": 2.8477,
+      "step": 160
+    },
+    {
+      "epoch": 0.31094049904030713,
+      "grad_norm": 2.6344621181488037,
+      "learning_rate": 9.703272984822946e-05,
+      "loss": 3.079,
+      "step": 162
+    },
+    {
+      "epoch": 0.31477927063339733,
+      "grad_norm": 2.1119256019592285,
+      "learning_rate": 9.69239808043799e-05,
+      "loss": 2.9445,
+      "step": 164
+    },
+    {
+      "epoch": 0.31861804222648754,
+      "grad_norm": 2.3171935081481934,
+      "learning_rate": 9.681333784527944e-05,
+      "loss": 3.0269,
+      "step": 166
+    },
+    {
+      "epoch": 0.32245681381957775,
+      "grad_norm": 2.5069351196289062,
+      "learning_rate": 9.67008054366274e-05,
+      "loss": 2.9887,
+      "step": 168
+    },
+    {
+      "epoch": 0.32629558541266795,
+      "grad_norm": 2.182694673538208,
+      "learning_rate": 9.658638812038379e-05,
+      "loss": 3.104,
+      "step": 170
+    },
+    {
+      "epoch": 0.33013435700575816,
+      "grad_norm": 2.539505958557129,
+      "learning_rate": 9.647009051458604e-05,
+      "loss": 2.959,
+      "step": 172
+    },
+    {
+      "epoch": 0.33397312859884837,
+      "grad_norm": 2.111574649810791,
+      "learning_rate": 9.635191731316262e-05,
+      "loss": 2.9279,
+      "step": 174
+    },
+    {
+      "epoch": 0.3378119001919386,
+      "grad_norm": 2.2139837741851807,
+      "learning_rate": 9.623187328574357e-05,
+      "loss": 2.9781,
+      "step": 176
+    },
+    {
+      "epoch": 0.3416506717850288,
+      "grad_norm": 2.3730077743530273,
+      "learning_rate": 9.6109963277468e-05,
+      "loss": 2.9521,
+      "step": 178
+    },
+    {
+      "epoch": 0.345489443378119,
+      "grad_norm": 2.147397756576538,
+      "learning_rate": 9.598619220878853e-05,
+      "loss": 2.9534,
+      "step": 180
+    },
+    {
+      "epoch": 0.3493282149712092,
+      "grad_norm": 2.1337695121765137,
+      "learning_rate": 9.586056507527266e-05,
+      "loss": 2.9549,
+      "step": 182
+    },
+    {
+      "epoch": 0.3531669865642994,
+      "grad_norm": 2.397371530532837,
+      "learning_rate": 9.573308694740121e-05,
+      "loss": 2.9714,
+      "step": 184
+    },
+    {
+      "epoch": 0.3570057581573896,
+      "grad_norm": 2.208829164505005,
+      "learning_rate": 9.560376297036362e-05,
+      "loss": 2.9773,
+      "step": 186
+    },
+    {
+      "epoch": 0.36084452975047987,
+      "grad_norm": 2.0512075424194336,
+      "learning_rate": 9.547259836385031e-05,
+      "loss": 2.9261,
+      "step": 188
+    },
+    {
+      "epoch": 0.3646833013435701,
+      "grad_norm": 2.486699104309082,
+      "learning_rate": 9.533959842184195e-05,
+      "loss": 2.9401,
+      "step": 190
+    },
+    {
+      "epoch": 0.3685220729366603,
+      "grad_norm": 2.0567739009857178,
+      "learning_rate": 9.520476851239588e-05,
+      "loss": 3.0689,
+      "step": 192
+    },
+    {
+      "epoch": 0.3723608445297505,
+      "grad_norm": 2.1673007011413574,
+      "learning_rate": 9.506811407742937e-05,
+      "loss": 3.022,
+      "step": 194
+    },
+    {
+      "epoch": 0.3761996161228407,
+      "grad_norm": 2.1367907524108887,
+      "learning_rate": 9.492964063250001e-05,
+      "loss": 2.89,
+      "step": 196
+    },
+    {
+      "epoch": 0.3800383877159309,
+      "grad_norm": 2.0380022525787354,
+      "learning_rate": 9.478935376658307e-05,
+      "loss": 2.9577,
+      "step": 198
+    },
+    {
+      "epoch": 0.3838771593090211,
+      "grad_norm": 2.535660743713379,
+      "learning_rate": 9.4647259141846e-05,
+      "loss": 2.9808,
+      "step": 200
+    },
+    {
+      "epoch": 0.3877159309021113,
+      "grad_norm": 2.2464702129364014,
+      "learning_rate": 9.450336249341976e-05,
+      "loss": 3.0819,
+      "step": 202
+    },
+    {
+      "epoch": 0.3915547024952015,
+      "grad_norm": 2.316953659057617,
+      "learning_rate": 9.435766962916748e-05,
+      "loss": 2.9961,
+      "step": 204
+    },
+    {
+      "epoch": 0.39539347408829173,
+      "grad_norm": 2.124931573867798,
+      "learning_rate": 9.421018642944997e-05,
+      "loss": 2.9755,
+      "step": 206
+    },
+    {
+      "epoch": 0.39923224568138194,
+      "grad_norm": 2.299165964126587,
+      "learning_rate": 9.406091884688836e-05,
+      "loss": 2.9174,
+      "step": 208
+    },
+    {
+      "epoch": 0.40307101727447214,
+      "grad_norm": 2.156485080718994,
+      "learning_rate": 9.390987290612395e-05,
+      "loss": 2.9381,
+      "step": 210
+    },
+    {
+      "epoch": 0.4069097888675624,
+      "grad_norm": 2.207650899887085,
+      "learning_rate": 9.375705470357493e-05,
+      "loss": 2.893,
+      "step": 212
+    },
+    {
+      "epoch": 0.4107485604606526,
+      "grad_norm": 2.3216702938079834,
+      "learning_rate": 9.360247040719039e-05,
+      "loss": 2.9004,
+      "step": 214
+    },
+    {
+      "epoch": 0.4145873320537428,
+      "grad_norm": 2.2900643348693848,
+      "learning_rate": 9.344612625620132e-05,
+      "loss": 2.9238,
+      "step": 216
+    },
+    {
+      "epoch": 0.418426103646833,
+      "grad_norm": 2.157139539718628,
+      "learning_rate": 9.32880285608689e-05,
+      "loss": 2.8885,
+      "step": 218
+    },
+    {
+      "epoch": 0.42226487523992323,
+      "grad_norm": 2.0238444805145264,
+      "learning_rate": 9.312818370222962e-05,
+      "loss": 2.8743,
+      "step": 220
+    },
+    {
+      "epoch": 0.42610364683301344,
+      "grad_norm": 2.537557363510132,
+      "learning_rate": 9.296659813183794e-05,
+      "loss": 2.9497,
+      "step": 222
+    },
+    {
+      "epoch": 0.42994241842610365,
+      "grad_norm": 2.218683958053589,
+      "learning_rate": 9.280327837150571e-05,
+      "loss": 2.9384,
+      "step": 224
+    },
+    {
+      "epoch": 0.43378119001919385,
+      "grad_norm": 1.8829752206802368,
+      "learning_rate": 9.26382310130391e-05,
+      "loss": 2.9037,
+      "step": 226
+    },
+    {
+      "epoch": 0.43761996161228406,
+      "grad_norm": 2.2257378101348877,
+      "learning_rate": 9.247146271797245e-05,
+      "loss": 2.9468,
+      "step": 228
+    },
+    {
+      "epoch": 0.44145873320537427,
+      "grad_norm": 1.9686921834945679,
+      "learning_rate": 9.23029802172994e-05,
+      "loss": 2.8394,
+      "step": 230
+    },
+    {
+      "epoch": 0.44529750479846447,
+      "grad_norm": 2.367706298828125,
+      "learning_rate": 9.213279031120129e-05,
+      "loss": 2.7104,
+      "step": 232
+    },
+    {
+      "epoch": 0.4491362763915547,
+      "grad_norm": 2.439753532409668,
+      "learning_rate": 9.19608998687726e-05,
+      "loss": 2.9991,
+      "step": 234
+    },
+    {
+      "epoch": 0.45297504798464494,
+      "grad_norm": 1.997983694076538,
+      "learning_rate": 9.178731582774381e-05,
+      "loss": 2.887,
+      "step": 236
+    },
+    {
+      "epoch": 0.45681381957773515,
+      "grad_norm": 2.371289014816284,
+      "learning_rate": 9.161204519420126e-05,
+      "loss": 2.8716,
+      "step": 238
+    },
+    {
+      "epoch": 0.46065259117082535,
+      "grad_norm": 1.9359811544418335,
+      "learning_rate": 9.143509504230449e-05,
+      "loss": 2.952,
+      "step": 240
+    },
+    {
+      "epoch": 0.46449136276391556,
+      "grad_norm": 2.3096532821655273,
+      "learning_rate": 9.125647251400067e-05,
+      "loss": 2.84,
+      "step": 242
+    },
+    {
+      "epoch": 0.46833013435700577,
+      "grad_norm": 2.2196881771087646,
+      "learning_rate": 9.107618481873632e-05,
+      "loss": 2.8153,
+      "step": 244
+    },
+    {
+      "epoch": 0.472168905950096,
+      "grad_norm": 1.984487533569336,
+      "learning_rate": 9.089423923316636e-05,
+      "loss": 2.9233,
+      "step": 246
+    },
+    {
+      "epoch": 0.4760076775431862,
+      "grad_norm": 2.1207079887390137,
+      "learning_rate": 9.07106431008604e-05,
+      "loss": 2.7975,
+      "step": 248
+    },
+    {
+      "epoch": 0.4798464491362764,
+      "grad_norm": 1.9502159357070923,
+      "learning_rate": 9.052540383200633e-05,
+      "loss": 2.7933,
+      "step": 250
+    },
+    {
+      "epoch": 0.4836852207293666,
+      "grad_norm": 2.3264806270599365,
+      "learning_rate": 9.033852890311128e-05,
+      "loss": 2.948,
+      "step": 252
+    },
+    {
+      "epoch": 0.4875239923224568,
+      "grad_norm": 2.3154518604278564,
+      "learning_rate": 9.01500258566998e-05,
+      "loss": 2.8546,
+      "step": 254
+    },
+    {
+      "epoch": 0.491362763915547,
+      "grad_norm": 2.2312159538269043,
+      "learning_rate": 8.995990230100949e-05,
+      "loss": 2.8168,
+      "step": 256
+    },
+    {
+      "epoch": 0.4952015355086372,
+      "grad_norm": 2.6638643741607666,
+      "learning_rate": 8.976816590968387e-05,
+      "loss": 2.9337,
+      "step": 258
+    },
+    {
+      "epoch": 0.4990403071017274,
+      "grad_norm": 2.1371665000915527,
+      "learning_rate": 8.957482442146272e-05,
+      "loss": 2.8714,
+      "step": 260
+    },
+    {
+      "epoch": 0.5028790786948176,
+      "grad_norm": 1.9727420806884766,
+      "learning_rate": 8.937988563986963e-05,
+      "loss": 2.7677,
+      "step": 262
+    },
+    {
+      "epoch": 0.5067178502879078,
+      "grad_norm": 2.2152533531188965,
+      "learning_rate": 8.918335743289718e-05,
+      "loss": 2.9224,
+      "step": 264
+    },
+    {
+      "epoch": 0.510556621880998,
+      "grad_norm": 2.115664005279541,
+      "learning_rate": 8.898524773268926e-05,
+      "loss": 2.8969,
+      "step": 266
+    },
+    {
+      "epoch": 0.5143953934740882,
+      "grad_norm": 2.238097667694092,
+      "learning_rate": 8.8785564535221e-05,
+      "loss": 2.8914,
+      "step": 268
+    },
+    {
+      "epoch": 0.5182341650671785,
+      "grad_norm": 2.088069200515747,
+      "learning_rate": 8.858431589997597e-05,
+      "loss": 2.841,
+      "step": 270
+    },
+    {
+      "epoch": 0.5220729366602687,
+      "grad_norm": 2.0779953002929688,
+      "learning_rate": 8.838150994962093e-05,
+      "loss": 2.9019,
+      "step": 272
+    },
+    {
+      "epoch": 0.525911708253359,
+      "grad_norm": 2.2514681816101074,
+      "learning_rate": 8.817715486967803e-05,
+      "loss": 2.9179,
+      "step": 274
+    },
+    {
+      "epoch": 0.5297504798464492,
+      "grad_norm": 2.3873977661132812,
+      "learning_rate": 8.797125890819428e-05,
+      "loss": 2.8158,
+      "step": 276
+    },
+    {
+      "epoch": 0.5335892514395394,
+      "grad_norm": 1.8718311786651611,
+      "learning_rate": 8.776383037540888e-05,
+      "loss": 2.8838,
+      "step": 278
+    },
+    {
+      "epoch": 0.5374280230326296,
+      "grad_norm": 1.7923882007598877,
+      "learning_rate": 8.755487764341757e-05,
+      "loss": 2.8069,
+      "step": 280
+    },
+    {
+      "epoch": 0.5412667946257198,
+      "grad_norm": 2.197343587875366,
+      "learning_rate": 8.734440914583486e-05,
+      "loss": 2.8006,
+      "step": 282
+    },
+    {
+      "epoch": 0.54510556621881,
+      "grad_norm": 1.9336212873458862,
+      "learning_rate": 8.713243337745365e-05,
+      "loss": 2.8515,
+      "step": 284
+    },
+    {
+      "epoch": 0.5489443378119002,
+      "grad_norm": 2.2000906467437744,
+      "learning_rate": 8.691895889390227e-05,
+      "loss": 2.8156,
+      "step": 286
+    },
+    {
+      "epoch": 0.5527831094049904,
+      "grad_norm": 1.9124946594238281,
+      "learning_rate": 8.670399431129926e-05,
+      "loss": 2.6808,
+      "step": 288
+    },
+    {
+      "epoch": 0.5566218809980806,
+      "grad_norm": 2.50055193901062,
+      "learning_rate": 8.648754830590552e-05,
+      "loss": 2.9538,
+      "step": 290
+    },
+    {
+      "epoch": 0.5604606525911708,
+      "grad_norm": 2.0405330657958984,
+      "learning_rate": 8.626962961377423e-05,
+      "loss": 2.9034,
+      "step": 292
+    },
+    {
+      "epoch": 0.564299424184261,
+      "grad_norm": 2.101141929626465,
+      "learning_rate": 8.605024703039816e-05,
+      "loss": 2.8311,
+      "step": 294
+    },
+    {
+      "epoch": 0.5681381957773513,
+      "grad_norm": 1.957920789718628,
+      "learning_rate": 8.582940941035476e-05,
+      "loss": 2.8421,
+      "step": 296
+    },
+    {
+      "epoch": 0.5719769673704415,
+      "grad_norm": 2.24116849899292,
+      "learning_rate": 8.560712566694862e-05,
+      "loss": 2.9505,
+      "step": 298
+    },
+    {
+      "epoch": 0.5758157389635317,
+      "grad_norm": 2.053492546081543,
+      "learning_rate": 8.538340477185192e-05,
+      "loss": 2.8465,
+      "step": 300
+    },
+    {
+      "epoch": 0.5796545105566219,
+      "grad_norm": 2.3354341983795166,
+      "learning_rate": 8.515825575474221e-05,
+      "loss": 2.8588,
+      "step": 302
+    },
+    {
+      "epoch": 0.5834932821497121,
+      "grad_norm": 2.1598827838897705,
+      "learning_rate": 8.493168770293792e-05,
+      "loss": 2.8756,
+      "step": 304
+    },
+    {
+      "epoch": 0.5873320537428023,
+      "grad_norm": 2.1858606338500977,
+      "learning_rate": 8.47037097610317e-05,
+      "loss": 2.8049,
+      "step": 306
+    },
+    {
+      "epoch": 0.5911708253358925,
+      "grad_norm": 2.0640292167663574,
+      "learning_rate": 8.447433113052123e-05,
+      "loss": 2.8209,
+      "step": 308
+    },
+    {
+      "epoch": 0.5950095969289827,
+      "grad_norm": 2.2367477416992188,
+      "learning_rate": 8.42435610694379e-05,
+      "loss": 2.8078,
+      "step": 310
+    },
+    {
+      "epoch": 0.5988483685220729,
+      "grad_norm": 2.088447093963623,
+      "learning_rate": 8.401140889197305e-05,
+      "loss": 2.7268,
+      "step": 312
+    },
+    {
+      "epoch": 0.6026871401151631,
+      "grad_norm": 2.095773458480835,
+      "learning_rate": 8.377788396810223e-05,
+      "loss": 2.7654,
+      "step": 314
+    },
+    {
+      "epoch": 0.6065259117082533,
+      "grad_norm": 2.2851994037628174,
+      "learning_rate": 8.354299572320679e-05,
+      "loss": 2.7514,
+      "step": 316
+    },
+    {
+      "epoch": 0.6103646833013435,
+      "grad_norm": 1.7963942289352417,
+      "learning_rate": 8.330675363769356e-05,
+      "loss": 2.8477,
+      "step": 318
+    },
+    {
+      "epoch": 0.6142034548944337,
+      "grad_norm": 2.051821708679199,
+      "learning_rate": 8.306916724661224e-05,
+      "loss": 2.7684,
+      "step": 320
+    },
+    {
+      "epoch": 0.6180422264875239,
+      "grad_norm": 2.041626214981079,
+      "learning_rate": 8.283024613927054e-05,
+      "loss": 2.7836,
+      "step": 322
+    },
+    {
+      "epoch": 0.6218809980806143,
+      "grad_norm": 2.3092660903930664,
+      "learning_rate": 8.258999995884705e-05,
+      "loss": 2.722,
+      "step": 324
+    },
+    {
+      "epoch": 0.6257197696737045,
+      "grad_norm": 2.1960623264312744,
+      "learning_rate": 8.234843840200217e-05,
+      "loss": 2.7858,
+      "step": 326
+    },
+    {
+      "epoch": 0.6295585412667947,
+      "grad_norm": 1.9009817838668823,
+      "learning_rate": 8.210557121848663e-05,
+      "loss": 2.7208,
+      "step": 328
+    },
+    {
+      "epoch": 0.6333973128598849,
+      "grad_norm": 2.1019928455352783,
+      "learning_rate": 8.186140821074801e-05,
+      "loss": 2.9432,
+      "step": 330
+    },
+    {
+      "epoch": 0.6372360844529751,
+      "grad_norm": 1.9978450536727905,
+      "learning_rate": 8.161595923353516e-05,
+      "loss": 2.8016,
+      "step": 332
+    },
+    {
+      "epoch": 0.6410748560460653,
+      "grad_norm": 2.3676860332489014,
+      "learning_rate": 8.136923419350031e-05,
+      "loss": 2.8101,
+      "step": 334
+    },
+    {
+      "epoch": 0.6449136276391555,
+      "grad_norm": 2.006544589996338,
+      "learning_rate": 8.112124304879937e-05,
+      "loss": 2.7789,
+      "step": 336
+    },
+    {
+      "epoch": 0.6487523992322457,
+      "grad_norm": 2.0482800006866455,
+      "learning_rate": 8.087199580868996e-05,
+      "loss": 2.7815,
+      "step": 338
+    },
+    {
+      "epoch": 0.6525911708253359,
+      "grad_norm": 2.2319834232330322,
+      "learning_rate": 8.062150253312735e-05,
+      "loss": 2.8728,
+      "step": 340
+    },
+    {
+      "epoch": 0.6564299424184261,
+      "grad_norm": 3.3775405883789062,
+      "learning_rate": 8.03697733323585e-05,
+      "loss": 2.814,
+      "step": 342
+    },
+    {
+      "epoch": 0.6602687140115163,
+      "grad_norm": 2.242619514465332,
+      "learning_rate": 8.011681836651402e-05,
+      "loss": 2.8066,
+      "step": 344
+    },
+    {
+      "epoch": 0.6641074856046065,
+      "grad_norm": 2.016690254211426,
+      "learning_rate": 7.986264784519801e-05,
+      "loss": 2.7449,
+      "step": 346
+    },
+    {
+      "epoch": 0.6679462571976967,
+      "grad_norm": 2.1939826011657715,
+      "learning_rate": 7.960727202707605e-05,
+      "loss": 2.8091,
+      "step": 348
+    },
+    {
+      "epoch": 0.6717850287907869,
+      "grad_norm": 3.5264973640441895,
+      "learning_rate": 7.935070121946115e-05,
+      "loss": 2.8325,
+      "step": 350
+    },
+    {
+      "epoch": 0.6756238003838771,
+      "grad_norm": 2.000185489654541,
+      "learning_rate": 7.909294577789766e-05,
+      "loss": 2.789,
+      "step": 352
+    },
+    {
+      "epoch": 0.6794625719769674,
+      "grad_norm": 2.2280194759368896,
+      "learning_rate": 7.883401610574336e-05,
+      "loss": 2.8581,
+      "step": 354
+    },
+    {
+      "epoch": 0.6833013435700576,
+      "grad_norm": 1.8479573726654053,
+      "learning_rate": 7.857392265374963e-05,
+      "loss": 2.8131,
+      "step": 356
+    },
+    {
+      "epoch": 0.6871401151631478,
+      "grad_norm": 2.1635046005249023,
+      "learning_rate": 7.83126759196395e-05,
+      "loss": 2.8089,
+      "step": 358
+    },
+    {
+      "epoch": 0.690978886756238,
+      "grad_norm": 2.1297736167907715,
+      "learning_rate": 7.805028644768407e-05,
+      "loss": 2.8082,
+      "step": 360
+    },
+    {
+      "epoch": 0.6948176583493282,
+      "grad_norm": 1.9042068719863892,
+      "learning_rate": 7.778676482827686e-05,
+      "loss": 2.8027,
+      "step": 362
+    },
+    {
+      "epoch": 0.6986564299424184,
+      "grad_norm": 1.8313944339752197,
+      "learning_rate": 7.752212169750641e-05,
+      "loss": 2.8438,
+      "step": 364
+    },
+    {
+      "epoch": 0.7024952015355086,
+      "grad_norm": 2.036674976348877,
+      "learning_rate": 7.725636773672693e-05,
+      "loss": 2.8047,
+      "step": 366
+    },
+    {
+      "epoch": 0.7063339731285988,
+      "grad_norm": 2.086843490600586,
+      "learning_rate": 7.69895136721273e-05,
+      "loss": 2.6723,
+      "step": 368
+    },
+    {
+      "epoch": 0.710172744721689,
+      "grad_norm": 2.046743154525757,
+      "learning_rate": 7.672157027429803e-05,
+      "loss": 2.7864,
+      "step": 370
+    },
+    {
+      "epoch": 0.7140115163147792,
+      "grad_norm": 2.1428844928741455,
+      "learning_rate": 7.645254835779657e-05,
+      "loss": 2.841,
+      "step": 372
+    },
+    {
+      "epoch": 0.7178502879078695,
+      "grad_norm": 2.015566825866699,
+      "learning_rate": 7.618245878071091e-05,
+      "loss": 2.8675,
+      "step": 374
+    },
+    {
+      "epoch": 0.7216890595009597,
+      "grad_norm": 2.0343127250671387,
+      "learning_rate": 7.591131244422118e-05,
+      "loss": 2.8092,
+      "step": 376
+    },
+    {
+      "epoch": 0.72552783109405,
+      "grad_norm": 1.9547837972640991,
+      "learning_rate": 7.563912029215983e-05,
+      "loss": 2.7245,
+      "step": 378
+    },
+    {
+      "epoch": 0.7293666026871402,
+      "grad_norm": 1.8891335725784302,
+      "learning_rate": 7.536589331056976e-05,
+      "loss": 2.8459,
+      "step": 380
+    },
+    {
+      "epoch": 0.7332053742802304,
+      "grad_norm": 2.191988706588745,
+      "learning_rate": 7.509164252726107e-05,
+      "loss": 2.8502,
+      "step": 382
+    },
+    {
+      "epoch": 0.7370441458733206,
+      "grad_norm": 1.8953111171722412,
+      "learning_rate": 7.481637901136577e-05,
+      "loss": 2.7664,
+      "step": 384
+    },
+    {
+      "epoch": 0.7408829174664108,
+      "grad_norm": 1.8514469861984253,
+      "learning_rate": 7.454011387289127e-05,
+      "loss": 2.7179,
+      "step": 386
+    },
+    {
+      "epoch": 0.744721689059501,
+      "grad_norm": 1.9623198509216309,
+      "learning_rate": 7.42628582622717e-05,
+      "loss": 2.65,
+      "step": 388
+    },
+    {
+      "epoch": 0.7485604606525912,
+      "grad_norm": 1.9796512126922607,
+      "learning_rate": 7.398462336991801e-05,
+      "loss": 2.6551,
+      "step": 390
+    },
+    {
+      "epoch": 0.7523992322456814,
+      "grad_norm": 1.9456619024276733,
+      "learning_rate": 7.370542042576634e-05,
+      "loss": 2.7129,
+      "step": 392
+    },
+    {
+      "epoch": 0.7562380038387716,
+      "grad_norm": 2.237915277481079,
+      "learning_rate": 7.342526069882465e-05,
+      "loss": 2.8557,
+      "step": 394
+    },
+    {
+      "epoch": 0.7600767754318618,
+      "grad_norm": 1.9474000930786133,
+      "learning_rate": 7.314415549671795e-05,
+      "loss": 2.8152,
+      "step": 396
+    },
+    {
+      "epoch": 0.763915547024952,
+      "grad_norm": 2.2595555782318115,
+      "learning_rate": 7.286211616523193e-05,
+      "loss": 2.8161,
+      "step": 398
+    },
+    {
+      "epoch": 0.7677543186180422,
+      "grad_norm": 2.072669267654419,
+      "learning_rate": 7.257915408785498e-05,
+      "loss": 2.8623,
+      "step": 400
+    },
+    {
+      "epoch": 0.7715930902111324,
+      "grad_norm": 2.018782138824463,
+      "learning_rate": 7.229528068531882e-05,
+      "loss": 2.7831,
+      "step": 402
+    },
+    {
+      "epoch": 0.7754318618042226,
+      "grad_norm": 1.9737513065338135,
+      "learning_rate": 7.201050741513736e-05,
+      "loss": 2.6135,
+      "step": 404
+    },
+    {
+      "epoch": 0.7792706333973128,
+      "grad_norm": 1.8319437503814697,
+      "learning_rate": 7.17248457711445e-05,
+      "loss": 2.783,
+      "step": 406
+    },
+    {
+      "epoch": 0.783109404990403,
+      "grad_norm": 1.8370659351348877,
+      "learning_rate": 7.14383072830301e-05,
+      "loss": 2.7643,
+      "step": 408
+    },
+    {
+      "epoch": 0.7869481765834933,
+      "grad_norm": 2.0122272968292236,
+      "learning_rate": 7.115090351587454e-05,
+      "loss": 2.7853,
+      "step": 410
+    },
+    {
+      "epoch": 0.7907869481765835,
+      "grad_norm": 2.0002529621124268,
+      "learning_rate": 7.086264606968214e-05,
+      "loss": 2.7595,
+      "step": 412
+    },
+    {
+      "epoch": 0.7946257197696737,
+      "grad_norm": 2.0362918376922607,
+      "learning_rate": 7.05735465789128e-05,
+      "loss": 2.9258,
+      "step": 414
+    },
+    {
+      "epoch": 0.7984644913627639,
+      "grad_norm": 2.0595998764038086,
+      "learning_rate": 7.028361671201245e-05,
+      "loss": 2.7771,
+      "step": 416
+    },
+    {
+      "epoch": 0.8023032629558541,
+      "grad_norm": 2.0576417446136475,
+      "learning_rate": 6.99928681709422e-05,
+      "loss": 2.7221,
+      "step": 418
+    },
+    {
+      "epoch": 0.8061420345489443,
+      "grad_norm": 1.9781322479248047,
+      "learning_rate": 6.970131269070592e-05,
+      "loss": 2.8678,
+      "step": 420
+    },
+    {
+      "epoch": 0.8099808061420346,
+      "grad_norm": 2.429964780807495,
+      "learning_rate": 6.940896203887659e-05,
+      "loss": 2.7395,
+      "step": 422
+    },
+    {
+      "epoch": 0.8138195777351248,
+      "grad_norm": 2.1173887252807617,
+      "learning_rate": 6.911582801512147e-05,
+      "loss": 2.8196,
+      "step": 424
+    },
+    {
+      "epoch": 0.817658349328215,
+      "grad_norm": 2.0920073986053467,
+      "learning_rate": 6.882192245072569e-05,
+      "loss": 2.7426,
+      "step": 426
+    },
+    {
+      "epoch": 0.8214971209213052,
+      "grad_norm": 1.986169695854187,
+      "learning_rate": 6.852725720811488e-05,
+      "loss": 2.7175,
+      "step": 428
+    },
+    {
+      "epoch": 0.8253358925143954,
+      "grad_norm": 1.837952733039856,
+      "learning_rate": 6.823184418037624e-05,
+      "loss": 2.7456,
+      "step": 430
+    },
+    {
+      "epoch": 0.8291746641074856,
+      "grad_norm": 2.084582567214966,
+      "learning_rate": 6.793569529077864e-05,
+      "loss": 2.6918,
+      "step": 432
+    },
+    {
+      "epoch": 0.8330134357005758,
+      "grad_norm": 1.9959920644760132,
+      "learning_rate": 6.76388224922913e-05,
+      "loss": 2.7062,
+      "step": 434
+    },
+    {
+      "epoch": 0.836852207293666,
+      "grad_norm": 2.0004842281341553,
+      "learning_rate": 6.734123776710137e-05,
+      "loss": 2.7469,
+      "step": 436
+    },
+    {
+      "epoch": 0.8406909788867563,
+      "grad_norm": 2.0607264041900635,
+      "learning_rate": 6.704295312613037e-05,
+      "loss": 2.6908,
+      "step": 438
+    },
+    {
+      "epoch": 0.8445297504798465,
+      "grad_norm": 1.8044137954711914,
+      "learning_rate": 6.674398060854931e-05,
+      "loss": 2.7281,
+      "step": 440
+    },
+    {
+      "epoch": 0.8483685220729367,
+      "grad_norm": 1.741652488708496,
+      "learning_rate": 6.644433228129287e-05,
+      "loss": 2.7622,
+      "step": 442
+    },
+    {
+      "epoch": 0.8522072936660269,
+      "grad_norm": 2.165587902069092,
+      "learning_rate": 6.614402023857232e-05,
+      "loss": 2.8081,
+      "step": 444
+    },
+    {
+      "epoch": 0.8560460652591171,
+      "grad_norm": 1.9901096820831299,
+      "learning_rate": 6.584305660138734e-05,
+      "loss": 2.8187,
+      "step": 446
+    },
+    {
+      "epoch": 0.8598848368522073,
+      "grad_norm": 1.9680149555206299,
+      "learning_rate": 6.554145351703689e-05,
+      "loss": 2.7813,
+      "step": 448
+    },
+    {
+      "epoch": 0.8637236084452975,
+      "grad_norm": 2.150981903076172,
+      "learning_rate": 6.523922315862887e-05,
+      "loss": 2.6663,
+      "step": 450
+    },
+    {
+      "epoch": 0.8675623800383877,
+      "grad_norm": 2.4717190265655518,
+      "learning_rate": 6.493637772458879e-05,
+      "loss": 2.7571,
+      "step": 452
+    },
+    {
+      "epoch": 0.8714011516314779,
+      "grad_norm": 2.0950613021850586,
+      "learning_rate": 6.463292943816746e-05,
+      "loss": 2.8617,
+      "step": 454
+    },
+    {
+      "epoch": 0.8752399232245681,
+      "grad_norm": 1.7509244680404663,
+      "learning_rate": 6.432889054694764e-05,
+      "loss": 2.7157,
+      "step": 456
+    },
+    {
+      "epoch": 0.8790786948176583,
+      "grad_norm": 3.837655544281006,
+      "learning_rate": 6.402427332234965e-05,
+      "loss": 2.6557,
+      "step": 458
+    },
+    {
+      "epoch": 0.8829174664107485,
+      "grad_norm": 1.9463653564453125,
+      "learning_rate": 6.371909005913618e-05,
+      "loss": 2.7937,
+      "step": 460
+    },
+    {
+      "epoch": 0.8867562380038387,
+      "grad_norm": 1.785074234008789,
+      "learning_rate": 6.341335307491595e-05,
+      "loss": 2.7041,
+      "step": 462
+    },
+    {
+      "epoch": 0.8905950095969289,
+      "grad_norm": 2.044428586959839,
+      "learning_rate": 6.310707470964668e-05,
+      "loss": 2.7131,
+      "step": 464
+    },
+    {
+      "epoch": 0.8944337811900192,
+      "grad_norm": 2.0175092220306396,
+      "learning_rate": 6.280026732513689e-05,
+      "loss": 2.6959,
+      "step": 466
+    },
+    {
+      "epoch": 0.8982725527831094,
+      "grad_norm": 1.9430804252624512,
+      "learning_rate": 6.249294330454705e-05,
+      "loss": 2.7563,
+      "step": 468
+    },
+    {
+      "epoch": 0.9021113243761996,
+      "grad_norm": 1.999422311782837,
+      "learning_rate": 6.21851150518898e-05,
+      "loss": 2.7608,
+      "step": 470
+    },
+    {
+      "epoch": 0.9059500959692899,
+      "grad_norm": 2.7273123264312744,
+      "learning_rate": 6.187679499152919e-05,
+      "loss": 2.7552,
+      "step": 472
+    },
+    {
+      "epoch": 0.9097888675623801,
+      "grad_norm": 1.9456199407577515,
+      "learning_rate": 6.15679955676794e-05,
+      "loss": 2.689,
+      "step": 474
+    },
+    {
+      "epoch": 0.9136276391554703,
+      "grad_norm": 1.8011109828948975,
+      "learning_rate": 6.125872924390226e-05,
+      "loss": 2.6342,
+      "step": 476
+    },
+    {
+      "epoch": 0.9174664107485605,
+      "grad_norm": 1.790582299232483,
+      "learning_rate": 6.094900850260439e-05,
+      "loss": 2.7611,
+      "step": 478
+    },
+    {
+      "epoch": 0.9213051823416507,
+      "grad_norm": 2.047144889831543,
+      "learning_rate": 6.0638845844533254e-05,
+      "loss": 2.7965,
+      "step": 480
+    },
+    {
+      "epoch": 0.9251439539347409,
+      "grad_norm": 1.7579201459884644,
+      "learning_rate": 6.032825378827273e-05,
+      "loss": 2.7516,
+      "step": 482
+    },
+    {
+      "epoch": 0.9289827255278311,
+      "grad_norm": 1.806973934173584,
+      "learning_rate": 6.0017244869737746e-05,
+      "loss": 2.7111,
+      "step": 484
+    },
+    {
+      "epoch": 0.9328214971209213,
+      "grad_norm": 1.781657099723816,
+      "learning_rate": 5.970583164166838e-05,
+      "loss": 2.7412,
+      "step": 486
+    },
+    {
+      "epoch": 0.9366602687140115,
+      "grad_norm": 1.904548168182373,
+      "learning_rate": 5.9394026673123157e-05,
+      "loss": 2.7388,
+      "step": 488
+    },
+    {
+      "epoch": 0.9404990403071017,
+      "grad_norm": 1.8767707347869873,
+      "learning_rate": 5.908184254897182e-05,
+      "loss": 2.694,
+      "step": 490
+    },
+    {
+      "epoch": 0.944337811900192,
+      "grad_norm": 1.9801278114318848,
+      "learning_rate": 5.8769291869387335e-05,
+      "loss": 2.7866,
+      "step": 492
+    },
+    {
+      "epoch": 0.9481765834932822,
+      "grad_norm": 2.026012897491455,
+      "learning_rate": 5.845638724933728e-05,
+      "loss": 2.7375,
+      "step": 494
+    },
+    {
+      "epoch": 0.9520153550863724,
+      "grad_norm": 1.8350569009780884,
+      "learning_rate": 5.814314131807486e-05,
+      "loss": 2.7364,
+      "step": 496
+    },
+    {
+      "epoch": 0.9558541266794626,
+      "grad_norm": 1.9296355247497559,
+      "learning_rate": 5.7829566718628946e-05,
+      "loss": 2.682,
+      "step": 498
+    },
+    {
+      "epoch": 0.9596928982725528,
+      "grad_norm": 1.8304519653320312,
+      "learning_rate": 5.751567610729398e-05,
+      "loss": 2.6421,
+      "step": 500
+    },
+    {
+      "epoch": 0.963531669865643,
+      "grad_norm": 1.8141051530838013,
+      "learning_rate": 5.7201482153119015e-05,
+      "loss": 2.6552,
+      "step": 502
+    },
+    {
+      "epoch": 0.9673704414587332,
+      "grad_norm": 1.9663299322128296,
+      "learning_rate": 5.6886997537396494e-05,
+      "loss": 2.6086,
+      "step": 504
+    },
+    {
+      "epoch": 0.9712092130518234,
+      "grad_norm": 1.9281374216079712,
+      "learning_rate": 5.6572234953150305e-05,
+      "loss": 2.7493,
+      "step": 506
+    },
+    {
+      "epoch": 0.9750479846449136,
+      "grad_norm": 1.7819768190383911,
+      "learning_rate": 5.6257207104623524e-05,
+      "loss": 2.6959,
+      "step": 508
+    },
+    {
+      "epoch": 0.9788867562380038,
+      "grad_norm": 1.8469796180725098,
+      "learning_rate": 5.594192670676568e-05,
+      "loss": 2.6471,
+      "step": 510
+    },
+    {
+      "epoch": 0.982725527831094,
+      "grad_norm": 1.8483726978302002,
+      "learning_rate": 5.5626406484719506e-05,
+      "loss": 2.699,
+      "step": 512
+    },
+    {
+      "epoch": 0.9865642994241842,
+      "grad_norm": 1.8236881494522095,
+      "learning_rate": 5.5310659173307366e-05,
+      "loss": 2.746,
+      "step": 514
+    },
+    {
+      "epoch": 0.9904030710172744,
+      "grad_norm": 1.7808964252471924,
+      "learning_rate": 5.499469751651728e-05,
+      "loss": 2.7885,
+      "step": 516
+    },
+    {
+      "epoch": 0.9942418426103646,
+      "grad_norm": 1.828331708908081,
+      "learning_rate": 5.467853426698853e-05,
+      "loss": 2.7437,
+      "step": 518
+    },
+    {
+      "epoch": 0.9980806142034548,
+      "grad_norm": 1.7730071544647217,
+      "learning_rate": 5.43621821854969e-05,
+      "loss": 2.6718,
+      "step": 520
+    },
+    {
+      "epoch": 1.0019193857965452,
+      "grad_norm": 1.7469446659088135,
+      "learning_rate": 5.4045654040439776e-05,
+      "loss": 2.6736,
+      "step": 522
+    },
+    {
+      "epoch": 1.0057581573896353,
+      "grad_norm": 1.751255989074707,
+      "learning_rate": 5.3728962607320654e-05,
+      "loss": 2.4882,
+      "step": 524
+    },
+    {
+      "epoch": 1.0095969289827256,
+      "grad_norm": 1.642378330230713,
+      "learning_rate": 5.341212066823355e-05,
+      "loss": 2.4544,
+      "step": 526
+    },
+    {
+      "epoch": 1.0134357005758157,
+      "grad_norm": 1.7672804594039917,
+      "learning_rate": 5.309514101134715e-05,
+      "loss": 2.4559,
+      "step": 528
+    },
+    {
+      "epoch": 1.017274472168906,
+      "grad_norm": 1.9962915182113647,
+      "learning_rate": 5.2778036430388554e-05,
+      "loss": 2.2846,
+      "step": 530
+    },
+    {
+      "epoch": 1.021113243761996,
+      "grad_norm": 1.837528109550476,
+      "learning_rate": 5.246081972412702e-05,
+      "loss": 2.5115,
+      "step": 532
+    },
+    {
+      "epoch": 1.0249520153550864,
+      "grad_norm": 1.781503438949585,
+      "learning_rate": 5.214350369585731e-05,
+      "loss": 2.4044,
+      "step": 534
+    },
+    {
+      "epoch": 1.0287907869481765,
+      "grad_norm": 2.2169623374938965,
+      "learning_rate": 5.182610115288295e-05,
+      "loss": 2.4961,
+      "step": 536
+    },
+    {
+      "epoch": 1.0326295585412668,
+      "grad_norm": 1.8776298761367798,
+      "learning_rate": 5.150862490599934e-05,
+      "loss": 2.4781,
+      "step": 538
+    },
+    {
+      "epoch": 1.036468330134357,
+      "grad_norm": 1.710525631904602,
+      "learning_rate": 5.119108776897664e-05,
+      "loss": 2.5032,
+      "step": 540
+    },
+    {
+      "epoch": 1.0403071017274472,
+      "grad_norm": 1.8963370323181152,
+      "learning_rate": 5.0873502558042665e-05,
+      "loss": 2.3898,
+      "step": 542
+    },
+    {
+      "epoch": 1.0441458733205373,
+      "grad_norm": 1.9298845529556274,
+      "learning_rate": 5.055588209136548e-05,
+      "loss": 2.5535,
+      "step": 544
+    },
+    {
+      "epoch": 1.0479846449136276,
+      "grad_norm": 1.697884440422058,
+      "learning_rate": 5.0238239188536215e-05,
+      "loss": 2.3346,
+      "step": 546
+    },
+    {
+      "epoch": 1.051823416506718,
+      "grad_norm": 1.7197191715240479,
+      "learning_rate": 4.992058667005149e-05,
+      "loss": 2.5456,
+      "step": 548
+    },
+    {
+      "epoch": 1.055662188099808,
+      "grad_norm": 1.8300470113754272,
+      "learning_rate": 4.9602937356796014e-05,
+      "loss": 2.3898,
+      "step": 550
+    },
+    {
+      "epoch": 1.0595009596928984,
+      "grad_norm": 1.945226788520813,
+      "learning_rate": 4.92853040695252e-05,
+      "loss": 2.4404,
+      "step": 552
+    },
+    {
+      "epoch": 1.0633397312859885,
+      "grad_norm": 1.791162371635437,
+      "learning_rate": 4.896769962834754e-05,
+      "loss": 2.3909,
+      "step": 554
+    },
+    {
+      "epoch": 1.0671785028790788,
+      "grad_norm": 1.9331914186477661,
+      "learning_rate": 4.865013685220733e-05,
+      "loss": 2.4176,
+      "step": 556
+    },
+    {
+      "epoch": 1.0710172744721689,
+      "grad_norm": 2.0219547748565674,
+      "learning_rate": 4.83326285583672e-05,
+      "loss": 2.5047,
+      "step": 558
+    },
+    {
+      "epoch": 1.0748560460652592,
+      "grad_norm": 2.1636030673980713,
+      "learning_rate": 4.801518756189074e-05,
+      "loss": 2.3333,
+      "step": 560
+    },
+    {
+      "epoch": 1.0786948176583493,
+      "grad_norm": 1.8409587144851685,
+      "learning_rate": 4.76978266751254e-05,
+      "loss": 2.5018,
+      "step": 562
+    },
+    {
+      "epoch": 1.0825335892514396,
+      "grad_norm": 1.8520073890686035,
+      "learning_rate": 4.7380558707185284e-05,
+      "loss": 2.2061,
+      "step": 564
+    },
+    {
+      "epoch": 1.0863723608445297,
+      "grad_norm": 1.7372022867202759,
+      "learning_rate": 4.7063396463434104e-05,
+      "loss": 2.4486,
+      "step": 566
+    },
+    {
+      "epoch": 1.09021113243762,
+      "grad_norm": 1.938421607017517,
+      "learning_rate": 4.674635274496847e-05,
+      "loss": 2.5271,
+      "step": 568
+    },
+    {
+      "epoch": 1.0940499040307101,
+      "grad_norm": 1.6374491453170776,
+      "learning_rate": 4.642944034810113e-05,
+      "loss": 2.4359,
+      "step": 570
+    },
+    {
+      "epoch": 1.0978886756238004,
+      "grad_norm": 1.8235459327697754,
+      "learning_rate": 4.61126720638445e-05,
+      "loss": 2.2379,
+      "step": 572
+    },
+    {
+      "epoch": 1.1017274472168905,
+      "grad_norm": 1.8577545881271362,
+      "learning_rate": 4.5796060677394444e-05,
+      "loss": 2.3655,
+      "step": 574
+    },
+    {
+      "epoch": 1.1055662188099808,
+      "grad_norm": 1.7996071577072144,
+      "learning_rate": 4.547961896761419e-05,
+      "loss": 2.4081,
+      "step": 576
+    },
+    {
+      "epoch": 1.109404990403071,
+      "grad_norm": 1.731217384338379,
+      "learning_rate": 4.51633597065186e-05,
+      "loss": 2.355,
+      "step": 578
+    },
+    {
+      "epoch": 1.1132437619961613,
+      "grad_norm": 1.9481264352798462,
+      "learning_rate": 4.484729565875865e-05,
+      "loss": 2.3707,
+      "step": 580
+    },
+    {
+      "epoch": 1.1170825335892514,
+      "grad_norm": 1.8329133987426758,
+      "learning_rate": 4.4531439581106295e-05,
+      "loss": 2.3278,
+      "step": 582
+    },
+    {
+      "epoch": 1.1209213051823417,
+      "grad_norm": 1.7967745065689087,
+      "learning_rate": 4.421580422193943e-05,
+      "loss": 2.3924,
+      "step": 584
+    },
+    {
+      "epoch": 1.1247600767754318,
+      "grad_norm": 1.763254165649414,
+      "learning_rate": 4.390040232072755e-05,
+      "loss": 2.4151,
+      "step": 586
+    },
+    {
+      "epoch": 1.128598848368522,
+      "grad_norm": 1.6523996591567993,
+      "learning_rate": 4.358524660751746e-05,
+      "loss": 2.2619,
+      "step": 588
+    },
+    {
+      "epoch": 1.1324376199616122,
+      "grad_norm": 1.8677583932876587,
+      "learning_rate": 4.327034980241941e-05,
+      "loss": 2.4496,
+      "step": 590
+    },
+    {
+      "epoch": 1.1362763915547025,
+      "grad_norm": 2.021632432937622,
+      "learning_rate": 4.2955724615093844e-05,
+      "loss": 2.4942,
+      "step": 592
+    },
+    {
+      "epoch": 1.1401151631477928,
+      "grad_norm": 1.7808256149291992,
+      "learning_rate": 4.264138374423835e-05,
+      "loss": 2.2299,
+      "step": 594
+    },
+    {
+      "epoch": 1.143953934740883,
+      "grad_norm": 1.8527741432189941,
+      "learning_rate": 4.232733987707505e-05,
+      "loss": 2.4934,
+      "step": 596
+    },
+    {
+      "epoch": 1.147792706333973,
+      "grad_norm": 1.8033527135849,
+      "learning_rate": 4.2013605688838656e-05,
+      "loss": 2.3155,
+      "step": 598
+    },
+    {
+      "epoch": 1.1516314779270633,
+      "grad_norm": 1.6837481260299683,
+      "learning_rate": 4.1700193842264814e-05,
+      "loss": 2.4002,
+      "step": 600
+    },
+    {
+      "epoch": 1.1554702495201536,
+      "grad_norm": 1.6659997701644897,
+      "learning_rate": 4.138711698707899e-05,
+      "loss": 2.3382,
+      "step": 602
+    },
+    {
+      "epoch": 1.1593090211132437,
+      "grad_norm": 1.766233205795288,
+      "learning_rate": 4.107438775948598e-05,
+      "loss": 2.3369,
+      "step": 604
+    },
+    {
+      "epoch": 1.163147792706334,
+      "grad_norm": 1.7691415548324585,
+      "learning_rate": 4.076201878165985e-05,
+      "loss": 2.3288,
+      "step": 606
+    },
+    {
+      "epoch": 1.1669865642994242,
+      "grad_norm": 1.904175043106079,
+      "learning_rate": 4.0450022661234496e-05,
+      "loss": 2.4264,
+      "step": 608
+    },
+    {
+      "epoch": 1.1708253358925145,
+      "grad_norm": 1.8914724588394165,
+      "learning_rate": 4.0138411990794786e-05,
+      "loss": 2.4262,
+      "step": 610
+    },
+    {
+      "epoch": 1.1746641074856046,
+      "grad_norm": 1.8268029689788818,
+      "learning_rate": 3.982719934736832e-05,
+      "loss": 2.2978,
+      "step": 612
+    },
+    {
+      "epoch": 1.1785028790786949,
+      "grad_norm": 2.129481077194214,
+      "learning_rate": 3.9516397291917754e-05,
+      "loss": 2.4071,
+      "step": 614
+    },
+    {
+      "epoch": 1.182341650671785,
+      "grad_norm": 1.8245187997817993,
+      "learning_rate": 3.920601836883389e-05,
+      "loss": 2.196,
+      "step": 616
+    },
+    {
+      "epoch": 1.1861804222648753,
+      "grad_norm": 1.8345937728881836,
+      "learning_rate": 3.889607510542936e-05,
+      "loss": 2.3043,
+      "step": 618
+    },
+    {
+      "epoch": 1.1900191938579654,
+      "grad_norm": 2.040839672088623,
+      "learning_rate": 3.858658001143289e-05,
+      "loss": 2.4619,
+      "step": 620
+    },
+    {
+      "epoch": 1.1938579654510557,
+      "grad_norm": 1.7487252950668335,
+      "learning_rate": 3.82775455784846e-05,
+      "loss": 2.3152,
+      "step": 622
+    },
+    {
+      "epoch": 1.1976967370441458,
+      "grad_norm": 1.8239243030548096,
+      "learning_rate": 3.796898427963163e-05,
+      "loss": 2.4258,
+      "step": 624
+    },
+    {
+      "epoch": 1.2015355086372361,
+      "grad_norm": 1.7934041023254395,
+      "learning_rate": 3.7660908568824805e-05,
+      "loss": 2.5038,
+      "step": 626
+    },
+    {
+      "epoch": 1.2053742802303262,
+      "grad_norm": 1.8060568571090698,
+      "learning_rate": 3.735333088041596e-05,
+      "loss": 2.3785,
+      "step": 628
+    },
+    {
+      "epoch": 1.2092130518234165,
+      "grad_norm": 1.7633274793624878,
+      "learning_rate": 3.704626362865612e-05,
+      "loss": 2.4583,
+      "step": 630
+    },
+    {
+      "epoch": 1.2130518234165066,
+      "grad_norm": 1.7965344190597534,
+      "learning_rate": 3.673971920719431e-05,
+      "loss": 2.3948,
+      "step": 632
+    },
+    {
+      "epoch": 1.216890595009597,
+      "grad_norm": 1.8032282590866089,
+      "learning_rate": 3.643370998857748e-05,
+      "loss": 2.4077,
+      "step": 634
+    },
+    {
+      "epoch": 1.220729366602687,
+      "grad_norm": 1.7689719200134277,
+      "learning_rate": 3.612824832375109e-05,
+      "loss": 2.19,
+      "step": 636
+    },
+    {
+      "epoch": 1.2245681381957774,
+      "grad_norm": 1.732125997543335,
+      "learning_rate": 3.5823346541560495e-05,
+      "loss": 2.2006,
+      "step": 638
+    },
+    {
+      "epoch": 1.2284069097888675,
+      "grad_norm": 1.9386759996414185,
+      "learning_rate": 3.551901694825352e-05,
+      "loss": 2.329,
+      "step": 640
+    },
+    {
+      "epoch": 1.2322456813819578,
+      "grad_norm": 1.8066078424453735,
+      "learning_rate": 3.521527182698365e-05,
+      "loss": 2.3122,
+      "step": 642
+    },
+    {
+      "epoch": 1.236084452975048,
+      "grad_norm": 1.6884784698486328,
+      "learning_rate": 3.491212343731428e-05,
+      "loss": 2.5158,
+      "step": 644
+    },
+    {
+      "epoch": 1.2399232245681382,
+      "grad_norm": 1.7934367656707764,
+      "learning_rate": 3.4609584014723904e-05,
+      "loss": 2.3961,
+      "step": 646
+    },
+    {
+      "epoch": 1.2437619961612283,
+      "grad_norm": 1.9283524751663208,
+      "learning_rate": 3.430766577011231e-05,
+      "loss": 2.3866,
+      "step": 648
+    },
+    {
+      "epoch": 1.2476007677543186,
+      "grad_norm": 1.7128459215164185,
+      "learning_rate": 3.400638088930766e-05,
+      "loss": 2.4371,
+      "step": 650
+    },
+    {
+      "epoch": 1.251439539347409,
+      "grad_norm": 2.000603199005127,
+      "learning_rate": 3.370574153257474e-05,
+      "loss": 2.568,
+      "step": 652
+    },
+    {
+      "epoch": 1.255278310940499,
+      "grad_norm": 1.77023184299469,
+      "learning_rate": 3.340575983412412e-05,
+      "loss": 2.3062,
+      "step": 654
+    },
+    {
+      "epoch": 1.2591170825335891,
+      "grad_norm": 1.7273350954055786,
+      "learning_rate": 3.3106447901622324e-05,
+      "loss": 2.4323,
+      "step": 656
+    },
+    {
+      "epoch": 1.2629558541266794,
+      "grad_norm": 1.8371007442474365,
+      "learning_rate": 3.280781781570328e-05,
+      "loss": 2.2954,
+      "step": 658
+    },
+    {
+      "epoch": 1.2667946257197698,
+      "grad_norm": 1.7804073095321655,
+      "learning_rate": 3.250988162948067e-05,
+      "loss": 2.4928,
+      "step": 660
+    },
+    {
+      "epoch": 1.2706333973128598,
+      "grad_norm": 1.7903341054916382,
+      "learning_rate": 3.221265136806139e-05,
+      "loss": 2.3969,
+      "step": 662
+    },
+    {
+      "epoch": 1.2744721689059502,
+      "grad_norm": 1.8445686101913452,
+      "learning_rate": 3.1916139028060315e-05,
+      "loss": 2.3781,
+      "step": 664
+    },
+    {
+      "epoch": 1.2783109404990403,
+      "grad_norm": 1.95808744430542,
+      "learning_rate": 3.162035657711604e-05,
+      "loss": 2.5179,
+      "step": 666
+    },
+    {
+      "epoch": 1.2821497120921306,
+      "grad_norm": 1.6103456020355225,
+      "learning_rate": 3.132531595340779e-05,
+      "loss": 2.4816,
+      "step": 668
+    },
+    {
+      "epoch": 1.2859884836852207,
+      "grad_norm": 1.7648108005523682,
+      "learning_rate": 3.1031029065173704e-05,
+      "loss": 2.5183,
+      "step": 670
+    },
+    {
+      "epoch": 1.289827255278311,
+      "grad_norm": 1.740958333015442,
+      "learning_rate": 3.073750779023014e-05,
+      "loss": 2.4515,
+      "step": 672
+    },
+    {
+      "epoch": 1.293666026871401,
+      "grad_norm": 1.647611141204834,
+      "learning_rate": 3.0444763975492208e-05,
+      "loss": 2.5238,
+      "step": 674
+    },
+    {
+      "epoch": 1.2975047984644914,
+      "grad_norm": 1.8491877317428589,
+      "learning_rate": 3.0152809436495734e-05,
+      "loss": 2.2241,
+      "step": 676
+    },
+    {
+      "epoch": 1.3013435700575815,
+      "grad_norm": 1.7405929565429688,
+      "learning_rate": 2.9861655956920286e-05,
+      "loss": 2.2462,
+      "step": 678
+    },
+    {
+      "epoch": 1.3051823416506718,
+      "grad_norm": 1.785489559173584,
+      "learning_rate": 2.95713152881136e-05,
+      "loss": 2.4231,
+      "step": 680
+    },
+    {
+      "epoch": 1.309021113243762,
+      "grad_norm": 1.6834338903427124,
+      "learning_rate": 2.9281799148617262e-05,
+      "loss": 2.3913,
+      "step": 682
+    },
+    {
+      "epoch": 1.3128598848368522,
+      "grad_norm": 1.8619253635406494,
+      "learning_rate": 2.8993119223693755e-05,
+      "loss": 2.4604,
+      "step": 684
+    },
+    {
+      "epoch": 1.3166986564299425,
+      "grad_norm": 2.087561845779419,
+      "learning_rate": 2.8705287164854755e-05,
+      "loss": 2.4383,
+      "step": 686
+    },
+    {
+      "epoch": 1.3205374280230326,
+      "grad_norm": 1.7334223985671997,
+      "learning_rate": 2.8418314589391027e-05,
+      "loss": 2.3927,
+      "step": 688
+    },
+    {
+      "epoch": 1.3243761996161227,
+      "grad_norm": 1.8354735374450684,
+      "learning_rate": 2.8132213079903336e-05,
+      "loss": 2.4723,
+      "step": 690
+    },
+    {
+      "epoch": 1.328214971209213,
+      "grad_norm": 1.8210691213607788,
+      "learning_rate": 2.7846994183835074e-05,
+      "loss": 2.4963,
+      "step": 692
+    },
+    {
+      "epoch": 1.3320537428023034,
+      "grad_norm": 1.834930419921875,
+      "learning_rate": 2.7562669413006148e-05,
+      "loss": 2.2875,
+      "step": 694
+    },
+    {
+      "epoch": 1.3358925143953935,
+      "grad_norm": 1.7620229721069336,
+      "learning_rate": 2.7279250243148412e-05,
+      "loss": 2.2999,
+      "step": 696
+    },
+    {
+      "epoch": 1.3397312859884836,
+      "grad_norm": 1.7829735279083252,
+      "learning_rate": 2.6996748113442394e-05,
+      "loss": 2.2558,
+      "step": 698
+    },
+    {
+      "epoch": 1.3435700575815739,
+      "grad_norm": 1.8787167072296143,
+      "learning_rate": 2.6715174426055662e-05,
+      "loss": 2.4712,
+      "step": 700
+    },
+    {
+      "epoch": 1.3474088291746642,
+      "grad_norm": 1.7032021284103394,
+      "learning_rate": 2.6434540545682584e-05,
+      "loss": 2.335,
+      "step": 702
+    },
+    {
+      "epoch": 1.3512476007677543,
+      "grad_norm": 1.6801146268844604,
+      "learning_rate": 2.6154857799085643e-05,
+      "loss": 2.2266,
+      "step": 704
+    },
+    {
+      "epoch": 1.3550863723608444,
+      "grad_norm": 1.741087555885315,
+      "learning_rate": 2.5876137474638323e-05,
+      "loss": 2.4525,
+      "step": 706
+    },
+    {
+      "epoch": 1.3589251439539347,
+      "grad_norm": 1.8621306419372559,
+      "learning_rate": 2.55983908218694e-05,
+      "loss": 2.4463,
+      "step": 708
+    },
+    {
+      "epoch": 1.362763915547025,
+      "grad_norm": 1.6690336465835571,
+      "learning_rate": 2.5321629051008976e-05,
+      "loss": 2.2719,
+      "step": 710
+    },
+    {
+      "epoch": 1.3666026871401151,
+      "grad_norm": 1.6526907682418823,
+      "learning_rate": 2.5045863332535945e-05,
+      "loss": 2.1614,
+      "step": 712
+    },
+    {
+      "epoch": 1.3704414587332054,
+      "grad_norm": 1.6786959171295166,
+      "learning_rate": 2.4771104796727273e-05,
+      "loss": 2.3446,
+      "step": 714
+    },
+    {
+      "epoch": 1.3742802303262955,
+      "grad_norm": 1.7497304677963257,
+      "learning_rate": 2.449736453320854e-05,
+      "loss": 2.4118,
+      "step": 716
+    },
+    {
+      "epoch": 1.3781190019193859,
+      "grad_norm": 2.005884885787964,
+      "learning_rate": 2.4224653590506607e-05,
+      "loss": 2.3735,
+      "step": 718
+    },
+    {
+      "epoch": 1.381957773512476,
+      "grad_norm": 1.688806176185608,
+      "learning_rate": 2.3952982975603496e-05,
+      "loss": 2.3046,
+      "step": 720
+    },
+    {
+      "epoch": 1.3857965451055663,
+      "grad_norm": 1.6230769157409668,
+      "learning_rate": 2.3682363653492178e-05,
+      "loss": 2.4163,
+      "step": 722
+    },
+    {
+      "epoch": 1.3896353166986564,
+      "grad_norm": 1.842087745666504,
+      "learning_rate": 2.3412806546734055e-05,
+      "loss": 2.3663,
+      "step": 724
+    },
+    {
+      "epoch": 1.3934740882917467,
+      "grad_norm": 1.8735792636871338,
+      "learning_rate": 2.3144322535018127e-05,
+      "loss": 2.4054,
+      "step": 726
+    },
+    {
+      "epoch": 1.3973128598848368,
+      "grad_norm": 1.7510662078857422,
+      "learning_rate": 2.2876922454721694e-05,
+      "loss": 2.2667,
+      "step": 728
+    },
+    {
+      "epoch": 1.401151631477927,
+      "grad_norm": 1.65615975856781,
+      "learning_rate": 2.261061709847327e-05,
+      "loss": 2.4111,
+      "step": 730
+    },
+    {
+      "epoch": 1.4049904030710172,
+      "grad_norm": 1.7408407926559448,
+      "learning_rate": 2.2345417214716748e-05,
+      "loss": 2.3616,
+      "step": 732
+    },
+    {
+      "epoch": 1.4088291746641075,
+      "grad_norm": 1.7720000743865967,
+      "learning_rate": 2.208133350727764e-05,
+      "loss": 2.3627,
+      "step": 734
+    },
+    {
+      "epoch": 1.4126679462571978,
+      "grad_norm": 1.731445550918579,
+      "learning_rate": 2.1818376634931153e-05,
+      "loss": 2.453,
+      "step": 736
+    },
+    {
+      "epoch": 1.416506717850288,
+      "grad_norm": 1.8124655485153198,
+      "learning_rate": 2.1556557210971847e-05,
+      "loss": 2.3809,
+      "step": 738
+    },
+    {
+      "epoch": 1.420345489443378,
+      "grad_norm": 1.607484221458435,
+      "learning_rate": 2.129588580278533e-05,
+      "loss": 2.3365,
+      "step": 740
+    },
+    {
+      "epoch": 1.4241842610364683,
+      "grad_norm": 2.7636168003082275,
+      "learning_rate": 2.103637293142175e-05,
+      "loss": 2.3025,
+      "step": 742
+    },
+    {
+      "epoch": 1.4280230326295587,
+      "grad_norm": 1.7648687362670898,
+      "learning_rate": 2.077802907117119e-05,
+      "loss": 2.4187,
+      "step": 744
+    },
+    {
+      "epoch": 1.4318618042226487,
+      "grad_norm": 1.7737313508987427,
+      "learning_rate": 2.0520864649140765e-05,
+      "loss": 2.3574,
+      "step": 746
+    },
+    {
+      "epoch": 1.4357005758157388,
+      "grad_norm": 1.7417478561401367,
+      "learning_rate": 2.0264890044833995e-05,
+      "loss": 2.4239,
+      "step": 748
+    },
+    {
+      "epoch": 1.4395393474088292,
+      "grad_norm": 1.69158136844635,
+      "learning_rate": 2.0010115589731613e-05,
+      "loss": 2.2593,
+      "step": 750
+    },
+    {
+      "epoch": 1.4433781190019195,
+      "grad_norm": 1.8541730642318726,
+      "learning_rate": 1.9756551566874837e-05,
+      "loss": 2.3546,
+      "step": 752
+    },
+    {
+      "epoch": 1.4472168905950096,
+      "grad_norm": 2.1249778270721436,
+      "learning_rate": 1.9504208210450127e-05,
+      "loss": 2.1447,
+      "step": 754
+    },
+    {
+      "epoch": 1.4510556621880997,
+      "grad_norm": 1.7692455053329468,
+      "learning_rate": 1.9253095705376218e-05,
+      "loss": 2.4882,
+      "step": 756
+    },
+    {
+      "epoch": 1.45489443378119,
+      "grad_norm": 1.693304181098938,
+      "learning_rate": 1.9003224186893e-05,
+      "loss": 2.361,
+      "step": 758
+    },
+    {
+      "epoch": 1.4587332053742803,
+      "grad_norm": 1.657742977142334,
+      "learning_rate": 1.875460374015253e-05,
+      "loss": 2.3439,
+      "step": 760
+    },
+    {
+      "epoch": 1.4625719769673704,
+      "grad_norm": 1.624610424041748,
+      "learning_rate": 1.850724439981186e-05,
+      "loss": 2.3045,
+      "step": 762
+    },
+    {
+      "epoch": 1.4664107485604607,
+      "grad_norm": 1.7179319858551025,
+      "learning_rate": 1.82611561496281e-05,
+      "loss": 2.4564,
+      "step": 764
+    },
+    {
+      "epoch": 1.4702495201535508,
+      "grad_norm": 1.7965707778930664,
+      "learning_rate": 1.801634892205545e-05,
+      "loss": 2.2801,
+      "step": 766
+    },
+    {
+      "epoch": 1.4740882917466411,
+      "grad_norm": 1.5514065027236938,
+      "learning_rate": 1.7772832597844285e-05,
+      "loss": 2.2783,
+      "step": 768
+    },
+    {
+      "epoch": 1.4779270633397312,
+      "grad_norm": 1.741755723953247,
+      "learning_rate": 1.753061700564243e-05,
+      "loss": 2.4353,
+      "step": 770
+    },
+    {
+      "epoch": 1.4817658349328215,
+      "grad_norm": 1.7593576908111572,
+      "learning_rate": 1.728971192159836e-05,
+      "loss": 2.3306,
+      "step": 772
+    },
+    {
+      "epoch": 1.4856046065259116,
+      "grad_norm": 1.7794172763824463,
+      "learning_rate": 1.705012706896668e-05,
+      "loss": 2.2498,
+      "step": 774
+    },
+    {
+      "epoch": 1.489443378119002,
+      "grad_norm": 1.7410024404525757,
+      "learning_rate": 1.6811872117715673e-05,
+      "loss": 2.3822,
+      "step": 776
+    },
+    {
+      "epoch": 1.493282149712092,
+      "grad_norm": 1.7072910070419312,
+      "learning_rate": 1.6574956684137045e-05,
+      "loss": 2.4064,
+      "step": 778
+    },
+    {
+      "epoch": 1.4971209213051824,
+      "grad_norm": 1.7460384368896484,
+      "learning_rate": 1.633939033045766e-05,
+      "loss": 2.3913,
+      "step": 780
+    },
+    {
+      "epoch": 1.5009596928982725,
+      "grad_norm": 1.6655436754226685,
+      "learning_rate": 1.610518256445382e-05,
+      "loss": 2.4221,
+      "step": 782
+    },
+    {
+      "epoch": 1.5047984644913628,
+      "grad_norm": 1.8322985172271729,
+      "learning_rate": 1.5872342839067306e-05,
+      "loss": 2.3166,
+      "step": 784
+    },
+    {
+      "epoch": 1.508637236084453,
+      "grad_norm": 1.7603063583374023,
+      "learning_rate": 1.5640880552023955e-05,
+      "loss": 2.4783,
+      "step": 786
+    },
+    {
+      "epoch": 1.5124760076775432,
+      "grad_norm": 1.6278369426727295,
+      "learning_rate": 1.541080504545433e-05,
+      "loss": 2.3013,
+      "step": 788
+    },
+    {
+      "epoch": 1.5163147792706333,
+      "grad_norm": 1.6899868249893188,
+      "learning_rate": 1.5182125605516707e-05,
+      "loss": 2.4585,
+      "step": 790
+    },
+    {
+      "epoch": 1.5201535508637236,
+      "grad_norm": 1.6993070840835571,
+      "learning_rate": 1.4954851462022118e-05,
+      "loss": 2.3507,
+      "step": 792
+    },
+    {
+      "epoch": 1.523992322456814,
+      "grad_norm": 1.6464179754257202,
+      "learning_rate": 1.4728991788062052e-05,
+      "loss": 2.1691,
+      "step": 794
+    },
+    {
+      "epoch": 1.527831094049904,
+      "grad_norm": 1.8518978357315063,
+      "learning_rate": 1.4504555699638034e-05,
+      "loss": 2.2251,
+      "step": 796
+    },
+    {
+      "epoch": 1.5316698656429941,
+      "grad_norm": 1.743229627609253,
+      "learning_rate": 1.4281552255293739e-05,
+      "loss": 2.3597,
+      "step": 798
+    },
+    {
+      "epoch": 1.5355086372360844,
+      "grad_norm": 1.66657292842865,
+      "learning_rate": 1.405999045574945e-05,
+      "loss": 2.2971,
+      "step": 800
+    },
+    {
+      "epoch": 1.5393474088291748,
+      "grad_norm": 1.5789369344711304,
+      "learning_rate": 1.3839879243538678e-05,
+      "loss": 2.0556,
+      "step": 802
+    },
+    {
+      "epoch": 1.5431861804222649,
+      "grad_norm": 1.6403902769088745,
+      "learning_rate": 1.362122750264727e-05,
+      "loss": 2.3706,
+      "step": 804
+    },
+    {
+      "epoch": 1.547024952015355,
+      "grad_norm": 1.8980406522750854,
+      "learning_rate": 1.3404044058154835e-05,
+      "loss": 2.3991,
+      "step": 806
+    },
+    {
+      "epoch": 1.5508637236084453,
+      "grad_norm": 1.9013497829437256,
+      "learning_rate": 1.3188337675878609e-05,
+      "loss": 2.2111,
+      "step": 808
+    },
+    {
+      "epoch": 1.5547024952015356,
+      "grad_norm": 1.6037538051605225,
+      "learning_rate": 1.2974117062019503e-05,
+      "loss": 2.41,
+      "step": 810
+    },
+    {
+      "epoch": 1.5585412667946257,
+      "grad_norm": 4.165694236755371,
+      "learning_rate": 1.2761390862810907e-05,
+      "loss": 2.4047,
+      "step": 812
+    },
+    {
+      "epoch": 1.5623800383877158,
+      "grad_norm": 1.740267276763916,
+      "learning_rate": 1.2550167664169566e-05,
+      "loss": 2.4278,
+      "step": 814
+    },
+    {
+      "epoch": 1.566218809980806,
+      "grad_norm": 1.6863207817077637,
+      "learning_rate": 1.2340455991349093e-05,
+      "loss": 2.3615,
+      "step": 816
+    },
+    {
+      "epoch": 1.5700575815738964,
+      "grad_norm": 1.8030253648757935,
+      "learning_rate": 1.2132264308595875e-05,
+      "loss": 2.427,
+      "step": 818
+    },
+    {
+      "epoch": 1.5738963531669867,
+      "grad_norm": 1.6316559314727783,
+      "learning_rate": 1.1925601018807497e-05,
+      "loss": 2.4522,
+      "step": 820
+    },
+    {
+      "epoch": 1.5777351247600768,
+      "grad_norm": 1.7768834829330444,
+      "learning_rate": 1.1720474463193443e-05,
+      "loss": 2.3969,
+      "step": 822
+    },
+    {
+      "epoch": 1.581573896353167,
+      "grad_norm": 1.5759806632995605,
+      "learning_rate": 1.1516892920938626e-05,
+      "loss": 2.4005,
+      "step": 824
+    },
+    {
+      "epoch": 1.5854126679462572,
+      "grad_norm": 1.9386920928955078,
+      "learning_rate": 1.131486460886908e-05,
+      "loss": 2.3769,
+      "step": 826
+    },
+    {
+      "epoch": 1.5892514395393476,
+      "grad_norm": 1.6804122924804688,
+      "learning_rate": 1.1114397681120387e-05,
+      "loss": 2.3608,
+      "step": 828
+    },
+    {
+      "epoch": 1.5930902111324377,
+      "grad_norm": 1.7132877111434937,
+      "learning_rate": 1.0915500228808523e-05,
+      "loss": 2.3858,
+      "step": 830
+    },
+    {
+      "epoch": 1.5969289827255277,
+      "grad_norm": 1.7644052505493164,
+      "learning_rate": 1.0718180279703371e-05,
+      "loss": 2.2794,
+      "step": 832
+    },
+    {
+      "epoch": 1.600767754318618,
+      "grad_norm": 1.820474624633789,
+      "learning_rate": 1.0522445797904607e-05,
+      "loss": 2.3528,
+      "step": 834
+    },
+    {
+      "epoch": 1.6046065259117084,
+      "grad_norm": 1.6057413816452026,
+      "learning_rate": 1.0328304683520306e-05,
+      "loss": 2.3739,
+      "step": 836
+    },
+    {
+      "epoch": 1.6084452975047985,
+      "grad_norm": 1.5365757942199707,
+      "learning_rate": 1.0135764772348106e-05,
+      "loss": 2.3543,
+      "step": 838
+    },
+    {
+      "epoch": 1.6122840690978886,
+      "grad_norm": 1.714257001876831,
+      "learning_rate": 9.944833835558887e-06,
+      "loss": 2.4455,
+      "step": 840
+    },
+    {
+      "epoch": 1.616122840690979,
+      "grad_norm": 1.6838536262512207,
+      "learning_rate": 9.755519579383204e-06,
+      "loss": 2.3978,
+      "step": 842
+    },
+    {
+      "epoch": 1.6199616122840692,
+      "grad_norm": 1.6596792936325073,
+      "learning_rate": 9.567829644800141e-06,
+      "loss": 2.4081,
+      "step": 844
+    },
+    {
+      "epoch": 1.6238003838771593,
+      "grad_norm": 1.6361202001571655,
+      "learning_rate": 9.381771607229e-06,
+      "loss": 2.4115,
+      "step": 846
+    },
+    {
+      "epoch": 1.6276391554702494,
+      "grad_norm": 1.811923861503601,
+      "learning_rate": 9.197352976223495e-06,
+      "loss": 2.3388,
+      "step": 848
+    },
+    {
+      "epoch": 1.6314779270633397,
+      "grad_norm": 1.5978856086730957,
+      "learning_rate": 9.014581195168725e-06,
+      "loss": 2.3515,
+      "step": 850
+    },
+    {
+      "epoch": 1.63531669865643,
+      "grad_norm": 1.7055366039276123,
+      "learning_rate": 8.83346364098061e-06,
+      "loss": 2.2939,
+      "step": 852
+    },
+    {
+      "epoch": 1.6391554702495201,
+      "grad_norm": 1.6278856992721558,
+      "learning_rate": 8.654007623808335e-06,
+      "loss": 2.4505,
+      "step": 854
+    },
+    {
+      "epoch": 1.6429942418426102,
+      "grad_norm": 1.6890220642089844,
+      "learning_rate": 8.476220386739153e-06,
+      "loss": 2.2088,
+      "step": 856
+    },
+    {
+      "epoch": 1.6468330134357005,
+      "grad_norm": 1.6934252977371216,
+      "learning_rate": 8.30010910550611e-06,
+      "loss": 2.0931,
+      "step": 858
+    },
+    {
+      "epoch": 1.6506717850287909,
+      "grad_norm": 1.6027623414993286,
+      "learning_rate": 8.125680888198396e-06,
+      "loss": 2.4532,
+      "step": 860
+    },
+    {
+      "epoch": 1.654510556621881,
+      "grad_norm": 1.7157975435256958,
+      "learning_rate": 7.95294277497452e-06,
+      "loss": 2.4376,
+      "step": 862
+    },
+    {
+      "epoch": 1.658349328214971,
+      "grad_norm": 1.7207980155944824,
+      "learning_rate": 7.781901737778013e-06,
+      "loss": 2.341,
+      "step": 864
+    },
+    {
+      "epoch": 1.6621880998080614,
+      "grad_norm": 1.831322193145752,
+      "learning_rate": 7.612564680056233e-06,
+      "loss": 2.316,
+      "step": 866
+    },
+    {
+      "epoch": 1.6660268714011517,
+      "grad_norm": 1.679987907409668,
+      "learning_rate": 7.4449384364815464e-06,
+      "loss": 2.2654,
+      "step": 868
+    },
+    {
+      "epoch": 1.669865642994242,
+      "grad_norm": 1.6710460186004639,
+      "learning_rate": 7.2790297726755716e-06,
+      "loss": 2.224,
+      "step": 870
+    },
+    {
+      "epoch": 1.673704414587332,
+      "grad_norm": 1.658496379852295,
+      "learning_rate": 7.114845384936108e-06,
+      "loss": 2.2764,
+      "step": 872
+    },
+    {
+      "epoch": 1.6775431861804222,
+      "grad_norm": 1.7485684156417847,
+      "learning_rate": 6.9523918999668255e-06,
+      "loss": 2.215,
+      "step": 874
+    },
+    {
+      "epoch": 1.6813819577735125,
+      "grad_norm": 1.6568515300750732,
+      "learning_rate": 6.7916758746098155e-06,
+      "loss": 2.3058,
+      "step": 876
+    },
+    {
+      "epoch": 1.6852207293666028,
+      "grad_norm": 1.7417012453079224,
+      "learning_rate": 6.632703795580946e-06,
+      "loss": 2.4268,
+      "step": 878
+    },
+    {
+      "epoch": 1.689059500959693,
+      "grad_norm": 1.7897471189498901,
+      "learning_rate": 6.475482079208112e-06,
+      "loss": 2.2536,
+      "step": 880
+    },
+    {
+      "epoch": 1.692898272552783,
+      "grad_norm": 1.6039488315582275,
+      "learning_rate": 6.320017071172113e-06,
+      "loss": 2.2135,
+      "step": 882
+    },
+    {
+      "epoch": 1.6967370441458733,
+      "grad_norm": 1.661766529083252,
+      "learning_rate": 6.1663150462507034e-06,
+      "loss": 2.325,
+      "step": 884
+    },
+    {
+      "epoch": 1.7005758157389637,
+      "grad_norm": 1.6283842325210571,
+      "learning_rate": 6.014382208065234e-06,
+      "loss": 2.3323,
+      "step": 886
+    },
+    {
+      "epoch": 1.7044145873320538,
+      "grad_norm": 2.2004783153533936,
+      "learning_rate": 5.864224688830283e-06,
+      "loss": 2.1856,
+      "step": 888
+    },
+    {
+      "epoch": 1.7082533589251438,
+      "grad_norm": 1.6094887256622314,
+      "learning_rate": 5.715848549106145e-06,
+      "loss": 2.2182,
+      "step": 890
+    },
+    {
+      "epoch": 1.7120921305182342,
+      "grad_norm": 1.6314260959625244,
+      "learning_rate": 5.569259777554287e-06,
+      "loss": 2.3467,
+      "step": 892
+    },
+    {
+      "epoch": 1.7159309021113245,
+      "grad_norm": 1.8429559469223022,
+      "learning_rate": 5.424464290695496e-06,
+      "loss": 2.349,
+      "step": 894
+    },
+    {
+      "epoch": 1.7197696737044146,
+      "grad_norm": 1.661424994468689,
+      "learning_rate": 5.281467932671253e-06,
+      "loss": 2.3159,
+      "step": 896
+    },
+    {
+      "epoch": 1.7236084452975047,
+      "grad_norm": 1.6067779064178467,
+      "learning_rate": 5.140276475007711e-06,
+      "loss": 2.4294,
+      "step": 898
+    },
+    {
+      "epoch": 1.727447216890595,
+      "grad_norm": 1.7527132034301758,
+      "learning_rate": 5.000895616382828e-06,
+      "loss": 2.2977,
+      "step": 900
+    },
+    {
+      "epoch": 1.7312859884836853,
+      "grad_norm": 1.639435052871704,
+      "learning_rate": 4.863330982396319e-06,
+      "loss": 2.2522,
+      "step": 902
+    },
+    {
+      "epoch": 1.7351247600767754,
+      "grad_norm": 1.676002860069275,
+      "learning_rate": 4.727588125342669e-06,
+      "loss": 2.2221,
+      "step": 904
+    },
+    {
+      "epoch": 1.7389635316698655,
+      "grad_norm": 1.7567448616027832,
+      "learning_rate": 4.593672523986936e-06,
+      "loss": 2.2736,
+      "step": 906
+    },
+    {
+      "epoch": 1.7428023032629558,
+      "grad_norm": 1.7621455192565918,
+      "learning_rate": 4.461589583343678e-06,
+      "loss": 2.2736,
+      "step": 908
+    },
+    {
+      "epoch": 1.7466410748560461,
+      "grad_norm": 1.6953126192092896,
+      "learning_rate": 4.331344634458812e-06,
+      "loss": 2.2319,
+      "step": 910
+    },
+    {
+      "epoch": 1.7504798464491362,
+      "grad_norm": 1.6162375211715698,
+      "learning_rate": 4.2029429341943984e-06,
+      "loss": 2.3247,
+      "step": 912
+    },
+    {
+      "epoch": 1.7543186180422263,
+      "grad_norm": 1.6382193565368652,
+      "learning_rate": 4.076389665016522e-06,
+      "loss": 2.3777,
+      "step": 914
+    },
+    {
+      "epoch": 1.7581573896353166,
+      "grad_norm": 1.6324814558029175,
+      "learning_rate": 3.951689934786068e-06,
+      "loss": 2.3113,
+      "step": 916
+    },
+    {
+      "epoch": 1.761996161228407,
+      "grad_norm": 1.6445063352584839,
+      "learning_rate": 3.828848776552596e-06,
+      "loss": 2.2427,
+      "step": 918
+    },
+    {
+      "epoch": 1.7658349328214973,
+      "grad_norm": 1.810166835784912,
+      "learning_rate": 3.707871148351183e-06,
+      "loss": 2.3464,
+      "step": 920
+    },
+    {
+      "epoch": 1.7696737044145874,
+      "grad_norm": 1.614855408668518,
+      "learning_rate": 3.5887619330023435e-06,
+      "loss": 2.3454,
+      "step": 922
+    },
+    {
+      "epoch": 1.7735124760076775,
+      "grad_norm": 1.5908379554748535,
+      "learning_rate": 3.4715259379148656e-06,
+      "loss": 2.3322,
+      "step": 924
+    },
+    {
+      "epoch": 1.7773512476007678,
+      "grad_norm": 1.5997949838638306,
+      "learning_rate": 3.3561678948919095e-06,
+      "loss": 2.2104,
+      "step": 926
+    },
+    {
+      "epoch": 1.781190019193858,
+      "grad_norm": 1.713915467262268,
+      "learning_rate": 3.2426924599399056e-06,
+      "loss": 2.1869,
+      "step": 928
+    },
+    {
+      "epoch": 1.7850287907869482,
+      "grad_norm": 1.5691741704940796,
+      "learning_rate": 3.131104213080688e-06,
+      "loss": 2.4053,
+      "step": 930
+    },
+    {
+      "epoch": 1.7888675623800383,
+      "grad_norm": 1.6705538034439087,
+      "learning_rate": 3.021407658166636e-06,
+      "loss": 2.2513,
+      "step": 932
+    },
+    {
+      "epoch": 1.7927063339731286,
+      "grad_norm": 1.6354550123214722,
+      "learning_rate": 2.913607222698905e-06,
+      "loss": 2.287,
+      "step": 934
+    },
+    {
+      "epoch": 1.796545105566219,
+      "grad_norm": 1.6781665086746216,
+      "learning_rate": 2.807707257648662e-06,
+      "loss": 2.3081,
+      "step": 936
+    },
+    {
+      "epoch": 1.800383877159309,
+      "grad_norm": 1.6056350469589233,
+      "learning_rate": 2.703712037281564e-06,
+      "loss": 2.3788,
+      "step": 938
+    },
+    {
+      "epoch": 1.8042226487523991,
+      "grad_norm": 1.5860586166381836,
+      "learning_rate": 2.601625758985182e-06,
+      "loss": 2.3995,
+      "step": 940
+    },
+    {
+      "epoch": 1.8080614203454894,
+      "grad_norm": 1.6104470491409302,
+      "learning_rate": 2.5014525430995916e-06,
+      "loss": 2.2181,
+      "step": 942
+    },
+    {
+      "epoch": 1.8119001919385798,
+      "grad_norm": 1.8218193054199219,
+      "learning_rate": 2.4031964327511313e-06,
+      "loss": 2.3648,
+      "step": 944
+    },
+    {
+      "epoch": 1.8157389635316699,
+      "grad_norm": 1.7795993089675903,
+      "learning_rate": 2.3068613936891137e-06,
+      "loss": 2.4016,
+      "step": 946
+    },
+    {
+      "epoch": 1.81957773512476,
+      "grad_norm": 1.5306187868118286,
+      "learning_rate": 2.2124513141258574e-06,
+      "loss": 2.3352,
+      "step": 948
+    },
+    {
+      "epoch": 1.8234165067178503,
+      "grad_norm": 1.6413097381591797,
+      "learning_rate": 2.1199700045797077e-06,
+      "loss": 2.3186,
+      "step": 950
+    },
+    {
+      "epoch": 1.8272552783109406,
+      "grad_norm": 1.7125797271728516,
+      "learning_rate": 2.0294211977212317e-06,
+      "loss": 2.3836,
+      "step": 952
+    },
+    {
+      "epoch": 1.8310940499040307,
+      "grad_norm": 1.6402167081832886,
+      "learning_rate": 1.9408085482225945e-06,
+      "loss": 2.429,
+      "step": 954
+    },
+    {
+      "epoch": 1.8349328214971208,
+      "grad_norm": 1.6203242540359497,
+      "learning_rate": 1.8541356326100433e-06,
+      "loss": 2.3133,
+      "step": 956
+    },
+    {
+      "epoch": 1.838771593090211,
+      "grad_norm": 1.5762218236923218,
+      "learning_rate": 1.7694059491195013e-06,
+      "loss": 2.3937,
+      "step": 958
+    },
+    {
+      "epoch": 1.8426103646833014,
+      "grad_norm": 1.627780556678772,
+      "learning_rate": 1.6866229175554748e-06,
+      "loss": 2.2514,
+      "step": 960
+    },
+    {
+      "epoch": 1.8464491362763915,
+      "grad_norm": 1.5594689846038818,
+      "learning_rate": 1.6057898791529303e-06,
+      "loss": 2.2433,
+      "step": 962
+    },
+    {
+      "epoch": 1.8502879078694816,
+      "grad_norm": 1.5747395753860474,
+      "learning_rate": 1.526910096442491e-06,
+      "loss": 2.1411,
+      "step": 964
+    },
+    {
+      "epoch": 1.854126679462572,
+      "grad_norm": 1.5389329195022583,
+      "learning_rate": 1.4499867531187373e-06,
+      "loss": 2.1453,
+      "step": 966
+    },
+    {
+      "epoch": 1.8579654510556622,
+      "grad_norm": 1.7601927518844604,
+      "learning_rate": 1.3750229539117144e-06,
+      "loss": 2.2826,
+      "step": 968
+    },
+    {
+      "epoch": 1.8618042226487526,
+      "grad_norm": 1.5119316577911377,
+      "learning_rate": 1.3020217244616273e-06,
+      "loss": 2.274,
+      "step": 970
+    },
+    {
+      "epoch": 1.8656429942418427,
+      "grad_norm": 1.647141456604004,
+      "learning_rate": 1.2309860111967053e-06,
+      "loss": 2.3202,
+      "step": 972
+    },
+    {
+      "epoch": 1.8694817658349328,
+      "grad_norm": 1.5036982297897339,
+      "learning_rate": 1.1619186812142857e-06,
+      "loss": 2.3904,
+      "step": 974
+    },
+    {
+      "epoch": 1.873320537428023,
+      "grad_norm": 1.653581142425537,
+      "learning_rate": 1.094822522165101e-06,
+      "loss": 2.472,
+      "step": 976
+    },
+    {
+      "epoch": 1.8771593090211134,
+      "grad_norm": 1.705700159072876,
+      "learning_rate": 1.0297002421407797e-06,
+      "loss": 2.3804,
+      "step": 978
+    },
+    {
+      "epoch": 1.8809980806142035,
+      "grad_norm": 1.621009349822998,
+      "learning_rate": 9.665544695645012e-07,
+      "loss": 2.3232,
+      "step": 980
+    },
+    {
+      "epoch": 1.8848368522072936,
+      "grad_norm": 1.651867151260376,
+      "learning_rate": 9.053877530849464e-07,
+      "loss": 2.2534,
+      "step": 982
+    },
+    {
+      "epoch": 1.888675623800384,
+      "grad_norm": 1.7189369201660156,
+      "learning_rate": 8.462025614734192e-07,
+      "loss": 2.3764,
+      "step": 984
+    },
+    {
+      "epoch": 1.8925143953934742,
+      "grad_norm": 1.6047943830490112,
+      "learning_rate": 7.890012835242044e-07,
+      "loss": 2.4137,
+      "step": 986
+    },
+    {
+      "epoch": 1.8963531669865643,
+      "grad_norm": 1.6477607488632202,
+      "learning_rate": 7.337862279581332e-07,
+      "loss": 2.3862,
+      "step": 988
+    },
+    {
+      "epoch": 1.9001919385796544,
+      "grad_norm": 1.6669402122497559,
+      "learning_rate": 6.805596233294576e-07,
+      "loss": 2.3233,
+      "step": 990
+    },
+    {
+      "epoch": 1.9040307101727447,
+      "grad_norm": 1.8017224073410034,
+      "learning_rate": 6.293236179358175e-07,
+      "loss": 2.4253,
+      "step": 992
+    },
+    {
+      "epoch": 1.907869481765835,
+      "grad_norm": 1.5671507120132446,
+      "learning_rate": 5.800802797316152e-07,
+      "loss": 2.1269,
+      "step": 994
+    },
+    {
+      "epoch": 1.9117082533589251,
+      "grad_norm": 1.6347520351409912,
+      "learning_rate": 5.328315962444874e-07,
+      "loss": 2.4157,
+      "step": 996
+    },
+    {
+      "epoch": 1.9155470249520152,
+      "grad_norm": 2.142366409301758,
+      "learning_rate": 4.875794744951423e-07,
+      "loss": 2.0941,
+      "step": 998
+    },
+    {
+      "epoch": 1.9193857965451055,
+      "grad_norm": 1.5768301486968994,
+      "learning_rate": 4.4432574092032053e-07,
+      "loss": 2.1746,
+      "step": 1000
+    },
+    {
+      "epoch": 1.9232245681381959,
+      "grad_norm": 1.6327540874481201,
+      "learning_rate": 4.0307214129914896e-07,
+      "loss": 2.3644,
+      "step": 1002
+    },
+    {
+      "epoch": 1.927063339731286,
+      "grad_norm": 1.6046617031097412,
+      "learning_rate": 3.6382034068263015e-07,
+      "loss": 2.2846,
+      "step": 1004
+    },
+    {
+      "epoch": 1.930902111324376,
+      "grad_norm": 1.6890329122543335,
+      "learning_rate": 3.265719233264575e-07,
+      "loss": 2.3874,
+      "step": 1006
+    },
+    {
+      "epoch": 1.9347408829174664,
+      "grad_norm": 1.6540732383728027,
+      "learning_rate": 2.913283926270771e-07,
+      "loss": 2.3438,
+      "step": 1008
+    },
+    {
+      "epoch": 1.9385796545105567,
+      "grad_norm": 1.6627967357635498,
+      "learning_rate": 2.580911710609923e-07,
+      "loss": 2.4765,
+      "step": 1010
+    },
+    {
+      "epoch": 1.9424184261036468,
+      "grad_norm": 1.579916000366211,
+      "learning_rate": 2.2686160012735936e-07,
+      "loss": 2.1976,
+      "step": 1012
+    },
+    {
+      "epoch": 1.9462571976967369,
+      "grad_norm": 1.5479332208633423,
+      "learning_rate": 1.9764094029385284e-07,
+      "loss": 2.2302,
+      "step": 1014
+    },
+    {
+      "epoch": 1.9500959692898272,
+      "grad_norm": 1.639022707939148,
+      "learning_rate": 1.704303709457733e-07,
+      "loss": 2.2169,
+      "step": 1016
+    },
+    {
+      "epoch": 1.9539347408829175,
+      "grad_norm": 1.752821445465088,
+      "learning_rate": 1.452309903384519e-07,
+      "loss": 2.2365,
+      "step": 1018
+    },
+    {
+      "epoch": 1.9577735124760078,
+      "grad_norm": 1.6994763612747192,
+      "learning_rate": 1.220438155529302e-07,
+      "loss": 2.4001,
+      "step": 1020
+    },
+    {
+      "epoch": 1.961612284069098,
+      "grad_norm": 1.6137231588363647,
+      "learning_rate": 1.0086978245490985e-07,
+      "loss": 2.3455,
+      "step": 1022
+    },
+    {
+      "epoch": 1.965451055662188,
+      "grad_norm": 1.6784782409667969,
+      "learning_rate": 8.170974565696599e-08,
+      "loss": 2.3519,
+      "step": 1024
+    },
+    {
+      "epoch": 1.9692898272552783,
+      "grad_norm": 1.6100566387176514,
+      "learning_rate": 6.456447848406932e-08,
+      "loss": 2.3392,
+      "step": 1026
+    },
+    {
+      "epoch": 1.9731285988483687,
+      "grad_norm": 1.6168123483657837,
+      "learning_rate": 4.94346729423556e-08,
+      "loss": 2.3783,
+      "step": 1028
+    },
+    {
+      "epoch": 1.9769673704414588,
+      "grad_norm": 1.681522250175476,
+      "learning_rate": 3.632093969121453e-08,
+      "loss": 2.3799,
+      "step": 1030
+    },
+    {
+      "epoch": 1.9808061420345489,
+      "grad_norm": 1.6089167594909668,
+      "learning_rate": 2.522380801863733e-08,
+      "loss": 2.3185,
+      "step": 1032
+    },
+    {
+      "epoch": 1.9846449136276392,
+      "grad_norm": 1.5647084712982178,
+      "learning_rate": 1.6143725819850463e-08,
+      "loss": 2.1239,
+      "step": 1034
+    },
+    {
+      "epoch": 1.9884836852207295,
+      "grad_norm": 1.6926655769348145,
+      "learning_rate": 9.081059579235662e-09,
+      "loss": 2.2994,
+      "step": 1036
+    },
+    {
+      "epoch": 1.9923224568138196,
+      "grad_norm": 1.770678162574768,
+      "learning_rate": 4.036094355541753e-09,
+      "loss": 2.3264,
+      "step": 1038
+    },
+    {
+      "epoch": 1.9961612284069097,
+      "grad_norm": 1.6762254238128662,
+      "learning_rate": 1.0090337703771991e-09,
+      "loss": 2.3756,
+      "step": 1040
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 2.1386659145355225,
+      "learning_rate": 0.0,
+      "loss": 2.1659,
+      "step": 1042
+    }
+  ],
+  "logging_steps": 2,
+  "max_steps": 1042,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 2,
+  "save_steps": 521,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.762124413890724e+17,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoint-1042/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:72490e88c6bf2877f2bc15eaa6e0f01e604bb0f1022b7d74641eda9e89f52e21
+size 5432

checkpoint-521/config.json

@@ -0,0 +1,27 @@
+{
+  "_name_or_path": "/mnt/parscratch/users/acp23ay/private/models/OLMo-2-1124-7B-ary_Arab-coli-align/",
+  "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-521/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-521/model-00001-of-00006.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a305afc649f45d162e2f827b89ffbaddc810191d3583c4eee71f0af47f866d44
+size 4950070472

checkpoint-521/model-00002-of-00006.safetensors

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

checkpoint-521/model-00003-of-00006.safetensors

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

checkpoint-521/model-00004-of-00006.safetensors

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

checkpoint-521/model-00005-of-00006.safetensors

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

checkpoint-521/model-00006-of-00006.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:98ad5a47a89446b52eb3df402508a9b33824a77548d1c1661b361fd833984634
+size 4614509080

checkpoint-521/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-521/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:31e49c6ce16d71f5bb170d27079a05062cadcfbb8ffbffe6898e80af87957c95
+size 13055369358

checkpoint-521/rng_state.pth

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

checkpoint-521/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0783d630f8ce8aef70d72f47e1602b69a36f3f4929db1a03884530798b9b38d7
+size 1064

checkpoint-521/trainer_state.json

@@ -0,0 +1,1853 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "eval_steps": 500,
+  "global_step": 521,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.003838771593090211,
+      "grad_norm": 39.1374626159668,
+      "learning_rate": 3.7735849056603773e-06,
+      "loss": 5.6737,
+      "step": 2
+    },
+    {
+      "epoch": 0.007677543186180422,
+      "grad_norm": 24.925661087036133,
+      "learning_rate": 7.547169811320755e-06,
+      "loss": 5.612,
+      "step": 4
+    },
+    {
+      "epoch": 0.011516314779270634,
+      "grad_norm": 12.562352180480957,
+      "learning_rate": 1.1320754716981132e-05,
+      "loss": 5.0919,
+      "step": 6
+    },
+    {
+      "epoch": 0.015355086372360844,
+      "grad_norm": 7.011171817779541,
+      "learning_rate": 1.509433962264151e-05,
+      "loss": 4.6621,
+      "step": 8
+    },
+    {
+      "epoch": 0.019193857965451054,
+      "grad_norm": 5.962992191314697,
+      "learning_rate": 1.8867924528301888e-05,
+      "loss": 4.6095,
+      "step": 10
+    },
+    {
+      "epoch": 0.023032629558541268,
+      "grad_norm": 6.030173301696777,
+      "learning_rate": 2.2641509433962265e-05,
+      "loss": 4.6055,
+      "step": 12
+    },
+    {
+      "epoch": 0.026871401151631478,
+      "grad_norm": 5.021393775939941,
+      "learning_rate": 2.641509433962264e-05,
+      "loss": 4.2814,
+      "step": 14
+    },
+    {
+      "epoch": 0.030710172744721688,
+      "grad_norm": 4.441908359527588,
+      "learning_rate": 3.018867924528302e-05,
+      "loss": 4.2399,
+      "step": 16
+    },
+    {
+      "epoch": 0.0345489443378119,
+      "grad_norm": 3.59828782081604,
+      "learning_rate": 3.39622641509434e-05,
+      "loss": 4.1171,
+      "step": 18
+    },
+    {
+      "epoch": 0.03838771593090211,
+      "grad_norm": 3.9370954036712646,
+      "learning_rate": 3.7735849056603776e-05,
+      "loss": 4.1206,
+      "step": 20
+    },
+    {
+      "epoch": 0.04222648752399232,
+      "grad_norm": 2.9339046478271484,
+      "learning_rate": 4.150943396226415e-05,
+      "loss": 3.9446,
+      "step": 22
+    },
+    {
+      "epoch": 0.046065259117082535,
+      "grad_norm": 3.2044146060943604,
+      "learning_rate": 4.528301886792453e-05,
+      "loss": 3.7939,
+      "step": 24
+    },
+    {
+      "epoch": 0.04990403071017274,
+      "grad_norm": 3.1344337463378906,
+      "learning_rate": 4.9056603773584906e-05,
+      "loss": 3.7661,
+      "step": 26
+    },
+    {
+      "epoch": 0.053742802303262956,
+      "grad_norm": 2.632190704345703,
+      "learning_rate": 5.283018867924528e-05,
+      "loss": 3.7593,
+      "step": 28
+    },
+    {
+      "epoch": 0.05758157389635317,
+      "grad_norm": 2.875645399093628,
+      "learning_rate": 5.660377358490566e-05,
+      "loss": 3.458,
+      "step": 30
+    },
+    {
+      "epoch": 0.061420345489443376,
+      "grad_norm": 2.202479600906372,
+      "learning_rate": 6.037735849056604e-05,
+      "loss": 3.7317,
+      "step": 32
+    },
+    {
+      "epoch": 0.06525911708253358,
+      "grad_norm": 3.269009590148926,
+      "learning_rate": 6.415094339622641e-05,
+      "loss": 3.6346,
+      "step": 34
+    },
+    {
+      "epoch": 0.0690978886756238,
+      "grad_norm": 2.55442476272583,
+      "learning_rate": 6.79245283018868e-05,
+      "loss": 3.5382,
+      "step": 36
+    },
+    {
+      "epoch": 0.07293666026871401,
+      "grad_norm": 2.7956719398498535,
+      "learning_rate": 7.169811320754717e-05,
+      "loss": 3.6172,
+      "step": 38
+    },
+    {
+      "epoch": 0.07677543186180422,
+      "grad_norm": 2.430014133453369,
+      "learning_rate": 7.547169811320755e-05,
+      "loss": 3.4135,
+      "step": 40
+    },
+    {
+      "epoch": 0.08061420345489444,
+      "grad_norm": 2.5124785900115967,
+      "learning_rate": 7.924528301886794e-05,
+      "loss": 3.565,
+      "step": 42
+    },
+    {
+      "epoch": 0.08445297504798464,
+      "grad_norm": 2.733428955078125,
+      "learning_rate": 8.30188679245283e-05,
+      "loss": 3.4902,
+      "step": 44
+    },
+    {
+      "epoch": 0.08829174664107485,
+      "grad_norm": 2.421130657196045,
+      "learning_rate": 8.679245283018869e-05,
+      "loss": 3.4303,
+      "step": 46
+    },
+    {
+      "epoch": 0.09213051823416507,
+      "grad_norm": 2.4804866313934326,
+      "learning_rate": 9.056603773584906e-05,
+      "loss": 3.2669,
+      "step": 48
+    },
+    {
+      "epoch": 0.09596928982725528,
+      "grad_norm": 2.927652359008789,
+      "learning_rate": 9.433962264150944e-05,
+      "loss": 3.4791,
+      "step": 50
+    },
+    {
+      "epoch": 0.09980806142034548,
+      "grad_norm": 2.6607537269592285,
+      "learning_rate": 9.811320754716981e-05,
+      "loss": 3.5444,
+      "step": 52
+    },
+    {
+      "epoch": 0.1036468330134357,
+      "grad_norm": 2.635265827178955,
+      "learning_rate": 9.999974774092106e-05,
+      "loss": 3.2915,
+      "step": 54
+    },
+    {
+      "epoch": 0.10748560460652591,
+      "grad_norm": 2.650139093399048,
+      "learning_rate": 9.999772968356182e-05,
+      "loss": 3.3209,
+      "step": 56
+    },
+    {
+      "epoch": 0.11132437619961612,
+      "grad_norm": 3.0350704193115234,
+      "learning_rate": 9.999369365029487e-05,
+      "loss": 3.3741,
+      "step": 58
+    },
+    {
+      "epoch": 0.11516314779270634,
+      "grad_norm": 2.821930170059204,
+      "learning_rate": 9.998763980401998e-05,
+      "loss": 3.3345,
+      "step": 60
+    },
+    {
+      "epoch": 0.11900191938579655,
+      "grad_norm": 2.58774471282959,
+      "learning_rate": 9.997956838907852e-05,
+      "loss": 3.3121,
+      "step": 62
+    },
+    {
+      "epoch": 0.12284069097888675,
+      "grad_norm": 2.615741491317749,
+      "learning_rate": 9.996947973124372e-05,
+      "loss": 3.4007,
+      "step": 64
+    },
+    {
+      "epoch": 0.12667946257197696,
+      "grad_norm": 2.759786605834961,
+      "learning_rate": 9.995737423770745e-05,
+      "loss": 3.1918,
+      "step": 66
+    },
+    {
+      "epoch": 0.13051823416506717,
+      "grad_norm": 2.426225423812866,
+      "learning_rate": 9.994325239706377e-05,
+      "loss": 3.252,
+      "step": 68
+    },
+    {
+      "epoch": 0.1343570057581574,
+      "grad_norm": 2.713718891143799,
+      "learning_rate": 9.992711477928926e-05,
+      "loss": 3.125,
+      "step": 70
+    },
+    {
+      "epoch": 0.1381957773512476,
+      "grad_norm": 2.7484843730926514,
+      "learning_rate": 9.990896203571995e-05,
+      "loss": 3.361,
+      "step": 72
+    },
+    {
+      "epoch": 0.1420345489443378,
+      "grad_norm": 2.6632022857666016,
+      "learning_rate": 9.988879489902509e-05,
+      "loss": 3.2453,
+      "step": 74
+    },
+    {
+      "epoch": 0.14587332053742802,
+      "grad_norm": 2.7256522178649902,
+      "learning_rate": 9.986661418317759e-05,
+      "loss": 3.2655,
+      "step": 76
+    },
+    {
+      "epoch": 0.14971209213051823,
+      "grad_norm": 2.542839288711548,
+      "learning_rate": 9.984242078342107e-05,
+      "loss": 3.317,
+      "step": 78
+    },
+    {
+      "epoch": 0.15355086372360843,
+      "grad_norm": 2.6361520290374756,
+      "learning_rate": 9.981621567623385e-05,
+      "loss": 3.2138,
+      "step": 80
+    },
+    {
+      "epoch": 0.15738963531669867,
+      "grad_norm": 2.4558825492858887,
+      "learning_rate": 9.978799991928944e-05,
+      "loss": 3.2643,
+      "step": 82
+    },
+    {
+      "epoch": 0.16122840690978887,
+      "grad_norm": 2.361485719680786,
+      "learning_rate": 9.97577746514139e-05,
+      "loss": 3.2489,
+      "step": 84
+    },
+    {
+      "epoch": 0.16506717850287908,
+      "grad_norm": 2.5965819358825684,
+      "learning_rate": 9.972554109253989e-05,
+      "loss": 3.2573,
+      "step": 86
+    },
+    {
+      "epoch": 0.1689059500959693,
+      "grad_norm": 2.2212257385253906,
+      "learning_rate": 9.969130054365736e-05,
+      "loss": 3.1332,
+      "step": 88
+    },
+    {
+      "epoch": 0.1727447216890595,
+      "grad_norm": 2.350938320159912,
+      "learning_rate": 9.965505438676115e-05,
+      "loss": 3.1953,
+      "step": 90
+    },
+    {
+      "epoch": 0.1765834932821497,
+      "grad_norm": 2.1297547817230225,
+      "learning_rate": 9.961680408479507e-05,
+      "loss": 3.0385,
+      "step": 92
+    },
+    {
+      "epoch": 0.18042226487523993,
+      "grad_norm": 2.6407082080841064,
+      "learning_rate": 9.957655118159304e-05,
+      "loss": 3.1181,
+      "step": 94
+    },
+    {
+      "epoch": 0.18426103646833014,
+      "grad_norm": 2.049705982208252,
+      "learning_rate": 9.953429730181653e-05,
+      "loss": 3.1154,
+      "step": 96
+    },
+    {
+      "epoch": 0.18809980806142035,
+      "grad_norm": 2.397480010986328,
+      "learning_rate": 9.949004415088927e-05,
+      "loss": 3.1141,
+      "step": 98
+    },
+    {
+      "epoch": 0.19193857965451055,
+      "grad_norm": 2.1459343433380127,
+      "learning_rate": 9.944379351492818e-05,
+      "loss": 3.0847,
+      "step": 100
+    },
+    {
+      "epoch": 0.19577735124760076,
+      "grad_norm": 2.409905433654785,
+      "learning_rate": 9.939554726067141e-05,
+      "loss": 3.1939,
+      "step": 102
+    },
+    {
+      "epoch": 0.19961612284069097,
+      "grad_norm": 2.6216790676116943,
+      "learning_rate": 9.934530733540294e-05,
+      "loss": 3.1399,
+      "step": 104
+    },
+    {
+      "epoch": 0.2034548944337812,
+      "grad_norm": 2.3735368251800537,
+      "learning_rate": 9.929307576687403e-05,
+      "loss": 3.1325,
+      "step": 106
+    },
+    {
+      "epoch": 0.2072936660268714,
+      "grad_norm": 2.61091685295105,
+      "learning_rate": 9.923885466322135e-05,
+      "loss": 3.2417,
+      "step": 108
+    },
+    {
+      "epoch": 0.21113243761996162,
+      "grad_norm": 2.302191734313965,
+      "learning_rate": 9.918264621288187e-05,
+      "loss": 3.192,
+      "step": 110
+    },
+    {
+      "epoch": 0.21497120921305182,
+      "grad_norm": 2.6899170875549316,
+      "learning_rate": 9.912445268450458e-05,
+      "loss": 3.1271,
+      "step": 112
+    },
+    {
+      "epoch": 0.21880998080614203,
+      "grad_norm": 2.581143856048584,
+      "learning_rate": 9.906427642685889e-05,
+      "loss": 3.1607,
+      "step": 114
+    },
+    {
+      "epoch": 0.22264875239923224,
+      "grad_norm": 2.510422706604004,
+      "learning_rate": 9.900211986873985e-05,
+      "loss": 3.159,
+      "step": 116
+    },
+    {
+      "epoch": 0.22648752399232247,
+      "grad_norm": 2.3196523189544678,
+      "learning_rate": 9.89379855188701e-05,
+      "loss": 3.0774,
+      "step": 118
+    },
+    {
+      "epoch": 0.23032629558541268,
+      "grad_norm": 2.4071953296661377,
+      "learning_rate": 9.887187596579866e-05,
+      "loss": 3.0611,
+      "step": 120
+    },
+    {
+      "epoch": 0.23416506717850288,
+      "grad_norm": 2.733366012573242,
+      "learning_rate": 9.880379387779637e-05,
+      "loss": 3.0284,
+      "step": 122
+    },
+    {
+      "epoch": 0.2380038387715931,
+      "grad_norm": 2.3916118144989014,
+      "learning_rate": 9.873374200274826e-05,
+      "loss": 3.0748,
+      "step": 124
+    },
+    {
+      "epoch": 0.2418426103646833,
+      "grad_norm": 13.829148292541504,
+      "learning_rate": 9.866172316804266e-05,
+      "loss": 3.1429,
+      "step": 126
+    },
+    {
+      "epoch": 0.2456813819577735,
+      "grad_norm": 2.5056142807006836,
+      "learning_rate": 9.8587740280457e-05,
+      "loss": 2.9744,
+      "step": 128
+    },
+    {
+      "epoch": 0.2495201535508637,
+      "grad_norm": 2.326772689819336,
+      "learning_rate": 9.851179632604057e-05,
+      "loss": 3.136,
+      "step": 130
+    },
+    {
+      "epoch": 0.2533589251439539,
+      "grad_norm": 2.530068874359131,
+      "learning_rate": 9.843389436999396e-05,
+      "loss": 3.0951,
+      "step": 132
+    },
+    {
+      "epoch": 0.2571976967370441,
+      "grad_norm": 2.2792017459869385,
+      "learning_rate": 9.835403755654536e-05,
+      "loss": 3.1231,
+      "step": 134
+    },
+    {
+      "epoch": 0.26103646833013433,
+      "grad_norm": 2.3225743770599365,
+      "learning_rate": 9.827222910882358e-05,
+      "loss": 3.1742,
+      "step": 136
+    },
+    {
+      "epoch": 0.2648752399232246,
+      "grad_norm": 2.5904088020324707,
+      "learning_rate": 9.818847232872815e-05,
+      "loss": 3.0991,
+      "step": 138
+    },
+    {
+      "epoch": 0.2687140115163148,
+      "grad_norm": 2.1772799491882324,
+      "learning_rate": 9.81027705967958e-05,
+      "loss": 3.055,
+      "step": 140
+    },
+    {
+      "epoch": 0.272552783109405,
+      "grad_norm": 2.2711241245269775,
+      "learning_rate": 9.801512737206422e-05,
+      "loss": 3.1628,
+      "step": 142
+    },
+    {
+      "epoch": 0.2763915547024952,
+      "grad_norm": 2.514880895614624,
+      "learning_rate": 9.792554619193235e-05,
+      "loss": 2.9655,
+      "step": 144
+    },
+    {
+      "epoch": 0.2802303262955854,
+      "grad_norm": 2.381726026535034,
+      "learning_rate": 9.783403067201762e-05,
+      "loss": 2.9727,
+      "step": 146
+    },
+    {
+      "epoch": 0.2840690978886756,
+      "grad_norm": 2.410158395767212,
+      "learning_rate": 9.774058450601002e-05,
+      "loss": 3.114,
+      "step": 148
+    },
+    {
+      "epoch": 0.28790786948176583,
+      "grad_norm": 2.2826764583587646,
+      "learning_rate": 9.764521146552309e-05,
+      "loss": 3.0518,
+      "step": 150
+    },
+    {
+      "epoch": 0.29174664107485604,
+      "grad_norm": 2.6153564453125,
+      "learning_rate": 9.754791539994152e-05,
+      "loss": 3.0053,
+      "step": 152
+    },
+    {
+      "epoch": 0.29558541266794625,
+      "grad_norm": 2.96071195602417,
+      "learning_rate": 9.744870023626599e-05,
+      "loss": 2.9478,
+      "step": 154
+    },
+    {
+      "epoch": 0.29942418426103645,
+      "grad_norm": 2.3832027912139893,
+      "learning_rate": 9.734756997895448e-05,
+      "loss": 3.029,
+      "step": 156
+    },
+    {
+      "epoch": 0.30326295585412666,
+      "grad_norm": 2.4208216667175293,
+      "learning_rate": 9.724452870976085e-05,
+      "loss": 3.1009,
+      "step": 158
+    },
+    {
+      "epoch": 0.30710172744721687,
+      "grad_norm": 2.337291717529297,
+      "learning_rate": 9.713958058756985e-05,
+      "loss": 2.8477,
+      "step": 160
+    },
+    {
+      "epoch": 0.31094049904030713,
+      "grad_norm": 2.6344621181488037,
+      "learning_rate": 9.703272984822946e-05,
+      "loss": 3.079,
+      "step": 162
+    },
+    {
+      "epoch": 0.31477927063339733,
+      "grad_norm": 2.1119256019592285,
+      "learning_rate": 9.69239808043799e-05,
+      "loss": 2.9445,
+      "step": 164
+    },
+    {
+      "epoch": 0.31861804222648754,
+      "grad_norm": 2.3171935081481934,
+      "learning_rate": 9.681333784527944e-05,
+      "loss": 3.0269,
+      "step": 166
+    },
+    {
+      "epoch": 0.32245681381957775,
+      "grad_norm": 2.5069351196289062,
+      "learning_rate": 9.67008054366274e-05,
+      "loss": 2.9887,
+      "step": 168
+    },
+    {
+      "epoch": 0.32629558541266795,
+      "grad_norm": 2.182694673538208,
+      "learning_rate": 9.658638812038379e-05,
+      "loss": 3.104,
+      "step": 170
+    },
+    {
+      "epoch": 0.33013435700575816,
+      "grad_norm": 2.539505958557129,
+      "learning_rate": 9.647009051458604e-05,
+      "loss": 2.959,
+      "step": 172
+    },
+    {
+      "epoch": 0.33397312859884837,
+      "grad_norm": 2.111574649810791,
+      "learning_rate": 9.635191731316262e-05,
+      "loss": 2.9279,
+      "step": 174
+    },
+    {
+      "epoch": 0.3378119001919386,
+      "grad_norm": 2.2139837741851807,
+      "learning_rate": 9.623187328574357e-05,
+      "loss": 2.9781,
+      "step": 176
+    },
+    {
+      "epoch": 0.3416506717850288,
+      "grad_norm": 2.3730077743530273,
+      "learning_rate": 9.6109963277468e-05,
+      "loss": 2.9521,
+      "step": 178
+    },
+    {
+      "epoch": 0.345489443378119,
+      "grad_norm": 2.147397756576538,
+      "learning_rate": 9.598619220878853e-05,
+      "loss": 2.9534,
+      "step": 180
+    },
+    {
+      "epoch": 0.3493282149712092,
+      "grad_norm": 2.1337695121765137,
+      "learning_rate": 9.586056507527266e-05,
+      "loss": 2.9549,
+      "step": 182
+    },
+    {
+      "epoch": 0.3531669865642994,
+      "grad_norm": 2.397371530532837,
+      "learning_rate": 9.573308694740121e-05,
+      "loss": 2.9714,
+      "step": 184
+    },
+    {
+      "epoch": 0.3570057581573896,
+      "grad_norm": 2.208829164505005,
+      "learning_rate": 9.560376297036362e-05,
+      "loss": 2.9773,
+      "step": 186
+    },
+    {
+      "epoch": 0.36084452975047987,
+      "grad_norm": 2.0512075424194336,
+      "learning_rate": 9.547259836385031e-05,
+      "loss": 2.9261,
+      "step": 188
+    },
+    {
+      "epoch": 0.3646833013435701,
+      "grad_norm": 2.486699104309082,
+      "learning_rate": 9.533959842184195e-05,
+      "loss": 2.9401,
+      "step": 190
+    },
+    {
+      "epoch": 0.3685220729366603,
+      "grad_norm": 2.0567739009857178,
+      "learning_rate": 9.520476851239588e-05,
+      "loss": 3.0689,
+      "step": 192
+    },
+    {
+      "epoch": 0.3723608445297505,
+      "grad_norm": 2.1673007011413574,
+      "learning_rate": 9.506811407742937e-05,
+      "loss": 3.022,
+      "step": 194
+    },
+    {
+      "epoch": 0.3761996161228407,
+      "grad_norm": 2.1367907524108887,
+      "learning_rate": 9.492964063250001e-05,
+      "loss": 2.89,
+      "step": 196
+    },
+    {
+      "epoch": 0.3800383877159309,
+      "grad_norm": 2.0380022525787354,
+      "learning_rate": 9.478935376658307e-05,
+      "loss": 2.9577,
+      "step": 198
+    },
+    {
+      "epoch": 0.3838771593090211,
+      "grad_norm": 2.535660743713379,
+      "learning_rate": 9.4647259141846e-05,
+      "loss": 2.9808,
+      "step": 200
+    },
+    {
+      "epoch": 0.3877159309021113,
+      "grad_norm": 2.2464702129364014,
+      "learning_rate": 9.450336249341976e-05,
+      "loss": 3.0819,
+      "step": 202
+    },
+    {
+      "epoch": 0.3915547024952015,
+      "grad_norm": 2.316953659057617,
+      "learning_rate": 9.435766962916748e-05,
+      "loss": 2.9961,
+      "step": 204
+    },
+    {
+      "epoch": 0.39539347408829173,
+      "grad_norm": 2.124931573867798,
+      "learning_rate": 9.421018642944997e-05,
+      "loss": 2.9755,
+      "step": 206
+    },
+    {
+      "epoch": 0.39923224568138194,
+      "grad_norm": 2.299165964126587,
+      "learning_rate": 9.406091884688836e-05,
+      "loss": 2.9174,
+      "step": 208
+    },
+    {
+      "epoch": 0.40307101727447214,
+      "grad_norm": 2.156485080718994,
+      "learning_rate": 9.390987290612395e-05,
+      "loss": 2.9381,
+      "step": 210
+    },
+    {
+      "epoch": 0.4069097888675624,
+      "grad_norm": 2.207650899887085,
+      "learning_rate": 9.375705470357493e-05,
+      "loss": 2.893,
+      "step": 212
+    },
+    {
+      "epoch": 0.4107485604606526,
+      "grad_norm": 2.3216702938079834,
+      "learning_rate": 9.360247040719039e-05,
+      "loss": 2.9004,
+      "step": 214
+    },
+    {
+      "epoch": 0.4145873320537428,
+      "grad_norm": 2.2900643348693848,
+      "learning_rate": 9.344612625620132e-05,
+      "loss": 2.9238,
+      "step": 216
+    },
+    {
+      "epoch": 0.418426103646833,
+      "grad_norm": 2.157139539718628,
+      "learning_rate": 9.32880285608689e-05,
+      "loss": 2.8885,
+      "step": 218
+    },
+    {
+      "epoch": 0.42226487523992323,
+      "grad_norm": 2.0238444805145264,
+      "learning_rate": 9.312818370222962e-05,
+      "loss": 2.8743,
+      "step": 220
+    },
+    {
+      "epoch": 0.42610364683301344,
+      "grad_norm": 2.537557363510132,
+      "learning_rate": 9.296659813183794e-05,
+      "loss": 2.9497,
+      "step": 222
+    },
+    {
+      "epoch": 0.42994241842610365,
+      "grad_norm": 2.218683958053589,
+      "learning_rate": 9.280327837150571e-05,
+      "loss": 2.9384,
+      "step": 224
+    },
+    {
+      "epoch": 0.43378119001919385,
+      "grad_norm": 1.8829752206802368,
+      "learning_rate": 9.26382310130391e-05,
+      "loss": 2.9037,
+      "step": 226
+    },
+    {
+      "epoch": 0.43761996161228406,
+      "grad_norm": 2.2257378101348877,
+      "learning_rate": 9.247146271797245e-05,
+      "loss": 2.9468,
+      "step": 228
+    },
+    {
+      "epoch": 0.44145873320537427,
+      "grad_norm": 1.9686921834945679,
+      "learning_rate": 9.23029802172994e-05,
+      "loss": 2.8394,
+      "step": 230
+    },
+    {
+      "epoch": 0.44529750479846447,
+      "grad_norm": 2.367706298828125,
+      "learning_rate": 9.213279031120129e-05,
+      "loss": 2.7104,
+      "step": 232
+    },
+    {
+      "epoch": 0.4491362763915547,
+      "grad_norm": 2.439753532409668,
+      "learning_rate": 9.19608998687726e-05,
+      "loss": 2.9991,
+      "step": 234
+    },
+    {
+      "epoch": 0.45297504798464494,
+      "grad_norm": 1.997983694076538,
+      "learning_rate": 9.178731582774381e-05,
+      "loss": 2.887,
+      "step": 236
+    },
+    {
+      "epoch": 0.45681381957773515,
+      "grad_norm": 2.371289014816284,
+      "learning_rate": 9.161204519420126e-05,
+      "loss": 2.8716,
+      "step": 238
+    },
+    {
+      "epoch": 0.46065259117082535,
+      "grad_norm": 1.9359811544418335,
+      "learning_rate": 9.143509504230449e-05,
+      "loss": 2.952,
+      "step": 240
+    },
+    {
+      "epoch": 0.46449136276391556,
+      "grad_norm": 2.3096532821655273,
+      "learning_rate": 9.125647251400067e-05,
+      "loss": 2.84,
+      "step": 242
+    },
+    {
+      "epoch": 0.46833013435700577,
+      "grad_norm": 2.2196881771087646,
+      "learning_rate": 9.107618481873632e-05,
+      "loss": 2.8153,
+      "step": 244
+    },
+    {
+      "epoch": 0.472168905950096,
+      "grad_norm": 1.984487533569336,
+      "learning_rate": 9.089423923316636e-05,
+      "loss": 2.9233,
+      "step": 246
+    },
+    {
+      "epoch": 0.4760076775431862,
+      "grad_norm": 2.1207079887390137,
+      "learning_rate": 9.07106431008604e-05,
+      "loss": 2.7975,
+      "step": 248
+    },
+    {
+      "epoch": 0.4798464491362764,
+      "grad_norm": 1.9502159357070923,
+      "learning_rate": 9.052540383200633e-05,
+      "loss": 2.7933,
+      "step": 250
+    },
+    {
+      "epoch": 0.4836852207293666,
+      "grad_norm": 2.3264806270599365,
+      "learning_rate": 9.033852890311128e-05,
+      "loss": 2.948,
+      "step": 252
+    },
+    {
+      "epoch": 0.4875239923224568,
+      "grad_norm": 2.3154518604278564,
+      "learning_rate": 9.01500258566998e-05,
+      "loss": 2.8546,
+      "step": 254
+    },
+    {
+      "epoch": 0.491362763915547,
+      "grad_norm": 2.2312159538269043,
+      "learning_rate": 8.995990230100949e-05,
+      "loss": 2.8168,
+      "step": 256
+    },
+    {
+      "epoch": 0.4952015355086372,
+      "grad_norm": 2.6638643741607666,
+      "learning_rate": 8.976816590968387e-05,
+      "loss": 2.9337,
+      "step": 258
+    },
+    {
+      "epoch": 0.4990403071017274,
+      "grad_norm": 2.1371665000915527,
+      "learning_rate": 8.957482442146272e-05,
+      "loss": 2.8714,
+      "step": 260
+    },
+    {
+      "epoch": 0.5028790786948176,
+      "grad_norm": 1.9727420806884766,
+      "learning_rate": 8.937988563986963e-05,
+      "loss": 2.7677,
+      "step": 262
+    },
+    {
+      "epoch": 0.5067178502879078,
+      "grad_norm": 2.2152533531188965,
+      "learning_rate": 8.918335743289718e-05,
+      "loss": 2.9224,
+      "step": 264
+    },
+    {
+      "epoch": 0.510556621880998,
+      "grad_norm": 2.115664005279541,
+      "learning_rate": 8.898524773268926e-05,
+      "loss": 2.8969,
+      "step": 266
+    },
+    {
+      "epoch": 0.5143953934740882,
+      "grad_norm": 2.238097667694092,
+      "learning_rate": 8.8785564535221e-05,
+      "loss": 2.8914,
+      "step": 268
+    },
+    {
+      "epoch": 0.5182341650671785,
+      "grad_norm": 2.088069200515747,
+      "learning_rate": 8.858431589997597e-05,
+      "loss": 2.841,
+      "step": 270
+    },
+    {
+      "epoch": 0.5220729366602687,
+      "grad_norm": 2.0779953002929688,
+      "learning_rate": 8.838150994962093e-05,
+      "loss": 2.9019,
+      "step": 272
+    },
+    {
+      "epoch": 0.525911708253359,
+      "grad_norm": 2.2514681816101074,
+      "learning_rate": 8.817715486967803e-05,
+      "loss": 2.9179,
+      "step": 274
+    },
+    {
+      "epoch": 0.5297504798464492,
+      "grad_norm": 2.3873977661132812,
+      "learning_rate": 8.797125890819428e-05,
+      "loss": 2.8158,
+      "step": 276
+    },
+    {
+      "epoch": 0.5335892514395394,
+      "grad_norm": 1.8718311786651611,
+      "learning_rate": 8.776383037540888e-05,
+      "loss": 2.8838,
+      "step": 278
+    },
+    {
+      "epoch": 0.5374280230326296,
+      "grad_norm": 1.7923882007598877,
+      "learning_rate": 8.755487764341757e-05,
+      "loss": 2.8069,
+      "step": 280
+    },
+    {
+      "epoch": 0.5412667946257198,
+      "grad_norm": 2.197343587875366,
+      "learning_rate": 8.734440914583486e-05,
+      "loss": 2.8006,
+      "step": 282
+    },
+    {
+      "epoch": 0.54510556621881,
+      "grad_norm": 1.9336212873458862,
+      "learning_rate": 8.713243337745365e-05,
+      "loss": 2.8515,
+      "step": 284
+    },
+    {
+      "epoch": 0.5489443378119002,
+      "grad_norm": 2.2000906467437744,
+      "learning_rate": 8.691895889390227e-05,
+      "loss": 2.8156,
+      "step": 286
+    },
+    {
+      "epoch": 0.5527831094049904,
+      "grad_norm": 1.9124946594238281,
+      "learning_rate": 8.670399431129926e-05,
+      "loss": 2.6808,
+      "step": 288
+    },
+    {
+      "epoch": 0.5566218809980806,
+      "grad_norm": 2.50055193901062,
+      "learning_rate": 8.648754830590552e-05,
+      "loss": 2.9538,
+      "step": 290
+    },
+    {
+      "epoch": 0.5604606525911708,
+      "grad_norm": 2.0405330657958984,
+      "learning_rate": 8.626962961377423e-05,
+      "loss": 2.9034,
+      "step": 292
+    },
+    {
+      "epoch": 0.564299424184261,
+      "grad_norm": 2.101141929626465,
+      "learning_rate": 8.605024703039816e-05,
+      "loss": 2.8311,
+      "step": 294
+    },
+    {
+      "epoch": 0.5681381957773513,
+      "grad_norm": 1.957920789718628,
+      "learning_rate": 8.582940941035476e-05,
+      "loss": 2.8421,
+      "step": 296
+    },
+    {
+      "epoch": 0.5719769673704415,
+      "grad_norm": 2.24116849899292,
+      "learning_rate": 8.560712566694862e-05,
+      "loss": 2.9505,
+      "step": 298
+    },
+    {
+      "epoch": 0.5758157389635317,
+      "grad_norm": 2.053492546081543,
+      "learning_rate": 8.538340477185192e-05,
+      "loss": 2.8465,
+      "step": 300
+    },
+    {
+      "epoch": 0.5796545105566219,
+      "grad_norm": 2.3354341983795166,
+      "learning_rate": 8.515825575474221e-05,
+      "loss": 2.8588,
+      "step": 302
+    },
+    {
+      "epoch": 0.5834932821497121,
+      "grad_norm": 2.1598827838897705,
+      "learning_rate": 8.493168770293792e-05,
+      "loss": 2.8756,
+      "step": 304
+    },
+    {
+      "epoch": 0.5873320537428023,
+      "grad_norm": 2.1858606338500977,
+      "learning_rate": 8.47037097610317e-05,
+      "loss": 2.8049,
+      "step": 306
+    },
+    {
+      "epoch": 0.5911708253358925,
+      "grad_norm": 2.0640292167663574,
+      "learning_rate": 8.447433113052123e-05,
+      "loss": 2.8209,
+      "step": 308
+    },
+    {
+      "epoch": 0.5950095969289827,
+      "grad_norm": 2.2367477416992188,
+      "learning_rate": 8.42435610694379e-05,
+      "loss": 2.8078,
+      "step": 310
+    },
+    {
+      "epoch": 0.5988483685220729,
+      "grad_norm": 2.088447093963623,
+      "learning_rate": 8.401140889197305e-05,
+      "loss": 2.7268,
+      "step": 312
+    },
+    {
+      "epoch": 0.6026871401151631,
+      "grad_norm": 2.095773458480835,
+      "learning_rate": 8.377788396810223e-05,
+      "loss": 2.7654,
+      "step": 314
+    },
+    {
+      "epoch": 0.6065259117082533,
+      "grad_norm": 2.2851994037628174,
+      "learning_rate": 8.354299572320679e-05,
+      "loss": 2.7514,
+      "step": 316
+    },
+    {
+      "epoch": 0.6103646833013435,
+      "grad_norm": 1.7963942289352417,
+      "learning_rate": 8.330675363769356e-05,
+      "loss": 2.8477,
+      "step": 318
+    },
+    {
+      "epoch": 0.6142034548944337,
+      "grad_norm": 2.051821708679199,
+      "learning_rate": 8.306916724661224e-05,
+      "loss": 2.7684,
+      "step": 320
+    },
+    {
+      "epoch": 0.6180422264875239,
+      "grad_norm": 2.041626214981079,
+      "learning_rate": 8.283024613927054e-05,
+      "loss": 2.7836,
+      "step": 322
+    },
+    {
+      "epoch": 0.6218809980806143,
+      "grad_norm": 2.3092660903930664,
+      "learning_rate": 8.258999995884705e-05,
+      "loss": 2.722,
+      "step": 324
+    },
+    {
+      "epoch": 0.6257197696737045,
+      "grad_norm": 2.1960623264312744,
+      "learning_rate": 8.234843840200217e-05,
+      "loss": 2.7858,
+      "step": 326
+    },
+    {
+      "epoch": 0.6295585412667947,
+      "grad_norm": 1.9009817838668823,
+      "learning_rate": 8.210557121848663e-05,
+      "loss": 2.7208,
+      "step": 328
+    },
+    {
+      "epoch": 0.6333973128598849,
+      "grad_norm": 2.1019928455352783,
+      "learning_rate": 8.186140821074801e-05,
+      "loss": 2.9432,
+      "step": 330
+    },
+    {
+      "epoch": 0.6372360844529751,
+      "grad_norm": 1.9978450536727905,
+      "learning_rate": 8.161595923353516e-05,
+      "loss": 2.8016,
+      "step": 332
+    },
+    {
+      "epoch": 0.6410748560460653,
+      "grad_norm": 2.3676860332489014,
+      "learning_rate": 8.136923419350031e-05,
+      "loss": 2.8101,
+      "step": 334
+    },
+    {
+      "epoch": 0.6449136276391555,
+      "grad_norm": 2.006544589996338,
+      "learning_rate": 8.112124304879937e-05,
+      "loss": 2.7789,
+      "step": 336
+    },
+    {
+      "epoch": 0.6487523992322457,
+      "grad_norm": 2.0482800006866455,
+      "learning_rate": 8.087199580868996e-05,
+      "loss": 2.7815,
+      "step": 338
+    },
+    {
+      "epoch": 0.6525911708253359,
+      "grad_norm": 2.2319834232330322,
+      "learning_rate": 8.062150253312735e-05,
+      "loss": 2.8728,
+      "step": 340
+    },
+    {
+      "epoch": 0.6564299424184261,
+      "grad_norm": 3.3775405883789062,
+      "learning_rate": 8.03697733323585e-05,
+      "loss": 2.814,
+      "step": 342
+    },
+    {
+      "epoch": 0.6602687140115163,
+      "grad_norm": 2.242619514465332,
+      "learning_rate": 8.011681836651402e-05,
+      "loss": 2.8066,
+      "step": 344
+    },
+    {
+      "epoch": 0.6641074856046065,
+      "grad_norm": 2.016690254211426,
+      "learning_rate": 7.986264784519801e-05,
+      "loss": 2.7449,
+      "step": 346
+    },
+    {
+      "epoch": 0.6679462571976967,
+      "grad_norm": 2.1939826011657715,
+      "learning_rate": 7.960727202707605e-05,
+      "loss": 2.8091,
+      "step": 348
+    },
+    {
+      "epoch": 0.6717850287907869,
+      "grad_norm": 3.5264973640441895,
+      "learning_rate": 7.935070121946115e-05,
+      "loss": 2.8325,
+      "step": 350
+    },
+    {
+      "epoch": 0.6756238003838771,
+      "grad_norm": 2.000185489654541,
+      "learning_rate": 7.909294577789766e-05,
+      "loss": 2.789,
+      "step": 352
+    },
+    {
+      "epoch": 0.6794625719769674,
+      "grad_norm": 2.2280194759368896,
+      "learning_rate": 7.883401610574336e-05,
+      "loss": 2.8581,
+      "step": 354
+    },
+    {
+      "epoch": 0.6833013435700576,
+      "grad_norm": 1.8479573726654053,
+      "learning_rate": 7.857392265374963e-05,
+      "loss": 2.8131,
+      "step": 356
+    },
+    {
+      "epoch": 0.6871401151631478,
+      "grad_norm": 2.1635046005249023,
+      "learning_rate": 7.83126759196395e-05,
+      "loss": 2.8089,
+      "step": 358
+    },
+    {
+      "epoch": 0.690978886756238,
+      "grad_norm": 2.1297736167907715,
+      "learning_rate": 7.805028644768407e-05,
+      "loss": 2.8082,
+      "step": 360
+    },
+    {
+      "epoch": 0.6948176583493282,
+      "grad_norm": 1.9042068719863892,
+      "learning_rate": 7.778676482827686e-05,
+      "loss": 2.8027,
+      "step": 362
+    },
+    {
+      "epoch": 0.6986564299424184,
+      "grad_norm": 1.8313944339752197,
+      "learning_rate": 7.752212169750641e-05,
+      "loss": 2.8438,
+      "step": 364
+    },
+    {
+      "epoch": 0.7024952015355086,
+      "grad_norm": 2.036674976348877,
+      "learning_rate": 7.725636773672693e-05,
+      "loss": 2.8047,
+      "step": 366
+    },
+    {
+      "epoch": 0.7063339731285988,
+      "grad_norm": 2.086843490600586,
+      "learning_rate": 7.69895136721273e-05,
+      "loss": 2.6723,
+      "step": 368
+    },
+    {
+      "epoch": 0.710172744721689,
+      "grad_norm": 2.046743154525757,
+      "learning_rate": 7.672157027429803e-05,
+      "loss": 2.7864,
+      "step": 370
+    },
+    {
+      "epoch": 0.7140115163147792,
+      "grad_norm": 2.1428844928741455,
+      "learning_rate": 7.645254835779657e-05,
+      "loss": 2.841,
+      "step": 372
+    },
+    {
+      "epoch": 0.7178502879078695,
+      "grad_norm": 2.015566825866699,
+      "learning_rate": 7.618245878071091e-05,
+      "loss": 2.8675,
+      "step": 374
+    },
+    {
+      "epoch": 0.7216890595009597,
+      "grad_norm": 2.0343127250671387,
+      "learning_rate": 7.591131244422118e-05,
+      "loss": 2.8092,
+      "step": 376
+    },
+    {
+      "epoch": 0.72552783109405,
+      "grad_norm": 1.9547837972640991,
+      "learning_rate": 7.563912029215983e-05,
+      "loss": 2.7245,
+      "step": 378
+    },
+    {
+      "epoch": 0.7293666026871402,
+      "grad_norm": 1.8891335725784302,
+      "learning_rate": 7.536589331056976e-05,
+      "loss": 2.8459,
+      "step": 380
+    },
+    {
+      "epoch": 0.7332053742802304,
+      "grad_norm": 2.191988706588745,
+      "learning_rate": 7.509164252726107e-05,
+      "loss": 2.8502,
+      "step": 382
+    },
+    {
+      "epoch": 0.7370441458733206,
+      "grad_norm": 1.8953111171722412,
+      "learning_rate": 7.481637901136577e-05,
+      "loss": 2.7664,
+      "step": 384
+    },
+    {
+      "epoch": 0.7408829174664108,
+      "grad_norm": 1.8514469861984253,
+      "learning_rate": 7.454011387289127e-05,
+      "loss": 2.7179,
+      "step": 386
+    },
+    {
+      "epoch": 0.744721689059501,
+      "grad_norm": 1.9623198509216309,
+      "learning_rate": 7.42628582622717e-05,
+      "loss": 2.65,
+      "step": 388
+    },
+    {
+      "epoch": 0.7485604606525912,
+      "grad_norm": 1.9796512126922607,
+      "learning_rate": 7.398462336991801e-05,
+      "loss": 2.6551,
+      "step": 390
+    },
+    {
+      "epoch": 0.7523992322456814,
+      "grad_norm": 1.9456619024276733,
+      "learning_rate": 7.370542042576634e-05,
+      "loss": 2.7129,
+      "step": 392
+    },
+    {
+      "epoch": 0.7562380038387716,
+      "grad_norm": 2.237915277481079,
+      "learning_rate": 7.342526069882465e-05,
+      "loss": 2.8557,
+      "step": 394
+    },
+    {
+      "epoch": 0.7600767754318618,
+      "grad_norm": 1.9474000930786133,
+      "learning_rate": 7.314415549671795e-05,
+      "loss": 2.8152,
+      "step": 396
+    },
+    {
+      "epoch": 0.763915547024952,
+      "grad_norm": 2.2595555782318115,
+      "learning_rate": 7.286211616523193e-05,
+      "loss": 2.8161,
+      "step": 398
+    },
+    {
+      "epoch": 0.7677543186180422,
+      "grad_norm": 2.072669267654419,
+      "learning_rate": 7.257915408785498e-05,
+      "loss": 2.8623,
+      "step": 400
+    },
+    {
+      "epoch": 0.7715930902111324,
+      "grad_norm": 2.018782138824463,
+      "learning_rate": 7.229528068531882e-05,
+      "loss": 2.7831,
+      "step": 402
+    },
+    {
+      "epoch": 0.7754318618042226,
+      "grad_norm": 1.9737513065338135,
+      "learning_rate": 7.201050741513736e-05,
+      "loss": 2.6135,
+      "step": 404
+    },
+    {
+      "epoch": 0.7792706333973128,
+      "grad_norm": 1.8319437503814697,
+      "learning_rate": 7.17248457711445e-05,
+      "loss": 2.783,
+      "step": 406
+    },
+    {
+      "epoch": 0.783109404990403,
+      "grad_norm": 1.8370659351348877,
+      "learning_rate": 7.14383072830301e-05,
+      "loss": 2.7643,
+      "step": 408
+    },
+    {
+      "epoch": 0.7869481765834933,
+      "grad_norm": 2.0122272968292236,
+      "learning_rate": 7.115090351587454e-05,
+      "loss": 2.7853,
+      "step": 410
+    },
+    {
+      "epoch": 0.7907869481765835,
+      "grad_norm": 2.0002529621124268,
+      "learning_rate": 7.086264606968214e-05,
+      "loss": 2.7595,
+      "step": 412
+    },
+    {
+      "epoch": 0.7946257197696737,
+      "grad_norm": 2.0362918376922607,
+      "learning_rate": 7.05735465789128e-05,
+      "loss": 2.9258,
+      "step": 414
+    },
+    {
+      "epoch": 0.7984644913627639,
+      "grad_norm": 2.0595998764038086,
+      "learning_rate": 7.028361671201245e-05,
+      "loss": 2.7771,
+      "step": 416
+    },
+    {
+      "epoch": 0.8023032629558541,
+      "grad_norm": 2.0576417446136475,
+      "learning_rate": 6.99928681709422e-05,
+      "loss": 2.7221,
+      "step": 418
+    },
+    {
+      "epoch": 0.8061420345489443,
+      "grad_norm": 1.9781322479248047,
+      "learning_rate": 6.970131269070592e-05,
+      "loss": 2.8678,
+      "step": 420
+    },
+    {
+      "epoch": 0.8099808061420346,
+      "grad_norm": 2.429964780807495,
+      "learning_rate": 6.940896203887659e-05,
+      "loss": 2.7395,
+      "step": 422
+    },
+    {
+      "epoch": 0.8138195777351248,
+      "grad_norm": 2.1173887252807617,
+      "learning_rate": 6.911582801512147e-05,
+      "loss": 2.8196,
+      "step": 424
+    },
+    {
+      "epoch": 0.817658349328215,
+      "grad_norm": 2.0920073986053467,
+      "learning_rate": 6.882192245072569e-05,
+      "loss": 2.7426,
+      "step": 426
+    },
+    {
+      "epoch": 0.8214971209213052,
+      "grad_norm": 1.986169695854187,
+      "learning_rate": 6.852725720811488e-05,
+      "loss": 2.7175,
+      "step": 428
+    },
+    {
+      "epoch": 0.8253358925143954,
+      "grad_norm": 1.837952733039856,
+      "learning_rate": 6.823184418037624e-05,
+      "loss": 2.7456,
+      "step": 430
+    },
+    {
+      "epoch": 0.8291746641074856,
+      "grad_norm": 2.084582567214966,
+      "learning_rate": 6.793569529077864e-05,
+      "loss": 2.6918,
+      "step": 432
+    },
+    {
+      "epoch": 0.8330134357005758,
+      "grad_norm": 1.9959920644760132,
+      "learning_rate": 6.76388224922913e-05,
+      "loss": 2.7062,
+      "step": 434
+    },
+    {
+      "epoch": 0.836852207293666,
+      "grad_norm": 2.0004842281341553,
+      "learning_rate": 6.734123776710137e-05,
+      "loss": 2.7469,
+      "step": 436
+    },
+    {
+      "epoch": 0.8406909788867563,
+      "grad_norm": 2.0607264041900635,
+      "learning_rate": 6.704295312613037e-05,
+      "loss": 2.6908,
+      "step": 438
+    },
+    {
+      "epoch": 0.8445297504798465,
+      "grad_norm": 1.8044137954711914,
+      "learning_rate": 6.674398060854931e-05,
+      "loss": 2.7281,
+      "step": 440
+    },
+    {
+      "epoch": 0.8483685220729367,
+      "grad_norm": 1.741652488708496,
+      "learning_rate": 6.644433228129287e-05,
+      "loss": 2.7622,
+      "step": 442
+    },
+    {
+      "epoch": 0.8522072936660269,
+      "grad_norm": 2.165587902069092,
+      "learning_rate": 6.614402023857232e-05,
+      "loss": 2.8081,
+      "step": 444
+    },
+    {
+      "epoch": 0.8560460652591171,
+      "grad_norm": 1.9901096820831299,
+      "learning_rate": 6.584305660138734e-05,
+      "loss": 2.8187,
+      "step": 446
+    },
+    {
+      "epoch": 0.8598848368522073,
+      "grad_norm": 1.9680149555206299,
+      "learning_rate": 6.554145351703689e-05,
+      "loss": 2.7813,
+      "step": 448
+    },
+    {
+      "epoch": 0.8637236084452975,
+      "grad_norm": 2.150981903076172,
+      "learning_rate": 6.523922315862887e-05,
+      "loss": 2.6663,
+      "step": 450
+    },
+    {
+      "epoch": 0.8675623800383877,
+      "grad_norm": 2.4717190265655518,
+      "learning_rate": 6.493637772458879e-05,
+      "loss": 2.7571,
+      "step": 452
+    },
+    {
+      "epoch": 0.8714011516314779,
+      "grad_norm": 2.0950613021850586,
+      "learning_rate": 6.463292943816746e-05,
+      "loss": 2.8617,
+      "step": 454
+    },
+    {
+      "epoch": 0.8752399232245681,
+      "grad_norm": 1.7509244680404663,
+      "learning_rate": 6.432889054694764e-05,
+      "loss": 2.7157,
+      "step": 456
+    },
+    {
+      "epoch": 0.8790786948176583,
+      "grad_norm": 3.837655544281006,
+      "learning_rate": 6.402427332234965e-05,
+      "loss": 2.6557,
+      "step": 458
+    },
+    {
+      "epoch": 0.8829174664107485,
+      "grad_norm": 1.9463653564453125,
+      "learning_rate": 6.371909005913618e-05,
+      "loss": 2.7937,
+      "step": 460
+    },
+    {
+      "epoch": 0.8867562380038387,
+      "grad_norm": 1.785074234008789,
+      "learning_rate": 6.341335307491595e-05,
+      "loss": 2.7041,
+      "step": 462
+    },
+    {
+      "epoch": 0.8905950095969289,
+      "grad_norm": 2.044428586959839,
+      "learning_rate": 6.310707470964668e-05,
+      "loss": 2.7131,
+      "step": 464
+    },
+    {
+      "epoch": 0.8944337811900192,
+      "grad_norm": 2.0175092220306396,
+      "learning_rate": 6.280026732513689e-05,
+      "loss": 2.6959,
+      "step": 466
+    },
+    {
+      "epoch": 0.8982725527831094,
+      "grad_norm": 1.9430804252624512,
+      "learning_rate": 6.249294330454705e-05,
+      "loss": 2.7563,
+      "step": 468
+    },
+    {
+      "epoch": 0.9021113243761996,
+      "grad_norm": 1.999422311782837,
+      "learning_rate": 6.21851150518898e-05,
+      "loss": 2.7608,
+      "step": 470
+    },
+    {
+      "epoch": 0.9059500959692899,
+      "grad_norm": 2.7273123264312744,
+      "learning_rate": 6.187679499152919e-05,
+      "loss": 2.7552,
+      "step": 472
+    },
+    {
+      "epoch": 0.9097888675623801,
+      "grad_norm": 1.9456199407577515,
+      "learning_rate": 6.15679955676794e-05,
+      "loss": 2.689,
+      "step": 474
+    },
+    {
+      "epoch": 0.9136276391554703,
+      "grad_norm": 1.8011109828948975,
+      "learning_rate": 6.125872924390226e-05,
+      "loss": 2.6342,
+      "step": 476
+    },
+    {
+      "epoch": 0.9174664107485605,
+      "grad_norm": 1.790582299232483,
+      "learning_rate": 6.094900850260439e-05,
+      "loss": 2.7611,
+      "step": 478
+    },
+    {
+      "epoch": 0.9213051823416507,
+      "grad_norm": 2.047144889831543,
+      "learning_rate": 6.0638845844533254e-05,
+      "loss": 2.7965,
+      "step": 480
+    },
+    {
+      "epoch": 0.9251439539347409,
+      "grad_norm": 1.7579201459884644,
+      "learning_rate": 6.032825378827273e-05,
+      "loss": 2.7516,
+      "step": 482
+    },
+    {
+      "epoch": 0.9289827255278311,
+      "grad_norm": 1.806973934173584,
+      "learning_rate": 6.0017244869737746e-05,
+      "loss": 2.7111,
+      "step": 484
+    },
+    {
+      "epoch": 0.9328214971209213,
+      "grad_norm": 1.781657099723816,
+      "learning_rate": 5.970583164166838e-05,
+      "loss": 2.7412,
+      "step": 486
+    },
+    {
+      "epoch": 0.9366602687140115,
+      "grad_norm": 1.904548168182373,
+      "learning_rate": 5.9394026673123157e-05,
+      "loss": 2.7388,
+      "step": 488
+    },
+    {
+      "epoch": 0.9404990403071017,
+      "grad_norm": 1.8767707347869873,
+      "learning_rate": 5.908184254897182e-05,
+      "loss": 2.694,
+      "step": 490
+    },
+    {
+      "epoch": 0.944337811900192,
+      "grad_norm": 1.9801278114318848,
+      "learning_rate": 5.8769291869387335e-05,
+      "loss": 2.7866,
+      "step": 492
+    },
+    {
+      "epoch": 0.9481765834932822,
+      "grad_norm": 2.026012897491455,
+      "learning_rate": 5.845638724933728e-05,
+      "loss": 2.7375,
+      "step": 494
+    },
+    {
+      "epoch": 0.9520153550863724,
+      "grad_norm": 1.8350569009780884,
+      "learning_rate": 5.814314131807486e-05,
+      "loss": 2.7364,
+      "step": 496
+    },
+    {
+      "epoch": 0.9558541266794626,
+      "grad_norm": 1.9296355247497559,
+      "learning_rate": 5.7829566718628946e-05,
+      "loss": 2.682,
+      "step": 498
+    },
+    {
+      "epoch": 0.9596928982725528,
+      "grad_norm": 1.8304519653320312,
+      "learning_rate": 5.751567610729398e-05,
+      "loss": 2.6421,
+      "step": 500
+    },
+    {
+      "epoch": 0.963531669865643,
+      "grad_norm": 1.8141051530838013,
+      "learning_rate": 5.7201482153119015e-05,
+      "loss": 2.6552,
+      "step": 502
+    },
+    {
+      "epoch": 0.9673704414587332,
+      "grad_norm": 1.9663299322128296,
+      "learning_rate": 5.6886997537396494e-05,
+      "loss": 2.6086,
+      "step": 504
+    },
+    {
+      "epoch": 0.9712092130518234,
+      "grad_norm": 1.9281374216079712,
+      "learning_rate": 5.6572234953150305e-05,
+      "loss": 2.7493,
+      "step": 506
+    },
+    {
+      "epoch": 0.9750479846449136,
+      "grad_norm": 1.7819768190383911,
+      "learning_rate": 5.6257207104623524e-05,
+      "loss": 2.6959,
+      "step": 508
+    },
+    {
+      "epoch": 0.9788867562380038,
+      "grad_norm": 1.8469796180725098,
+      "learning_rate": 5.594192670676568e-05,
+      "loss": 2.6471,
+      "step": 510
+    },
+    {
+      "epoch": 0.982725527831094,
+      "grad_norm": 1.8483726978302002,
+      "learning_rate": 5.5626406484719506e-05,
+      "loss": 2.699,
+      "step": 512
+    },
+    {
+      "epoch": 0.9865642994241842,
+      "grad_norm": 1.8236881494522095,
+      "learning_rate": 5.5310659173307366e-05,
+      "loss": 2.746,
+      "step": 514
+    },
+    {
+      "epoch": 0.9904030710172744,
+      "grad_norm": 1.7808964252471924,
+      "learning_rate": 5.499469751651728e-05,
+      "loss": 2.7885,
+      "step": 516
+    },
+    {
+      "epoch": 0.9942418426103646,
+      "grad_norm": 1.828331708908081,
+      "learning_rate": 5.467853426698853e-05,
+      "loss": 2.7437,
+      "step": 518
+    },
+    {
+      "epoch": 0.9980806142034548,
+      "grad_norm": 1.7730071544647217,
+      "learning_rate": 5.43621821854969e-05,
+      "loss": 2.6718,
+      "step": 520
+    }
+  ],
+  "logging_steps": 2,
+  "max_steps": 1042,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 2,
+  "save_steps": 521,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 8.81062206945362e+16,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoint-521/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:72490e88c6bf2877f2bc15eaa6e0f01e604bb0f1022b7d74641eda9e89f52e21
+size 5432

config.json

@@ -0,0 +1,27 @@
+{
+  "_name_or_path": "/mnt/parscratch/users/acp23ay/private/models/OLMo-2-1124-7B-ary_Arab-coli-align/",
+  "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 sha256:3d4bce1a45701e55a95ce02f30455c3413855ce255ec0102192770b2d8d34e30
+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 sha256:16c198b2a120e6355187683d79784abf4b91d01c842ff312d327a8fc4b1d694b
+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 sha256:72490e88c6bf2877f2bc15eaa6e0f01e604bb0f1022b7d74641eda9e89f52e21
+size 5432