Hugging Face's logo

sdoouyangboya
/
llava-finetued-dpo

Model card Files Community
sdoouyangboya commited on
Commit
b30864f
·
verified ·
1 Parent(s): 022a111

Upload 18 files

Browse files
Files changed (18) hide show
  1. checkpoint-100-llava/config.json +74 -0
  2. checkpoint-100-llava/generation_config.json +6 -0
  3. checkpoint-100-llava/global_step100/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt +3 -0
  4. checkpoint-100-llava/global_step100/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt +3 -0
  5. checkpoint-100-llava/global_step100/zero_pp_rank_0_mp_rank_00_model_states.pt +3 -0
  6. checkpoint-100-llava/global_step100/zero_pp_rank_1_mp_rank_00_model_states.pt +3 -0
  7. checkpoint-100-llava/latest +1 -0
  8. checkpoint-100-llava/pytorch_model-00001-of-00002.bin +3 -0
  9. checkpoint-100-llava/pytorch_model-00002-of-00002.bin +3 -0
  10. checkpoint-100-llava/pytorch_model.bin.index.json +725 -0
  11. checkpoint-100-llava/rng_state_0.pth +3 -0
  12. checkpoint-100-llava/rng_state_1.pth +3 -0
  13. checkpoint-100-llava/special_tokens_map.json +24 -0
  14. checkpoint-100-llava/tokenizer.model +3 -0
  15. checkpoint-100-llava/tokenizer_config.json +72 -0
  16. checkpoint-100-llava/trainer_state.json +1616 -0
  17. checkpoint-100-llava/training_args.bin +3 -0
  18. checkpoint-100-llava/zero_to_fp32.py +578 -0
checkpoint-100-llava/config.json ADDED
@@ -0,0 +1,74 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "/mnt/datascience3/Boya/bouyang/llava-ftmodel\u2014twoepoch",
3
+ "architectures": [
4
+ "LlavaLlamaForCausalLM"
5
+ ],
6
+ "attention_bias": false,
7
+ "attention_dropout": 0.0,
8
+ "bos_token_id": 1,
9
+ "eos_token_id": 2,
10
+ "freeze_mm_mlp_adapter": false,
11
+ "freeze_mm_vision_resampler": false,
12
+ "hidden_act": "silu",
13
+ "hidden_size": 4096,
14
+ "image_aspect_ratio": "pad",
15
+ "image_crop_resolution": 224,
16
+ "image_grid_pinpoints": [
17
+ [
18
+ 336,
19
+ 672
20
+ ],
21
+ [
22
+ 672,
23
+ 336
24
+ ],
25
+ [
26
+ 672,
27
+ 672
28
+ ],
29
+ [
30
+ 1008,
31
+ 336
32
+ ],
33
+ [
34
+ 336,
35
+ 1008
36
+ ]
37
+ ],
38
+ "image_split_resolution": 224,
39
+ "initializer_range": 0.02,
40
+ "intermediate_size": 14336,
41
+ "max_position_embeddings": 32768,
42
+ "mm_hidden_size": 1024,
43
+ "mm_patch_merge_type": "spatial_unpad",
44
+ "mm_projector_lr": null,
45
+ "mm_projector_type": "mlp2x_gelu",
46
+ "mm_resampler_type": null,
47
+ "mm_use_im_patch_token": false,
48
+ "mm_use_im_start_end": false,
49
+ "mm_vision_select_feature": "patch",
50
+ "mm_vision_select_layer": -2,
51
+ "mm_vision_tower": "model/clip-vit-large-patch14-336",
52
+ "mm_vision_tower_lr": 2e-06,
53
+ "model_type": "llava",
54
+ "num_attention_heads": 32,
55
+ "num_hidden_layers": 32,
56
+ "num_key_value_heads": 8,
57
+ "pad_token_id": 0,
58
+ "pretraining_tp": 1,
59
+ "rms_norm_eps": 1e-05,
60
+ "rope_scaling": null,
61
+ "rope_theta": 1000000.0,
62
+ "sliding_window": null,
63
+ "tie_word_embeddings": false,
64
+ "tokenizer_model_max_length": 2048,
65
+ "tokenizer_padding_side": "right",
66
+ "torch_dtype": "bfloat16",
67
+ "transformers_version": "4.31.0",
68
+ "tune_mm_mlp_adapter": false,
69
+ "tune_mm_vision_resampler": false,
70
+ "unfreeze_mm_vision_tower": true,
71
+ "use_cache": false,
72
+ "use_mm_proj": true,
73
+ "vocab_size": 32000
74
+ }
checkpoint-100-llava/generation_config.json ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 1,
4
+ "eos_token_id": 2,
5
+ "transformers_version": "4.31.0"
6
+ }
checkpoint-100-llava/global_step100/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ff002d1e4a4e3cc2d3c7fe2671492e90cfaf397a76670ee248981a983bc7d05b
3
+ size 912314871
checkpoint-100-llava/global_step100/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9681badb4470e3025d375f8ff3d0bcdb500b2ccd81202f19458a6d537a385ca7
3
+ size 912314871
checkpoint-100-llava/global_step100/zero_pp_rank_0_mp_rank_00_model_states.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:98c1696e97171c09831ddd43c846d0420dc259a3d19b5218a5daf3a6f523b392
3
+ size 1019478016
checkpoint-100-llava/global_step100/zero_pp_rank_1_mp_rank_00_model_states.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1ccacd3ab037412a6d5dc89780f55977911fa99094506b278c92258827d9cefe
3
+ size 1018691584
checkpoint-100-llava/latest ADDED
@@ -0,0 +1 @@
 
 
1
+ global_step100
checkpoint-100-llava/pytorch_model-00001-of-00002.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d1e7e0b672c0130bb7806a1c46ec886a31850ba4a9a53f22c04fb7eca7557471
3
+ size 1018429440
checkpoint-100-llava/pytorch_model-00002-of-00002.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fdb29e9f6dfe8381efa76701aca1249d37c0564adfd884518d9d77edca3d04ce
3
+ size 1020002304
checkpoint-100-llava/pytorch_model.bin.index.json ADDED
@@ -0,0 +1,725 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "metadata": {
3
+ "total_size": 15132442624
4
+ },
5
+ "weight_map": {
6
+ "lm_head.weight": "pytorch_model-00002-of-00002.bin",
7
+ "model.embed_tokens.weight": "pytorch_model-00001-of-00002.bin",
8
+ "model.layers.0.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
9
+ "model.layers.0.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
10
+ "model.layers.0.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
11
+ "model.layers.0.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
12
+ "model.layers.0.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
13
+ "model.layers.0.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
14
+ "model.layers.0.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
15
+ "model.layers.0.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
16
+ "model.layers.0.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
17
+ "model.layers.0.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
18
+ "model.layers.1.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
19
+ "model.layers.1.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
20
+ "model.layers.1.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
21
+ "model.layers.1.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
22
+ "model.layers.1.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
23
+ "model.layers.1.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
24
+ "model.layers.1.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
25
+ "model.layers.1.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
26
+ "model.layers.1.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
27
+ "model.layers.1.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
28
+ "model.layers.10.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
29
+ "model.layers.10.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
30
+ "model.layers.10.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
31
+ "model.layers.10.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
32
+ "model.layers.10.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
33
+ "model.layers.10.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
34
+ "model.layers.10.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
35
+ "model.layers.10.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
36
+ "model.layers.10.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
37
+ "model.layers.10.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
38
+ "model.layers.11.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
39
+ "model.layers.11.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
40
+ "model.layers.11.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
41
+ "model.layers.11.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
42
+ "model.layers.11.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
43
+ "model.layers.11.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
44
+ "model.layers.11.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
45
+ "model.layers.11.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
46
+ "model.layers.11.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
47
+ "model.layers.11.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
48
+ "model.layers.12.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
49
+ "model.layers.12.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
50
+ "model.layers.12.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
51
+ "model.layers.12.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
52
+ "model.layers.12.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
53
+ "model.layers.12.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
54
+ "model.layers.12.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
55
+ "model.layers.12.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
56
+ "model.layers.12.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
57
+ "model.layers.12.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
58
+ "model.layers.13.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
59
+ "model.layers.13.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
60
+ "model.layers.13.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
61
+ "model.layers.13.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
62
+ "model.layers.13.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
63
+ "model.layers.13.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
64
+ "model.layers.13.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
65
+ "model.layers.13.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
66
+ "model.layers.13.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
67
+ "model.layers.13.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
68
+ "model.layers.14.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
69
+ "model.layers.14.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
70
+ "model.layers.14.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
71
+ "model.layers.14.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
72
+ "model.layers.14.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
73
+ "model.layers.14.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
74
+ "model.layers.14.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
75
+ "model.layers.14.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
76
+ "model.layers.14.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
77
+ "model.layers.14.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
78
+ "model.layers.15.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
79
+ "model.layers.15.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
80
+ "model.layers.15.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
81
+ "model.layers.15.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
82
+ "model.layers.15.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
83
+ "model.layers.15.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
84
+ "model.layers.15.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
85
+ "model.layers.15.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
86
+ "model.layers.15.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
87
+ "model.layers.15.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
88
+ "model.layers.16.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
89
+ "model.layers.16.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
90
+ "model.layers.16.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
91
+ "model.layers.16.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
92
+ "model.layers.16.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
93
+ "model.layers.16.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
94
+ "model.layers.16.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
95
+ "model.layers.16.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
96
+ "model.layers.16.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
97
+ "model.layers.16.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
98
+ "model.layers.17.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
99
+ "model.layers.17.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
100
+ "model.layers.17.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
101
+ "model.layers.17.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
102
+ "model.layers.17.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
103
+ "model.layers.17.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
104
+ "model.layers.17.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
105
+ "model.layers.17.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
106
+ "model.layers.17.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
107
+ "model.layers.17.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
108
+ "model.layers.18.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
109
+ "model.layers.18.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
110
+ "model.layers.18.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
111
+ "model.layers.18.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
112
+ "model.layers.18.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
113
+ "model.layers.18.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
114
+ "model.layers.18.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
115
+ "model.layers.18.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
116
+ "model.layers.18.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
117
+ "model.layers.18.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
118
+ "model.layers.19.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
119
+ "model.layers.19.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
120
+ "model.layers.19.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
121
+ "model.layers.19.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
122
+ "model.layers.19.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
123
+ "model.layers.19.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
124
+ "model.layers.19.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
125
+ "model.layers.19.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
126
+ "model.layers.19.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
127
+ "model.layers.19.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
128
+ "model.layers.2.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
129
+ "model.layers.2.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
130
+ "model.layers.2.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
131
+ "model.layers.2.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
132
+ "model.layers.2.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
133
+ "model.layers.2.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
134
+ "model.layers.2.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
135
+ "model.layers.2.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
136
+ "model.layers.2.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
137
+ "model.layers.2.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
138
+ "model.layers.20.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
139
+ "model.layers.20.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
140
+ "model.layers.20.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
141
+ "model.layers.20.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
142
+ "model.layers.20.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
143
+ "model.layers.20.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
144
+ "model.layers.20.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
145
+ "model.layers.20.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
146
+ "model.layers.20.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
147
+ "model.layers.20.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
148
+ "model.layers.21.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
149
+ "model.layers.21.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
150
+ "model.layers.21.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
151
+ "model.layers.21.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
152
+ "model.layers.21.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
153
+ "model.layers.21.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
154
+ "model.layers.21.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
155
+ "model.layers.21.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
156
+ "model.layers.21.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
157
+ "model.layers.21.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
158
+ "model.layers.22.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
159
+ "model.layers.22.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
160
+ "model.layers.22.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
161
+ "model.layers.22.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
162
+ "model.layers.22.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
163
+ "model.layers.22.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
164
+ "model.layers.22.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
165
+ "model.layers.22.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
166
+ "model.layers.22.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
167
+ "model.layers.22.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
168
+ "model.layers.23.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
169
+ "model.layers.23.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
170
+ "model.layers.23.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
171
+ "model.layers.23.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
172
+ "model.layers.23.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
173
+ "model.layers.23.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
174
+ "model.layers.23.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
175
+ "model.layers.23.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
176
+ "model.layers.23.self_attn.rotary_emb.inv_freq": "pytorch_model-00002-of-00002.bin",
177
+ "model.layers.23.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
178
+ "model.layers.24.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
179
+ "model.layers.24.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
180
+ "model.layers.24.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
181
+ "model.layers.24.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
182
+ "model.layers.24.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
183
+ "model.layers.24.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
184
+ "model.layers.24.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
185
+ "model.layers.24.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
186
+ "model.layers.24.self_attn.rotary_emb.inv_freq": "pytorch_model-00002-of-00002.bin",
187
+ "model.layers.24.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
188
+ "model.layers.25.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
189
+ "model.layers.25.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
190
+ "model.layers.25.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
191
+ "model.layers.25.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
192
+ "model.layers.25.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
193
+ "model.layers.25.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
194
+ "model.layers.25.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
195
+ "model.layers.25.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
196
+ "model.layers.25.self_attn.rotary_emb.inv_freq": "pytorch_model-00002-of-00002.bin",
197
+ "model.layers.25.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
198
+ "model.layers.26.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
199
+ "model.layers.26.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
200
+ "model.layers.26.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
201
+ "model.layers.26.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
202
+ "model.layers.26.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
203
+ "model.layers.26.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
204
+ "model.layers.26.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
205
+ "model.layers.26.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
206
+ "model.layers.26.self_attn.rotary_emb.inv_freq": "pytorch_model-00002-of-00002.bin",
207
+ "model.layers.26.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
208
+ "model.layers.27.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
209
+ "model.layers.27.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
210
+ "model.layers.27.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
211
+ "model.layers.27.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
212
+ "model.layers.27.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
213
+ "model.layers.27.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
214
+ "model.layers.27.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
215
+ "model.layers.27.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
216
+ "model.layers.27.self_attn.rotary_emb.inv_freq": "pytorch_model-00002-of-00002.bin",
217
+ "model.layers.27.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
218
+ "model.layers.28.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
219
+ "model.layers.28.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
220
+ "model.layers.28.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
221
+ "model.layers.28.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
222
+ "model.layers.28.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
223
+ "model.layers.28.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
224
+ "model.layers.28.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
225
+ "model.layers.28.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
226
+ "model.layers.28.self_attn.rotary_emb.inv_freq": "pytorch_model-00002-of-00002.bin",
227
+ "model.layers.28.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
228
+ "model.layers.29.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
229
+ "model.layers.29.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
230
+ "model.layers.29.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
231
+ "model.layers.29.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
232
+ "model.layers.29.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
233
+ "model.layers.29.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
234
+ "model.layers.29.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
235
+ "model.layers.29.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
236
+ "model.layers.29.self_attn.rotary_emb.inv_freq": "pytorch_model-00002-of-00002.bin",
237
+ "model.layers.29.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
238
+ "model.layers.3.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
239
+ "model.layers.3.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
240
+ "model.layers.3.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
241
+ "model.layers.3.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
242
+ "model.layers.3.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
243
+ "model.layers.3.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
244
+ "model.layers.3.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
245
+ "model.layers.3.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
246
+ "model.layers.3.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
247
+ "model.layers.3.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
248
+ "model.layers.30.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
249
+ "model.layers.30.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
250
+ "model.layers.30.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
251
+ "model.layers.30.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
252
+ "model.layers.30.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
253
+ "model.layers.30.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
254
+ "model.layers.30.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
255
+ "model.layers.30.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
256
+ "model.layers.30.self_attn.rotary_emb.inv_freq": "pytorch_model-00002-of-00002.bin",
257
+ "model.layers.30.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
258
+ "model.layers.31.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
259
+ "model.layers.31.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
260
+ "model.layers.31.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
261
+ "model.layers.31.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
262
+ "model.layers.31.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
263
+ "model.layers.31.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
264
+ "model.layers.31.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
265
+ "model.layers.31.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
266
+ "model.layers.31.self_attn.rotary_emb.inv_freq": "pytorch_model-00002-of-00002.bin",
267
+ "model.layers.31.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
268
+ "model.layers.4.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
269
+ "model.layers.4.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
270
+ "model.layers.4.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
271
+ "model.layers.4.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
272
+ "model.layers.4.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
273
+ "model.layers.4.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
274
+ "model.layers.4.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
275
+ "model.layers.4.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
276
+ "model.layers.4.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
277
+ "model.layers.4.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
278
+ "model.layers.5.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
279
+ "model.layers.5.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
280
+ "model.layers.5.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
281
+ "model.layers.5.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
282
+ "model.layers.5.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
283
+ "model.layers.5.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
284
+ "model.layers.5.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
285
+ "model.layers.5.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
286
+ "model.layers.5.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
287
+ "model.layers.5.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
288
+ "model.layers.6.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
289
+ "model.layers.6.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
290
+ "model.layers.6.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
291
+ "model.layers.6.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
292
+ "model.layers.6.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
293
+ "model.layers.6.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
294
+ "model.layers.6.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
295
+ "model.layers.6.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
296
+ "model.layers.6.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
297
+ "model.layers.6.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
298
+ "model.layers.7.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
299
+ "model.layers.7.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
300
+ "model.layers.7.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
301
+ "model.layers.7.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
302
+ "model.layers.7.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
303
+ "model.layers.7.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
304
+ "model.layers.7.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
305
+ "model.layers.7.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
306
+ "model.layers.7.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
307
+ "model.layers.7.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
308
+ "model.layers.8.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
309
+ "model.layers.8.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
310
+ "model.layers.8.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
311
+ "model.layers.8.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
312
+ "model.layers.8.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
313
+ "model.layers.8.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
314
+ "model.layers.8.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
315
+ "model.layers.8.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
316
+ "model.layers.8.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
317
+ "model.layers.8.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
318
+ "model.layers.9.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
319
+ "model.layers.9.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
320
+ "model.layers.9.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
321
+ "model.layers.9.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
322
+ "model.layers.9.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
323
+ "model.layers.9.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
324
+ "model.layers.9.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
325
+ "model.layers.9.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
326
+ "model.layers.9.self_attn.rotary_emb.inv_freq": "pytorch_model-00001-of-00002.bin",
327
+ "model.layers.9.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
328
+ "model.mm_projector.0.bias": "pytorch_model-00002-of-00002.bin",
329
+ "model.mm_projector.0.weight": "pytorch_model-00002-of-00002.bin",
330
+ "model.mm_projector.2.bias": "pytorch_model-00002-of-00002.bin",
331
+ "model.mm_projector.2.weight": "pytorch_model-00002-of-00002.bin",
332
+ "model.norm.weight": "pytorch_model-00002-of-00002.bin",
333
+ "model.vision_tower.vision_tower.vision_model.embeddings.class_embedding": "pytorch_model-00002-of-00002.bin",
334
+ "model.vision_tower.vision_tower.vision_model.embeddings.patch_embedding.weight": "pytorch_model-00002-of-00002.bin",
335
+ "model.vision_tower.vision_tower.vision_model.embeddings.position_embedding.weight": "pytorch_model-00002-of-00002.bin",
336
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.0.layer_norm1.bias": "pytorch_model-00002-of-00002.bin",
337
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.0.layer_norm1.weight": "pytorch_model-00002-of-00002.bin",
338
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.0.layer_norm2.bias": "pytorch_model-00002-of-00002.bin",
339
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.0.layer_norm2.weight": "pytorch_model-00002-of-00002.bin",
340
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.0.mlp.fc1.bias": "pytorch_model-00002-of-00002.bin",
341
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.0.mlp.fc1.weight": "pytorch_model-00002-of-00002.bin",
342
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.0.mlp.fc2.bias": "pytorch_model-00002-of-00002.bin",
343
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.0.mlp.fc2.weight": "pytorch_model-00002-of-00002.bin",
344
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.0.self_attn.k_proj.bias": "pytorch_model-00002-of-00002.bin",
345
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.0.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
346
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.0.self_attn.out_proj.bias": "pytorch_model-00002-of-00002.bin",
347
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.0.self_attn.out_proj.weight": "pytorch_model-00002-of-00002.bin",
348
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.0.self_attn.q_proj.bias": "pytorch_model-00002-of-00002.bin",
349
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.0.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
350
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.0.self_attn.v_proj.bias": "pytorch_model-00002-of-00002.bin",
351
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.0.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
352
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.1.layer_norm1.bias": "pytorch_model-00002-of-00002.bin",
353
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.1.layer_norm1.weight": "pytorch_model-00002-of-00002.bin",
354
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.1.layer_norm2.bias": "pytorch_model-00002-of-00002.bin",
355
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.1.layer_norm2.weight": "pytorch_model-00002-of-00002.bin",
356
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.1.mlp.fc1.bias": "pytorch_model-00002-of-00002.bin",
357
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.1.mlp.fc1.weight": "pytorch_model-00002-of-00002.bin",
358
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.1.mlp.fc2.bias": "pytorch_model-00002-of-00002.bin",
359
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.1.mlp.fc2.weight": "pytorch_model-00002-of-00002.bin",
360
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.1.self_attn.k_proj.bias": "pytorch_model-00002-of-00002.bin",
361
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.1.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
362
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.1.self_attn.out_proj.bias": "pytorch_model-00002-of-00002.bin",
363
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.1.self_attn.out_proj.weight": "pytorch_model-00002-of-00002.bin",
364
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.1.self_attn.q_proj.bias": "pytorch_model-00002-of-00002.bin",
365
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.1.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
366
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.1.self_attn.v_proj.bias": "pytorch_model-00002-of-00002.bin",
367
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.1.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
368
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.10.layer_norm1.bias": "pytorch_model-00002-of-00002.bin",
369
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.10.layer_norm1.weight": "pytorch_model-00002-of-00002.bin",
370
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.10.layer_norm2.bias": "pytorch_model-00002-of-00002.bin",
371
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.10.layer_norm2.weight": "pytorch_model-00002-of-00002.bin",
372
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.10.mlp.fc1.bias": "pytorch_model-00002-of-00002.bin",
373
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.10.mlp.fc1.weight": "pytorch_model-00002-of-00002.bin",
374
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.10.mlp.fc2.bias": "pytorch_model-00002-of-00002.bin",
375
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.10.mlp.fc2.weight": "pytorch_model-00002-of-00002.bin",
376
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.10.self_attn.k_proj.bias": "pytorch_model-00002-of-00002.bin",
377
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.10.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
378
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.10.self_attn.out_proj.bias": "pytorch_model-00002-of-00002.bin",
379
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.10.self_attn.out_proj.weight": "pytorch_model-00002-of-00002.bin",
380
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.10.self_attn.q_proj.bias": "pytorch_model-00002-of-00002.bin",
381
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.10.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
382
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.10.self_attn.v_proj.bias": "pytorch_model-00002-of-00002.bin",
383
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.10.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
384
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.11.layer_norm1.bias": "pytorch_model-00002-of-00002.bin",
385
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.11.layer_norm1.weight": "pytorch_model-00002-of-00002.bin",
386
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.11.layer_norm2.bias": "pytorch_model-00002-of-00002.bin",
387
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.11.layer_norm2.weight": "pytorch_model-00002-of-00002.bin",
388
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.11.mlp.fc1.bias": "pytorch_model-00002-of-00002.bin",
389
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.11.mlp.fc1.weight": "pytorch_model-00002-of-00002.bin",
390
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.11.mlp.fc2.bias": "pytorch_model-00002-of-00002.bin",
391
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.11.mlp.fc2.weight": "pytorch_model-00002-of-00002.bin",
392
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.11.self_attn.k_proj.bias": "pytorch_model-00002-of-00002.bin",
393
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.11.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
394
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.11.self_attn.out_proj.bias": "pytorch_model-00002-of-00002.bin",
395
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.11.self_attn.out_proj.weight": "pytorch_model-00002-of-00002.bin",
396
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.11.self_attn.q_proj.bias": "pytorch_model-00002-of-00002.bin",
397
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.11.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
398
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.11.self_attn.v_proj.bias": "pytorch_model-00002-of-00002.bin",
399
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.11.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
400
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.12.layer_norm1.bias": "pytorch_model-00002-of-00002.bin",
401
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.12.layer_norm1.weight": "pytorch_model-00002-of-00002.bin",
402
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.12.layer_norm2.bias": "pytorch_model-00002-of-00002.bin",
403
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.12.layer_norm2.weight": "pytorch_model-00002-of-00002.bin",
404
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.12.mlp.fc1.bias": "pytorch_model-00002-of-00002.bin",
405
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.12.mlp.fc1.weight": "pytorch_model-00002-of-00002.bin",
406
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.12.mlp.fc2.bias": "pytorch_model-00002-of-00002.bin",
407
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.12.mlp.fc2.weight": "pytorch_model-00002-of-00002.bin",
408
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.12.self_attn.k_proj.bias": "pytorch_model-00002-of-00002.bin",
409
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.12.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
410
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.12.self_attn.out_proj.bias": "pytorch_model-00002-of-00002.bin",
411
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.12.self_attn.out_proj.weight": "pytorch_model-00002-of-00002.bin",
412
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.12.self_attn.q_proj.bias": "pytorch_model-00002-of-00002.bin",
413
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.12.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
414
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.12.self_attn.v_proj.bias": "pytorch_model-00002-of-00002.bin",
415
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.12.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
416
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.13.layer_norm1.bias": "pytorch_model-00002-of-00002.bin",
417
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.13.layer_norm1.weight": "pytorch_model-00002-of-00002.bin",
418
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.13.layer_norm2.bias": "pytorch_model-00002-of-00002.bin",
419
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.13.layer_norm2.weight": "pytorch_model-00002-of-00002.bin",
420
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.13.mlp.fc1.bias": "pytorch_model-00002-of-00002.bin",
421
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.13.mlp.fc1.weight": "pytorch_model-00002-of-00002.bin",
422
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.13.mlp.fc2.bias": "pytorch_model-00002-of-00002.bin",
423
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.13.mlp.fc2.weight": "pytorch_model-00002-of-00002.bin",
424
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.13.self_attn.k_proj.bias": "pytorch_model-00002-of-00002.bin",
425
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.13.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
426
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.13.self_attn.out_proj.bias": "pytorch_model-00002-of-00002.bin",
427
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.13.self_attn.out_proj.weight": "pytorch_model-00002-of-00002.bin",
428
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.13.self_attn.q_proj.bias": "pytorch_model-00002-of-00002.bin",
429
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.13.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
430
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.13.self_attn.v_proj.bias": "pytorch_model-00002-of-00002.bin",
431
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.13.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
432
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.14.layer_norm1.bias": "pytorch_model-00002-of-00002.bin",
433
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.14.layer_norm1.weight": "pytorch_model-00002-of-00002.bin",
434
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.14.layer_norm2.bias": "pytorch_model-00002-of-00002.bin",
435
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.14.layer_norm2.weight": "pytorch_model-00002-of-00002.bin",
436
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.14.mlp.fc1.bias": "pytorch_model-00002-of-00002.bin",
437
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.14.mlp.fc1.weight": "pytorch_model-00002-of-00002.bin",
438
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.14.mlp.fc2.bias": "pytorch_model-00002-of-00002.bin",
439
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.14.mlp.fc2.weight": "pytorch_model-00002-of-00002.bin",
440
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.14.self_attn.k_proj.bias": "pytorch_model-00002-of-00002.bin",
441
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.14.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
442
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.14.self_attn.out_proj.bias": "pytorch_model-00002-of-00002.bin",
443
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.14.self_attn.out_proj.weight": "pytorch_model-00002-of-00002.bin",
444
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.14.self_attn.q_proj.bias": "pytorch_model-00002-of-00002.bin",
445
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.14.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
446
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.14.self_attn.v_proj.bias": "pytorch_model-00002-of-00002.bin",
447
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.14.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
448
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.15.layer_norm1.bias": "pytorch_model-00002-of-00002.bin",
449
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.15.layer_norm1.weight": "pytorch_model-00002-of-00002.bin",
450
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.15.layer_norm2.bias": "pytorch_model-00002-of-00002.bin",
451
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.15.layer_norm2.weight": "pytorch_model-00002-of-00002.bin",
452
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.15.mlp.fc1.bias": "pytorch_model-00002-of-00002.bin",
453
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.15.mlp.fc1.weight": "pytorch_model-00002-of-00002.bin",
454
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.15.mlp.fc2.bias": "pytorch_model-00002-of-00002.bin",
455
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.15.mlp.fc2.weight": "pytorch_model-00002-of-00002.bin",
456
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.15.self_attn.k_proj.bias": "pytorch_model-00002-of-00002.bin",
457
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.15.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
458
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.15.self_attn.out_proj.bias": "pytorch_model-00002-of-00002.bin",
459
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.15.self_attn.out_proj.weight": "pytorch_model-00002-of-00002.bin",
460
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.15.self_attn.q_proj.bias": "pytorch_model-00002-of-00002.bin",
461
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.15.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
462
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.15.self_attn.v_proj.bias": "pytorch_model-00002-of-00002.bin",
463
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.15.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
464
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.16.layer_norm1.bias": "pytorch_model-00002-of-00002.bin",
465
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.16.layer_norm1.weight": "pytorch_model-00002-of-00002.bin",
466
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.16.layer_norm2.bias": "pytorch_model-00002-of-00002.bin",
467
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.16.layer_norm2.weight": "pytorch_model-00002-of-00002.bin",
468
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.16.mlp.fc1.bias": "pytorch_model-00002-of-00002.bin",
469
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.16.mlp.fc1.weight": "pytorch_model-00002-of-00002.bin",
470
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.16.mlp.fc2.bias": "pytorch_model-00002-of-00002.bin",
471
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.16.mlp.fc2.weight": "pytorch_model-00002-of-00002.bin",
472
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.16.self_attn.k_proj.bias": "pytorch_model-00002-of-00002.bin",
473
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.16.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
474
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.16.self_attn.out_proj.bias": "pytorch_model-00002-of-00002.bin",
475
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.16.self_attn.out_proj.weight": "pytorch_model-00002-of-00002.bin",
476
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.16.self_attn.q_proj.bias": "pytorch_model-00002-of-00002.bin",
477
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.16.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
478
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.16.self_attn.v_proj.bias": "pytorch_model-00002-of-00002.bin",
479
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.16.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
480
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.17.layer_norm1.bias": "pytorch_model-00002-of-00002.bin",
481
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.17.layer_norm1.weight": "pytorch_model-00002-of-00002.bin",
482
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.17.layer_norm2.bias": "pytorch_model-00002-of-00002.bin",
483
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.17.layer_norm2.weight": "pytorch_model-00002-of-00002.bin",
484
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.17.mlp.fc1.bias": "pytorch_model-00002-of-00002.bin",
485
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.17.mlp.fc1.weight": "pytorch_model-00002-of-00002.bin",
486
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.17.mlp.fc2.bias": "pytorch_model-00002-of-00002.bin",
487
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.17.mlp.fc2.weight": "pytorch_model-00002-of-00002.bin",
488
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.17.self_attn.k_proj.bias": "pytorch_model-00002-of-00002.bin",
489
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.17.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
490
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.17.self_attn.out_proj.bias": "pytorch_model-00002-of-00002.bin",
491
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.17.self_attn.out_proj.weight": "pytorch_model-00002-of-00002.bin",
492
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.17.self_attn.q_proj.bias": "pytorch_model-00002-of-00002.bin",
493
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.17.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
494
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.17.self_attn.v_proj.bias": "pytorch_model-00002-of-00002.bin",
495
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.17.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
496
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.18.layer_norm1.bias": "pytorch_model-00002-of-00002.bin",
497
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.18.layer_norm1.weight": "pytorch_model-00002-of-00002.bin",
498
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.18.layer_norm2.bias": "pytorch_model-00002-of-00002.bin",
499
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.18.layer_norm2.weight": "pytorch_model-00002-of-00002.bin",
500
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.18.mlp.fc1.bias": "pytorch_model-00002-of-00002.bin",
501
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.18.mlp.fc1.weight": "pytorch_model-00002-of-00002.bin",
502
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.18.mlp.fc2.bias": "pytorch_model-00002-of-00002.bin",
503
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.18.mlp.fc2.weight": "pytorch_model-00002-of-00002.bin",
504
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.18.self_attn.k_proj.bias": "pytorch_model-00002-of-00002.bin",
505
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.18.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
506
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.18.self_attn.out_proj.bias": "pytorch_model-00002-of-00002.bin",
507
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.18.self_attn.out_proj.weight": "pytorch_model-00002-of-00002.bin",
508
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.18.self_attn.q_proj.bias": "pytorch_model-00002-of-00002.bin",
509
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.18.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
510
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.18.self_attn.v_proj.bias": "pytorch_model-00002-of-00002.bin",
511
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.18.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
512
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.19.layer_norm1.bias": "pytorch_model-00002-of-00002.bin",
513
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.19.layer_norm1.weight": "pytorch_model-00002-of-00002.bin",
514
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.19.layer_norm2.bias": "pytorch_model-00002-of-00002.bin",
515
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.19.layer_norm2.weight": "pytorch_model-00002-of-00002.bin",
516
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.19.mlp.fc1.bias": "pytorch_model-00002-of-00002.bin",
517
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.19.mlp.fc1.weight": "pytorch_model-00002-of-00002.bin",
518
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.19.mlp.fc2.bias": "pytorch_model-00002-of-00002.bin",
519
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.19.mlp.fc2.weight": "pytorch_model-00002-of-00002.bin",
520
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.19.self_attn.k_proj.bias": "pytorch_model-00002-of-00002.bin",
521
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.19.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
522
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.19.self_attn.out_proj.bias": "pytorch_model-00002-of-00002.bin",
523
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.19.self_attn.out_proj.weight": "pytorch_model-00002-of-00002.bin",
524
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.19.self_attn.q_proj.bias": "pytorch_model-00002-of-00002.bin",
525
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.19.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
526
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.19.self_attn.v_proj.bias": "pytorch_model-00002-of-00002.bin",
527
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.19.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
528
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.2.layer_norm1.bias": "pytorch_model-00002-of-00002.bin",
529
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.2.layer_norm1.weight": "pytorch_model-00002-of-00002.bin",
530
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.2.layer_norm2.bias": "pytorch_model-00002-of-00002.bin",
531
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.2.layer_norm2.weight": "pytorch_model-00002-of-00002.bin",
532
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.2.mlp.fc1.bias": "pytorch_model-00002-of-00002.bin",
533
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.2.mlp.fc1.weight": "pytorch_model-00002-of-00002.bin",
534
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.2.mlp.fc2.bias": "pytorch_model-00002-of-00002.bin",
535
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.2.mlp.fc2.weight": "pytorch_model-00002-of-00002.bin",
536
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.2.self_attn.k_proj.bias": "pytorch_model-00002-of-00002.bin",
537
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.2.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
538
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.2.self_attn.out_proj.bias": "pytorch_model-00002-of-00002.bin",
539
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.2.self_attn.out_proj.weight": "pytorch_model-00002-of-00002.bin",
540
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.2.self_attn.q_proj.bias": "pytorch_model-00002-of-00002.bin",
541
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.2.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
542
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.2.self_attn.v_proj.bias": "pytorch_model-00002-of-00002.bin",
543
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.2.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
544
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.20.layer_norm1.bias": "pytorch_model-00002-of-00002.bin",
545
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.20.layer_norm1.weight": "pytorch_model-00002-of-00002.bin",
546
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.20.layer_norm2.bias": "pytorch_model-00002-of-00002.bin",
547
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.20.layer_norm2.weight": "pytorch_model-00002-of-00002.bin",
548
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.20.mlp.fc1.bias": "pytorch_model-00002-of-00002.bin",
549
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.20.mlp.fc1.weight": "pytorch_model-00002-of-00002.bin",
550
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.20.mlp.fc2.bias": "pytorch_model-00002-of-00002.bin",
551
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.20.mlp.fc2.weight": "pytorch_model-00002-of-00002.bin",
552
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.20.self_attn.k_proj.bias": "pytorch_model-00002-of-00002.bin",
553
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.20.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
554
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.20.self_attn.out_proj.bias": "pytorch_model-00002-of-00002.bin",
555
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.20.self_attn.out_proj.weight": "pytorch_model-00002-of-00002.bin",
556
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.20.self_attn.q_proj.bias": "pytorch_model-00002-of-00002.bin",
557
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.20.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
558
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.20.self_attn.v_proj.bias": "pytorch_model-00002-of-00002.bin",
559
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.20.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
560
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.21.layer_norm1.bias": "pytorch_model-00002-of-00002.bin",
561
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.21.layer_norm1.weight": "pytorch_model-00002-of-00002.bin",
562
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.21.layer_norm2.bias": "pytorch_model-00002-of-00002.bin",
563
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.21.layer_norm2.weight": "pytorch_model-00002-of-00002.bin",
564
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.21.mlp.fc1.bias": "pytorch_model-00002-of-00002.bin",
565
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.21.mlp.fc1.weight": "pytorch_model-00002-of-00002.bin",
566
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.21.mlp.fc2.bias": "pytorch_model-00002-of-00002.bin",
567
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.21.mlp.fc2.weight": "pytorch_model-00002-of-00002.bin",
568
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.21.self_attn.k_proj.bias": "pytorch_model-00002-of-00002.bin",
569
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.21.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
570
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.21.self_attn.out_proj.bias": "pytorch_model-00002-of-00002.bin",
571
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.21.self_attn.out_proj.weight": "pytorch_model-00002-of-00002.bin",
572
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.21.self_attn.q_proj.bias": "pytorch_model-00002-of-00002.bin",
573
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.21.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
574
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.21.self_attn.v_proj.bias": "pytorch_model-00002-of-00002.bin",
575
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.21.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
576
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.22.layer_norm1.bias": "pytorch_model-00002-of-00002.bin",
577
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.22.layer_norm1.weight": "pytorch_model-00002-of-00002.bin",
578
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.22.layer_norm2.bias": "pytorch_model-00002-of-00002.bin",
579
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.22.layer_norm2.weight": "pytorch_model-00002-of-00002.bin",
580
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.22.mlp.fc1.bias": "pytorch_model-00002-of-00002.bin",
581
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.22.mlp.fc1.weight": "pytorch_model-00002-of-00002.bin",
582
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.22.mlp.fc2.bias": "pytorch_model-00002-of-00002.bin",
583
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.22.mlp.fc2.weight": "pytorch_model-00002-of-00002.bin",
584
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.22.self_attn.k_proj.bias": "pytorch_model-00002-of-00002.bin",
585
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.22.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
586
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.22.self_attn.out_proj.bias": "pytorch_model-00002-of-00002.bin",
587
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.22.self_attn.out_proj.weight": "pytorch_model-00002-of-00002.bin",
588
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.22.self_attn.q_proj.bias": "pytorch_model-00002-of-00002.bin",
589
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.22.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
590
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.22.self_attn.v_proj.bias": "pytorch_model-00002-of-00002.bin",
591
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.22.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
592
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.23.layer_norm1.bias": "pytorch_model-00002-of-00002.bin",
593
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.23.layer_norm1.weight": "pytorch_model-00002-of-00002.bin",
594
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.23.layer_norm2.bias": "pytorch_model-00002-of-00002.bin",
595
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.23.layer_norm2.weight": "pytorch_model-00002-of-00002.bin",
596
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.23.mlp.fc1.bias": "pytorch_model-00002-of-00002.bin",
597
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.23.mlp.fc1.weight": "pytorch_model-00002-of-00002.bin",
598
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.23.mlp.fc2.bias": "pytorch_model-00002-of-00002.bin",
599
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.23.mlp.fc2.weight": "pytorch_model-00002-of-00002.bin",
600
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.23.self_attn.k_proj.bias": "pytorch_model-00002-of-00002.bin",
601
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.23.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
602
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.23.self_attn.out_proj.bias": "pytorch_model-00002-of-00002.bin",
603
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.23.self_attn.out_proj.weight": "pytorch_model-00002-of-00002.bin",
604
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.23.self_attn.q_proj.bias": "pytorch_model-00002-of-00002.bin",
605
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.23.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
606
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.23.self_attn.v_proj.bias": "pytorch_model-00002-of-00002.bin",
607
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.23.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
608
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.3.layer_norm1.bias": "pytorch_model-00002-of-00002.bin",
609
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.3.layer_norm1.weight": "pytorch_model-00002-of-00002.bin",
610
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.3.layer_norm2.bias": "pytorch_model-00002-of-00002.bin",
611
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.3.layer_norm2.weight": "pytorch_model-00002-of-00002.bin",
612
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.3.mlp.fc1.bias": "pytorch_model-00002-of-00002.bin",
613
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.3.mlp.fc1.weight": "pytorch_model-00002-of-00002.bin",
614
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.3.mlp.fc2.bias": "pytorch_model-00002-of-00002.bin",
615
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.3.mlp.fc2.weight": "pytorch_model-00002-of-00002.bin",
616
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.3.self_attn.k_proj.bias": "pytorch_model-00002-of-00002.bin",
617
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.3.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
618
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.3.self_attn.out_proj.bias": "pytorch_model-00002-of-00002.bin",
619
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.3.self_attn.out_proj.weight": "pytorch_model-00002-of-00002.bin",
620
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.3.self_attn.q_proj.bias": "pytorch_model-00002-of-00002.bin",
621
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.3.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
622
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.3.self_attn.v_proj.bias": "pytorch_model-00002-of-00002.bin",
623
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.3.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
624
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.4.layer_norm1.bias": "pytorch_model-00002-of-00002.bin",
625
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.4.layer_norm1.weight": "pytorch_model-00002-of-00002.bin",
626
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.4.layer_norm2.bias": "pytorch_model-00002-of-00002.bin",
627
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.4.layer_norm2.weight": "pytorch_model-00002-of-00002.bin",
628
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.4.mlp.fc1.bias": "pytorch_model-00002-of-00002.bin",
629
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.4.mlp.fc1.weight": "pytorch_model-00002-of-00002.bin",
630
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.4.mlp.fc2.bias": "pytorch_model-00002-of-00002.bin",
631
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.4.mlp.fc2.weight": "pytorch_model-00002-of-00002.bin",
632
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.4.self_attn.k_proj.bias": "pytorch_model-00002-of-00002.bin",
633
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.4.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
634
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.4.self_attn.out_proj.bias": "pytorch_model-00002-of-00002.bin",
635
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.4.self_attn.out_proj.weight": "pytorch_model-00002-of-00002.bin",
636
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.4.self_attn.q_proj.bias": "pytorch_model-00002-of-00002.bin",
637
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.4.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
638
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.4.self_attn.v_proj.bias": "pytorch_model-00002-of-00002.bin",
639
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.4.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
640
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.5.layer_norm1.bias": "pytorch_model-00002-of-00002.bin",
641
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.5.layer_norm1.weight": "pytorch_model-00002-of-00002.bin",
642
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.5.layer_norm2.bias": "pytorch_model-00002-of-00002.bin",
643
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.5.layer_norm2.weight": "pytorch_model-00002-of-00002.bin",
644
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.5.mlp.fc1.bias": "pytorch_model-00002-of-00002.bin",
645
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.5.mlp.fc1.weight": "pytorch_model-00002-of-00002.bin",
646
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.5.mlp.fc2.bias": "pytorch_model-00002-of-00002.bin",
647
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.5.mlp.fc2.weight": "pytorch_model-00002-of-00002.bin",
648
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.5.self_attn.k_proj.bias": "pytorch_model-00002-of-00002.bin",
649
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.5.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
650
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.5.self_attn.out_proj.bias": "pytorch_model-00002-of-00002.bin",
651
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.5.self_attn.out_proj.weight": "pytorch_model-00002-of-00002.bin",
652
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.5.self_attn.q_proj.bias": "pytorch_model-00002-of-00002.bin",
653
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.5.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
654
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.5.self_attn.v_proj.bias": "pytorch_model-00002-of-00002.bin",
655
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.5.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
656
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.6.layer_norm1.bias": "pytorch_model-00002-of-00002.bin",
657
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.6.layer_norm1.weight": "pytorch_model-00002-of-00002.bin",
658
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.6.layer_norm2.bias": "pytorch_model-00002-of-00002.bin",
659
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.6.layer_norm2.weight": "pytorch_model-00002-of-00002.bin",
660
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.6.mlp.fc1.bias": "pytorch_model-00002-of-00002.bin",
661
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.6.mlp.fc1.weight": "pytorch_model-00002-of-00002.bin",
662
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.6.mlp.fc2.bias": "pytorch_model-00002-of-00002.bin",
663
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.6.mlp.fc2.weight": "pytorch_model-00002-of-00002.bin",
664
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.6.self_attn.k_proj.bias": "pytorch_model-00002-of-00002.bin",
665
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.6.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
666
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.6.self_attn.out_proj.bias": "pytorch_model-00002-of-00002.bin",
667
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.6.self_attn.out_proj.weight": "pytorch_model-00002-of-00002.bin",
668
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.6.self_attn.q_proj.bias": "pytorch_model-00002-of-00002.bin",
669
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.6.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
670
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.6.self_attn.v_proj.bias": "pytorch_model-00002-of-00002.bin",
671
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.6.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
672
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.7.layer_norm1.bias": "pytorch_model-00002-of-00002.bin",
673
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.7.layer_norm1.weight": "pytorch_model-00002-of-00002.bin",
674
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.7.layer_norm2.bias": "pytorch_model-00002-of-00002.bin",
675
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.7.layer_norm2.weight": "pytorch_model-00002-of-00002.bin",
676
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.7.mlp.fc1.bias": "pytorch_model-00002-of-00002.bin",
677
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.7.mlp.fc1.weight": "pytorch_model-00002-of-00002.bin",
678
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.7.mlp.fc2.bias": "pytorch_model-00002-of-00002.bin",
679
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.7.mlp.fc2.weight": "pytorch_model-00002-of-00002.bin",
680
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.7.self_attn.k_proj.bias": "pytorch_model-00002-of-00002.bin",
681
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.7.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
682
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.7.self_attn.out_proj.bias": "pytorch_model-00002-of-00002.bin",
683
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.7.self_attn.out_proj.weight": "pytorch_model-00002-of-00002.bin",
684
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.7.self_attn.q_proj.bias": "pytorch_model-00002-of-00002.bin",
685
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.7.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
686
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.7.self_attn.v_proj.bias": "pytorch_model-00002-of-00002.bin",
687
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.7.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
688
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.8.layer_norm1.bias": "pytorch_model-00002-of-00002.bin",
689
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.8.layer_norm1.weight": "pytorch_model-00002-of-00002.bin",
690
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.8.layer_norm2.bias": "pytorch_model-00002-of-00002.bin",
691
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.8.layer_norm2.weight": "pytorch_model-00002-of-00002.bin",
692
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.8.mlp.fc1.bias": "pytorch_model-00002-of-00002.bin",
693
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.8.mlp.fc1.weight": "pytorch_model-00002-of-00002.bin",
694
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.8.mlp.fc2.bias": "pytorch_model-00002-of-00002.bin",
695
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.8.mlp.fc2.weight": "pytorch_model-00002-of-00002.bin",
696
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.8.self_attn.k_proj.bias": "pytorch_model-00002-of-00002.bin",
697
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.8.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
698
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.8.self_attn.out_proj.bias": "pytorch_model-00002-of-00002.bin",
699
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.8.self_attn.out_proj.weight": "pytorch_model-00002-of-00002.bin",
700
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.8.self_attn.q_proj.bias": "pytorch_model-00002-of-00002.bin",
701
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.8.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
702
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.8.self_attn.v_proj.bias": "pytorch_model-00002-of-00002.bin",
703
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.8.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
704
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.9.layer_norm1.bias": "pytorch_model-00002-of-00002.bin",
705
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.9.layer_norm1.weight": "pytorch_model-00002-of-00002.bin",
706
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.9.layer_norm2.bias": "pytorch_model-00002-of-00002.bin",
707
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.9.layer_norm2.weight": "pytorch_model-00002-of-00002.bin",
708
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.9.mlp.fc1.bias": "pytorch_model-00002-of-00002.bin",
709
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.9.mlp.fc1.weight": "pytorch_model-00002-of-00002.bin",
710
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.9.mlp.fc2.bias": "pytorch_model-00002-of-00002.bin",
711
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.9.mlp.fc2.weight": "pytorch_model-00002-of-00002.bin",
712
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.9.self_attn.k_proj.bias": "pytorch_model-00002-of-00002.bin",
713
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.9.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
714
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.9.self_attn.out_proj.bias": "pytorch_model-00002-of-00002.bin",
715
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.9.self_attn.out_proj.weight": "pytorch_model-00002-of-00002.bin",
716
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.9.self_attn.q_proj.bias": "pytorch_model-00002-of-00002.bin",
717
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.9.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
718
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.9.self_attn.v_proj.bias": "pytorch_model-00002-of-00002.bin",
719
+ "model.vision_tower.vision_tower.vision_model.encoder.layers.9.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
720
+ "model.vision_tower.vision_tower.vision_model.post_layernorm.bias": "pytorch_model-00002-of-00002.bin",
721
+ "model.vision_tower.vision_tower.vision_model.post_layernorm.weight": "pytorch_model-00002-of-00002.bin",
722
+ "model.vision_tower.vision_tower.vision_model.pre_layrnorm.bias": "pytorch_model-00002-of-00002.bin",
723
+ "model.vision_tower.vision_tower.vision_model.pre_layrnorm.weight": "pytorch_model-00002-of-00002.bin"
724
+ }
725
+ }
checkpoint-100-llava/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:70c0b62bd953493d0675c34feecb6cb4564541e1efc60bb1704ce38255044908
3
+ size 15607
checkpoint-100-llava/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:241736953f399f6929c6ad898eaf28227b020331cba8da2c020aaaff41594b10
3
+ size 15607
checkpoint-100-llava/special_tokens_map.json ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token": {
3
+ "content": "<s>",
4
+ "lstrip": false,
5
+ "normalized": false,
6
+ "rstrip": false,
7
+ "single_word": false
8
+ },
9
+ "eos_token": {
10
+ "content": "</s>",
11
+ "lstrip": false,
12
+ "normalized": false,
13
+ "rstrip": false,
14
+ "single_word": false
15
+ },
16
+ "pad_token": "<unk>",
17
+ "unk_token": {
18
+ "content": "<unk>",
19
+ "lstrip": false,
20
+ "normalized": false,
21
+ "rstrip": false,
22
+ "single_word": false
23
+ }
24
+ }
checkpoint-100-llava/tokenizer.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dadfd56d766715c61d2ef780a525ab43b8e6da4de6865bda3d95fdef5e134055
3
+ size 493443
checkpoint-100-llava/tokenizer_config.json ADDED
@@ -0,0 +1,72 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_bos_token": true,
3
+ "add_eos_token": false,
4
+ "added_tokens_decoder": {
5
+ "0": {
6
+ "content": "<unk>",
7
+ "lstrip": false,
8
+ "normalized": false,
9
+ "rstrip": false,
10
+ "single_word": false,
11
+ "special": true
12
+ },
13
+ "1": {
14
+ "content": "<s>",
15
+ "lstrip": false,
16
+ "normalized": false,
17
+ "rstrip": false,
18
+ "single_word": false,
19
+ "special": true
20
+ },
21
+ "2": {
22
+ "content": "</s>",
23
+ "lstrip": false,
24
+ "normalized": false,
25
+ "rstrip": false,
26
+ "single_word": false,
27
+ "special": true
28
+ }
29
+ },
30
+ "additional_special_tokens": [],
31
+ "bos_token": {
32
+ "__type": "AddedToken",
33
+ "content": "<s>",
34
+ "lstrip": false,
35
+ "normalized": true,
36
+ "rstrip": false,
37
+ "single_word": false
38
+ },
39
+ "chat_template": "{{ bos_token }}{% for message in messages %}{% if (message['role'] == 'user') != (loop.index0 % 2 == 0) %}{{ raise_exception('Conversation roles must alternate user/assistant/user/assistant/...') }}{% endif %}{% if message['role'] == 'user' %}{{ '[INST] ' + message['content'] + ' [/INST]' }}{% elif message['role'] == 'assistant' %}{{ message['content'] + eos_token}}{% else %}{{ raise_exception('Only user and assistant roles are supported!') }}{% endif %}{% endfor %}",
40
+ "clean_up_tokenization_spaces": false,
41
+ "eos_token": {
42
+ "__type": "AddedToken",
43
+ "content": "</s>",
44
+ "lstrip": false,
45
+ "normalized": true,
46
+ "rstrip": false,
47
+ "single_word": false
48
+ },
49
+ "legacy": true,
50
+ "model_max_length": 2048,
51
+ "pad_token": {
52
+ "__type": "AddedToken",
53
+ "content": "<unk>",
54
+ "lstrip": false,
55
+ "normalized": true,
56
+ "rstrip": false,
57
+ "single_word": false
58
+ },
59
+ "padding_side": "right",
60
+ "sp_model_kwargs": {},
61
+ "spaces_between_special_tokens": false,
62
+ "tokenizer_class": "LlamaTokenizer",
63
+ "unk_token": {
64
+ "__type": "AddedToken",
65
+ "content": "<unk>",
66
+ "lstrip": false,
67
+ "normalized": true,
68
+ "rstrip": false,
69
+ "single_word": false
70
+ },
71
+ "use_default_system_prompt": false
72
+ }
checkpoint-100-llava/trainer_state.json ADDED
@@ -0,0 +1,1616 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 0.6379585326953748,
5
+ "global_step": 100,
6
+ "is_hyper_param_search": false,
7
+ "is_local_process_zero": true,
8
+ "is_world_process_zero": true,
9
+ "log_history": [
10
+ {
11
+ "epoch": 0.01,
12
+ "learning_rate": 4e-07,
13
+ "logits/chosen": -2.9071130752563477,
14
+ "logits/rejected": -2.8750061988830566,
15
+ "loss": 0.6931,
16
+ "policy_logps/chosen": -127.82667541503906,
17
+ "policy_logps/rejected": -130.1011505126953,
18
+ "referece_logps/chosen": -127.82667541503906,
19
+ "referece_logps/rejected": -130.1011505126953,
20
+ "rewards/accuracies": 0.0,
21
+ "rewards/chosen": 0.0,
22
+ "rewards/margins": 0.0,
23
+ "rewards/rejected": 0.0,
24
+ "step": 1
25
+ },
26
+ {
27
+ "epoch": 0.01,
28
+ "learning_rate": 8e-07,
29
+ "logits/chosen": -2.8690595626831055,
30
+ "logits/rejected": -2.921374797821045,
31
+ "loss": 0.6931,
32
+ "policy_logps/chosen": -127.44483947753906,
33
+ "policy_logps/rejected": -118.97954559326172,
34
+ "referece_logps/chosen": -127.44483947753906,
35
+ "referece_logps/rejected": -118.97954559326172,
36
+ "rewards/accuracies": 0.0,
37
+ "rewards/chosen": 0.0,
38
+ "rewards/margins": 0.0,
39
+ "rewards/rejected": 0.0,
40
+ "step": 2
41
+ },
42
+ {
43
+ "epoch": 0.02,
44
+ "learning_rate": 1.2e-06,
45
+ "logits/chosen": -2.876248359680176,
46
+ "logits/rejected": -2.9442975521087646,
47
+ "loss": 0.6931,
48
+ "policy_logps/chosen": -151.38050842285156,
49
+ "policy_logps/rejected": -104.58517456054688,
50
+ "referece_logps/chosen": -151.3734893798828,
51
+ "referece_logps/rejected": -104.58135986328125,
52
+ "rewards/accuracies": 0.3125,
53
+ "rewards/chosen": -0.0007032513385638595,
54
+ "rewards/margins": -0.00032021405058912933,
55
+ "rewards/rejected": -0.0003830373170785606,
56
+ "step": 3
57
+ },
58
+ {
59
+ "epoch": 0.03,
60
+ "learning_rate": 1.6e-06,
61
+ "logits/chosen": -2.923973560333252,
62
+ "logits/rejected": -2.9425337314605713,
63
+ "loss": 0.6924,
64
+ "policy_logps/chosen": -121.4244613647461,
65
+ "policy_logps/rejected": -117.06787109375,
66
+ "referece_logps/chosen": -121.42953491210938,
67
+ "referece_logps/rejected": -117.06060028076172,
68
+ "rewards/accuracies": 0.5625,
69
+ "rewards/chosen": 0.0005075454828329384,
70
+ "rewards/margins": 0.0012342334957793355,
71
+ "rewards/rejected": -0.0007266878965310752,
72
+ "step": 4
73
+ },
74
+ {
75
+ "epoch": 0.03,
76
+ "learning_rate": 2e-06,
77
+ "logits/chosen": -2.896630048751831,
78
+ "logits/rejected": -2.874107837677002,
79
+ "loss": 0.6925,
80
+ "policy_logps/chosen": -112.3115234375,
81
+ "policy_logps/rejected": -117.28302764892578,
82
+ "referece_logps/chosen": -112.33646392822266,
83
+ "referece_logps/rejected": -117.27299499511719,
84
+ "rewards/accuracies": 0.625,
85
+ "rewards/chosen": 0.0024944781325757504,
86
+ "rewards/margins": 0.003496956778690219,
87
+ "rewards/rejected": -0.0010024786461144686,
88
+ "step": 5
89
+ },
90
+ {
91
+ "epoch": 0.04,
92
+ "learning_rate": 1.999783578606323e-06,
93
+ "logits/chosen": -2.8770463466644287,
94
+ "logits/rejected": -2.893326759338379,
95
+ "loss": 0.6908,
96
+ "policy_logps/chosen": -111.89071655273438,
97
+ "policy_logps/rejected": -110.08009338378906,
98
+ "referece_logps/chosen": -111.92901611328125,
99
+ "referece_logps/rejected": -110.06503295898438,
100
+ "rewards/accuracies": 0.75,
101
+ "rewards/chosen": 0.00383090996183455,
102
+ "rewards/margins": 0.005336493253707886,
103
+ "rewards/rejected": -0.0015055835247039795,
104
+ "step": 6
105
+ },
106
+ {
107
+ "epoch": 0.04,
108
+ "learning_rate": 1.999134408101731e-06,
109
+ "logits/chosen": -2.8742122650146484,
110
+ "logits/rejected": -2.9248428344726562,
111
+ "loss": 0.6872,
112
+ "policy_logps/chosen": -160.53836059570312,
113
+ "policy_logps/rejected": -137.42295837402344,
114
+ "referece_logps/chosen": -160.57460021972656,
115
+ "referece_logps/rejected": -137.33425903320312,
116
+ "rewards/accuracies": 0.75,
117
+ "rewards/chosen": 0.003623390104621649,
118
+ "rewards/margins": 0.01249313447624445,
119
+ "rewards/rejected": -0.008869742974638939,
120
+ "step": 7
121
+ },
122
+ {
123
+ "epoch": 0.05,
124
+ "learning_rate": 1.998052769474995e-06,
125
+ "logits/chosen": -2.9328360557556152,
126
+ "logits/rejected": -2.975874185562134,
127
+ "loss": 0.6898,
128
+ "policy_logps/chosen": -75.984619140625,
129
+ "policy_logps/rejected": -74.07447814941406,
130
+ "referece_logps/chosen": -76.07889556884766,
131
+ "referece_logps/rejected": -74.01802825927734,
132
+ "rewards/accuracies": 0.75,
133
+ "rewards/chosen": 0.009427506476640701,
134
+ "rewards/margins": 0.015072083100676537,
135
+ "rewards/rejected": -0.0056445784866809845,
136
+ "step": 8
137
+ },
138
+ {
139
+ "epoch": 0.06,
140
+ "learning_rate": 1.9965391309055927e-06,
141
+ "logits/chosen": -2.882821798324585,
142
+ "logits/rejected": -2.901142120361328,
143
+ "loss": 0.6773,
144
+ "policy_logps/chosen": -107.12052917480469,
145
+ "policy_logps/rejected": -93.86226654052734,
146
+ "referece_logps/chosen": -107.38616943359375,
147
+ "referece_logps/rejected": -93.67011260986328,
148
+ "rewards/accuracies": 0.9375,
149
+ "rewards/chosen": 0.026563584804534912,
150
+ "rewards/margins": 0.04577912390232086,
151
+ "rewards/rejected": -0.019215542823076248,
152
+ "step": 9
153
+ },
154
+ {
155
+ "epoch": 0.06,
156
+ "learning_rate": 1.9945941475610623e-06,
157
+ "logits/chosen": -2.8708529472351074,
158
+ "logits/rejected": -2.9397594928741455,
159
+ "loss": 0.6735,
160
+ "policy_logps/chosen": -123.71336364746094,
161
+ "policy_logps/rejected": -100.81864166259766,
162
+ "referece_logps/chosen": -124.01136779785156,
163
+ "referece_logps/rejected": -100.63360595703125,
164
+ "rewards/accuracies": 1.0,
165
+ "rewards/chosen": 0.029799818992614746,
166
+ "rewards/margins": 0.048302434384822845,
167
+ "rewards/rejected": -0.0185026116669178,
168
+ "step": 10
169
+ },
170
+ {
171
+ "epoch": 0.07,
172
+ "learning_rate": 1.992218661313415e-06,
173
+ "logits/chosen": -2.8874049186706543,
174
+ "logits/rejected": -2.9123711585998535,
175
+ "loss": 0.6729,
176
+ "policy_logps/chosen": -104.46468353271484,
177
+ "policy_logps/rejected": -98.05982971191406,
178
+ "referece_logps/chosen": -104.58216857910156,
179
+ "referece_logps/rejected": -97.82801055908203,
180
+ "rewards/accuracies": 0.8125,
181
+ "rewards/chosen": 0.011748719029128551,
182
+ "rewards/margins": 0.03493000194430351,
183
+ "rewards/rejected": -0.023181283846497536,
184
+ "step": 11
185
+ },
186
+ {
187
+ "epoch": 0.08,
188
+ "learning_rate": 1.98941370037474e-06,
189
+ "logits/chosen": -2.9089269638061523,
190
+ "logits/rejected": -2.887115478515625,
191
+ "loss": 0.67,
192
+ "policy_logps/chosen": -125.26738739013672,
193
+ "policy_logps/rejected": -118.79678344726562,
194
+ "referece_logps/chosen": -125.5167007446289,
195
+ "referece_logps/rejected": -118.42857360839844,
196
+ "rewards/accuracies": 0.875,
197
+ "rewards/chosen": 0.024931641295552254,
198
+ "rewards/margins": 0.06175263226032257,
199
+ "rewards/rejected": -0.036820992827415466,
200
+ "step": 12
201
+ },
202
+ {
203
+ "epoch": 0.08,
204
+ "learning_rate": 1.986180478852149e-06,
205
+ "logits/chosen": -2.8961265087127686,
206
+ "logits/rejected": -2.85491943359375,
207
+ "loss": 0.6635,
208
+ "policy_logps/chosen": -152.51437377929688,
209
+ "policy_logps/rejected": -152.42971801757812,
210
+ "referece_logps/chosen": -152.56588745117188,
211
+ "referece_logps/rejected": -152.2041015625,
212
+ "rewards/accuracies": 0.5625,
213
+ "rewards/chosen": 0.005150413140654564,
214
+ "rewards/margins": 0.027712417766451836,
215
+ "rewards/rejected": -0.02256200462579727,
216
+ "step": 13
217
+ },
218
+ {
219
+ "epoch": 0.09,
220
+ "learning_rate": 1.982520396222257e-06,
221
+ "logits/chosen": -2.8753738403320312,
222
+ "logits/rejected": -2.912999153137207,
223
+ "loss": 0.6472,
224
+ "policy_logps/chosen": -101.14571380615234,
225
+ "policy_logps/rejected": -99.60147094726562,
226
+ "referece_logps/chosen": -101.51478576660156,
227
+ "referece_logps/rejected": -98.9403076171875,
228
+ "rewards/accuracies": 0.875,
229
+ "rewards/chosen": 0.03690744936466217,
230
+ "rewards/margins": 0.10302485525608063,
231
+ "rewards/rejected": -0.06611741334199905,
232
+ "step": 14
233
+ },
234
+ {
235
+ "epoch": 0.1,
236
+ "learning_rate": 1.978435036725432e-06,
237
+ "logits/chosen": -2.9012084007263184,
238
+ "logits/rejected": -2.8953423500061035,
239
+ "loss": 0.6372,
240
+ "policy_logps/chosen": -126.27330017089844,
241
+ "policy_logps/rejected": -128.28578186035156,
242
+ "referece_logps/chosen": -126.68486785888672,
243
+ "referece_logps/rejected": -127.51297760009766,
244
+ "rewards/accuracies": 1.0,
245
+ "rewards/chosen": 0.04115738719701767,
246
+ "rewards/margins": 0.11843809485435486,
247
+ "rewards/rejected": -0.07728070020675659,
248
+ "step": 15
249
+ },
250
+ {
251
+ "epoch": 0.1,
252
+ "learning_rate": 1.9739261686800657e-06,
253
+ "logits/chosen": -2.9024605751037598,
254
+ "logits/rejected": -2.8853814601898193,
255
+ "loss": 0.6414,
256
+ "policy_logps/chosen": -114.59141540527344,
257
+ "policy_logps/rejected": -128.96331787109375,
258
+ "referece_logps/chosen": -114.98219299316406,
259
+ "referece_logps/rejected": -128.28781127929688,
260
+ "rewards/accuracies": 0.875,
261
+ "rewards/chosen": 0.03907782956957817,
262
+ "rewards/margins": 0.1066286489367485,
263
+ "rewards/rejected": -0.06755081564188004,
264
+ "step": 16
265
+ },
266
+ {
267
+ "epoch": 0.11,
268
+ "learning_rate": 1.968995743717171e-06,
269
+ "logits/chosen": -2.912461757659912,
270
+ "logits/rejected": -2.9493064880371094,
271
+ "loss": 0.6429,
272
+ "policy_logps/chosen": -111.6058349609375,
273
+ "policy_logps/rejected": -109.18556213378906,
274
+ "referece_logps/chosen": -111.89552307128906,
275
+ "referece_logps/rejected": -108.55181884765625,
276
+ "rewards/accuracies": 0.75,
277
+ "rewards/chosen": 0.028969965875148773,
278
+ "rewards/margins": 0.09234414994716644,
279
+ "rewards/rejected": -0.06337418407201767,
280
+ "step": 17
281
+ },
282
+ {
283
+ "epoch": 0.11,
284
+ "learning_rate": 1.9636458959356316e-06,
285
+ "logits/chosen": -2.8974556922912598,
286
+ "logits/rejected": -2.89353084564209,
287
+ "loss": 0.6528,
288
+ "policy_logps/chosen": -138.28262329101562,
289
+ "policy_logps/rejected": -128.17782592773438,
290
+ "referece_logps/chosen": -138.3295440673828,
291
+ "referece_logps/rejected": -127.26476287841797,
292
+ "rewards/accuracies": 0.875,
293
+ "rewards/chosen": 0.004691671580076218,
294
+ "rewards/margins": 0.09599801898002625,
295
+ "rewards/rejected": -0.09130635112524033,
296
+ "step": 18
297
+ },
298
+ {
299
+ "epoch": 0.12,
300
+ "learning_rate": 1.9578789409784727e-06,
301
+ "logits/chosen": -2.8959522247314453,
302
+ "logits/rejected": -2.9369704723358154,
303
+ "loss": 0.6446,
304
+ "policy_logps/chosen": -111.41603088378906,
305
+ "policy_logps/rejected": -104.68195343017578,
306
+ "referece_logps/chosen": -111.60549926757812,
307
+ "referece_logps/rejected": -104.04905700683594,
308
+ "rewards/accuracies": 0.8125,
309
+ "rewards/chosen": 0.018947793170809746,
310
+ "rewards/margins": 0.08223824948072433,
311
+ "rewards/rejected": -0.06329045444726944,
312
+ "step": 19
313
+ },
314
+ {
315
+ "epoch": 0.13,
316
+ "learning_rate": 1.951697375030553e-06,
317
+ "logits/chosen": -2.853641986846924,
318
+ "logits/rejected": -2.8619699478149414,
319
+ "loss": 0.6466,
320
+ "policy_logps/chosen": -145.3450927734375,
321
+ "policy_logps/rejected": -134.5753631591797,
322
+ "referece_logps/chosen": -145.5601348876953,
323
+ "referece_logps/rejected": -133.66200256347656,
324
+ "rewards/accuracies": 0.8125,
325
+ "rewards/chosen": 0.021503955125808716,
326
+ "rewards/margins": 0.11283906549215317,
327
+ "rewards/rejected": -0.09133510291576385,
328
+ "step": 20
329
+ },
330
+ {
331
+ "epoch": 0.13,
332
+ "learning_rate": 1.9451038737381077e-06,
333
+ "logits/chosen": -2.93542218208313,
334
+ "logits/rejected": -2.9347047805786133,
335
+ "loss": 0.6346,
336
+ "policy_logps/chosen": -97.80740356445312,
337
+ "policy_logps/rejected": -92.36180877685547,
338
+ "referece_logps/chosen": -97.90306091308594,
339
+ "referece_logps/rejected": -91.67369079589844,
340
+ "rewards/accuracies": 0.8125,
341
+ "rewards/chosen": 0.009566396474838257,
342
+ "rewards/margins": 0.07837802916765213,
343
+ "rewards/rejected": -0.06881163269281387,
344
+ "step": 21
345
+ },
346
+ {
347
+ "epoch": 0.14,
348
+ "learning_rate": 1.9381012910506143e-06,
349
+ "logits/chosen": -2.8574209213256836,
350
+ "logits/rejected": -2.860379219055176,
351
+ "loss": 0.6102,
352
+ "policy_logps/chosen": -152.8387908935547,
353
+ "policy_logps/rejected": -156.60391235351562,
354
+ "referece_logps/chosen": -152.914306640625,
355
+ "referece_logps/rejected": -155.2355499267578,
356
+ "rewards/accuracies": 0.9375,
357
+ "rewards/chosen": 0.007550956681370735,
358
+ "rewards/margins": 0.14438626170158386,
359
+ "rewards/rejected": -0.13683530688285828,
360
+ "step": 22
361
+ },
362
+ {
363
+ "epoch": 0.15,
364
+ "learning_rate": 1.9306926579854817e-06,
365
+ "logits/chosen": -2.8623623847961426,
366
+ "logits/rejected": -2.8780465126037598,
367
+ "loss": 0.6306,
368
+ "policy_logps/chosen": -143.302490234375,
369
+ "policy_logps/rejected": -127.38278198242188,
370
+ "referece_logps/chosen": -143.29205322265625,
371
+ "referece_logps/rejected": -126.24747467041016,
372
+ "rewards/accuracies": 0.875,
373
+ "rewards/chosen": -0.0010449867695569992,
374
+ "rewards/margins": 0.11248550564050674,
375
+ "rewards/rejected": -0.11353050917387009,
376
+ "step": 23
377
+ },
378
+ {
379
+ "epoch": 0.15,
380
+ "learning_rate": 1.922881181316097e-06,
381
+ "logits/chosen": -2.949070930480957,
382
+ "logits/rejected": -2.954594135284424,
383
+ "loss": 0.6002,
384
+ "policy_logps/chosen": -74.232177734375,
385
+ "policy_logps/rejected": -73.76258850097656,
386
+ "referece_logps/chosen": -75.20740509033203,
387
+ "referece_logps/rejected": -72.94302368164062,
388
+ "rewards/accuracies": 0.875,
389
+ "rewards/chosen": 0.09752248972654343,
390
+ "rewards/margins": 0.17947959899902344,
391
+ "rewards/rejected": -0.08195710927248001,
392
+ "step": 24
393
+ },
394
+ {
395
+ "epoch": 0.16,
396
+ "learning_rate": 1.9146702421837946e-06,
397
+ "logits/chosen": -2.8681116104125977,
398
+ "logits/rejected": -2.8957810401916504,
399
+ "loss": 0.612,
400
+ "policy_logps/chosen": -127.76103210449219,
401
+ "policy_logps/rejected": -125.47801208496094,
402
+ "referece_logps/chosen": -128.02197265625,
403
+ "referece_logps/rejected": -124.08172607421875,
404
+ "rewards/accuracies": 0.8125,
405
+ "rewards/chosen": 0.026093529537320137,
406
+ "rewards/margins": 0.1657221019268036,
407
+ "rewards/rejected": -0.1396285593509674,
408
+ "step": 25
409
+ },
410
+ {
411
+ "epoch": 0.17,
412
+ "learning_rate": 1.906063394634356e-06,
413
+ "logits/chosen": -2.862689256668091,
414
+ "logits/rejected": -2.877995252609253,
415
+ "loss": 0.6398,
416
+ "policy_logps/chosen": -120.70744323730469,
417
+ "policy_logps/rejected": -113.76991271972656,
418
+ "referece_logps/chosen": -120.96971130371094,
419
+ "referece_logps/rejected": -112.71368408203125,
420
+ "rewards/accuracies": 0.875,
421
+ "rewards/chosen": 0.026226602494716644,
422
+ "rewards/margins": 0.13184988498687744,
423
+ "rewards/rejected": -0.1056232899427414,
424
+ "step": 26
425
+ },
426
+ {
427
+ "epoch": 0.17,
428
+ "learning_rate": 1.897064364079664e-06,
429
+ "logits/chosen": -2.9357306957244873,
430
+ "logits/rejected": -2.976917028427124,
431
+ "loss": 0.5904,
432
+ "policy_logps/chosen": -105.41529846191406,
433
+ "policy_logps/rejected": -96.86430358886719,
434
+ "referece_logps/chosen": -106.42874145507812,
435
+ "referece_logps/rejected": -95.49075317382812,
436
+ "rewards/accuracies": 0.875,
437
+ "rewards/chosen": 0.10134478658437729,
438
+ "rewards/margins": 0.23869961500167847,
439
+ "rewards/rejected": -0.13735483586788177,
440
+ "step": 27
441
+ },
442
+ {
443
+ "epoch": 0.18,
444
+ "learning_rate": 1.8876770456851876e-06,
445
+ "logits/chosen": -2.8511600494384766,
446
+ "logits/rejected": -2.888685941696167,
447
+ "loss": 0.5817,
448
+ "policy_logps/chosen": -135.38853454589844,
449
+ "policy_logps/rejected": -136.4590301513672,
450
+ "referece_logps/chosen": -135.95562744140625,
451
+ "referece_logps/rejected": -134.50482177734375,
452
+ "rewards/accuracies": 0.9375,
453
+ "rewards/chosen": 0.05670913681387901,
454
+ "rewards/margins": 0.2521297335624695,
455
+ "rewards/rejected": -0.19542059302330017,
456
+ "step": 28
457
+ },
458
+ {
459
+ "epoch": 0.19,
460
+ "learning_rate": 1.8779055026839868e-06,
461
+ "logits/chosen": -2.9084866046905518,
462
+ "logits/rejected": -2.9267258644104004,
463
+ "loss": 0.6042,
464
+ "policy_logps/chosen": -134.9838409423828,
465
+ "policy_logps/rejected": -114.98860931396484,
466
+ "referece_logps/chosen": -135.53085327148438,
467
+ "referece_logps/rejected": -113.39382934570312,
468
+ "rewards/accuracies": 0.9375,
469
+ "rewards/chosen": 0.05469997972249985,
470
+ "rewards/margins": 0.21417750418186188,
471
+ "rewards/rejected": -0.15947751700878143,
472
+ "step": 29
473
+ },
474
+ {
475
+ "epoch": 0.19,
476
+ "learning_rate": 1.8677539646179705e-06,
477
+ "logits/chosen": -2.88179349899292,
478
+ "logits/rejected": -2.929935932159424,
479
+ "loss": 0.5827,
480
+ "policy_logps/chosen": -163.58660888671875,
481
+ "policy_logps/rejected": -131.79258728027344,
482
+ "referece_logps/chosen": -163.90985107421875,
483
+ "referece_logps/rejected": -129.81314086914062,
484
+ "rewards/accuracies": 0.8125,
485
+ "rewards/chosen": 0.0323248989880085,
486
+ "rewards/margins": 0.23026807606220245,
487
+ "rewards/rejected": -0.19794318079948425,
488
+ "step": 30
489
+ },
490
+ {
491
+ "epoch": 0.2,
492
+ "learning_rate": 1.8572268255071718e-06,
493
+ "logits/chosen": -2.9416298866271973,
494
+ "logits/rejected": -2.961357831954956,
495
+ "loss": 0.5977,
496
+ "policy_logps/chosen": -98.88802337646484,
497
+ "policy_logps/rejected": -94.90333557128906,
498
+ "referece_logps/chosen": -99.61119842529297,
499
+ "referece_logps/rejected": -93.1702651977539,
500
+ "rewards/accuracies": 0.875,
501
+ "rewards/chosen": 0.07231828570365906,
502
+ "rewards/margins": 0.24562585353851318,
503
+ "rewards/rejected": -0.17330753803253174,
504
+ "step": 31
505
+ },
506
+ {
507
+ "epoch": 0.2,
508
+ "learning_rate": 1.8463286419478252e-06,
509
+ "logits/chosen": -2.9383907318115234,
510
+ "logits/rejected": -2.880384922027588,
511
+ "loss": 0.5772,
512
+ "policy_logps/chosen": -118.6290283203125,
513
+ "policy_logps/rejected": -126.94461822509766,
514
+ "referece_logps/chosen": -118.98780822753906,
515
+ "referece_logps/rejected": -125.21376037597656,
516
+ "rewards/accuracies": 0.9375,
517
+ "rewards/chosen": 0.03587843477725983,
518
+ "rewards/margins": 0.20896492898464203,
519
+ "rewards/rejected": -0.173086479306221,
520
+ "step": 32
521
+ },
522
+ {
523
+ "epoch": 0.21,
524
+ "learning_rate": 1.835064131140081e-06,
525
+ "logits/chosen": -2.908090829849243,
526
+ "logits/rejected": -2.9141459465026855,
527
+ "loss": 0.5723,
528
+ "policy_logps/chosen": -132.4810333251953,
529
+ "policy_logps/rejected": -130.16603088378906,
530
+ "referece_logps/chosen": -132.7151336669922,
531
+ "referece_logps/rejected": -127.14649963378906,
532
+ "rewards/accuracies": 1.0,
533
+ "rewards/chosen": 0.02341010421514511,
534
+ "rewards/margins": 0.32536280155181885,
535
+ "rewards/rejected": -0.3019527196884155,
536
+ "step": 33
537
+ },
538
+ {
539
+ "epoch": 0.22,
540
+ "learning_rate": 1.8234381688461941e-06,
541
+ "logits/chosen": -2.9611878395080566,
542
+ "logits/rejected": -2.966728448867798,
543
+ "loss": 0.5734,
544
+ "policy_logps/chosen": -119.45679473876953,
545
+ "policy_logps/rejected": -118.76419830322266,
546
+ "referece_logps/chosen": -118.86337280273438,
547
+ "referece_logps/rejected": -115.91516876220703,
548
+ "rewards/accuracies": 0.8125,
549
+ "rewards/chosen": -0.059341806918382645,
550
+ "rewards/margins": 0.22556202113628387,
551
+ "rewards/rejected": -0.28490379452705383,
552
+ "step": 34
553
+ },
554
+ {
555
+ "epoch": 0.22,
556
+ "learning_rate": 1.8114557872800905e-06,
557
+ "logits/chosen": -2.967761993408203,
558
+ "logits/rejected": -2.900844097137451,
559
+ "loss": 0.5633,
560
+ "policy_logps/chosen": -130.78848266601562,
561
+ "policy_logps/rejected": -142.6917724609375,
562
+ "referece_logps/chosen": -130.37750244140625,
563
+ "referece_logps/rejected": -139.14178466796875,
564
+ "rewards/accuracies": 0.9375,
565
+ "rewards/chosen": -0.041097551584243774,
566
+ "rewards/margins": 0.31390050053596497,
567
+ "rewards/rejected": -0.35499805212020874,
568
+ "step": 35
569
+ },
570
+ {
571
+ "epoch": 0.23,
572
+ "learning_rate": 1.7991221729292058e-06,
573
+ "logits/chosen": -2.8585638999938965,
574
+ "logits/rejected": -2.90903377532959,
575
+ "loss": 0.5649,
576
+ "policy_logps/chosen": -137.8734893798828,
577
+ "policy_logps/rejected": -127.91438293457031,
578
+ "referece_logps/chosen": -137.9367218017578,
579
+ "referece_logps/rejected": -125.2115478515625,
580
+ "rewards/accuracies": 0.9375,
581
+ "rewards/chosen": 0.0063229575753211975,
582
+ "rewards/margins": 0.27660617232322693,
583
+ "rewards/rejected": -0.27028322219848633,
584
+ "step": 36
585
+ },
586
+ {
587
+ "epoch": 0.24,
588
+ "learning_rate": 1.7864426643095536e-06,
589
+ "logits/chosen": -2.9357798099517822,
590
+ "logits/rejected": -2.900618553161621,
591
+ "loss": 0.5592,
592
+ "policy_logps/chosen": -151.2121124267578,
593
+ "policy_logps/rejected": -134.10804748535156,
594
+ "referece_logps/chosen": -150.83583068847656,
595
+ "referece_logps/rejected": -131.0611572265625,
596
+ "rewards/accuracies": 0.875,
597
+ "rewards/chosen": -0.03762848675251007,
598
+ "rewards/margins": 0.26705947518348694,
599
+ "rewards/rejected": -0.3046879470348358,
600
+ "step": 37
601
+ },
602
+ {
603
+ "epoch": 0.24,
604
+ "learning_rate": 1.7734227496549878e-06,
605
+ "logits/chosen": -2.9032468795776367,
606
+ "logits/rejected": -2.9027702808380127,
607
+ "loss": 0.5684,
608
+ "policy_logps/chosen": -104.47734069824219,
609
+ "policy_logps/rejected": -106.55696105957031,
610
+ "referece_logps/chosen": -105.3730697631836,
611
+ "referece_logps/rejected": -104.14241027832031,
612
+ "rewards/accuracies": 0.9375,
613
+ "rewards/chosen": 0.0895722359418869,
614
+ "rewards/margins": 0.3310272991657257,
615
+ "rewards/rejected": -0.241455078125,
616
+ "step": 38
617
+ },
618
+ {
619
+ "epoch": 0.25,
620
+ "learning_rate": 1.7600680645416582e-06,
621
+ "logits/chosen": -2.9712843894958496,
622
+ "logits/rejected": -2.894430637359619,
623
+ "loss": 0.5499,
624
+ "policy_logps/chosen": -126.83047485351562,
625
+ "policy_logps/rejected": -138.54771423339844,
626
+ "referece_logps/chosen": -126.56389617919922,
627
+ "referece_logps/rejected": -135.42103576660156,
628
+ "rewards/accuracies": 0.875,
629
+ "rewards/chosen": -0.026657823473215103,
630
+ "rewards/margins": 0.2860097885131836,
631
+ "rewards/rejected": -0.312667578458786,
632
+ "step": 39
633
+ },
634
+ {
635
+ "epoch": 0.26,
636
+ "learning_rate": 1.7463843894486936e-06,
637
+ "logits/chosen": -2.942542552947998,
638
+ "logits/rejected": -2.9863080978393555,
639
+ "loss": 0.5534,
640
+ "policy_logps/chosen": -93.51473999023438,
641
+ "policy_logps/rejected": -94.46737670898438,
642
+ "referece_logps/chosen": -93.94270324707031,
643
+ "referece_logps/rejected": -91.81558227539062,
644
+ "rewards/accuracies": 0.9375,
645
+ "rewards/chosen": 0.04279506206512451,
646
+ "rewards/margins": 0.3079749345779419,
647
+ "rewards/rejected": -0.2651798725128174,
648
+ "step": 40
649
+ },
650
+ {
651
+ "epoch": 0.26,
652
+ "learning_rate": 1.7323776472561625e-06,
653
+ "logits/chosen": -2.9191946983337402,
654
+ "logits/rejected": -2.935122013092041,
655
+ "loss": 0.5682,
656
+ "policy_logps/chosen": -128.17813110351562,
657
+ "policy_logps/rejected": -139.06613159179688,
658
+ "referece_logps/chosen": -127.54570770263672,
659
+ "referece_logps/rejected": -135.201904296875,
660
+ "rewards/accuracies": 0.875,
661
+ "rewards/chosen": -0.06324195861816406,
662
+ "rewards/margins": 0.32317861914634705,
663
+ "rewards/rejected": -0.3864205777645111,
664
+ "step": 41
665
+ },
666
+ {
667
+ "epoch": 0.27,
668
+ "learning_rate": 1.7180539006813969e-06,
669
+ "logits/chosen": -2.920085906982422,
670
+ "logits/rejected": -2.906773567199707,
671
+ "loss": 0.5454,
672
+ "policy_logps/chosen": -126.83687591552734,
673
+ "policy_logps/rejected": -114.17430877685547,
674
+ "referece_logps/chosen": -126.77012634277344,
675
+ "referece_logps/rejected": -110.9568862915039,
676
+ "rewards/accuracies": 0.875,
677
+ "rewards/chosen": -0.006674099713563919,
678
+ "rewards/margins": 0.31506818532943726,
679
+ "rewards/rejected": -0.3217422664165497,
680
+ "step": 42
681
+ },
682
+ {
683
+ "epoch": 0.27,
684
+ "learning_rate": 1.7034193496547902e-06,
685
+ "logits/chosen": -2.8620564937591553,
686
+ "logits/rejected": -2.9222469329833984,
687
+ "loss": 0.534,
688
+ "policy_logps/chosen": -125.01854705810547,
689
+ "policy_logps/rejected": -116.95589447021484,
690
+ "referece_logps/chosen": -125.00530242919922,
691
+ "referece_logps/rejected": -113.22078704833984,
692
+ "rewards/accuracies": 0.9375,
693
+ "rewards/chosen": -0.0013249870389699936,
694
+ "rewards/margins": 0.3721860647201538,
695
+ "rewards/rejected": -0.37351107597351074,
696
+ "step": 43
697
+ },
698
+ {
699
+ "epoch": 0.28,
700
+ "learning_rate": 1.6884803286362e-06,
701
+ "logits/chosen": -2.8868565559387207,
702
+ "logits/rejected": -2.91511869430542,
703
+ "loss": 0.5459,
704
+ "policy_logps/chosen": -150.9369354248047,
705
+ "policy_logps/rejected": -144.1173858642578,
706
+ "referece_logps/chosen": -150.93002319335938,
707
+ "referece_logps/rejected": -139.8741455078125,
708
+ "rewards/accuracies": 0.875,
709
+ "rewards/chosen": -0.0006910450756549835,
710
+ "rewards/margins": 0.42363405227661133,
711
+ "rewards/rejected": -0.4243250787258148,
712
+ "step": 44
713
+ },
714
+ {
715
+ "epoch": 0.29,
716
+ "learning_rate": 1.673243303873124e-06,
717
+ "logits/chosen": -2.9313180446624756,
718
+ "logits/rejected": -2.9371540546417236,
719
+ "loss": 0.5569,
720
+ "policy_logps/chosen": -122.50914764404297,
721
+ "policy_logps/rejected": -110.94712829589844,
722
+ "referece_logps/chosen": -122.10940551757812,
723
+ "referece_logps/rejected": -108.29950714111328,
724
+ "rewards/accuracies": 0.6875,
725
+ "rewards/chosen": -0.03997454792261124,
726
+ "rewards/margins": 0.22478806972503662,
727
+ "rewards/rejected": -0.26476261019706726,
728
+ "step": 45
729
+ },
730
+ {
731
+ "epoch": 0.29,
732
+ "learning_rate": 1.6577148706018328e-06,
733
+ "logits/chosen": -2.894786834716797,
734
+ "logits/rejected": -2.977565288543701,
735
+ "loss": 0.5775,
736
+ "policy_logps/chosen": -114.12425231933594,
737
+ "policy_logps/rejected": -113.44467163085938,
738
+ "referece_logps/chosen": -112.93356323242188,
739
+ "referece_logps/rejected": -110.42334747314453,
740
+ "rewards/accuracies": 0.8125,
741
+ "rewards/chosen": -0.11906924843788147,
742
+ "rewards/margins": 0.18306350708007812,
743
+ "rewards/rejected": -0.3021327555179596,
744
+ "step": 46
745
+ },
746
+ {
747
+ "epoch": 0.3,
748
+ "learning_rate": 1.6419017501926656e-06,
749
+ "logits/chosen": -2.90000581741333,
750
+ "logits/rejected": -2.9254188537597656,
751
+ "loss": 0.5176,
752
+ "policy_logps/chosen": -128.19607543945312,
753
+ "policy_logps/rejected": -122.17625427246094,
754
+ "referece_logps/chosen": -128.63787841796875,
755
+ "referece_logps/rejected": -118.01435852050781,
756
+ "rewards/accuracies": 0.9375,
757
+ "rewards/chosen": 0.04418013244867325,
758
+ "rewards/margins": 0.4603692889213562,
759
+ "rewards/rejected": -0.41618919372558594,
760
+ "step": 47
761
+ },
762
+ {
763
+ "epoch": 0.31,
764
+ "learning_rate": 1.6258107872407374e-06,
765
+ "logits/chosen": -2.9072225093841553,
766
+ "logits/rejected": -2.9065234661102295,
767
+ "loss": 0.5243,
768
+ "policy_logps/chosen": -125.48599243164062,
769
+ "policy_logps/rejected": -129.7559814453125,
770
+ "referece_logps/chosen": -126.17718505859375,
771
+ "referece_logps/rejected": -127.1695785522461,
772
+ "rewards/accuracies": 0.875,
773
+ "rewards/chosen": 0.06911970674991608,
774
+ "rewards/margins": 0.32776087522506714,
775
+ "rewards/rejected": -0.25864115357398987,
776
+ "step": 48
777
+ },
778
+ {
779
+ "epoch": 0.31,
780
+ "learning_rate": 1.6094489466033042e-06,
781
+ "logits/chosen": -2.917146682739258,
782
+ "logits/rejected": -2.9667577743530273,
783
+ "loss": 0.5559,
784
+ "policy_logps/chosen": -108.18141174316406,
785
+ "policy_logps/rejected": -99.39623260498047,
786
+ "referece_logps/chosen": -108.76286315917969,
787
+ "referece_logps/rejected": -96.53907775878906,
788
+ "rewards/accuracies": 0.875,
789
+ "rewards/chosen": 0.0581454373896122,
790
+ "rewards/margins": 0.34386110305786133,
791
+ "rewards/rejected": -0.28571566939353943,
792
+ "step": 49
793
+ },
794
+ {
795
+ "epoch": 0.32,
796
+ "learning_rate": 1.5928233103850727e-06,
797
+ "logits/chosen": -2.906874179840088,
798
+ "logits/rejected": -2.919196844100952,
799
+ "loss": 0.5364,
800
+ "policy_logps/chosen": -161.1468048095703,
801
+ "policy_logps/rejected": -143.97592163085938,
802
+ "referece_logps/chosen": -160.62081909179688,
803
+ "referece_logps/rejected": -140.28585815429688,
804
+ "rewards/accuracies": 0.9375,
805
+ "rewards/chosen": -0.052598677575588226,
806
+ "rewards/margins": 0.31640806794166565,
807
+ "rewards/rejected": -0.3690067529678345,
808
+ "step": 50
809
+ },
810
+ {
811
+ "epoch": 0.33,
812
+ "learning_rate": 1.575941074872766e-06,
813
+ "logits/chosen": -2.957002878189087,
814
+ "logits/rejected": -2.9672722816467285,
815
+ "loss": 0.5172,
816
+ "policy_logps/chosen": -119.42903900146484,
817
+ "policy_logps/rejected": -116.86624145507812,
818
+ "referece_logps/chosen": -120.13378143310547,
819
+ "referece_logps/rejected": -113.2657241821289,
820
+ "rewards/accuracies": 1.0,
821
+ "rewards/chosen": 0.07047442346811295,
822
+ "rewards/margins": 0.4305253326892853,
823
+ "rewards/rejected": -0.36005088686943054,
824
+ "step": 51
825
+ },
826
+ {
827
+ "epoch": 0.33,
828
+ "learning_rate": 1.5588095474202594e-06,
829
+ "logits/chosen": -2.8591084480285645,
830
+ "logits/rejected": -2.8590025901794434,
831
+ "loss": 0.4933,
832
+ "policy_logps/chosen": -157.062744140625,
833
+ "policy_logps/rejected": -151.7144775390625,
834
+ "referece_logps/chosen": -156.53114318847656,
835
+ "referece_logps/rejected": -147.15911865234375,
836
+ "rewards/accuracies": 0.9375,
837
+ "rewards/chosen": -0.05316000431776047,
838
+ "rewards/margins": 0.4023759067058563,
839
+ "rewards/rejected": -0.4555359482765198,
840
+ "step": 52
841
+ },
842
+ {
843
+ "epoch": 0.34,
844
+ "learning_rate": 1.5414361432856474e-06,
845
+ "logits/chosen": -2.9438529014587402,
846
+ "logits/rejected": -2.9987499713897705,
847
+ "loss": 0.4979,
848
+ "policy_logps/chosen": -117.94309997558594,
849
+ "policy_logps/rejected": -108.414306640625,
850
+ "referece_logps/chosen": -118.55531311035156,
851
+ "referece_logps/rejected": -104.54682922363281,
852
+ "rewards/accuracies": 0.9375,
853
+ "rewards/chosen": 0.061221349984407425,
854
+ "rewards/margins": 0.4479690194129944,
855
+ "rewards/rejected": -0.38674765825271606,
856
+ "step": 53
857
+ },
858
+ {
859
+ "epoch": 0.34,
860
+ "learning_rate": 1.5238283824216013e-06,
861
+ "logits/chosen": -2.9244260787963867,
862
+ "logits/rejected": -2.9373199939727783,
863
+ "loss": 0.4961,
864
+ "policy_logps/chosen": -119.873779296875,
865
+ "policy_logps/rejected": -108.0604248046875,
866
+ "referece_logps/chosen": -120.80511474609375,
867
+ "referece_logps/rejected": -104.79627990722656,
868
+ "rewards/accuracies": 0.9375,
869
+ "rewards/chosen": 0.09313352406024933,
870
+ "rewards/margins": 0.4195476770401001,
871
+ "rewards/rejected": -0.32641416788101196,
872
+ "step": 54
873
+ },
874
+ {
875
+ "epoch": 0.35,
876
+ "learning_rate": 1.5059938862204125e-06,
877
+ "logits/chosen": -2.920050621032715,
878
+ "logits/rejected": -2.9494502544403076,
879
+ "loss": 0.5339,
880
+ "policy_logps/chosen": -128.6884307861328,
881
+ "policy_logps/rejected": -124.91862487792969,
882
+ "referece_logps/chosen": -127.86734008789062,
883
+ "referece_logps/rejected": -120.62930297851562,
884
+ "rewards/accuracies": 0.875,
885
+ "rewards/chosen": -0.08210951089859009,
886
+ "rewards/margins": 0.34682315587997437,
887
+ "rewards/rejected": -0.42893266677856445,
888
+ "step": 55
889
+ },
890
+ {
891
+ "epoch": 0.36,
892
+ "learning_rate": 1.4879403742151283e-06,
893
+ "logits/chosen": -2.907794713973999,
894
+ "logits/rejected": -2.9134182929992676,
895
+ "loss": 0.5389,
896
+ "policy_logps/chosen": -129.15188598632812,
897
+ "policy_logps/rejected": -135.58035278320312,
898
+ "referece_logps/chosen": -127.82069396972656,
899
+ "referece_logps/rejected": -131.39085388183594,
900
+ "rewards/accuracies": 0.6875,
901
+ "rewards/chosen": -0.13311973214149475,
902
+ "rewards/margins": 0.28583019971847534,
903
+ "rewards/rejected": -0.4189499020576477,
904
+ "step": 56
905
+ },
906
+ {
907
+ "epoch": 0.36,
908
+ "learning_rate": 1.4696756607382058e-06,
909
+ "logits/chosen": -2.9352166652679443,
910
+ "logits/rejected": -2.9330966472625732,
911
+ "loss": 0.543,
912
+ "policy_logps/chosen": -125.51739501953125,
913
+ "policy_logps/rejected": -128.10951232910156,
914
+ "referece_logps/chosen": -124.24830627441406,
915
+ "referece_logps/rejected": -123.51809692382812,
916
+ "rewards/accuracies": 0.75,
917
+ "rewards/chosen": -0.12690886855125427,
918
+ "rewards/margins": 0.33223241567611694,
919
+ "rewards/rejected": -0.4591412842273712,
920
+ "step": 57
921
+ },
922
+ {
923
+ "epoch": 0.37,
924
+ "learning_rate": 1.4512076515391374e-06,
925
+ "logits/chosen": -2.941624164581299,
926
+ "logits/rejected": -2.9078333377838135,
927
+ "loss": 0.4758,
928
+ "policy_logps/chosen": -109.23553466796875,
929
+ "policy_logps/rejected": -104.69176483154297,
930
+ "referece_logps/chosen": -110.22058868408203,
931
+ "referece_logps/rejected": -100.0502700805664,
932
+ "rewards/accuracies": 1.0,
933
+ "rewards/chosen": 0.09850560873746872,
934
+ "rewards/margins": 0.5626559257507324,
935
+ "rewards/rejected": -0.4641503393650055,
936
+ "step": 58
937
+ },
938
+ {
939
+ "epoch": 0.38,
940
+ "learning_rate": 1.432544340362501e-06,
941
+ "logits/chosen": -2.972963809967041,
942
+ "logits/rejected": -2.9390883445739746,
943
+ "loss": 0.4904,
944
+ "policy_logps/chosen": -95.29005432128906,
945
+ "policy_logps/rejected": -121.40338897705078,
946
+ "referece_logps/chosen": -95.24687194824219,
947
+ "referece_logps/rejected": -116.24635314941406,
948
+ "rewards/accuracies": 0.9375,
949
+ "rewards/chosen": -0.004318548366427422,
950
+ "rewards/margins": 0.5113850235939026,
951
+ "rewards/rejected": -0.515703558921814,
952
+ "step": 59
953
+ },
954
+ {
955
+ "epoch": 0.38,
956
+ "learning_rate": 1.4136938054879282e-06,
957
+ "logits/chosen": -2.9395623207092285,
958
+ "logits/rejected": -2.953756809234619,
959
+ "loss": 0.5286,
960
+ "policy_logps/chosen": -122.0515365600586,
961
+ "policy_logps/rejected": -128.63021850585938,
962
+ "referece_logps/chosen": -120.7651596069336,
963
+ "referece_logps/rejected": -123.8788070678711,
964
+ "rewards/accuracies": 0.875,
965
+ "rewards/chosen": -0.1286384016275406,
966
+ "rewards/margins": 0.34650319814682007,
967
+ "rewards/rejected": -0.47514158487319946,
968
+ "step": 60
969
+ },
970
+ {
971
+ "epoch": 0.39,
972
+ "learning_rate": 1.3946642062334763e-06,
973
+ "logits/chosen": -2.9107141494750977,
974
+ "logits/rejected": -2.9332330226898193,
975
+ "loss": 0.4832,
976
+ "policy_logps/chosen": -121.9334716796875,
977
+ "policy_logps/rejected": -119.00639343261719,
978
+ "referece_logps/chosen": -122.2487564086914,
979
+ "referece_logps/rejected": -114.04248809814453,
980
+ "rewards/accuracies": 1.0,
981
+ "rewards/chosen": 0.03152900189161301,
982
+ "rewards/margins": 0.5279202461242676,
983
+ "rewards/rejected": -0.49639129638671875,
984
+ "step": 61
985
+ },
986
+ {
987
+ "epoch": 0.4,
988
+ "learning_rate": 1.37546377942393e-06,
989
+ "logits/chosen": -2.950266122817993,
990
+ "logits/rejected": -2.9494924545288086,
991
+ "loss": 0.5178,
992
+ "policy_logps/chosen": -104.2540283203125,
993
+ "policy_logps/rejected": -125.53955841064453,
994
+ "referece_logps/chosen": -104.66864776611328,
995
+ "referece_logps/rejected": -121.41145324707031,
996
+ "rewards/accuracies": 0.9375,
997
+ "rewards/chosen": 0.04146187752485275,
998
+ "rewards/margins": 0.45427215099334717,
999
+ "rewards/rejected": -0.4128102958202362,
1000
+ "step": 62
1001
+ },
1002
+ {
1003
+ "epoch": 0.4,
1004
+ "learning_rate": 1.3561008358255469e-06,
1005
+ "logits/chosen": -2.917292594909668,
1006
+ "logits/rejected": -2.952850818634033,
1007
+ "loss": 0.5263,
1008
+ "policy_logps/chosen": -118.34769439697266,
1009
+ "policy_logps/rejected": -112.80188751220703,
1010
+ "referece_logps/chosen": -118.5436782836914,
1011
+ "referece_logps/rejected": -109.08131408691406,
1012
+ "rewards/accuracies": 0.8125,
1013
+ "rewards/chosen": 0.01959807053208351,
1014
+ "rewards/margins": 0.39165574312210083,
1015
+ "rewards/rejected": -0.3720576763153076,
1016
+ "step": 63
1017
+ },
1018
+ {
1019
+ "epoch": 0.41,
1020
+ "learning_rate": 1.3365837565488062e-06,
1021
+ "logits/chosen": -2.92209529876709,
1022
+ "logits/rejected": -2.9668972492218018,
1023
+ "loss": 0.4813,
1024
+ "policy_logps/chosen": -161.57546997070312,
1025
+ "policy_logps/rejected": -137.75039672851562,
1026
+ "referece_logps/chosen": -161.21104431152344,
1027
+ "referece_logps/rejected": -130.89552307128906,
1028
+ "rewards/accuracies": 1.0,
1029
+ "rewards/chosen": -0.03644174337387085,
1030
+ "rewards/margins": 0.6490457057952881,
1031
+ "rewards/rejected": -0.6854873895645142,
1032
+ "step": 64
1033
+ },
1034
+ {
1035
+ "epoch": 0.41,
1036
+ "learning_rate": 1.3169209894207027e-06,
1037
+ "logits/chosen": -2.9616472721099854,
1038
+ "logits/rejected": -2.961239814758301,
1039
+ "loss": 0.4486,
1040
+ "policy_logps/chosen": -154.44467163085938,
1041
+ "policy_logps/rejected": -158.21372985839844,
1042
+ "referece_logps/chosen": -154.07650756835938,
1043
+ "referece_logps/rejected": -151.34902954101562,
1044
+ "rewards/accuracies": 0.9375,
1045
+ "rewards/chosen": -0.03681756183505058,
1046
+ "rewards/margins": 0.6496531367301941,
1047
+ "rewards/rejected": -0.6864707469940186,
1048
+ "step": 65
1049
+ },
1050
+ {
1051
+ "epoch": 0.42,
1052
+ "learning_rate": 1.2971210453281673e-06,
1053
+ "logits/chosen": -2.892361640930176,
1054
+ "logits/rejected": -2.8887600898742676,
1055
+ "loss": 0.4939,
1056
+ "policy_logps/chosen": -121.81805419921875,
1057
+ "policy_logps/rejected": -114.29660034179688,
1058
+ "referece_logps/chosen": -122.0481948852539,
1059
+ "referece_logps/rejected": -110.4261474609375,
1060
+ "rewards/accuracies": 0.9375,
1061
+ "rewards/chosen": 0.023014426231384277,
1062
+ "rewards/margins": 0.41005975008010864,
1063
+ "rewards/rejected": -0.38704538345336914,
1064
+ "step": 66
1065
+ },
1066
+ {
1067
+ "epoch": 0.43,
1068
+ "learning_rate": 1.2771924945341906e-06,
1069
+ "logits/chosen": -2.91727352142334,
1070
+ "logits/rejected": -2.922391891479492,
1071
+ "loss": 0.4868,
1072
+ "policy_logps/chosen": -111.62061309814453,
1073
+ "policy_logps/rejected": -106.5035400390625,
1074
+ "referece_logps/chosen": -112.6988525390625,
1075
+ "referece_logps/rejected": -101.50572967529297,
1076
+ "rewards/accuracies": 1.0,
1077
+ "rewards/chosen": 0.10782448947429657,
1078
+ "rewards/margins": 0.6076046228408813,
1079
+ "rewards/rejected": -0.49978014826774597,
1080
+ "step": 67
1081
+ },
1082
+ {
1083
+ "epoch": 0.43,
1084
+ "learning_rate": 1.257143962968246e-06,
1085
+ "logits/chosen": -2.9374663829803467,
1086
+ "logits/rejected": -2.9512386322021484,
1087
+ "loss": 0.5144,
1088
+ "policy_logps/chosen": -149.2052001953125,
1089
+ "policy_logps/rejected": -130.6556396484375,
1090
+ "referece_logps/chosen": -147.80715942382812,
1091
+ "referece_logps/rejected": -125.28959655761719,
1092
+ "rewards/accuracies": 0.75,
1093
+ "rewards/chosen": -0.13980263471603394,
1094
+ "rewards/margins": 0.39680200815200806,
1095
+ "rewards/rejected": -0.536604642868042,
1096
+ "step": 68
1097
+ },
1098
+ {
1099
+ "epoch": 0.44,
1100
+ "learning_rate": 1.236984128492619e-06,
1101
+ "logits/chosen": -2.9574966430664062,
1102
+ "logits/rejected": -2.9541845321655273,
1103
+ "loss": 0.4811,
1104
+ "policy_logps/chosen": -105.58607482910156,
1105
+ "policy_logps/rejected": -107.0927734375,
1106
+ "referece_logps/chosen": -106.00244903564453,
1107
+ "referece_logps/rejected": -103.1064453125,
1108
+ "rewards/accuracies": 0.9375,
1109
+ "rewards/chosen": 0.041637033224105835,
1110
+ "rewards/margins": 0.44027072191238403,
1111
+ "rewards/rejected": -0.3986337184906006,
1112
+ "step": 69
1113
+ },
1114
+ {
1115
+ "epoch": 0.45,
1116
+ "learning_rate": 1.2167217171462566e-06,
1117
+ "logits/chosen": -2.96567964553833,
1118
+ "logits/rejected": -2.9865453243255615,
1119
+ "loss": 0.4782,
1120
+ "policy_logps/chosen": -137.23434448242188,
1121
+ "policy_logps/rejected": -116.43749237060547,
1122
+ "referece_logps/chosen": -137.0900421142578,
1123
+ "referece_logps/rejected": -110.79519653320312,
1124
+ "rewards/accuracies": 0.875,
1125
+ "rewards/chosen": -0.014430008828639984,
1126
+ "rewards/margins": 0.5497984886169434,
1127
+ "rewards/rejected": -0.5642285346984863,
1128
+ "step": 70
1129
+ },
1130
+ {
1131
+ "epoch": 0.45,
1132
+ "learning_rate": 1.1963654993677643e-06,
1133
+ "logits/chosen": -2.9047865867614746,
1134
+ "logits/rejected": -2.9375927448272705,
1135
+ "loss": 0.4831,
1136
+ "policy_logps/chosen": -142.0609588623047,
1137
+ "policy_logps/rejected": -138.23785400390625,
1138
+ "referece_logps/chosen": -140.9311981201172,
1139
+ "referece_logps/rejected": -132.38949584960938,
1140
+ "rewards/accuracies": 0.9375,
1141
+ "rewards/chosen": -0.1129767894744873,
1142
+ "rewards/margins": 0.47185903787612915,
1143
+ "rewards/rejected": -0.5848358869552612,
1144
+ "step": 71
1145
+ },
1146
+ {
1147
+ "epoch": 0.46,
1148
+ "learning_rate": 1.1759242861991854e-06,
1149
+ "logits/chosen": -2.9874184131622314,
1150
+ "logits/rejected": -2.9638171195983887,
1151
+ "loss": 0.4436,
1152
+ "policy_logps/chosen": -111.93683624267578,
1153
+ "policy_logps/rejected": -113.12979888916016,
1154
+ "referece_logps/chosen": -111.51397705078125,
1155
+ "referece_logps/rejected": -107.90960693359375,
1156
+ "rewards/accuracies": 0.9375,
1157
+ "rewards/chosen": -0.042286355048418045,
1158
+ "rewards/margins": 0.4797336161136627,
1159
+ "rewards/rejected": -0.5220199227333069,
1160
+ "step": 72
1161
+ },
1162
+ {
1163
+ "epoch": 0.47,
1164
+ "learning_rate": 1.155406925472205e-06,
1165
+ "logits/chosen": -2.927311420440674,
1166
+ "logits/rejected": -2.921337366104126,
1167
+ "loss": 0.4284,
1168
+ "policy_logps/chosen": -147.83377075195312,
1169
+ "policy_logps/rejected": -148.07089233398438,
1170
+ "referece_logps/chosen": -147.18499755859375,
1171
+ "referece_logps/rejected": -140.68211364746094,
1172
+ "rewards/accuracies": 1.0,
1173
+ "rewards/chosen": -0.0648760199546814,
1174
+ "rewards/margins": 0.6740000247955322,
1175
+ "rewards/rejected": -0.7388760447502136,
1176
+ "step": 73
1177
+ },
1178
+ {
1179
+ "epoch": 0.47,
1180
+ "learning_rate": 1.1348222979784287e-06,
1181
+ "logits/chosen": -2.9232394695281982,
1182
+ "logits/rejected": -2.9430999755859375,
1183
+ "loss": 0.5002,
1184
+ "policy_logps/chosen": -143.42274475097656,
1185
+ "policy_logps/rejected": -143.17144775390625,
1186
+ "referece_logps/chosen": -141.336669921875,
1187
+ "referece_logps/rejected": -136.30712890625,
1188
+ "rewards/accuracies": 0.9375,
1189
+ "rewards/chosen": -0.20860746502876282,
1190
+ "rewards/margins": 0.4778253436088562,
1191
+ "rewards/rejected": -0.6864327788352966,
1192
+ "step": 74
1193
+ },
1194
+ {
1195
+ "epoch": 0.48,
1196
+ "learning_rate": 1.1141793136253986e-06,
1197
+ "logits/chosen": -2.8802056312561035,
1198
+ "logits/rejected": -2.9181809425354004,
1199
+ "loss": 0.5141,
1200
+ "policy_logps/chosen": -155.84146118164062,
1201
+ "policy_logps/rejected": -145.16802978515625,
1202
+ "referece_logps/chosen": -154.25274658203125,
1203
+ "referece_logps/rejected": -140.40045166015625,
1204
+ "rewards/accuracies": 0.8125,
1205
+ "rewards/chosen": -0.15887098014354706,
1206
+ "rewards/margins": 0.31788796186447144,
1207
+ "rewards/rejected": -0.4767589569091797,
1208
+ "step": 75
1209
+ },
1210
+ {
1211
+ "epoch": 0.48,
1212
+ "learning_rate": 1.09348690758e-06,
1213
+ "logits/chosen": -2.921013355255127,
1214
+ "logits/rejected": -2.949887275695801,
1215
+ "loss": 0.4978,
1216
+ "policy_logps/chosen": -140.55694580078125,
1217
+ "policy_logps/rejected": -137.7884521484375,
1218
+ "referece_logps/chosen": -139.3011474609375,
1219
+ "referece_logps/rejected": -130.94146728515625,
1220
+ "rewards/accuracies": 0.9375,
1221
+ "rewards/chosen": -0.12557877600193024,
1222
+ "rewards/margins": 0.5591215491294861,
1223
+ "rewards/rejected": -0.6847003102302551,
1224
+ "step": 76
1225
+ },
1226
+ {
1227
+ "epoch": 0.49,
1228
+ "learning_rate": 1.072754036400944e-06,
1229
+ "logits/chosen": -2.9696455001831055,
1230
+ "logits/rejected": -2.9857373237609863,
1231
+ "loss": 0.417,
1232
+ "policy_logps/chosen": -113.93096923828125,
1233
+ "policy_logps/rejected": -107.09919738769531,
1234
+ "referece_logps/chosen": -115.27886962890625,
1235
+ "referece_logps/rejected": -102.10234069824219,
1236
+ "rewards/accuracies": 1.0,
1237
+ "rewards/chosen": 0.1347896158695221,
1238
+ "rewards/margins": 0.6344748139381409,
1239
+ "rewards/rejected": -0.49968522787094116,
1240
+ "step": 77
1241
+ },
1242
+ {
1243
+ "epoch": 0.5,
1244
+ "learning_rate": 1.0519896741619803e-06,
1245
+ "logits/chosen": -2.919851303100586,
1246
+ "logits/rejected": -2.9218404293060303,
1247
+ "loss": 0.4688,
1248
+ "policy_logps/chosen": -155.71444702148438,
1249
+ "policy_logps/rejected": -144.62152099609375,
1250
+ "referece_logps/chosen": -155.02752685546875,
1251
+ "referece_logps/rejected": -138.43026733398438,
1252
+ "rewards/accuracies": 0.8125,
1253
+ "rewards/chosen": -0.0686924085021019,
1254
+ "rewards/margins": 0.5504311323165894,
1255
+ "rewards/rejected": -0.6191235184669495,
1256
+ "step": 78
1257
+ },
1258
+ {
1259
+ "epoch": 0.5,
1260
+ "learning_rate": 1.031202808567539e-06,
1261
+ "logits/chosen": -2.8995673656463623,
1262
+ "logits/rejected": -2.930441379547119,
1263
+ "loss": 0.4539,
1264
+ "policy_logps/chosen": -155.89486694335938,
1265
+ "policy_logps/rejected": -133.66708374023438,
1266
+ "referece_logps/chosen": -154.44532775878906,
1267
+ "referece_logps/rejected": -126.08793640136719,
1268
+ "rewards/accuracies": 0.9375,
1269
+ "rewards/chosen": -0.14495408535003662,
1270
+ "rewards/margins": 0.6129606366157532,
1271
+ "rewards/rejected": -0.7579147815704346,
1272
+ "step": 79
1273
+ },
1274
+ {
1275
+ "epoch": 0.51,
1276
+ "learning_rate": 1.0104024370624642e-06,
1277
+ "logits/chosen": -2.9370193481445312,
1278
+ "logits/rejected": -2.9806201457977295,
1279
+ "loss": 0.4106,
1280
+ "policy_logps/chosen": -134.17471313476562,
1281
+ "policy_logps/rejected": -112.07872772216797,
1282
+ "referece_logps/chosen": -134.83750915527344,
1283
+ "referece_logps/rejected": -105.02826690673828,
1284
+ "rewards/accuracies": 0.9375,
1285
+ "rewards/chosen": 0.06628021597862244,
1286
+ "rewards/margins": 0.7713260650634766,
1287
+ "rewards/rejected": -0.7050458192825317,
1288
+ "step": 80
1289
+ },
1290
+ {
1291
+ "epoch": 0.52,
1292
+ "learning_rate": 9.895975629375357e-07,
1293
+ "logits/chosen": -2.9489431381225586,
1294
+ "logits/rejected": -2.931128978729248,
1295
+ "loss": 0.4902,
1296
+ "policy_logps/chosen": -132.9930877685547,
1297
+ "policy_logps/rejected": -148.03472900390625,
1298
+ "referece_logps/chosen": -131.95632934570312,
1299
+ "referece_logps/rejected": -141.67227172851562,
1300
+ "rewards/accuracies": 0.9375,
1301
+ "rewards/chosen": -0.1036761999130249,
1302
+ "rewards/margins": 0.5325698852539062,
1303
+ "rewards/rejected": -0.6362460851669312,
1304
+ "step": 81
1305
+ },
1306
+ {
1307
+ "epoch": 0.52,
1308
+ "learning_rate": 9.687971914324607e-07,
1309
+ "logits/chosen": -2.911447525024414,
1310
+ "logits/rejected": -2.9363858699798584,
1311
+ "loss": 0.4865,
1312
+ "policy_logps/chosen": -136.85560607910156,
1313
+ "policy_logps/rejected": -110.8121337890625,
1314
+ "referece_logps/chosen": -135.94354248046875,
1315
+ "referece_logps/rejected": -105.58758544921875,
1316
+ "rewards/accuracies": 0.9375,
1317
+ "rewards/chosen": -0.09120647609233856,
1318
+ "rewards/margins": 0.4312480390071869,
1319
+ "rewards/rejected": -0.5224545001983643,
1320
+ "step": 82
1321
+ },
1322
+ {
1323
+ "epoch": 0.53,
1324
+ "learning_rate": 9.480103258380197e-07,
1325
+ "logits/chosen": -2.9371180534362793,
1326
+ "logits/rejected": -2.9503769874572754,
1327
+ "loss": 0.4557,
1328
+ "policy_logps/chosen": -150.19725036621094,
1329
+ "policy_logps/rejected": -149.049072265625,
1330
+ "referece_logps/chosen": -147.80014038085938,
1331
+ "referece_logps/rejected": -141.38333129882812,
1332
+ "rewards/accuracies": 0.9375,
1333
+ "rewards/chosen": -0.23970948159694672,
1334
+ "rewards/margins": 0.5268632173538208,
1335
+ "rewards/rejected": -0.7665727138519287,
1336
+ "step": 83
1337
+ },
1338
+ {
1339
+ "epoch": 0.54,
1340
+ "learning_rate": 9.272459635990562e-07,
1341
+ "logits/chosen": -2.914623737335205,
1342
+ "logits/rejected": -2.909700393676758,
1343
+ "loss": 0.4753,
1344
+ "policy_logps/chosen": -149.44277954101562,
1345
+ "policy_logps/rejected": -158.23593139648438,
1346
+ "referece_logps/chosen": -147.42015075683594,
1347
+ "referece_logps/rejected": -149.85214233398438,
1348
+ "rewards/accuracies": 0.875,
1349
+ "rewards/chosen": -0.2022620439529419,
1350
+ "rewards/margins": 0.6361156702041626,
1351
+ "rewards/rejected": -0.838377833366394,
1352
+ "step": 84
1353
+ },
1354
+ {
1355
+ "epoch": 0.54,
1356
+ "learning_rate": 9.065130924199998e-07,
1357
+ "logits/chosen": -2.964315414428711,
1358
+ "logits/rejected": -2.9234838485717773,
1359
+ "loss": 0.5092,
1360
+ "policy_logps/chosen": -133.36880493164062,
1361
+ "policy_logps/rejected": -146.08169555664062,
1362
+ "referece_logps/chosen": -132.2157440185547,
1363
+ "referece_logps/rejected": -140.08538818359375,
1364
+ "rewards/accuracies": 0.875,
1365
+ "rewards/chosen": -0.11530620604753494,
1366
+ "rewards/margins": 0.4843238890171051,
1367
+ "rewards/rejected": -0.5996301174163818,
1368
+ "step": 85
1369
+ },
1370
+ {
1371
+ "epoch": 0.55,
1372
+ "learning_rate": 8.858206863746017e-07,
1373
+ "logits/chosen": -2.909698486328125,
1374
+ "logits/rejected": -2.918955087661743,
1375
+ "loss": 0.4525,
1376
+ "policy_logps/chosen": -132.2825469970703,
1377
+ "policy_logps/rejected": -136.70962524414062,
1378
+ "referece_logps/chosen": -130.87799072265625,
1379
+ "referece_logps/rejected": -130.4683837890625,
1380
+ "rewards/accuracies": 0.8125,
1381
+ "rewards/chosen": -0.1404554843902588,
1382
+ "rewards/margins": 0.48366838693618774,
1383
+ "rewards/rejected": -0.6241238713264465,
1384
+ "step": 86
1385
+ },
1386
+ {
1387
+ "epoch": 0.56,
1388
+ "learning_rate": 8.651777020215712e-07,
1389
+ "logits/chosen": -2.8775806427001953,
1390
+ "logits/rejected": -2.876236915588379,
1391
+ "loss": 0.4829,
1392
+ "policy_logps/chosen": -138.22640991210938,
1393
+ "policy_logps/rejected": -135.67745971679688,
1394
+ "referece_logps/chosen": -137.68087768554688,
1395
+ "referece_logps/rejected": -129.38088989257812,
1396
+ "rewards/accuracies": 1.0,
1397
+ "rewards/chosen": -0.05455498397350311,
1398
+ "rewards/margins": 0.5751017928123474,
1399
+ "rewards/rejected": -0.6296567916870117,
1400
+ "step": 87
1401
+ },
1402
+ {
1403
+ "epoch": 0.56,
1404
+ "learning_rate": 8.445930745277951e-07,
1405
+ "logits/chosen": -2.9498825073242188,
1406
+ "logits/rejected": -2.9456593990325928,
1407
+ "loss": 0.4666,
1408
+ "policy_logps/chosen": -99.36846160888672,
1409
+ "policy_logps/rejected": -110.91441345214844,
1410
+ "referece_logps/chosen": -97.86239624023438,
1411
+ "referece_logps/rejected": -104.65141296386719,
1412
+ "rewards/accuracies": 0.875,
1413
+ "rewards/chosen": -0.15060700476169586,
1414
+ "rewards/margins": 0.47569307684898376,
1415
+ "rewards/rejected": -0.6263000965118408,
1416
+ "step": 88
1417
+ },
1418
+ {
1419
+ "epoch": 0.57,
1420
+ "learning_rate": 8.240757138008148e-07,
1421
+ "logits/chosen": -2.9100852012634277,
1422
+ "logits/rejected": -2.9250974655151367,
1423
+ "loss": 0.4921,
1424
+ "policy_logps/chosen": -144.23516845703125,
1425
+ "policy_logps/rejected": -140.12838745117188,
1426
+ "referece_logps/chosen": -142.4854736328125,
1427
+ "referece_logps/rejected": -133.91094970703125,
1428
+ "rewards/accuracies": 0.875,
1429
+ "rewards/chosen": -0.17496907711029053,
1430
+ "rewards/margins": 0.4467761516571045,
1431
+ "rewards/rejected": -0.621745228767395,
1432
+ "step": 89
1433
+ },
1434
+ {
1435
+ "epoch": 0.57,
1436
+ "learning_rate": 8.036345006322358e-07,
1437
+ "logits/chosen": -2.9227583408355713,
1438
+ "logits/rejected": -2.936587333679199,
1439
+ "loss": 0.4516,
1440
+ "policy_logps/chosen": -130.54054260253906,
1441
+ "policy_logps/rejected": -137.36415100097656,
1442
+ "referece_logps/chosen": -129.6831512451172,
1443
+ "referece_logps/rejected": -129.63467407226562,
1444
+ "rewards/accuracies": 0.9375,
1445
+ "rewards/chosen": -0.08573976159095764,
1446
+ "rewards/margins": 0.6872075796127319,
1447
+ "rewards/rejected": -0.772947371006012,
1448
+ "step": 90
1449
+ },
1450
+ {
1451
+ "epoch": 0.58,
1452
+ "learning_rate": 7.832782828537435e-07,
1453
+ "logits/chosen": -2.946342706680298,
1454
+ "logits/rejected": -2.9744181632995605,
1455
+ "loss": 0.4716,
1456
+ "policy_logps/chosen": -155.42410278320312,
1457
+ "policy_logps/rejected": -155.65284729003906,
1458
+ "referece_logps/chosen": -153.79293823242188,
1459
+ "referece_logps/rejected": -145.5734100341797,
1460
+ "rewards/accuracies": 0.9375,
1461
+ "rewards/chosen": -0.16311725974082947,
1462
+ "rewards/margins": 0.8448256850242615,
1463
+ "rewards/rejected": -1.0079429149627686,
1464
+ "step": 91
1465
+ },
1466
+ {
1467
+ "epoch": 0.59,
1468
+ "learning_rate": 7.630158715073812e-07,
1469
+ "logits/chosen": -2.9480092525482178,
1470
+ "logits/rejected": -2.971871852874756,
1471
+ "loss": 0.4588,
1472
+ "policy_logps/chosen": -117.96566772460938,
1473
+ "policy_logps/rejected": -120.06287384033203,
1474
+ "referece_logps/chosen": -117.35060119628906,
1475
+ "referece_logps/rejected": -114.15763092041016,
1476
+ "rewards/accuracies": 0.875,
1477
+ "rewards/chosen": -0.06150689721107483,
1478
+ "rewards/margins": 0.5290161371231079,
1479
+ "rewards/rejected": -0.5905230045318604,
1480
+ "step": 92
1481
+ },
1482
+ {
1483
+ "epoch": 0.59,
1484
+ "learning_rate": 7.428560370317541e-07,
1485
+ "logits/chosen": -2.9216105937957764,
1486
+ "logits/rejected": -2.973964214324951,
1487
+ "loss": 0.4124,
1488
+ "policy_logps/chosen": -118.88021850585938,
1489
+ "policy_logps/rejected": -116.6799087524414,
1490
+ "referece_logps/chosen": -118.77764892578125,
1491
+ "referece_logps/rejected": -110.41334533691406,
1492
+ "rewards/accuracies": 0.9375,
1493
+ "rewards/chosen": -0.010256083682179451,
1494
+ "rewards/margins": 0.6163986921310425,
1495
+ "rewards/rejected": -0.6266547441482544,
1496
+ "step": 93
1497
+ },
1498
+ {
1499
+ "epoch": 0.6,
1500
+ "learning_rate": 7.228075054658095e-07,
1501
+ "logits/chosen": -2.914088249206543,
1502
+ "logits/rejected": -2.949735164642334,
1503
+ "loss": 0.5151,
1504
+ "policy_logps/chosen": -156.11376953125,
1505
+ "policy_logps/rejected": -136.4941864013672,
1506
+ "referece_logps/chosen": -154.32318115234375,
1507
+ "referece_logps/rejected": -129.20169067382812,
1508
+ "rewards/accuracies": 0.875,
1509
+ "rewards/chosen": -0.17905890941619873,
1510
+ "rewards/margins": 0.5501898527145386,
1511
+ "rewards/rejected": -0.7292487621307373,
1512
+ "step": 94
1513
+ },
1514
+ {
1515
+ "epoch": 0.61,
1516
+ "learning_rate": 7.028789546718325e-07,
1517
+ "logits/chosen": -2.9285836219787598,
1518
+ "logits/rejected": -2.9289567470550537,
1519
+ "loss": 0.46,
1520
+ "policy_logps/chosen": -138.6322021484375,
1521
+ "policy_logps/rejected": -140.56346130371094,
1522
+ "referece_logps/chosen": -137.0457763671875,
1523
+ "referece_logps/rejected": -134.56263732910156,
1524
+ "rewards/accuracies": 0.9375,
1525
+ "rewards/chosen": -0.15864019095897675,
1526
+ "rewards/margins": 0.44144219160079956,
1527
+ "rewards/rejected": -0.6000823974609375,
1528
+ "step": 95
1529
+ },
1530
+ {
1531
+ "epoch": 0.61,
1532
+ "learning_rate": 6.830790105792973e-07,
1533
+ "logits/chosen": -2.9379961490631104,
1534
+ "logits/rejected": -2.9724767208099365,
1535
+ "loss": 0.4291,
1536
+ "policy_logps/chosen": -101.97422790527344,
1537
+ "policy_logps/rejected": -120.30264282226562,
1538
+ "referece_logps/chosen": -102.00912475585938,
1539
+ "referece_logps/rejected": -112.79841613769531,
1540
+ "rewards/accuracies": 0.9375,
1541
+ "rewards/chosen": 0.0034902598708868027,
1542
+ "rewards/margins": 0.7539129257202148,
1543
+ "rewards/rejected": -0.7504226565361023,
1544
+ "step": 96
1545
+ },
1546
+ {
1547
+ "epoch": 0.62,
1548
+ "learning_rate": 6.634162434511938e-07,
1549
+ "logits/chosen": -2.9047234058380127,
1550
+ "logits/rejected": -2.9031195640563965,
1551
+ "loss": 0.4965,
1552
+ "policy_logps/chosen": -143.34796142578125,
1553
+ "policy_logps/rejected": -137.45285034179688,
1554
+ "referece_logps/chosen": -140.2106170654297,
1555
+ "referece_logps/rejected": -128.766845703125,
1556
+ "rewards/accuracies": 0.875,
1557
+ "rewards/chosen": -0.3137343227863312,
1558
+ "rewards/margins": 0.5548651814460754,
1559
+ "rewards/rejected": -0.868599534034729,
1560
+ "step": 97
1561
+ },
1562
+ {
1563
+ "epoch": 0.63,
1564
+ "learning_rate": 6.43899164174453e-07,
1565
+ "logits/chosen": -2.9683728218078613,
1566
+ "logits/rejected": -2.9473183155059814,
1567
+ "loss": 0.4951,
1568
+ "policy_logps/chosen": -99.94713592529297,
1569
+ "policy_logps/rejected": -99.26773834228516,
1570
+ "referece_logps/chosen": -100.4857406616211,
1571
+ "referece_logps/rejected": -93.8021011352539,
1572
+ "rewards/accuracies": 0.9375,
1573
+ "rewards/chosen": 0.05386023223400116,
1574
+ "rewards/margins": 0.6004235148429871,
1575
+ "rewards/rejected": -0.5465632677078247,
1576
+ "step": 98
1577
+ },
1578
+ {
1579
+ "epoch": 0.63,
1580
+ "learning_rate": 6.245362205760703e-07,
1581
+ "logits/chosen": -2.912815570831299,
1582
+ "logits/rejected": -2.941206216812134,
1583
+ "loss": 0.4346,
1584
+ "policy_logps/chosen": -149.0521240234375,
1585
+ "policy_logps/rejected": -140.37326049804688,
1586
+ "referece_logps/chosen": -146.72096252441406,
1587
+ "referece_logps/rejected": -131.8609161376953,
1588
+ "rewards/accuracies": 0.875,
1589
+ "rewards/chosen": -0.23311704397201538,
1590
+ "rewards/margins": 0.6181185245513916,
1591
+ "rewards/rejected": -0.8512355089187622,
1592
+ "step": 99
1593
+ },
1594
+ {
1595
+ "epoch": 0.64,
1596
+ "learning_rate": 6.053357937665236e-07,
1597
+ "logits/chosen": -2.9297289848327637,
1598
+ "logits/rejected": -2.9020700454711914,
1599
+ "loss": 0.4962,
1600
+ "policy_logps/chosen": -130.0217742919922,
1601
+ "policy_logps/rejected": -139.19395446777344,
1602
+ "referece_logps/chosen": -128.82655334472656,
1603
+ "referece_logps/rejected": -132.361572265625,
1604
+ "rewards/accuracies": 0.9375,
1605
+ "rewards/chosen": -0.11952205747365952,
1606
+ "rewards/margins": 0.563715934753418,
1607
+ "rewards/rejected": -0.6832380294799805,
1608
+ "step": 100
1609
+ }
1610
+ ],
1611
+ "max_steps": 156,
1612
+ "num_train_epochs": 1,
1613
+ "total_flos": 0.0,
1614
+ "trial_name": null,
1615
+ "trial_params": null
1616
+ }
checkpoint-100-llava/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:43dc92799f8180dc0c0da8cb61b9160653e4cc8d8bfe02a1bbedc47f967e36ba
3
+ size 5755
checkpoint-100-llava/zero_to_fp32.py ADDED
@@ -0,0 +1,578 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env python
2
+
3
+ # Copyright (c) Microsoft Corporation.
4
+ # SPDX-License-Identifier: Apache-2.0
5
+
6
+ # DeepSpeed Team
7
+
8
+ # This script extracts fp32 consolidated weights from a zero 2 and 3 DeepSpeed checkpoints. It gets
9
+ # copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
10
+ # the future. Once extracted, the weights don't require DeepSpeed and can be used in any
11
+ # application.
12
+ #
13
+ # example: python zero_to_fp32.py . pytorch_model.bin
14
+
15
+ import argparse
16
+ import torch
17
+ import glob
18
+ import math
19
+ import os
20
+ import re
21
+ from collections import OrderedDict
22
+ from dataclasses import dataclass
23
+
24
+ # while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
25
+ # DeepSpeed data structures it has to be available in the current python environment.
26
+ from deepspeed.utils import logger
27
+ from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
28
+ FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
29
+ FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
30
+
31
+
32
+ @dataclass
33
+ class zero_model_state:
34
+ buffers: dict()
35
+ param_shapes: dict()
36
+ shared_params: list
37
+ ds_version: int
38
+ frozen_param_shapes: dict()
39
+ frozen_param_fragments: dict()
40
+
41
+
42
+ debug = 0
43
+
44
+ # load to cpu
45
+ device = torch.device('cpu')
46
+
47
+
48
+ def atoi(text):
49
+ return int(text) if text.isdigit() else text
50
+
51
+
52
+ def natural_keys(text):
53
+ '''
54
+ alist.sort(key=natural_keys) sorts in human order
55
+ http://nedbatchelder.com/blog/200712/human_sorting.html
56
+ (See Toothy's implementation in the comments)
57
+ '''
58
+ return [atoi(c) for c in re.split(r'(\d+)', text)]
59
+
60
+
61
+ def get_model_state_file(checkpoint_dir, zero_stage):
62
+ if not os.path.isdir(checkpoint_dir):
63
+ raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
64
+
65
+ # there should be only one file
66
+ if zero_stage == 2:
67
+ file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
68
+ elif zero_stage == 3:
69
+ file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
70
+
71
+ if not os.path.exists(file):
72
+ raise FileNotFoundError(f"can't find model states file at '{file}'")
73
+
74
+ return file
75
+
76
+
77
+ def get_checkpoint_files(checkpoint_dir, glob_pattern):
78
+ # XXX: need to test that this simple glob rule works for multi-node setup too
79
+ ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
80
+
81
+ if len(ckpt_files) == 0:
82
+ raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
83
+
84
+ return ckpt_files
85
+
86
+
87
+ def get_optim_files(checkpoint_dir):
88
+ return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
89
+
90
+
91
+ def get_model_state_files(checkpoint_dir):
92
+ return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
93
+
94
+
95
+ def parse_model_states(files):
96
+ zero_model_states = []
97
+ for file in files:
98
+ state_dict = torch.load(file, map_location=device)
99
+
100
+ if BUFFER_NAMES not in state_dict:
101
+ raise ValueError(f"{file} is not a model state checkpoint")
102
+ buffer_names = state_dict[BUFFER_NAMES]
103
+ if debug:
104
+ print("Found buffers:", buffer_names)
105
+
106
+ # recover just the buffers while restoring them to fp32 if they were saved in fp16
107
+ buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
108
+ param_shapes = state_dict[PARAM_SHAPES]
109
+
110
+ # collect parameters that are included in param_shapes
111
+ param_names = []
112
+ for s in param_shapes:
113
+ for name in s.keys():
114
+ param_names.append(name)
115
+
116
+ # update with frozen parameters
117
+ frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
118
+ if frozen_param_shapes is not None:
119
+ if debug:
120
+ print(f"Found frozen_param_shapes: {frozen_param_shapes}")
121
+ param_names += list(frozen_param_shapes.keys())
122
+
123
+ # handle shared params
124
+ shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
125
+
126
+ ds_version = state_dict.get(DS_VERSION, None)
127
+
128
+ frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
129
+
130
+ z_model_state = zero_model_state(buffers=buffers,
131
+ param_shapes=param_shapes,
132
+ shared_params=shared_params,
133
+ ds_version=ds_version,
134
+ frozen_param_shapes=frozen_param_shapes,
135
+ frozen_param_fragments=frozen_param_fragments)
136
+ zero_model_states.append(z_model_state)
137
+
138
+ return zero_model_states
139
+
140
+
141
+ def parse_optim_states(files, ds_checkpoint_dir):
142
+
143
+ total_files = len(files)
144
+ state_dicts = []
145
+ for f in files:
146
+ state_dicts.append(torch.load(f, map_location=device))
147
+
148
+ if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
149
+ raise ValueError(f"{files[0]} is not a zero checkpoint")
150
+ zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
151
+ world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
152
+
153
+ # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
154
+ # parameters can be different from data parallelism for non-expert parameters. So we can just
155
+ # use the max of the partition_count to get the dp world_size.
156
+
157
+ if type(world_size) is list:
158
+ world_size = max(world_size)
159
+
160
+ if world_size != total_files:
161
+ raise ValueError(
162
+ f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
163
+ "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
164
+ )
165
+
166
+ # the groups are named differently in each stage
167
+ if zero_stage == 2:
168
+ fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
169
+ elif zero_stage == 3:
170
+ fp32_groups_key = FP32_FLAT_GROUPS
171
+ else:
172
+ raise ValueError(f"unknown zero stage {zero_stage}")
173
+
174
+ if zero_stage == 2:
175
+ fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
176
+ elif zero_stage == 3:
177
+ # if there is more than one param group, there will be multiple flattened tensors - one
178
+ # flattened tensor per group - for simplicity merge them into a single tensor
179
+ #
180
+ # XXX: could make the script more memory efficient for when there are multiple groups - it
181
+ # will require matching the sub-lists of param_shapes for each param group flattened tensor
182
+
183
+ fp32_flat_groups = [
184
+ torch.cat(state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key], 0) for i in range(len(state_dicts))
185
+ ]
186
+
187
+ return zero_stage, world_size, fp32_flat_groups
188
+
189
+
190
+ def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir):
191
+ """
192
+ Returns fp32 state_dict reconstructed from ds checkpoint
193
+
194
+ Args:
195
+ - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
196
+
197
+ """
198
+ print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
199
+
200
+ optim_files = get_optim_files(ds_checkpoint_dir)
201
+ zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
202
+ print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
203
+
204
+ model_files = get_model_state_files(ds_checkpoint_dir)
205
+
206
+ zero_model_states = parse_model_states(model_files)
207
+ print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
208
+
209
+ if zero_stage == 2:
210
+ return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states)
211
+ elif zero_stage == 3:
212
+ return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states)
213
+
214
+
215
+ def _zero2_merge_frozen_params(state_dict, zero_model_states):
216
+ if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
217
+ return
218
+
219
+ frozen_param_shapes = zero_model_states[0].frozen_param_shapes
220
+ frozen_param_fragments = zero_model_states[0].frozen_param_fragments
221
+
222
+ if debug:
223
+ num_elem = sum(s.numel() for s in frozen_param_shapes.values())
224
+ print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
225
+
226
+ wanted_params = len(frozen_param_shapes)
227
+ wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
228
+ avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
229
+ print(f'Frozen params: Have {avail_numel} numels to process.')
230
+ print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
231
+
232
+ total_params = 0
233
+ total_numel = 0
234
+ for name, shape in frozen_param_shapes.items():
235
+ total_params += 1
236
+ unpartitioned_numel = shape.numel()
237
+ total_numel += unpartitioned_numel
238
+
239
+ state_dict[name] = frozen_param_fragments[name]
240
+
241
+ if debug:
242
+ print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
243
+
244
+ print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
245
+
246
+
247
+ def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
248
+ param_shapes = zero_model_states[0].param_shapes
249
+
250
+ # Reconstruction protocol:
251
+ #
252
+ # XXX: document this
253
+
254
+ if debug:
255
+ for i in range(world_size):
256
+ for j in range(len(fp32_flat_groups[0])):
257
+ print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
258
+
259
+ # XXX: memory usage doubles here (zero2)
260
+ num_param_groups = len(fp32_flat_groups[0])
261
+ merged_single_partition_of_fp32_groups = []
262
+ for i in range(num_param_groups):
263
+ merged_partitions = [sd[i] for sd in fp32_flat_groups]
264
+ full_single_fp32_vector = torch.cat(merged_partitions, 0)
265
+ merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
266
+ avail_numel = sum(
267
+ [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
268
+
269
+ if debug:
270
+ wanted_params = sum([len(shapes) for shapes in param_shapes])
271
+ wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
272
+ # not asserting if there is a mismatch due to possible padding
273
+ print(f"Have {avail_numel} numels to process.")
274
+ print(f"Need {wanted_numel} numels in {wanted_params} params.")
275
+
276
+ # params
277
+ # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
278
+ # out-of-core computing solution
279
+ total_numel = 0
280
+ total_params = 0
281
+ for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
282
+ offset = 0
283
+ avail_numel = full_single_fp32_vector.numel()
284
+ for name, shape in shapes.items():
285
+
286
+ unpartitioned_numel = shape.numel()
287
+ total_numel += unpartitioned_numel
288
+ total_params += 1
289
+
290
+ if debug:
291
+ print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
292
+ state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
293
+ offset += unpartitioned_numel
294
+
295
+ # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
296
+ # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
297
+ # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
298
+ # live optimizer object, so we are checking that the numbers are within the right range
299
+ align_to = 2 * world_size
300
+
301
+ def zero2_align(x):
302
+ return align_to * math.ceil(x / align_to)
303
+
304
+ if debug:
305
+ print(f"original offset={offset}, avail_numel={avail_numel}")
306
+
307
+ offset = zero2_align(offset)
308
+ avail_numel = zero2_align(avail_numel)
309
+
310
+ if debug:
311
+ print(f"aligned offset={offset}, avail_numel={avail_numel}")
312
+
313
+ # Sanity check
314
+ if offset != avail_numel:
315
+ raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
316
+
317
+ print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
318
+
319
+
320
+ def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states):
321
+ state_dict = OrderedDict()
322
+
323
+ # buffers
324
+ buffers = zero_model_states[0].buffers
325
+ state_dict.update(buffers)
326
+ if debug:
327
+ print(f"added {len(buffers)} buffers")
328
+
329
+ _zero2_merge_frozen_params(state_dict, zero_model_states)
330
+
331
+ _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
332
+
333
+ # recover shared parameters
334
+ for pair in zero_model_states[0].shared_params:
335
+ if pair[1] in state_dict:
336
+ state_dict[pair[0]] = state_dict[pair[1]]
337
+
338
+ return state_dict
339
+
340
+
341
+ def zero3_partitioned_param_info(unpartitioned_numel, world_size):
342
+ remainder = unpartitioned_numel % world_size
343
+ padding_numel = (world_size - remainder) if remainder else 0
344
+ partitioned_numel = math.ceil(unpartitioned_numel / world_size)
345
+ return partitioned_numel, padding_numel
346
+
347
+
348
+ def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
349
+ if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
350
+ return
351
+
352
+ if debug:
353
+ for i in range(world_size):
354
+ num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
355
+ print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
356
+
357
+ frozen_param_shapes = zero_model_states[0].frozen_param_shapes
358
+ wanted_params = len(frozen_param_shapes)
359
+ wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
360
+ avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
361
+ print(f'Frozen params: Have {avail_numel} numels to process.')
362
+ print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
363
+
364
+ total_params = 0
365
+ total_numel = 0
366
+ for name, shape in zero_model_states[0].frozen_param_shapes.items():
367
+ total_params += 1
368
+ unpartitioned_numel = shape.numel()
369
+ total_numel += unpartitioned_numel
370
+
371
+ param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
372
+ state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
373
+
374
+ partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
375
+
376
+ if debug:
377
+ print(
378
+ f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
379
+ )
380
+
381
+ print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
382
+
383
+
384
+ def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
385
+ param_shapes = zero_model_states[0].param_shapes
386
+ avail_numel = fp32_flat_groups[0].numel() * world_size
387
+ # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
388
+ # param, re-consolidating each param, while dealing with padding if any
389
+
390
+ # merge list of dicts, preserving order
391
+ param_shapes = {k: v for d in param_shapes for k, v in d.items()}
392
+
393
+ if debug:
394
+ for i in range(world_size):
395
+ print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
396
+
397
+ wanted_params = len(param_shapes)
398
+ wanted_numel = sum(shape.numel() for shape in param_shapes.values())
399
+ # not asserting if there is a mismatch due to possible padding
400
+ avail_numel = fp32_flat_groups[0].numel() * world_size
401
+ print(f"Trainable params: Have {avail_numel} numels to process.")
402
+ print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
403
+
404
+ # params
405
+ # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
406
+ # out-of-core computing solution
407
+ offset = 0
408
+ total_numel = 0
409
+ total_params = 0
410
+ for name, shape in param_shapes.items():
411
+
412
+ unpartitioned_numel = shape.numel()
413
+ total_numel += unpartitioned_numel
414
+ total_params += 1
415
+
416
+ partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
417
+
418
+ if debug:
419
+ print(
420
+ f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
421
+ )
422
+
423
+ # XXX: memory usage doubles here
424
+ state_dict[name] = torch.cat(
425
+ tuple(fp32_flat_groups[i].narrow(0, offset, partitioned_numel) for i in range(world_size)),
426
+ 0).narrow(0, 0, unpartitioned_numel).view(shape)
427
+ offset += partitioned_numel
428
+
429
+ offset *= world_size
430
+
431
+ # Sanity check
432
+ if offset != avail_numel:
433
+ raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
434
+
435
+ print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
436
+
437
+
438
+ def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states):
439
+ state_dict = OrderedDict()
440
+
441
+ # buffers
442
+ buffers = zero_model_states[0].buffers
443
+ state_dict.update(buffers)
444
+ if debug:
445
+ print(f"added {len(buffers)} buffers")
446
+
447
+ _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
448
+
449
+ _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
450
+
451
+ # recover shared parameters
452
+ for pair in zero_model_states[0].shared_params:
453
+ if pair[1] in state_dict:
454
+ state_dict[pair[0]] = state_dict[pair[1]]
455
+
456
+ return state_dict
457
+
458
+
459
+ def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None):
460
+ """
461
+ Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
462
+ ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
463
+ via a model hub.
464
+
465
+ Args:
466
+ - ``checkpoint_dir``: path to the desired checkpoint folder
467
+ - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
468
+
469
+ Returns:
470
+ - pytorch ``state_dict``
471
+
472
+ Note: this approach may not work if your application doesn't have sufficient free CPU memory and
473
+ you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
474
+ the checkpoint.
475
+
476
+ A typical usage might be ::
477
+
478
+ from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
479
+ # do the training and checkpoint saving
480
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
481
+ model = model.cpu() # move to cpu
482
+ model.load_state_dict(state_dict)
483
+ # submit to model hub or save the model to share with others
484
+
485
+ In this example the ``model`` will no longer be usable in the deepspeed context of the same
486
+ application. i.e. you will need to re-initialize the deepspeed engine, since
487
+ ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
488
+
489
+ If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
490
+
491
+ """
492
+ if tag is None:
493
+ latest_path = os.path.join(checkpoint_dir, 'latest')
494
+ if os.path.isfile(latest_path):
495
+ with open(latest_path, 'r') as fd:
496
+ tag = fd.read().strip()
497
+ else:
498
+ raise ValueError(f"Unable to find 'latest' file at {latest_path}")
499
+
500
+ ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
501
+
502
+ if not os.path.isdir(ds_checkpoint_dir):
503
+ raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
504
+
505
+ return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir)
506
+
507
+
508
+ def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None):
509
+ """
510
+ Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
511
+ loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
512
+
513
+ Args:
514
+ - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
515
+ - ``output_file``: path to the pytorch fp32 state_dict output file (e.g. path/pytorch_model.bin)
516
+ - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
517
+ """
518
+
519
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
520
+ print(f"Saving fp32 state dict to {output_file}")
521
+ torch.save(state_dict, output_file)
522
+
523
+
524
+ def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
525
+ """
526
+ 1. Put the provided model to cpu
527
+ 2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
528
+ 3. Load it into the provided model
529
+
530
+ Args:
531
+ - ``model``: the model object to update
532
+ - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
533
+ - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
534
+
535
+ Returns:
536
+ - ``model`: modified model
537
+
538
+ Make sure you have plenty of CPU memory available before you call this function. If you don't
539
+ have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
540
+ conveniently placed for you in the checkpoint folder.
541
+
542
+ A typical usage might be ::
543
+
544
+ from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
545
+ model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
546
+ # submit to model hub or save the model to share with others
547
+
548
+ Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
549
+ of the same application. i.e. you will need to re-initialize the deepspeed engine, since
550
+ ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
551
+
552
+ """
553
+ logger.info(f"Extracting fp32 weights")
554
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
555
+
556
+ logger.info(f"Overwriting model with fp32 weights")
557
+ model = model.cpu()
558
+ model.load_state_dict(state_dict, strict=False)
559
+
560
+ return model
561
+
562
+
563
+ if __name__ == "__main__":
564
+
565
+ parser = argparse.ArgumentParser()
566
+ parser.add_argument("checkpoint_dir",
567
+ type=str,
568
+ help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
569
+ parser.add_argument(
570
+ "output_file",
571
+ type=str,
572
+ help="path to the pytorch fp32 state_dict output file (e.g. path/checkpoint-12/pytorch_model.bin)")
573
+ parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
574
+ args = parser.parse_args()
575
+
576
+ debug = args.debug
577
+
578
+ convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir, args.output_file)