Hugging Face's logo

Cylingo
/
Xinyuan-QS-72B

Safetensors
cn
qwen2
fine-tuned
Model card Files
xet
Community
jztdd commited on
Commit
c962e57
·
verified ·
1 Parent(s): dd61649

Upload 10 files

Browse files
Files changed (10) hide show
  1. added_tokens.json +5 -0
  2. config.json +28 -0
  3. generation_config.json +6 -0
  4. merges.txt +0 -0
  5. model.safetensors.index.json +970 -0
  6. special_tokens_map.json +20 -0
  7. tokenizer_config.json +44 -0
  8. trainer_state.json +2604 -0
  9. training_args.bin +3 -0
  10. vocab.json +0 -0
added_tokens.json ADDED
@@ -0,0 +1,5 @@
 
 
 
 
 
 
1
+ {
2
+ "<|endoftext|>": 151643,
3
+ "<|im_end|>": 151645,
4
+ "<|im_start|>": 151644
5
+ }
config.json ADDED
@@ -0,0 +1,28 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "/nfs60/liuqi/models/Qwen2-72B",
3
+ "architectures": [
4
+ "Qwen2ForCausalLM"
5
+ ],
6
+ "attention_dropout": 0.0,
7
+ "bos_token_id": 151643,
8
+ "eos_token_id": 151643,
9
+ "hidden_act": "silu",
10
+ "hidden_size": 8192,
11
+ "initializer_range": 0.02,
12
+ "intermediate_size": 29568,
13
+ "max_position_embeddings": 131072,
14
+ "max_window_layers": 28,
15
+ "model_type": "qwen2",
16
+ "num_attention_heads": 64,
17
+ "num_hidden_layers": 80,
18
+ "num_key_value_heads": 8,
19
+ "rms_norm_eps": 1e-05,
20
+ "rope_theta": 1000000.0,
21
+ "sliding_window": 131072,
22
+ "tie_word_embeddings": false,
23
+ "torch_dtype": "bfloat16",
24
+ "transformers_version": "4.40.1",
25
+ "use_cache": false,
26
+ "use_sliding_window": false,
27
+ "vocab_size": 152064
28
+ }
generation_config.json ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token_id": 151643,
3
+ "eos_token_id": 151643,
4
+ "max_new_tokens": 2048,
5
+ "transformers_version": "4.40.1"
6
+ }
merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
model.safetensors.index.json ADDED
@@ -0,0 +1,970 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "metadata": {
3
+ "total_size": 145412407296
4
+ },
5
+ "weight_map": {
6
+ "lm_head.weight": "model-00031-of-00031.safetensors",
7
+ "model.embed_tokens.weight": "model-00001-of-00031.safetensors",
8
+ "model.layers.0.input_layernorm.weight": "model-00001-of-00031.safetensors",
9
+ "model.layers.0.mlp.down_proj.weight": "model-00001-of-00031.safetensors",
10
+ "model.layers.0.mlp.gate_proj.weight": "model-00001-of-00031.safetensors",
11
+ "model.layers.0.mlp.up_proj.weight": "model-00001-of-00031.safetensors",
12
+ "model.layers.0.post_attention_layernorm.weight": "model-00001-of-00031.safetensors",
13
+ "model.layers.0.self_attn.k_proj.bias": "model-00001-of-00031.safetensors",
14
+ "model.layers.0.self_attn.k_proj.weight": "model-00001-of-00031.safetensors",
15
+ "model.layers.0.self_attn.o_proj.weight": "model-00001-of-00031.safetensors",
16
+ "model.layers.0.self_attn.q_proj.bias": "model-00001-of-00031.safetensors",
17
+ "model.layers.0.self_attn.q_proj.weight": "model-00001-of-00031.safetensors",
18
+ "model.layers.0.self_attn.v_proj.bias": "model-00001-of-00031.safetensors",
19
+ "model.layers.0.self_attn.v_proj.weight": "model-00001-of-00031.safetensors",
20
+ "model.layers.1.input_layernorm.weight": "model-00002-of-00031.safetensors",
21
+ "model.layers.1.mlp.down_proj.weight": "model-00002-of-00031.safetensors",
22
+ "model.layers.1.mlp.gate_proj.weight": "model-00002-of-00031.safetensors",
23
+ "model.layers.1.mlp.up_proj.weight": "model-00002-of-00031.safetensors",
24
+ "model.layers.1.post_attention_layernorm.weight": "model-00002-of-00031.safetensors",
25
+ "model.layers.1.self_attn.k_proj.bias": "model-00001-of-00031.safetensors",
26
+ "model.layers.1.self_attn.k_proj.weight": "model-00001-of-00031.safetensors",
27
+ "model.layers.1.self_attn.o_proj.weight": "model-00001-of-00031.safetensors",
28
+ "model.layers.1.self_attn.q_proj.bias": "model-00001-of-00031.safetensors",
29
+ "model.layers.1.self_attn.q_proj.weight": "model-00001-of-00031.safetensors",
30
+ "model.layers.1.self_attn.v_proj.bias": "model-00001-of-00031.safetensors",
31
+ "model.layers.1.self_attn.v_proj.weight": "model-00001-of-00031.safetensors",
32
+ "model.layers.10.input_layernorm.weight": "model-00005-of-00031.safetensors",
33
+ "model.layers.10.mlp.down_proj.weight": "model-00005-of-00031.safetensors",
34
+ "model.layers.10.mlp.gate_proj.weight": "model-00005-of-00031.safetensors",
35
+ "model.layers.10.mlp.up_proj.weight": "model-00005-of-00031.safetensors",
36
+ "model.layers.10.post_attention_layernorm.weight": "model-00005-of-00031.safetensors",
37
+ "model.layers.10.self_attn.k_proj.bias": "model-00005-of-00031.safetensors",
38
+ "model.layers.10.self_attn.k_proj.weight": "model-00005-of-00031.safetensors",
39
+ "model.layers.10.self_attn.o_proj.weight": "model-00005-of-00031.safetensors",
40
+ "model.layers.10.self_attn.q_proj.bias": "model-00005-of-00031.safetensors",
41
+ "model.layers.10.self_attn.q_proj.weight": "model-00005-of-00031.safetensors",
42
+ "model.layers.10.self_attn.v_proj.bias": "model-00005-of-00031.safetensors",
43
+ "model.layers.10.self_attn.v_proj.weight": "model-00005-of-00031.safetensors",
44
+ "model.layers.11.input_layernorm.weight": "model-00005-of-00031.safetensors",
45
+ "model.layers.11.mlp.down_proj.weight": "model-00005-of-00031.safetensors",
46
+ "model.layers.11.mlp.gate_proj.weight": "model-00005-of-00031.safetensors",
47
+ "model.layers.11.mlp.up_proj.weight": "model-00005-of-00031.safetensors",
48
+ "model.layers.11.post_attention_layernorm.weight": "model-00005-of-00031.safetensors",
49
+ "model.layers.11.self_attn.k_proj.bias": "model-00005-of-00031.safetensors",
50
+ "model.layers.11.self_attn.k_proj.weight": "model-00005-of-00031.safetensors",
51
+ "model.layers.11.self_attn.o_proj.weight": "model-00005-of-00031.safetensors",
52
+ "model.layers.11.self_attn.q_proj.bias": "model-00005-of-00031.safetensors",
53
+ "model.layers.11.self_attn.q_proj.weight": "model-00005-of-00031.safetensors",
54
+ "model.layers.11.self_attn.v_proj.bias": "model-00005-of-00031.safetensors",
55
+ "model.layers.11.self_attn.v_proj.weight": "model-00005-of-00031.safetensors",
56
+ "model.layers.12.input_layernorm.weight": "model-00006-of-00031.safetensors",
57
+ "model.layers.12.mlp.down_proj.weight": "model-00006-of-00031.safetensors",
58
+ "model.layers.12.mlp.gate_proj.weight": "model-00006-of-00031.safetensors",
59
+ "model.layers.12.mlp.up_proj.weight": "model-00006-of-00031.safetensors",
60
+ "model.layers.12.post_attention_layernorm.weight": "model-00006-of-00031.safetensors",
61
+ "model.layers.12.self_attn.k_proj.bias": "model-00005-of-00031.safetensors",
62
+ "model.layers.12.self_attn.k_proj.weight": "model-00005-of-00031.safetensors",
63
+ "model.layers.12.self_attn.o_proj.weight": "model-00005-of-00031.safetensors",
64
+ "model.layers.12.self_attn.q_proj.bias": "model-00005-of-00031.safetensors",
65
+ "model.layers.12.self_attn.q_proj.weight": "model-00005-of-00031.safetensors",
66
+ "model.layers.12.self_attn.v_proj.bias": "model-00005-of-00031.safetensors",
67
+ "model.layers.12.self_attn.v_proj.weight": "model-00005-of-00031.safetensors",
68
+ "model.layers.13.input_layernorm.weight": "model-00006-of-00031.safetensors",
69
+ "model.layers.13.mlp.down_proj.weight": "model-00006-of-00031.safetensors",
70
+ "model.layers.13.mlp.gate_proj.weight": "model-00006-of-00031.safetensors",
71
+ "model.layers.13.mlp.up_proj.weight": "model-00006-of-00031.safetensors",
72
+ "model.layers.13.post_attention_layernorm.weight": "model-00006-of-00031.safetensors",
73
+ "model.layers.13.self_attn.k_proj.bias": "model-00006-of-00031.safetensors",
74
+ "model.layers.13.self_attn.k_proj.weight": "model-00006-of-00031.safetensors",
75
+ "model.layers.13.self_attn.o_proj.weight": "model-00006-of-00031.safetensors",
76
+ "model.layers.13.self_attn.q_proj.bias": "model-00006-of-00031.safetensors",
77
+ "model.layers.13.self_attn.q_proj.weight": "model-00006-of-00031.safetensors",
78
+ "model.layers.13.self_attn.v_proj.bias": "model-00006-of-00031.safetensors",
79
+ "model.layers.13.self_attn.v_proj.weight": "model-00006-of-00031.safetensors",
80
+ "model.layers.14.input_layernorm.weight": "model-00006-of-00031.safetensors",
81
+ "model.layers.14.mlp.down_proj.weight": "model-00006-of-00031.safetensors",
82
+ "model.layers.14.mlp.gate_proj.weight": "model-00006-of-00031.safetensors",
83
+ "model.layers.14.mlp.up_proj.weight": "model-00006-of-00031.safetensors",
84
+ "model.layers.14.post_attention_layernorm.weight": "model-00006-of-00031.safetensors",
85
+ "model.layers.14.self_attn.k_proj.bias": "model-00006-of-00031.safetensors",
86
+ "model.layers.14.self_attn.k_proj.weight": "model-00006-of-00031.safetensors",
87
+ "model.layers.14.self_attn.o_proj.weight": "model-00006-of-00031.safetensors",
88
+ "model.layers.14.self_attn.q_proj.bias": "model-00006-of-00031.safetensors",
89
+ "model.layers.14.self_attn.q_proj.weight": "model-00006-of-00031.safetensors",
90
+ "model.layers.14.self_attn.v_proj.bias": "model-00006-of-00031.safetensors",
91
+ "model.layers.14.self_attn.v_proj.weight": "model-00006-of-00031.safetensors",
92
+ "model.layers.15.input_layernorm.weight": "model-00007-of-00031.safetensors",
93
+ "model.layers.15.mlp.down_proj.weight": "model-00007-of-00031.safetensors",
94
+ "model.layers.15.mlp.gate_proj.weight": "model-00007-of-00031.safetensors",
95
+ "model.layers.15.mlp.up_proj.weight": "model-00007-of-00031.safetensors",
96
+ "model.layers.15.post_attention_layernorm.weight": "model-00007-of-00031.safetensors",
97
+ "model.layers.15.self_attn.k_proj.bias": "model-00007-of-00031.safetensors",
98
+ "model.layers.15.self_attn.k_proj.weight": "model-00007-of-00031.safetensors",
99
+ "model.layers.15.self_attn.o_proj.weight": "model-00007-of-00031.safetensors",
100
+ "model.layers.15.self_attn.q_proj.bias": "model-00007-of-00031.safetensors",
101
+ "model.layers.15.self_attn.q_proj.weight": "model-00007-of-00031.safetensors",
102
+ "model.layers.15.self_attn.v_proj.bias": "model-00007-of-00031.safetensors",
103
+ "model.layers.15.self_attn.v_proj.weight": "model-00007-of-00031.safetensors",
104
+ "model.layers.16.input_layernorm.weight": "model-00007-of-00031.safetensors",
105
+ "model.layers.16.mlp.down_proj.weight": "model-00007-of-00031.safetensors",
106
+ "model.layers.16.mlp.gate_proj.weight": "model-00007-of-00031.safetensors",
107
+ "model.layers.16.mlp.up_proj.weight": "model-00007-of-00031.safetensors",
108
+ "model.layers.16.post_attention_layernorm.weight": "model-00007-of-00031.safetensors",
109
+ "model.layers.16.self_attn.k_proj.bias": "model-00007-of-00031.safetensors",
110
+ "model.layers.16.self_attn.k_proj.weight": "model-00007-of-00031.safetensors",
111
+ "model.layers.16.self_attn.o_proj.weight": "model-00007-of-00031.safetensors",
112
+ "model.layers.16.self_attn.q_proj.bias": "model-00007-of-00031.safetensors",
113
+ "model.layers.16.self_attn.q_proj.weight": "model-00007-of-00031.safetensors",
114
+ "model.layers.16.self_attn.v_proj.bias": "model-00007-of-00031.safetensors",
115
+ "model.layers.16.self_attn.v_proj.weight": "model-00007-of-00031.safetensors",
116
+ "model.layers.17.input_layernorm.weight": "model-00008-of-00031.safetensors",
117
+ "model.layers.17.mlp.down_proj.weight": "model-00008-of-00031.safetensors",
118
+ "model.layers.17.mlp.gate_proj.weight": "model-00007-of-00031.safetensors",
119
+ "model.layers.17.mlp.up_proj.weight": "model-00007-of-00031.safetensors",
120
+ "model.layers.17.post_attention_layernorm.weight": "model-00008-of-00031.safetensors",
121
+ "model.layers.17.self_attn.k_proj.bias": "model-00007-of-00031.safetensors",
122
+ "model.layers.17.self_attn.k_proj.weight": "model-00007-of-00031.safetensors",
123
+ "model.layers.17.self_attn.o_proj.weight": "model-00007-of-00031.safetensors",
124
+ "model.layers.17.self_attn.q_proj.bias": "model-00007-of-00031.safetensors",
125
+ "model.layers.17.self_attn.q_proj.weight": "model-00007-of-00031.safetensors",
126
+ "model.layers.17.self_attn.v_proj.bias": "model-00007-of-00031.safetensors",
127
+ "model.layers.17.self_attn.v_proj.weight": "model-00007-of-00031.safetensors",
128
+ "model.layers.18.input_layernorm.weight": "model-00008-of-00031.safetensors",
129
+ "model.layers.18.mlp.down_proj.weight": "model-00008-of-00031.safetensors",
130
+ "model.layers.18.mlp.gate_proj.weight": "model-00008-of-00031.safetensors",
131
+ "model.layers.18.mlp.up_proj.weight": "model-00008-of-00031.safetensors",
132
+ "model.layers.18.post_attention_layernorm.weight": "model-00008-of-00031.safetensors",
133
+ "model.layers.18.self_attn.k_proj.bias": "model-00008-of-00031.safetensors",
134
+ "model.layers.18.self_attn.k_proj.weight": "model-00008-of-00031.safetensors",
135
+ "model.layers.18.self_attn.o_proj.weight": "model-00008-of-00031.safetensors",
136
+ "model.layers.18.self_attn.q_proj.bias": "model-00008-of-00031.safetensors",
137
+ "model.layers.18.self_attn.q_proj.weight": "model-00008-of-00031.safetensors",
138
+ "model.layers.18.self_attn.v_proj.bias": "model-00008-of-00031.safetensors",
139
+ "model.layers.18.self_attn.v_proj.weight": "model-00008-of-00031.safetensors",
140
+ "model.layers.19.input_layernorm.weight": "model-00008-of-00031.safetensors",
141
+ "model.layers.19.mlp.down_proj.weight": "model-00008-of-00031.safetensors",
142
+ "model.layers.19.mlp.gate_proj.weight": "model-00008-of-00031.safetensors",
143
+ "model.layers.19.mlp.up_proj.weight": "model-00008-of-00031.safetensors",
144
+ "model.layers.19.post_attention_layernorm.weight": "model-00008-of-00031.safetensors",
145
+ "model.layers.19.self_attn.k_proj.bias": "model-00008-of-00031.safetensors",
146
+ "model.layers.19.self_attn.k_proj.weight": "model-00008-of-00031.safetensors",
147
+ "model.layers.19.self_attn.o_proj.weight": "model-00008-of-00031.safetensors",
148
+ "model.layers.19.self_attn.q_proj.bias": "model-00008-of-00031.safetensors",
149
+ "model.layers.19.self_attn.q_proj.weight": "model-00008-of-00031.safetensors",
150
+ "model.layers.19.self_attn.v_proj.bias": "model-00008-of-00031.safetensors",
151
+ "model.layers.19.self_attn.v_proj.weight": "model-00008-of-00031.safetensors",
152
+ "model.layers.2.input_layernorm.weight": "model-00002-of-00031.safetensors",
153
+ "model.layers.2.mlp.down_proj.weight": "model-00002-of-00031.safetensors",
154
+ "model.layers.2.mlp.gate_proj.weight": "model-00002-of-00031.safetensors",
155
+ "model.layers.2.mlp.up_proj.weight": "model-00002-of-00031.safetensors",
156
+ "model.layers.2.post_attention_layernorm.weight": "model-00002-of-00031.safetensors",
157
+ "model.layers.2.self_attn.k_proj.bias": "model-00002-of-00031.safetensors",
158
+ "model.layers.2.self_attn.k_proj.weight": "model-00002-of-00031.safetensors",
159
+ "model.layers.2.self_attn.o_proj.weight": "model-00002-of-00031.safetensors",
160
+ "model.layers.2.self_attn.q_proj.bias": "model-00002-of-00031.safetensors",
161
+ "model.layers.2.self_attn.q_proj.weight": "model-00002-of-00031.safetensors",
162
+ "model.layers.2.self_attn.v_proj.bias": "model-00002-of-00031.safetensors",
163
+ "model.layers.2.self_attn.v_proj.weight": "model-00002-of-00031.safetensors",
164
+ "model.layers.20.input_layernorm.weight": "model-00009-of-00031.safetensors",
165
+ "model.layers.20.mlp.down_proj.weight": "model-00009-of-00031.safetensors",
166
+ "model.layers.20.mlp.gate_proj.weight": "model-00008-of-00031.safetensors",
167
+ "model.layers.20.mlp.up_proj.weight": "model-00009-of-00031.safetensors",
168
+ "model.layers.20.post_attention_layernorm.weight": "model-00009-of-00031.safetensors",
169
+ "model.layers.20.self_attn.k_proj.bias": "model-00008-of-00031.safetensors",
170
+ "model.layers.20.self_attn.k_proj.weight": "model-00008-of-00031.safetensors",
171
+ "model.layers.20.self_attn.o_proj.weight": "model-00008-of-00031.safetensors",
172
+ "model.layers.20.self_attn.q_proj.bias": "model-00008-of-00031.safetensors",
173
+ "model.layers.20.self_attn.q_proj.weight": "model-00008-of-00031.safetensors",
174
+ "model.layers.20.self_attn.v_proj.bias": "model-00008-of-00031.safetensors",
175
+ "model.layers.20.self_attn.v_proj.weight": "model-00008-of-00031.safetensors",
176
+ "model.layers.21.input_layernorm.weight": "model-00009-of-00031.safetensors",
177
+ "model.layers.21.mlp.down_proj.weight": "model-00009-of-00031.safetensors",
178
+ "model.layers.21.mlp.gate_proj.weight": "model-00009-of-00031.safetensors",
179
+ "model.layers.21.mlp.up_proj.weight": "model-00009-of-00031.safetensors",
180
+ "model.layers.21.post_attention_layernorm.weight": "model-00009-of-00031.safetensors",
181
+ "model.layers.21.self_attn.k_proj.bias": "model-00009-of-00031.safetensors",
182
+ "model.layers.21.self_attn.k_proj.weight": "model-00009-of-00031.safetensors",
183
+ "model.layers.21.self_attn.o_proj.weight": "model-00009-of-00031.safetensors",
184
+ "model.layers.21.self_attn.q_proj.bias": "model-00009-of-00031.safetensors",
185
+ "model.layers.21.self_attn.q_proj.weight": "model-00009-of-00031.safetensors",
186
+ "model.layers.21.self_attn.v_proj.bias": "model-00009-of-00031.safetensors",
187
+ "model.layers.21.self_attn.v_proj.weight": "model-00009-of-00031.safetensors",
188
+ "model.layers.22.input_layernorm.weight": "model-00009-of-00031.safetensors",
189
+ "model.layers.22.mlp.down_proj.weight": "model-00009-of-00031.safetensors",
190
+ "model.layers.22.mlp.gate_proj.weight": "model-00009-of-00031.safetensors",
191
+ "model.layers.22.mlp.up_proj.weight": "model-00009-of-00031.safetensors",
192
+ "model.layers.22.post_attention_layernorm.weight": "model-00009-of-00031.safetensors",
193
+ "model.layers.22.self_attn.k_proj.bias": "model-00009-of-00031.safetensors",
194
+ "model.layers.22.self_attn.k_proj.weight": "model-00009-of-00031.safetensors",
195
+ "model.layers.22.self_attn.o_proj.weight": "model-00009-of-00031.safetensors",
196
+ "model.layers.22.self_attn.q_proj.bias": "model-00009-of-00031.safetensors",
197
+ "model.layers.22.self_attn.q_proj.weight": "model-00009-of-00031.safetensors",
198
+ "model.layers.22.self_attn.v_proj.bias": "model-00009-of-00031.safetensors",
199
+ "model.layers.22.self_attn.v_proj.weight": "model-00009-of-00031.safetensors",
200
+ "model.layers.23.input_layernorm.weight": "model-00010-of-00031.safetensors",
201
+ "model.layers.23.mlp.down_proj.weight": "model-00010-of-00031.safetensors",
202
+ "model.layers.23.mlp.gate_proj.weight": "model-00010-of-00031.safetensors",
203
+ "model.layers.23.mlp.up_proj.weight": "model-00010-of-00031.safetensors",
204
+ "model.layers.23.post_attention_layernorm.weight": "model-00010-of-00031.safetensors",
205
+ "model.layers.23.self_attn.k_proj.bias": "model-00009-of-00031.safetensors",
206
+ "model.layers.23.self_attn.k_proj.weight": "model-00009-of-00031.safetensors",
207
+ "model.layers.23.self_attn.o_proj.weight": "model-00009-of-00031.safetensors",
208
+ "model.layers.23.self_attn.q_proj.bias": "model-00009-of-00031.safetensors",
209
+ "model.layers.23.self_attn.q_proj.weight": "model-00009-of-00031.safetensors",
210
+ "model.layers.23.self_attn.v_proj.bias": "model-00009-of-00031.safetensors",
211
+ "model.layers.23.self_attn.v_proj.weight": "model-00009-of-00031.safetensors",
212
+ "model.layers.24.input_layernorm.weight": "model-00010-of-00031.safetensors",
213
+ "model.layers.24.mlp.down_proj.weight": "model-00010-of-00031.safetensors",
214
+ "model.layers.24.mlp.gate_proj.weight": "model-00010-of-00031.safetensors",
215
+ "model.layers.24.mlp.up_proj.weight": "model-00010-of-00031.safetensors",
216
+ "model.layers.24.post_attention_layernorm.weight": "model-00010-of-00031.safetensors",
217
+ "model.layers.24.self_attn.k_proj.bias": "model-00010-of-00031.safetensors",
218
+ "model.layers.24.self_attn.k_proj.weight": "model-00010-of-00031.safetensors",
219
+ "model.layers.24.self_attn.o_proj.weight": "model-00010-of-00031.safetensors",
220
+ "model.layers.24.self_attn.q_proj.bias": "model-00010-of-00031.safetensors",
221
+ "model.layers.24.self_attn.q_proj.weight": "model-00010-of-00031.safetensors",
222
+ "model.layers.24.self_attn.v_proj.bias": "model-00010-of-00031.safetensors",
223
+ "model.layers.24.self_attn.v_proj.weight": "model-00010-of-00031.safetensors",
224
+ "model.layers.25.input_layernorm.weight": "model-00010-of-00031.safetensors",
225
+ "model.layers.25.mlp.down_proj.weight": "model-00010-of-00031.safetensors",
226
+ "model.layers.25.mlp.gate_proj.weight": "model-00010-of-00031.safetensors",
227
+ "model.layers.25.mlp.up_proj.weight": "model-00010-of-00031.safetensors",
228
+ "model.layers.25.post_attention_layernorm.weight": "model-00010-of-00031.safetensors",
229
+ "model.layers.25.self_attn.k_proj.bias": "model-00010-of-00031.safetensors",
230
+ "model.layers.25.self_attn.k_proj.weight": "model-00010-of-00031.safetensors",
231
+ "model.layers.25.self_attn.o_proj.weight": "model-00010-of-00031.safetensors",
232
+ "model.layers.25.self_attn.q_proj.bias": "model-00010-of-00031.safetensors",
233
+ "model.layers.25.self_attn.q_proj.weight": "model-00010-of-00031.safetensors",
234
+ "model.layers.25.self_attn.v_proj.bias": "model-00010-of-00031.safetensors",
235
+ "model.layers.25.self_attn.v_proj.weight": "model-00010-of-00031.safetensors",
236
+ "model.layers.26.input_layernorm.weight": "model-00011-of-00031.safetensors",
237
+ "model.layers.26.mlp.down_proj.weight": "model-00011-of-00031.safetensors",
238
+ "model.layers.26.mlp.gate_proj.weight": "model-00011-of-00031.safetensors",
239
+ "model.layers.26.mlp.up_proj.weight": "model-00011-of-00031.safetensors",
240
+ "model.layers.26.post_attention_layernorm.weight": "model-00011-of-00031.safetensors",
241
+ "model.layers.26.self_attn.k_proj.bias": "model-00011-of-00031.safetensors",
242
+ "model.layers.26.self_attn.k_proj.weight": "model-00011-of-00031.safetensors",
243
+ "model.layers.26.self_attn.o_proj.weight": "model-00011-of-00031.safetensors",
244
+ "model.layers.26.self_attn.q_proj.bias": "model-00011-of-00031.safetensors",
245
+ "model.layers.26.self_attn.q_proj.weight": "model-00011-of-00031.safetensors",
246
+ "model.layers.26.self_attn.v_proj.bias": "model-00011-of-00031.safetensors",
247
+ "model.layers.26.self_attn.v_proj.weight": "model-00011-of-00031.safetensors",
248
+ "model.layers.27.input_layernorm.weight": "model-00011-of-00031.safetensors",
249
+ "model.layers.27.mlp.down_proj.weight": "model-00011-of-00031.safetensors",
250
+ "model.layers.27.mlp.gate_proj.weight": "model-00011-of-00031.safetensors",
251
+ "model.layers.27.mlp.up_proj.weight": "model-00011-of-00031.safetensors",
252
+ "model.layers.27.post_attention_layernorm.weight": "model-00011-of-00031.safetensors",
253
+ "model.layers.27.self_attn.k_proj.bias": "model-00011-of-00031.safetensors",
254
+ "model.layers.27.self_attn.k_proj.weight": "model-00011-of-00031.safetensors",
255
+ "model.layers.27.self_attn.o_proj.weight": "model-00011-of-00031.safetensors",
256
+ "model.layers.27.self_attn.q_proj.bias": "model-00011-of-00031.safetensors",
257
+ "model.layers.27.self_attn.q_proj.weight": "model-00011-of-00031.safetensors",
258
+ "model.layers.27.self_attn.v_proj.bias": "model-00011-of-00031.safetensors",
259
+ "model.layers.27.self_attn.v_proj.weight": "model-00011-of-00031.safetensors",
260
+ "model.layers.28.input_layernorm.weight": "model-00012-of-00031.safetensors",
261
+ "model.layers.28.mlp.down_proj.weight": "model-00012-of-00031.safetensors",
262
+ "model.layers.28.mlp.gate_proj.weight": "model-00011-of-00031.safetensors",
263
+ "model.layers.28.mlp.up_proj.weight": "model-00011-of-00031.safetensors",
264
+ "model.layers.28.post_attention_layernorm.weight": "model-00012-of-00031.safetensors",
265
+ "model.layers.28.self_attn.k_proj.bias": "model-00011-of-00031.safetensors",
266
+ "model.layers.28.self_attn.k_proj.weight": "model-00011-of-00031.safetensors",
267
+ "model.layers.28.self_attn.o_proj.weight": "model-00011-of-00031.safetensors",
268
+ "model.layers.28.self_attn.q_proj.bias": "model-00011-of-00031.safetensors",
269
+ "model.layers.28.self_attn.q_proj.weight": "model-00011-of-00031.safetensors",
270
+ "model.layers.28.self_attn.v_proj.bias": "model-00011-of-00031.safetensors",
271
+ "model.layers.28.self_attn.v_proj.weight": "model-00011-of-00031.safetensors",
272
+ "model.layers.29.input_layernorm.weight": "model-00012-of-00031.safetensors",
273
+ "model.layers.29.mlp.down_proj.weight": "model-00012-of-00031.safetensors",
274
+ "model.layers.29.mlp.gate_proj.weight": "model-00012-of-00031.safetensors",
275
+ "model.layers.29.mlp.up_proj.weight": "model-00012-of-00031.safetensors",
276
+ "model.layers.29.post_attention_layernorm.weight": "model-00012-of-00031.safetensors",
277
+ "model.layers.29.self_attn.k_proj.bias": "model-00012-of-00031.safetensors",
278
+ "model.layers.29.self_attn.k_proj.weight": "model-00012-of-00031.safetensors",
279
+ "model.layers.29.self_attn.o_proj.weight": "model-00012-of-00031.safetensors",
280
+ "model.layers.29.self_attn.q_proj.bias": "model-00012-of-00031.safetensors",
281
+ "model.layers.29.self_attn.q_proj.weight": "model-00012-of-00031.safetensors",
282
+ "model.layers.29.self_attn.v_proj.bias": "model-00012-of-00031.safetensors",
283
+ "model.layers.29.self_attn.v_proj.weight": "model-00012-of-00031.safetensors",
284
+ "model.layers.3.input_layernorm.weight": "model-00002-of-00031.safetensors",
285
+ "model.layers.3.mlp.down_proj.weight": "model-00002-of-00031.safetensors",
286
+ "model.layers.3.mlp.gate_proj.weight": "model-00002-of-00031.safetensors",
287
+ "model.layers.3.mlp.up_proj.weight": "model-00002-of-00031.safetensors",
288
+ "model.layers.3.post_attention_layernorm.weight": "model-00002-of-00031.safetensors",
289
+ "model.layers.3.self_attn.k_proj.bias": "model-00002-of-00031.safetensors",
290
+ "model.layers.3.self_attn.k_proj.weight": "model-00002-of-00031.safetensors",
291
+ "model.layers.3.self_attn.o_proj.weight": "model-00002-of-00031.safetensors",
292
+ "model.layers.3.self_attn.q_proj.bias": "model-00002-of-00031.safetensors",
293
+ "model.layers.3.self_attn.q_proj.weight": "model-00002-of-00031.safetensors",
294
+ "model.layers.3.self_attn.v_proj.bias": "model-00002-of-00031.safetensors",
295
+ "model.layers.3.self_attn.v_proj.weight": "model-00002-of-00031.safetensors",
296
+ "model.layers.30.input_layernorm.weight": "model-00012-of-00031.safetensors",
297
+ "model.layers.30.mlp.down_proj.weight": "model-00012-of-00031.safetensors",
298
+ "model.layers.30.mlp.gate_proj.weight": "model-00012-of-00031.safetensors",
299
+ "model.layers.30.mlp.up_proj.weight": "model-00012-of-00031.safetensors",
300
+ "model.layers.30.post_attention_layernorm.weight": "model-00012-of-00031.safetensors",
301
+ "model.layers.30.self_attn.k_proj.bias": "model-00012-of-00031.safetensors",
302
+ "model.layers.30.self_attn.k_proj.weight": "model-00012-of-00031.safetensors",
303
+ "model.layers.30.self_attn.o_proj.weight": "model-00012-of-00031.safetensors",
304
+ "model.layers.30.self_attn.q_proj.bias": "model-00012-of-00031.safetensors",
305
+ "model.layers.30.self_attn.q_proj.weight": "model-00012-of-00031.safetensors",
306
+ "model.layers.30.self_attn.v_proj.bias": "model-00012-of-00031.safetensors",
307
+ "model.layers.30.self_attn.v_proj.weight": "model-00012-of-00031.safetensors",
308
+ "model.layers.31.input_layernorm.weight": "model-00013-of-00031.safetensors",
309
+ "model.layers.31.mlp.down_proj.weight": "model-00013-of-00031.safetensors",
310
+ "model.layers.31.mlp.gate_proj.weight": "model-00012-of-00031.safetensors",
311
+ "model.layers.31.mlp.up_proj.weight": "model-00013-of-00031.safetensors",
312
+ "model.layers.31.post_attention_layernorm.weight": "model-00013-of-00031.safetensors",
313
+ "model.layers.31.self_attn.k_proj.bias": "model-00012-of-00031.safetensors",
314
+ "model.layers.31.self_attn.k_proj.weight": "model-00012-of-00031.safetensors",
315
+ "model.layers.31.self_attn.o_proj.weight": "model-00012-of-00031.safetensors",
316
+ "model.layers.31.self_attn.q_proj.bias": "model-00012-of-00031.safetensors",
317
+ "model.layers.31.self_attn.q_proj.weight": "model-00012-of-00031.safetensors",
318
+ "model.layers.31.self_attn.v_proj.bias": "model-00012-of-00031.safetensors",
319
+ "model.layers.31.self_attn.v_proj.weight": "model-00012-of-00031.safetensors",
320
+ "model.layers.32.input_layernorm.weight": "model-00013-of-00031.safetensors",
321
+ "model.layers.32.mlp.down_proj.weight": "model-00013-of-00031.safetensors",
322
+ "model.layers.32.mlp.gate_proj.weight": "model-00013-of-00031.safetensors",
323
+ "model.layers.32.mlp.up_proj.weight": "model-00013-of-00031.safetensors",
324
+ "model.layers.32.post_attention_layernorm.weight": "model-00013-of-00031.safetensors",
325
+ "model.layers.32.self_attn.k_proj.bias": "model-00013-of-00031.safetensors",
326
+ "model.layers.32.self_attn.k_proj.weight": "model-00013-of-00031.safetensors",
327
+ "model.layers.32.self_attn.o_proj.weight": "model-00013-of-00031.safetensors",
328
+ "model.layers.32.self_attn.q_proj.bias": "model-00013-of-00031.safetensors",
329
+ "model.layers.32.self_attn.q_proj.weight": "model-00013-of-00031.safetensors",
330
+ "model.layers.32.self_attn.v_proj.bias": "model-00013-of-00031.safetensors",
331
+ "model.layers.32.self_attn.v_proj.weight": "model-00013-of-00031.safetensors",
332
+ "model.layers.33.input_layernorm.weight": "model-00013-of-00031.safetensors",
333
+ "model.layers.33.mlp.down_proj.weight": "model-00013-of-00031.safetensors",
334
+ "model.layers.33.mlp.gate_proj.weight": "model-00013-of-00031.safetensors",
335
+ "model.layers.33.mlp.up_proj.weight": "model-00013-of-00031.safetensors",
336
+ "model.layers.33.post_attention_layernorm.weight": "model-00013-of-00031.safetensors",
337
+ "model.layers.33.self_attn.k_proj.bias": "model-00013-of-00031.safetensors",
338
+ "model.layers.33.self_attn.k_proj.weight": "model-00013-of-00031.safetensors",
339
+ "model.layers.33.self_attn.o_proj.weight": "model-00013-of-00031.safetensors",
340
+ "model.layers.33.self_attn.q_proj.bias": "model-00013-of-00031.safetensors",
341
+ "model.layers.33.self_attn.q_proj.weight": "model-00013-of-00031.safetensors",
342
+ "model.layers.33.self_attn.v_proj.bias": "model-00013-of-00031.safetensors",
343
+ "model.layers.33.self_attn.v_proj.weight": "model-00013-of-00031.safetensors",
344
+ "model.layers.34.input_layernorm.weight": "model-00014-of-00031.safetensors",
345
+ "model.layers.34.mlp.down_proj.weight": "model-00014-of-00031.safetensors",
346
+ "model.layers.34.mlp.gate_proj.weight": "model-00014-of-00031.safetensors",
347
+ "model.layers.34.mlp.up_proj.weight": "model-00014-of-00031.safetensors",
348
+ "model.layers.34.post_attention_layernorm.weight": "model-00014-of-00031.safetensors",
349
+ "model.layers.34.self_attn.k_proj.bias": "model-00013-of-00031.safetensors",
350
+ "model.layers.34.self_attn.k_proj.weight": "model-00013-of-00031.safetensors",
351
+ "model.layers.34.self_attn.o_proj.weight": "model-00013-of-00031.safetensors",
352
+ "model.layers.34.self_attn.q_proj.bias": "model-00013-of-00031.safetensors",
353
+ "model.layers.34.self_attn.q_proj.weight": "model-00013-of-00031.safetensors",
354
+ "model.layers.34.self_attn.v_proj.bias": "model-00013-of-00031.safetensors",
355
+ "model.layers.34.self_attn.v_proj.weight": "model-00013-of-00031.safetensors",
356
+ "model.layers.35.input_layernorm.weight": "model-00014-of-00031.safetensors",
357
+ "model.layers.35.mlp.down_proj.weight": "model-00014-of-00031.safetensors",
358
+ "model.layers.35.mlp.gate_proj.weight": "model-00014-of-00031.safetensors",
359
+ "model.layers.35.mlp.up_proj.weight": "model-00014-of-00031.safetensors",
360
+ "model.layers.35.post_attention_layernorm.weight": "model-00014-of-00031.safetensors",
361
+ "model.layers.35.self_attn.k_proj.bias": "model-00014-of-00031.safetensors",
362
+ "model.layers.35.self_attn.k_proj.weight": "model-00014-of-00031.safetensors",
363
+ "model.layers.35.self_attn.o_proj.weight": "model-00014-of-00031.safetensors",
364
+ "model.layers.35.self_attn.q_proj.bias": "model-00014-of-00031.safetensors",
365
+ "model.layers.35.self_attn.q_proj.weight": "model-00014-of-00031.safetensors",
366
+ "model.layers.35.self_attn.v_proj.bias": "model-00014-of-00031.safetensors",
367
+ "model.layers.35.self_attn.v_proj.weight": "model-00014-of-00031.safetensors",
368
+ "model.layers.36.input_layernorm.weight": "model-00014-of-00031.safetensors",
369
+ "model.layers.36.mlp.down_proj.weight": "model-00014-of-00031.safetensors",
370
+ "model.layers.36.mlp.gate_proj.weight": "model-00014-of-00031.safetensors",
371
+ "model.layers.36.mlp.up_proj.weight": "model-00014-of-00031.safetensors",
372
+ "model.layers.36.post_attention_layernorm.weight": "model-00014-of-00031.safetensors",
373
+ "model.layers.36.self_attn.k_proj.bias": "model-00014-of-00031.safetensors",
374
+ "model.layers.36.self_attn.k_proj.weight": "model-00014-of-00031.safetensors",
375
+ "model.layers.36.self_attn.o_proj.weight": "model-00014-of-00031.safetensors",
376
+ "model.layers.36.self_attn.q_proj.bias": "model-00014-of-00031.safetensors",
377
+ "model.layers.36.self_attn.q_proj.weight": "model-00014-of-00031.safetensors",
378
+ "model.layers.36.self_attn.v_proj.bias": "model-00014-of-00031.safetensors",
379
+ "model.layers.36.self_attn.v_proj.weight": "model-00014-of-00031.safetensors",
380
+ "model.layers.37.input_layernorm.weight": "model-00015-of-00031.safetensors",
381
+ "model.layers.37.mlp.down_proj.weight": "model-00015-of-00031.safetensors",
382
+ "model.layers.37.mlp.gate_proj.weight": "model-00015-of-00031.safetensors",
383
+ "model.layers.37.mlp.up_proj.weight": "model-00015-of-00031.safetensors",
384
+ "model.layers.37.post_attention_layernorm.weight": "model-00015-of-00031.safetensors",
385
+ "model.layers.37.self_attn.k_proj.bias": "model-00015-of-00031.safetensors",
386
+ "model.layers.37.self_attn.k_proj.weight": "model-00015-of-00031.safetensors",
387
+ "model.layers.37.self_attn.o_proj.weight": "model-00015-of-00031.safetensors",
388
+ "model.layers.37.self_attn.q_proj.bias": "model-00015-of-00031.safetensors",
389
+ "model.layers.37.self_attn.q_proj.weight": "model-00015-of-00031.safetensors",
390
+ "model.layers.37.self_attn.v_proj.bias": "model-00015-of-00031.safetensors",
391
+ "model.layers.37.self_attn.v_proj.weight": "model-00015-of-00031.safetensors",
392
+ "model.layers.38.input_layernorm.weight": "model-00015-of-00031.safetensors",
393
+ "model.layers.38.mlp.down_proj.weight": "model-00015-of-00031.safetensors",
394
+ "model.layers.38.mlp.gate_proj.weight": "model-00015-of-00031.safetensors",
395
+ "model.layers.38.mlp.up_proj.weight": "model-00015-of-00031.safetensors",
396
+ "model.layers.38.post_attention_layernorm.weight": "model-00015-of-00031.safetensors",
397
+ "model.layers.38.self_attn.k_proj.bias": "model-00015-of-00031.safetensors",
398
+ "model.layers.38.self_attn.k_proj.weight": "model-00015-of-00031.safetensors",
399
+ "model.layers.38.self_attn.o_proj.weight": "model-00015-of-00031.safetensors",
400
+ "model.layers.38.self_attn.q_proj.bias": "model-00015-of-00031.safetensors",
401
+ "model.layers.38.self_attn.q_proj.weight": "model-00015-of-00031.safetensors",
402
+ "model.layers.38.self_attn.v_proj.bias": "model-00015-of-00031.safetensors",
403
+ "model.layers.38.self_attn.v_proj.weight": "model-00015-of-00031.safetensors",
404
+ "model.layers.39.input_layernorm.weight": "model-00016-of-00031.safetensors",
405
+ "model.layers.39.mlp.down_proj.weight": "model-00016-of-00031.safetensors",
406
+ "model.layers.39.mlp.gate_proj.weight": "model-00015-of-00031.safetensors",
407
+ "model.layers.39.mlp.up_proj.weight": "model-00015-of-00031.safetensors",
408
+ "model.layers.39.post_attention_layernorm.weight": "model-00016-of-00031.safetensors",
409
+ "model.layers.39.self_attn.k_proj.bias": "model-00015-of-00031.safetensors",
410
+ "model.layers.39.self_attn.k_proj.weight": "model-00015-of-00031.safetensors",
411
+ "model.layers.39.self_attn.o_proj.weight": "model-00015-of-00031.safetensors",
412
+ "model.layers.39.self_attn.q_proj.bias": "model-00015-of-00031.safetensors",
413
+ "model.layers.39.self_attn.q_proj.weight": "model-00015-of-00031.safetensors",
414
+ "model.layers.39.self_attn.v_proj.bias": "model-00015-of-00031.safetensors",
415
+ "model.layers.39.self_attn.v_proj.weight": "model-00015-of-00031.safetensors",
416
+ "model.layers.4.input_layernorm.weight": "model-00003-of-00031.safetensors",
417
+ "model.layers.4.mlp.down_proj.weight": "model-00003-of-00031.safetensors",
418
+ "model.layers.4.mlp.gate_proj.weight": "model-00003-of-00031.safetensors",
419
+ "model.layers.4.mlp.up_proj.weight": "model-00003-of-00031.safetensors",
420
+ "model.layers.4.post_attention_layernorm.weight": "model-00003-of-00031.safetensors",
421
+ "model.layers.4.self_attn.k_proj.bias": "model-00003-of-00031.safetensors",
422
+ "model.layers.4.self_attn.k_proj.weight": "model-00003-of-00031.safetensors",
423
+ "model.layers.4.self_attn.o_proj.weight": "model-00003-of-00031.safetensors",
424
+ "model.layers.4.self_attn.q_proj.bias": "model-00003-of-00031.safetensors",
425
+ "model.layers.4.self_attn.q_proj.weight": "model-00003-of-00031.safetensors",
426
+ "model.layers.4.self_attn.v_proj.bias": "model-00003-of-00031.safetensors",
427
+ "model.layers.4.self_attn.v_proj.weight": "model-00003-of-00031.safetensors",
428
+ "model.layers.40.input_layernorm.weight": "model-00016-of-00031.safetensors",
429
+ "model.layers.40.mlp.down_proj.weight": "model-00016-of-00031.safetensors",
430
+ "model.layers.40.mlp.gate_proj.weight": "model-00016-of-00031.safetensors",
431
+ "model.layers.40.mlp.up_proj.weight": "model-00016-of-00031.safetensors",
432
+ "model.layers.40.post_attention_layernorm.weight": "model-00016-of-00031.safetensors",
433
+ "model.layers.40.self_attn.k_proj.bias": "model-00016-of-00031.safetensors",
434
+ "model.layers.40.self_attn.k_proj.weight": "model-00016-of-00031.safetensors",
435
+ "model.layers.40.self_attn.o_proj.weight": "model-00016-of-00031.safetensors",
436
+ "model.layers.40.self_attn.q_proj.bias": "model-00016-of-00031.safetensors",
437
+ "model.layers.40.self_attn.q_proj.weight": "model-00016-of-00031.safetensors",
438
+ "model.layers.40.self_attn.v_proj.bias": "model-00016-of-00031.safetensors",
439
+ "model.layers.40.self_attn.v_proj.weight": "model-00016-of-00031.safetensors",
440
+ "model.layers.41.input_layernorm.weight": "model-00016-of-00031.safetensors",
441
+ "model.layers.41.mlp.down_proj.weight": "model-00016-of-00031.safetensors",
442
+ "model.layers.41.mlp.gate_proj.weight": "model-00016-of-00031.safetensors",
443
+ "model.layers.41.mlp.up_proj.weight": "model-00016-of-00031.safetensors",
444
+ "model.layers.41.post_attention_layernorm.weight": "model-00016-of-00031.safetensors",
445
+ "model.layers.41.self_attn.k_proj.bias": "model-00016-of-00031.safetensors",
446
+ "model.layers.41.self_attn.k_proj.weight": "model-00016-of-00031.safetensors",
447
+ "model.layers.41.self_attn.o_proj.weight": "model-00016-of-00031.safetensors",
448
+ "model.layers.41.self_attn.q_proj.bias": "model-00016-of-00031.safetensors",
449
+ "model.layers.41.self_attn.q_proj.weight": "model-00016-of-00031.safetensors",
450
+ "model.layers.41.self_attn.v_proj.bias": "model-00016-of-00031.safetensors",
451
+ "model.layers.41.self_attn.v_proj.weight": "model-00016-of-00031.safetensors",
452
+ "model.layers.42.input_layernorm.weight": "model-00017-of-00031.safetensors",
453
+ "model.layers.42.mlp.down_proj.weight": "model-00017-of-00031.safetensors",
454
+ "model.layers.42.mlp.gate_proj.weight": "model-00016-of-00031.safetensors",
455
+ "model.layers.42.mlp.up_proj.weight": "model-00017-of-00031.safetensors",
456
+ "model.layers.42.post_attention_layernorm.weight": "model-00017-of-00031.safetensors",
457
+ "model.layers.42.self_attn.k_proj.bias": "model-00016-of-00031.safetensors",
458
+ "model.layers.42.self_attn.k_proj.weight": "model-00016-of-00031.safetensors",
459
+ "model.layers.42.self_attn.o_proj.weight": "model-00016-of-00031.safetensors",
460
+ "model.layers.42.self_attn.q_proj.bias": "model-00016-of-00031.safetensors",
461
+ "model.layers.42.self_attn.q_proj.weight": "model-00016-of-00031.safetensors",
462
+ "model.layers.42.self_attn.v_proj.bias": "model-00016-of-00031.safetensors",
463
+ "model.layers.42.self_attn.v_proj.weight": "model-00016-of-00031.safetensors",
464
+ "model.layers.43.input_layernorm.weight": "model-00017-of-00031.safetensors",
465
+ "model.layers.43.mlp.down_proj.weight": "model-00017-of-00031.safetensors",
466
+ "model.layers.43.mlp.gate_proj.weight": "model-00017-of-00031.safetensors",
467
+ "model.layers.43.mlp.up_proj.weight": "model-00017-of-00031.safetensors",
468
+ "model.layers.43.post_attention_layernorm.weight": "model-00017-of-00031.safetensors",
469
+ "model.layers.43.self_attn.k_proj.bias": "model-00017-of-00031.safetensors",
470
+ "model.layers.43.self_attn.k_proj.weight": "model-00017-of-00031.safetensors",
471
+ "model.layers.43.self_attn.o_proj.weight": "model-00017-of-00031.safetensors",
472
+ "model.layers.43.self_attn.q_proj.bias": "model-00017-of-00031.safetensors",
473
+ "model.layers.43.self_attn.q_proj.weight": "model-00017-of-00031.safetensors",
474
+ "model.layers.43.self_attn.v_proj.bias": "model-00017-of-00031.safetensors",
475
+ "model.layers.43.self_attn.v_proj.weight": "model-00017-of-00031.safetensors",
476
+ "model.layers.44.input_layernorm.weight": "model-00017-of-00031.safetensors",
477
+ "model.layers.44.mlp.down_proj.weight": "model-00017-of-00031.safetensors",
478
+ "model.layers.44.mlp.gate_proj.weight": "model-00017-of-00031.safetensors",
479
+ "model.layers.44.mlp.up_proj.weight": "model-00017-of-00031.safetensors",
480
+ "model.layers.44.post_attention_layernorm.weight": "model-00017-of-00031.safetensors",
481
+ "model.layers.44.self_attn.k_proj.bias": "model-00017-of-00031.safetensors",
482
+ "model.layers.44.self_attn.k_proj.weight": "model-00017-of-00031.safetensors",
483
+ "model.layers.44.self_attn.o_proj.weight": "model-00017-of-00031.safetensors",
484
+ "model.layers.44.self_attn.q_proj.bias": "model-00017-of-00031.safetensors",
485
+ "model.layers.44.self_attn.q_proj.weight": "model-00017-of-00031.safetensors",
486
+ "model.layers.44.self_attn.v_proj.bias": "model-00017-of-00031.safetensors",
487
+ "model.layers.44.self_attn.v_proj.weight": "model-00017-of-00031.safetensors",
488
+ "model.layers.45.input_layernorm.weight": "model-00018-of-00031.safetensors",
489
+ "model.layers.45.mlp.down_proj.weight": "model-00018-of-00031.safetensors",
490
+ "model.layers.45.mlp.gate_proj.weight": "model-00018-of-00031.safetensors",
491
+ "model.layers.45.mlp.up_proj.weight": "model-00018-of-00031.safetensors",
492
+ "model.layers.45.post_attention_layernorm.weight": "model-00018-of-00031.safetensors",
493
+ "model.layers.45.self_attn.k_proj.bias": "model-00017-of-00031.safetensors",
494
+ "model.layers.45.self_attn.k_proj.weight": "model-00017-of-00031.safetensors",
495
+ "model.layers.45.self_attn.o_proj.weight": "model-00017-of-00031.safetensors",
496
+ "model.layers.45.self_attn.q_proj.bias": "model-00017-of-00031.safetensors",
497
+ "model.layers.45.self_attn.q_proj.weight": "model-00017-of-00031.safetensors",
498
+ "model.layers.45.self_attn.v_proj.bias": "model-00017-of-00031.safetensors",
499
+ "model.layers.45.self_attn.v_proj.weight": "model-00017-of-00031.safetensors",
500
+ "model.layers.46.input_layernorm.weight": "model-00018-of-00031.safetensors",
501
+ "model.layers.46.mlp.down_proj.weight": "model-00018-of-00031.safetensors",
502
+ "model.layers.46.mlp.gate_proj.weight": "model-00018-of-00031.safetensors",
503
+ "model.layers.46.mlp.up_proj.weight": "model-00018-of-00031.safetensors",
504
+ "model.layers.46.post_attention_layernorm.weight": "model-00018-of-00031.safetensors",
505
+ "model.layers.46.self_attn.k_proj.bias": "model-00018-of-00031.safetensors",
506
+ "model.layers.46.self_attn.k_proj.weight": "model-00018-of-00031.safetensors",
507
+ "model.layers.46.self_attn.o_proj.weight": "model-00018-of-00031.safetensors",
508
+ "model.layers.46.self_attn.q_proj.bias": "model-00018-of-00031.safetensors",
509
+ "model.layers.46.self_attn.q_proj.weight": "model-00018-of-00031.safetensors",
510
+ "model.layers.46.self_attn.v_proj.bias": "model-00018-of-00031.safetensors",
511
+ "model.layers.46.self_attn.v_proj.weight": "model-00018-of-00031.safetensors",
512
+ "model.layers.47.input_layernorm.weight": "model-00018-of-00031.safetensors",
513
+ "model.layers.47.mlp.down_proj.weight": "model-00018-of-00031.safetensors",
514
+ "model.layers.47.mlp.gate_proj.weight": "model-00018-of-00031.safetensors",
515
+ "model.layers.47.mlp.up_proj.weight": "model-00018-of-00031.safetensors",
516
+ "model.layers.47.post_attention_layernorm.weight": "model-00018-of-00031.safetensors",
517
+ "model.layers.47.self_attn.k_proj.bias": "model-00018-of-00031.safetensors",
518
+ "model.layers.47.self_attn.k_proj.weight": "model-00018-of-00031.safetensors",
519
+ "model.layers.47.self_attn.o_proj.weight": "model-00018-of-00031.safetensors",
520
+ "model.layers.47.self_attn.q_proj.bias": "model-00018-of-00031.safetensors",
521
+ "model.layers.47.self_attn.q_proj.weight": "model-00018-of-00031.safetensors",
522
+ "model.layers.47.self_attn.v_proj.bias": "model-00018-of-00031.safetensors",
523
+ "model.layers.47.self_attn.v_proj.weight": "model-00018-of-00031.safetensors",
524
+ "model.layers.48.input_layernorm.weight": "model-00019-of-00031.safetensors",
525
+ "model.layers.48.mlp.down_proj.weight": "model-00019-of-00031.safetensors",
526
+ "model.layers.48.mlp.gate_proj.weight": "model-00019-of-00031.safetensors",
527
+ "model.layers.48.mlp.up_proj.weight": "model-00019-of-00031.safetensors",
528
+ "model.layers.48.post_attention_layernorm.weight": "model-00019-of-00031.safetensors",
529
+ "model.layers.48.self_attn.k_proj.bias": "model-00019-of-00031.safetensors",
530
+ "model.layers.48.self_attn.k_proj.weight": "model-00019-of-00031.safetensors",
531
+ "model.layers.48.self_attn.o_proj.weight": "model-00019-of-00031.safetensors",
532
+ "model.layers.48.self_attn.q_proj.bias": "model-00019-of-00031.safetensors",
533
+ "model.layers.48.self_attn.q_proj.weight": "model-00019-of-00031.safetensors",
534
+ "model.layers.48.self_attn.v_proj.bias": "model-00019-of-00031.safetensors",
535
+ "model.layers.48.self_attn.v_proj.weight": "model-00019-of-00031.safetensors",
536
+ "model.layers.49.input_layernorm.weight": "model-00019-of-00031.safetensors",
537
+ "model.layers.49.mlp.down_proj.weight": "model-00019-of-00031.safetensors",
538
+ "model.layers.49.mlp.gate_proj.weight": "model-00019-of-00031.safetensors",
539
+ "model.layers.49.mlp.up_proj.weight": "model-00019-of-00031.safetensors",
540
+ "model.layers.49.post_attention_layernorm.weight": "model-00019-of-00031.safetensors",
541
+ "model.layers.49.self_attn.k_proj.bias": "model-00019-of-00031.safetensors",
542
+ "model.layers.49.self_attn.k_proj.weight": "model-00019-of-00031.safetensors",
543
+ "model.layers.49.self_attn.o_proj.weight": "model-00019-of-00031.safetensors",
544
+ "model.layers.49.self_attn.q_proj.bias": "model-00019-of-00031.safetensors",
545
+ "model.layers.49.self_attn.q_proj.weight": "model-00019-of-00031.safetensors",
546
+ "model.layers.49.self_attn.v_proj.bias": "model-00019-of-00031.safetensors",
547
+ "model.layers.49.self_attn.v_proj.weight": "model-00019-of-00031.safetensors",
548
+ "model.layers.5.input_layernorm.weight": "model-00003-of-00031.safetensors",
549
+ "model.layers.5.mlp.down_proj.weight": "model-00003-of-00031.safetensors",
550
+ "model.layers.5.mlp.gate_proj.weight": "model-00003-of-00031.safetensors",
551
+ "model.layers.5.mlp.up_proj.weight": "model-00003-of-00031.safetensors",
552
+ "model.layers.5.post_attention_layernorm.weight": "model-00003-of-00031.safetensors",
553
+ "model.layers.5.self_attn.k_proj.bias": "model-00003-of-00031.safetensors",
554
+ "model.layers.5.self_attn.k_proj.weight": "model-00003-of-00031.safetensors",
555
+ "model.layers.5.self_attn.o_proj.weight": "model-00003-of-00031.safetensors",
556
+ "model.layers.5.self_attn.q_proj.bias": "model-00003-of-00031.safetensors",
557
+ "model.layers.5.self_attn.q_proj.weight": "model-00003-of-00031.safetensors",
558
+ "model.layers.5.self_attn.v_proj.bias": "model-00003-of-00031.safetensors",
559
+ "model.layers.5.self_attn.v_proj.weight": "model-00003-of-00031.safetensors",
560
+ "model.layers.50.input_layernorm.weight": "model-00020-of-00031.safetensors",
561
+ "model.layers.50.mlp.down_proj.weight": "model-00020-of-00031.safetensors",
562
+ "model.layers.50.mlp.gate_proj.weight": "model-00019-of-00031.safetensors",
563
+ "model.layers.50.mlp.up_proj.weight": "model-00019-of-00031.safetensors",
564
+ "model.layers.50.post_attention_layernorm.weight": "model-00020-of-00031.safetensors",
565
+ "model.layers.50.self_attn.k_proj.bias": "model-00019-of-00031.safetensors",
566
+ "model.layers.50.self_attn.k_proj.weight": "model-00019-of-00031.safetensors",
567
+ "model.layers.50.self_attn.o_proj.weight": "model-00019-of-00031.safetensors",
568
+ "model.layers.50.self_attn.q_proj.bias": "model-00019-of-00031.safetensors",
569
+ "model.layers.50.self_attn.q_proj.weight": "model-00019-of-00031.safetensors",
570
+ "model.layers.50.self_attn.v_proj.bias": "model-00019-of-00031.safetensors",
571
+ "model.layers.50.self_attn.v_proj.weight": "model-00019-of-00031.safetensors",
572
+ "model.layers.51.input_layernorm.weight": "model-00020-of-00031.safetensors",
573
+ "model.layers.51.mlp.down_proj.weight": "model-00020-of-00031.safetensors",
574
+ "model.layers.51.mlp.gate_proj.weight": "model-00020-of-00031.safetensors",
575
+ "model.layers.51.mlp.up_proj.weight": "model-00020-of-00031.safetensors",
576
+ "model.layers.51.post_attention_layernorm.weight": "model-00020-of-00031.safetensors",
577
+ "model.layers.51.self_attn.k_proj.bias": "model-00020-of-00031.safetensors",
578
+ "model.layers.51.self_attn.k_proj.weight": "model-00020-of-00031.safetensors",
579
+ "model.layers.51.self_attn.o_proj.weight": "model-00020-of-00031.safetensors",
580
+ "model.layers.51.self_attn.q_proj.bias": "model-00020-of-00031.safetensors",
581
+ "model.layers.51.self_attn.q_proj.weight": "model-00020-of-00031.safetensors",
582
+ "model.layers.51.self_attn.v_proj.bias": "model-00020-of-00031.safetensors",
583
+ "model.layers.51.self_attn.v_proj.weight": "model-00020-of-00031.safetensors",
584
+ "model.layers.52.input_layernorm.weight": "model-00020-of-00031.safetensors",
585
+ "model.layers.52.mlp.down_proj.weight": "model-00020-of-00031.safetensors",
586
+ "model.layers.52.mlp.gate_proj.weight": "model-00020-of-00031.safetensors",
587
+ "model.layers.52.mlp.up_proj.weight": "model-00020-of-00031.safetensors",
588
+ "model.layers.52.post_attention_layernorm.weight": "model-00020-of-00031.safetensors",
589
+ "model.layers.52.self_attn.k_proj.bias": "model-00020-of-00031.safetensors",
590
+ "model.layers.52.self_attn.k_proj.weight": "model-00020-of-00031.safetensors",
591
+ "model.layers.52.self_attn.o_proj.weight": "model-00020-of-00031.safetensors",
592
+ "model.layers.52.self_attn.q_proj.bias": "model-00020-of-00031.safetensors",
593
+ "model.layers.52.self_attn.q_proj.weight": "model-00020-of-00031.safetensors",
594
+ "model.layers.52.self_attn.v_proj.bias": "model-00020-of-00031.safetensors",
595
+ "model.layers.52.self_attn.v_proj.weight": "model-00020-of-00031.safetensors",
596
+ "model.layers.53.input_layernorm.weight": "model-00021-of-00031.safetensors",
597
+ "model.layers.53.mlp.down_proj.weight": "model-00021-of-00031.safetensors",
598
+ "model.layers.53.mlp.gate_proj.weight": "model-00020-of-00031.safetensors",
599
+ "model.layers.53.mlp.up_proj.weight": "model-00021-of-00031.safetensors",
600
+ "model.layers.53.post_attention_layernorm.weight": "model-00021-of-00031.safetensors",
601
+ "model.layers.53.self_attn.k_proj.bias": "model-00020-of-00031.safetensors",
602
+ "model.layers.53.self_attn.k_proj.weight": "model-00020-of-00031.safetensors",
603
+ "model.layers.53.self_attn.o_proj.weight": "model-00020-of-00031.safetensors",
604
+ "model.layers.53.self_attn.q_proj.bias": "model-00020-of-00031.safetensors",
605
+ "model.layers.53.self_attn.q_proj.weight": "model-00020-of-00031.safetensors",
606
+ "model.layers.53.self_attn.v_proj.bias": "model-00020-of-00031.safetensors",
607
+ "model.layers.53.self_attn.v_proj.weight": "model-00020-of-00031.safetensors",
608
+ "model.layers.54.input_layernorm.weight": "model-00021-of-00031.safetensors",
609
+ "model.layers.54.mlp.down_proj.weight": "model-00021-of-00031.safetensors",
610
+ "model.layers.54.mlp.gate_proj.weight": "model-00021-of-00031.safetensors",
611
+ "model.layers.54.mlp.up_proj.weight": "model-00021-of-00031.safetensors",
612
+ "model.layers.54.post_attention_layernorm.weight": "model-00021-of-00031.safetensors",
613
+ "model.layers.54.self_attn.k_proj.bias": "model-00021-of-00031.safetensors",
614
+ "model.layers.54.self_attn.k_proj.weight": "model-00021-of-00031.safetensors",
615
+ "model.layers.54.self_attn.o_proj.weight": "model-00021-of-00031.safetensors",
616
+ "model.layers.54.self_attn.q_proj.bias": "model-00021-of-00031.safetensors",
617
+ "model.layers.54.self_attn.q_proj.weight": "model-00021-of-00031.safetensors",
618
+ "model.layers.54.self_attn.v_proj.bias": "model-00021-of-00031.safetensors",
619
+ "model.layers.54.self_attn.v_proj.weight": "model-00021-of-00031.safetensors",
620
+ "model.layers.55.input_layernorm.weight": "model-00021-of-00031.safetensors",
621
+ "model.layers.55.mlp.down_proj.weight": "model-00021-of-00031.safetensors",
622
+ "model.layers.55.mlp.gate_proj.weight": "model-00021-of-00031.safetensors",
623
+ "model.layers.55.mlp.up_proj.weight": "model-00021-of-00031.safetensors",
624
+ "model.layers.55.post_attention_layernorm.weight": "model-00021-of-00031.safetensors",
625
+ "model.layers.55.self_attn.k_proj.bias": "model-00021-of-00031.safetensors",
626
+ "model.layers.55.self_attn.k_proj.weight": "model-00021-of-00031.safetensors",
627
+ "model.layers.55.self_attn.o_proj.weight": "model-00021-of-00031.safetensors",
628
+ "model.layers.55.self_attn.q_proj.bias": "model-00021-of-00031.safetensors",
629
+ "model.layers.55.self_attn.q_proj.weight": "model-00021-of-00031.safetensors",
630
+ "model.layers.55.self_attn.v_proj.bias": "model-00021-of-00031.safetensors",
631
+ "model.layers.55.self_attn.v_proj.weight": "model-00021-of-00031.safetensors",
632
+ "model.layers.56.input_layernorm.weight": "model-00022-of-00031.safetensors",
633
+ "model.layers.56.mlp.down_proj.weight": "model-00022-of-00031.safetensors",
634
+ "model.layers.56.mlp.gate_proj.weight": "model-00022-of-00031.safetensors",
635
+ "model.layers.56.mlp.up_proj.weight": "model-00022-of-00031.safetensors",
636
+ "model.layers.56.post_attention_layernorm.weight": "model-00022-of-00031.safetensors",
637
+ "model.layers.56.self_attn.k_proj.bias": "model-00021-of-00031.safetensors",
638
+ "model.layers.56.self_attn.k_proj.weight": "model-00021-of-00031.safetensors",
639
+ "model.layers.56.self_attn.o_proj.weight": "model-00021-of-00031.safetensors",
640
+ "model.layers.56.self_attn.q_proj.bias": "model-00021-of-00031.safetensors",
641
+ "model.layers.56.self_attn.q_proj.weight": "model-00021-of-00031.safetensors",
642
+ "model.layers.56.self_attn.v_proj.bias": "model-00021-of-00031.safetensors",
643
+ "model.layers.56.self_attn.v_proj.weight": "model-00021-of-00031.safetensors",
644
+ "model.layers.57.input_layernorm.weight": "model-00022-of-00031.safetensors",
645
+ "model.layers.57.mlp.down_proj.weight": "model-00022-of-00031.safetensors",
646
+ "model.layers.57.mlp.gate_proj.weight": "model-00022-of-00031.safetensors",
647
+ "model.layers.57.mlp.up_proj.weight": "model-00022-of-00031.safetensors",
648
+ "model.layers.57.post_attention_layernorm.weight": "model-00022-of-00031.safetensors",
649
+ "model.layers.57.self_attn.k_proj.bias": "model-00022-of-00031.safetensors",
650
+ "model.layers.57.self_attn.k_proj.weight": "model-00022-of-00031.safetensors",
651
+ "model.layers.57.self_attn.o_proj.weight": "model-00022-of-00031.safetensors",
652
+ "model.layers.57.self_attn.q_proj.bias": "model-00022-of-00031.safetensors",
653
+ "model.layers.57.self_attn.q_proj.weight": "model-00022-of-00031.safetensors",
654
+ "model.layers.57.self_attn.v_proj.bias": "model-00022-of-00031.safetensors",
655
+ "model.layers.57.self_attn.v_proj.weight": "model-00022-of-00031.safetensors",
656
+ "model.layers.58.input_layernorm.weight": "model-00022-of-00031.safetensors",
657
+ "model.layers.58.mlp.down_proj.weight": "model-00022-of-00031.safetensors",
658
+ "model.layers.58.mlp.gate_proj.weight": "model-00022-of-00031.safetensors",
659
+ "model.layers.58.mlp.up_proj.weight": "model-00022-of-00031.safetensors",
660
+ "model.layers.58.post_attention_layernorm.weight": "model-00022-of-00031.safetensors",
661
+ "model.layers.58.self_attn.k_proj.bias": "model-00022-of-00031.safetensors",
662
+ "model.layers.58.self_attn.k_proj.weight": "model-00022-of-00031.safetensors",
663
+ "model.layers.58.self_attn.o_proj.weight": "model-00022-of-00031.safetensors",
664
+ "model.layers.58.self_attn.q_proj.bias": "model-00022-of-00031.safetensors",
665
+ "model.layers.58.self_attn.q_proj.weight": "model-00022-of-00031.safetensors",
666
+ "model.layers.58.self_attn.v_proj.bias": "model-00022-of-00031.safetensors",
667
+ "model.layers.58.self_attn.v_proj.weight": "model-00022-of-00031.safetensors",
668
+ "model.layers.59.input_layernorm.weight": "model-00023-of-00031.safetensors",
669
+ "model.layers.59.mlp.down_proj.weight": "model-00023-of-00031.safetensors",
670
+ "model.layers.59.mlp.gate_proj.weight": "model-00023-of-00031.safetensors",
671
+ "model.layers.59.mlp.up_proj.weight": "model-00023-of-00031.safetensors",
672
+ "model.layers.59.post_attention_layernorm.weight": "model-00023-of-00031.safetensors",
673
+ "model.layers.59.self_attn.k_proj.bias": "model-00023-of-00031.safetensors",
674
+ "model.layers.59.self_attn.k_proj.weight": "model-00023-of-00031.safetensors",
675
+ "model.layers.59.self_attn.o_proj.weight": "model-00023-of-00031.safetensors",
676
+ "model.layers.59.self_attn.q_proj.bias": "model-00023-of-00031.safetensors",
677
+ "model.layers.59.self_attn.q_proj.weight": "model-00023-of-00031.safetensors",
678
+ "model.layers.59.self_attn.v_proj.bias": "model-00023-of-00031.safetensors",
679
+ "model.layers.59.self_attn.v_proj.weight": "model-00023-of-00031.safetensors",
680
+ "model.layers.6.input_layernorm.weight": "model-00004-of-00031.safetensors",
681
+ "model.layers.6.mlp.down_proj.weight": "model-00004-of-00031.safetensors",
682
+ "model.layers.6.mlp.gate_proj.weight": "model-00003-of-00031.safetensors",
683
+ "model.layers.6.mlp.up_proj.weight": "model-00003-of-00031.safetensors",
684
+ "model.layers.6.post_attention_layernorm.weight": "model-00004-of-00031.safetensors",
685
+ "model.layers.6.self_attn.k_proj.bias": "model-00003-of-00031.safetensors",
686
+ "model.layers.6.self_attn.k_proj.weight": "model-00003-of-00031.safetensors",
687
+ "model.layers.6.self_attn.o_proj.weight": "model-00003-of-00031.safetensors",
688
+ "model.layers.6.self_attn.q_proj.bias": "model-00003-of-00031.safetensors",
689
+ "model.layers.6.self_attn.q_proj.weight": "model-00003-of-00031.safetensors",
690
+ "model.layers.6.self_attn.v_proj.bias": "model-00003-of-00031.safetensors",
691
+ "model.layers.6.self_attn.v_proj.weight": "model-00003-of-00031.safetensors",
692
+ "model.layers.60.input_layernorm.weight": "model-00023-of-00031.safetensors",
693
+ "model.layers.60.mlp.down_proj.weight": "model-00023-of-00031.safetensors",
694
+ "model.layers.60.mlp.gate_proj.weight": "model-00023-of-00031.safetensors",
695
+ "model.layers.60.mlp.up_proj.weight": "model-00023-of-00031.safetensors",
696
+ "model.layers.60.post_attention_layernorm.weight": "model-00023-of-00031.safetensors",
697
+ "model.layers.60.self_attn.k_proj.bias": "model-00023-of-00031.safetensors",
698
+ "model.layers.60.self_attn.k_proj.weight": "model-00023-of-00031.safetensors",
699
+ "model.layers.60.self_attn.o_proj.weight": "model-00023-of-00031.safetensors",
700
+ "model.layers.60.self_attn.q_proj.bias": "model-00023-of-00031.safetensors",
701
+ "model.layers.60.self_attn.q_proj.weight": "model-00023-of-00031.safetensors",
702
+ "model.layers.60.self_attn.v_proj.bias": "model-00023-of-00031.safetensors",
703
+ "model.layers.60.self_attn.v_proj.weight": "model-00023-of-00031.safetensors",
704
+ "model.layers.61.input_layernorm.weight": "model-00024-of-00031.safetensors",
705
+ "model.layers.61.mlp.down_proj.weight": "model-00024-of-00031.safetensors",
706
+ "model.layers.61.mlp.gate_proj.weight": "model-00023-of-00031.safetensors",
707
+ "model.layers.61.mlp.up_proj.weight": "model-00023-of-00031.safetensors",
708
+ "model.layers.61.post_attention_layernorm.weight": "model-00024-of-00031.safetensors",
709
+ "model.layers.61.self_attn.k_proj.bias": "model-00023-of-00031.safetensors",
710
+ "model.layers.61.self_attn.k_proj.weight": "model-00023-of-00031.safetensors",
711
+ "model.layers.61.self_attn.o_proj.weight": "model-00023-of-00031.safetensors",
712
+ "model.layers.61.self_attn.q_proj.bias": "model-00023-of-00031.safetensors",
713
+ "model.layers.61.self_attn.q_proj.weight": "model-00023-of-00031.safetensors",
714
+ "model.layers.61.self_attn.v_proj.bias": "model-00023-of-00031.safetensors",
715
+ "model.layers.61.self_attn.v_proj.weight": "model-00023-of-00031.safetensors",
716
+ "model.layers.62.input_layernorm.weight": "model-00024-of-00031.safetensors",
717
+ "model.layers.62.mlp.down_proj.weight": "model-00024-of-00031.safetensors",
718
+ "model.layers.62.mlp.gate_proj.weight": "model-00024-of-00031.safetensors",
719
+ "model.layers.62.mlp.up_proj.weight": "model-00024-of-00031.safetensors",
720
+ "model.layers.62.post_attention_layernorm.weight": "model-00024-of-00031.safetensors",
721
+ "model.layers.62.self_attn.k_proj.bias": "model-00024-of-00031.safetensors",
722
+ "model.layers.62.self_attn.k_proj.weight": "model-00024-of-00031.safetensors",
723
+ "model.layers.62.self_attn.o_proj.weight": "model-00024-of-00031.safetensors",
724
+ "model.layers.62.self_attn.q_proj.bias": "model-00024-of-00031.safetensors",
725
+ "model.layers.62.self_attn.q_proj.weight": "model-00024-of-00031.safetensors",
726
+ "model.layers.62.self_attn.v_proj.bias": "model-00024-of-00031.safetensors",
727
+ "model.layers.62.self_attn.v_proj.weight": "model-00024-of-00031.safetensors",
728
+ "model.layers.63.input_layernorm.weight": "model-00024-of-00031.safetensors",
729
+ "model.layers.63.mlp.down_proj.weight": "model-00024-of-00031.safetensors",
730
+ "model.layers.63.mlp.gate_proj.weight": "model-00024-of-00031.safetensors",
731
+ "model.layers.63.mlp.up_proj.weight": "model-00024-of-00031.safetensors",
732
+ "model.layers.63.post_attention_layernorm.weight": "model-00024-of-00031.safetensors",
733
+ "model.layers.63.self_attn.k_proj.bias": "model-00024-of-00031.safetensors",
734
+ "model.layers.63.self_attn.k_proj.weight": "model-00024-of-00031.safetensors",
735
+ "model.layers.63.self_attn.o_proj.weight": "model-00024-of-00031.safetensors",
736
+ "model.layers.63.self_attn.q_proj.bias": "model-00024-of-00031.safetensors",
737
+ "model.layers.63.self_attn.q_proj.weight": "model-00024-of-00031.safetensors",
738
+ "model.layers.63.self_attn.v_proj.bias": "model-00024-of-00031.safetensors",
739
+ "model.layers.63.self_attn.v_proj.weight": "model-00024-of-00031.safetensors",
740
+ "model.layers.64.input_layernorm.weight": "model-00025-of-00031.safetensors",
741
+ "model.layers.64.mlp.down_proj.weight": "model-00025-of-00031.safetensors",
742
+ "model.layers.64.mlp.gate_proj.weight": "model-00024-of-00031.safetensors",
743
+ "model.layers.64.mlp.up_proj.weight": "model-00025-of-00031.safetensors",
744
+ "model.layers.64.post_attention_layernorm.weight": "model-00025-of-00031.safetensors",
745
+ "model.layers.64.self_attn.k_proj.bias": "model-00024-of-00031.safetensors",
746
+ "model.layers.64.self_attn.k_proj.weight": "model-00024-of-00031.safetensors",
747
+ "model.layers.64.self_attn.o_proj.weight": "model-00024-of-00031.safetensors",
748
+ "model.layers.64.self_attn.q_proj.bias": "model-00024-of-00031.safetensors",
749
+ "model.layers.64.self_attn.q_proj.weight": "model-00024-of-00031.safetensors",
750
+ "model.layers.64.self_attn.v_proj.bias": "model-00024-of-00031.safetensors",
751
+ "model.layers.64.self_attn.v_proj.weight": "model-00024-of-00031.safetensors",
752
+ "model.layers.65.input_layernorm.weight": "model-00025-of-00031.safetensors",
753
+ "model.layers.65.mlp.down_proj.weight": "model-00025-of-00031.safetensors",
754
+ "model.layers.65.mlp.gate_proj.weight": "model-00025-of-00031.safetensors",
755
+ "model.layers.65.mlp.up_proj.weight": "model-00025-of-00031.safetensors",
756
+ "model.layers.65.post_attention_layernorm.weight": "model-00025-of-00031.safetensors",
757
+ "model.layers.65.self_attn.k_proj.bias": "model-00025-of-00031.safetensors",
758
+ "model.layers.65.self_attn.k_proj.weight": "model-00025-of-00031.safetensors",
759
+ "model.layers.65.self_attn.o_proj.weight": "model-00025-of-00031.safetensors",
760
+ "model.layers.65.self_attn.q_proj.bias": "model-00025-of-00031.safetensors",
761
+ "model.layers.65.self_attn.q_proj.weight": "model-00025-of-00031.safetensors",
762
+ "model.layers.65.self_attn.v_proj.bias": "model-00025-of-00031.safetensors",
763
+ "model.layers.65.self_attn.v_proj.weight": "model-00025-of-00031.safetensors",
764
+ "model.layers.66.input_layernorm.weight": "model-00025-of-00031.safetensors",
765
+ "model.layers.66.mlp.down_proj.weight": "model-00025-of-00031.safetensors",
766
+ "model.layers.66.mlp.gate_proj.weight": "model-00025-of-00031.safetensors",
767
+ "model.layers.66.mlp.up_proj.weight": "model-00025-of-00031.safetensors",
768
+ "model.layers.66.post_attention_layernorm.weight": "model-00025-of-00031.safetensors",
769
+ "model.layers.66.self_attn.k_proj.bias": "model-00025-of-00031.safetensors",
770
+ "model.layers.66.self_attn.k_proj.weight": "model-00025-of-00031.safetensors",
771
+ "model.layers.66.self_attn.o_proj.weight": "model-00025-of-00031.safetensors",
772
+ "model.layers.66.self_attn.q_proj.bias": "model-00025-of-00031.safetensors",
773
+ "model.layers.66.self_attn.q_proj.weight": "model-00025-of-00031.safetensors",
774
+ "model.layers.66.self_attn.v_proj.bias": "model-00025-of-00031.safetensors",
775
+ "model.layers.66.self_attn.v_proj.weight": "model-00025-of-00031.safetensors",
776
+ "model.layers.67.input_layernorm.weight": "model-00026-of-00031.safetensors",
777
+ "model.layers.67.mlp.down_proj.weight": "model-00026-of-00031.safetensors",
778
+ "model.layers.67.mlp.gate_proj.weight": "model-00026-of-00031.safetensors",
779
+ "model.layers.67.mlp.up_proj.weight": "model-00026-of-00031.safetensors",
780
+ "model.layers.67.post_attention_layernorm.weight": "model-00026-of-00031.safetensors",
781
+ "model.layers.67.self_attn.k_proj.bias": "model-00025-of-00031.safetensors",
782
+ "model.layers.67.self_attn.k_proj.weight": "model-00025-of-00031.safetensors",
783
+ "model.layers.67.self_attn.o_proj.weight": "model-00025-of-00031.safetensors",
784
+ "model.layers.67.self_attn.q_proj.bias": "model-00025-of-00031.safetensors",
785
+ "model.layers.67.self_attn.q_proj.weight": "model-00025-of-00031.safetensors",
786
+ "model.layers.67.self_attn.v_proj.bias": "model-00025-of-00031.safetensors",
787
+ "model.layers.67.self_attn.v_proj.weight": "model-00025-of-00031.safetensors",
788
+ "model.layers.68.input_layernorm.weight": "model-00026-of-00031.safetensors",
789
+ "model.layers.68.mlp.down_proj.weight": "model-00026-of-00031.safetensors",
790
+ "model.layers.68.mlp.gate_proj.weight": "model-00026-of-00031.safetensors",
791
+ "model.layers.68.mlp.up_proj.weight": "model-00026-of-00031.safetensors",
792
+ "model.layers.68.post_attention_layernorm.weight": "model-00026-of-00031.safetensors",
793
+ "model.layers.68.self_attn.k_proj.bias": "model-00026-of-00031.safetensors",
794
+ "model.layers.68.self_attn.k_proj.weight": "model-00026-of-00031.safetensors",
795
+ "model.layers.68.self_attn.o_proj.weight": "model-00026-of-00031.safetensors",
796
+ "model.layers.68.self_attn.q_proj.bias": "model-00026-of-00031.safetensors",
797
+ "model.layers.68.self_attn.q_proj.weight": "model-00026-of-00031.safetensors",
798
+ "model.layers.68.self_attn.v_proj.bias": "model-00026-of-00031.safetensors",
799
+ "model.layers.68.self_attn.v_proj.weight": "model-00026-of-00031.safetensors",
800
+ "model.layers.69.input_layernorm.weight": "model-00026-of-00031.safetensors",
801
+ "model.layers.69.mlp.down_proj.weight": "model-00026-of-00031.safetensors",
802
+ "model.layers.69.mlp.gate_proj.weight": "model-00026-of-00031.safetensors",
803
+ "model.layers.69.mlp.up_proj.weight": "model-00026-of-00031.safetensors",
804
+ "model.layers.69.post_attention_layernorm.weight": "model-00026-of-00031.safetensors",
805
+ "model.layers.69.self_attn.k_proj.bias": "model-00026-of-00031.safetensors",
806
+ "model.layers.69.self_attn.k_proj.weight": "model-00026-of-00031.safetensors",
807
+ "model.layers.69.self_attn.o_proj.weight": "model-00026-of-00031.safetensors",
808
+ "model.layers.69.self_attn.q_proj.bias": "model-00026-of-00031.safetensors",
809
+ "model.layers.69.self_attn.q_proj.weight": "model-00026-of-00031.safetensors",
810
+ "model.layers.69.self_attn.v_proj.bias": "model-00026-of-00031.safetensors",
811
+ "model.layers.69.self_attn.v_proj.weight": "model-00026-of-00031.safetensors",
812
+ "model.layers.7.input_layernorm.weight": "model-00004-of-00031.safetensors",
813
+ "model.layers.7.mlp.down_proj.weight": "model-00004-of-00031.safetensors",
814
+ "model.layers.7.mlp.gate_proj.weight": "model-00004-of-00031.safetensors",
815
+ "model.layers.7.mlp.up_proj.weight": "model-00004-of-00031.safetensors",
816
+ "model.layers.7.post_attention_layernorm.weight": "model-00004-of-00031.safetensors",
817
+ "model.layers.7.self_attn.k_proj.bias": "model-00004-of-00031.safetensors",
818
+ "model.layers.7.self_attn.k_proj.weight": "model-00004-of-00031.safetensors",
819
+ "model.layers.7.self_attn.o_proj.weight": "model-00004-of-00031.safetensors",
820
+ "model.layers.7.self_attn.q_proj.bias": "model-00004-of-00031.safetensors",
821
+ "model.layers.7.self_attn.q_proj.weight": "model-00004-of-00031.safetensors",
822
+ "model.layers.7.self_attn.v_proj.bias": "model-00004-of-00031.safetensors",
823
+ "model.layers.7.self_attn.v_proj.weight": "model-00004-of-00031.safetensors",
824
+ "model.layers.70.input_layernorm.weight": "model-00027-of-00031.safetensors",
825
+ "model.layers.70.mlp.down_proj.weight": "model-00027-of-00031.safetensors",
826
+ "model.layers.70.mlp.gate_proj.weight": "model-00027-of-00031.safetensors",
827
+ "model.layers.70.mlp.up_proj.weight": "model-00027-of-00031.safetensors",
828
+ "model.layers.70.post_attention_layernorm.weight": "model-00027-of-00031.safetensors",
829
+ "model.layers.70.self_attn.k_proj.bias": "model-00027-of-00031.safetensors",
830
+ "model.layers.70.self_attn.k_proj.weight": "model-00027-of-00031.safetensors",
831
+ "model.layers.70.self_attn.o_proj.weight": "model-00027-of-00031.safetensors",
832
+ "model.layers.70.self_attn.q_proj.bias": "model-00027-of-00031.safetensors",
833
+ "model.layers.70.self_attn.q_proj.weight": "model-00027-of-00031.safetensors",
834
+ "model.layers.70.self_attn.v_proj.bias": "model-00027-of-00031.safetensors",
835
+ "model.layers.70.self_attn.v_proj.weight": "model-00027-of-00031.safetensors",
836
+ "model.layers.71.input_layernorm.weight": "model-00027-of-00031.safetensors",
837
+ "model.layers.71.mlp.down_proj.weight": "model-00027-of-00031.safetensors",
838
+ "model.layers.71.mlp.gate_proj.weight": "model-00027-of-00031.safetensors",
839
+ "model.layers.71.mlp.up_proj.weight": "model-00027-of-00031.safetensors",
840
+ "model.layers.71.post_attention_layernorm.weight": "model-00027-of-00031.safetensors",
841
+ "model.layers.71.self_attn.k_proj.bias": "model-00027-of-00031.safetensors",
842
+ "model.layers.71.self_attn.k_proj.weight": "model-00027-of-00031.safetensors",
843
+ "model.layers.71.self_attn.o_proj.weight": "model-00027-of-00031.safetensors",
844
+ "model.layers.71.self_attn.q_proj.bias": "model-00027-of-00031.safetensors",
845
+ "model.layers.71.self_attn.q_proj.weight": "model-00027-of-00031.safetensors",
846
+ "model.layers.71.self_attn.v_proj.bias": "model-00027-of-00031.safetensors",
847
+ "model.layers.71.self_attn.v_proj.weight": "model-00027-of-00031.safetensors",
848
+ "model.layers.72.input_layernorm.weight": "model-00028-of-00031.safetensors",
849
+ "model.layers.72.mlp.down_proj.weight": "model-00028-of-00031.safetensors",
850
+ "model.layers.72.mlp.gate_proj.weight": "model-00027-of-00031.safetensors",
851
+ "model.layers.72.mlp.up_proj.weight": "model-00027-of-00031.safetensors",
852
+ "model.layers.72.post_attention_layernorm.weight": "model-00028-of-00031.safetensors",
853
+ "model.layers.72.self_attn.k_proj.bias": "model-00027-of-00031.safetensors",
854
+ "model.layers.72.self_attn.k_proj.weight": "model-00027-of-00031.safetensors",
855
+ "model.layers.72.self_attn.o_proj.weight": "model-00027-of-00031.safetensors",
856
+ "model.layers.72.self_attn.q_proj.bias": "model-00027-of-00031.safetensors",
857
+ "model.layers.72.self_attn.q_proj.weight": "model-00027-of-00031.safetensors",
858
+ "model.layers.72.self_attn.v_proj.bias": "model-00027-of-00031.safetensors",
859
+ "model.layers.72.self_attn.v_proj.weight": "model-00027-of-00031.safetensors",
860
+ "model.layers.73.input_layernorm.weight": "model-00028-of-00031.safetensors",
861
+ "model.layers.73.mlp.down_proj.weight": "model-00028-of-00031.safetensors",
862
+ "model.layers.73.mlp.gate_proj.weight": "model-00028-of-00031.safetensors",
863
+ "model.layers.73.mlp.up_proj.weight": "model-00028-of-00031.safetensors",
864
+ "model.layers.73.post_attention_layernorm.weight": "model-00028-of-00031.safetensors",
865
+ "model.layers.73.self_attn.k_proj.bias": "model-00028-of-00031.safetensors",
866
+ "model.layers.73.self_attn.k_proj.weight": "model-00028-of-00031.safetensors",
867
+ "model.layers.73.self_attn.o_proj.weight": "model-00028-of-00031.safetensors",
868
+ "model.layers.73.self_attn.q_proj.bias": "model-00028-of-00031.safetensors",
869
+ "model.layers.73.self_attn.q_proj.weight": "model-00028-of-00031.safetensors",
870
+ "model.layers.73.self_attn.v_proj.bias": "model-00028-of-00031.safetensors",
871
+ "model.layers.73.self_attn.v_proj.weight": "model-00028-of-00031.safetensors",
872
+ "model.layers.74.input_layernorm.weight": "model-00028-of-00031.safetensors",
873
+ "model.layers.74.mlp.down_proj.weight": "model-00028-of-00031.safetensors",
874
+ "model.layers.74.mlp.gate_proj.weight": "model-00028-of-00031.safetensors",
875
+ "model.layers.74.mlp.up_proj.weight": "model-00028-of-00031.safetensors",
876
+ "model.layers.74.post_attention_layernorm.weight": "model-00028-of-00031.safetensors",
877
+ "model.layers.74.self_attn.k_proj.bias": "model-00028-of-00031.safetensors",
878
+ "model.layers.74.self_attn.k_proj.weight": "model-00028-of-00031.safetensors",
879
+ "model.layers.74.self_attn.o_proj.weight": "model-00028-of-00031.safetensors",
880
+ "model.layers.74.self_attn.q_proj.bias": "model-00028-of-00031.safetensors",
881
+ "model.layers.74.self_attn.q_proj.weight": "model-00028-of-00031.safetensors",
882
+ "model.layers.74.self_attn.v_proj.bias": "model-00028-of-00031.safetensors",
883
+ "model.layers.74.self_attn.v_proj.weight": "model-00028-of-00031.safetensors",
884
+ "model.layers.75.input_layernorm.weight": "model-00029-of-00031.safetensors",
885
+ "model.layers.75.mlp.down_proj.weight": "model-00029-of-00031.safetensors",
886
+ "model.layers.75.mlp.gate_proj.weight": "model-00028-of-00031.safetensors",
887
+ "model.layers.75.mlp.up_proj.weight": "model-00029-of-00031.safetensors",
888
+ "model.layers.75.post_attention_layernorm.weight": "model-00029-of-00031.safetensors",
889
+ "model.layers.75.self_attn.k_proj.bias": "model-00028-of-00031.safetensors",
890
+ "model.layers.75.self_attn.k_proj.weight": "model-00028-of-00031.safetensors",
891
+ "model.layers.75.self_attn.o_proj.weight": "model-00028-of-00031.safetensors",
892
+ "model.layers.75.self_attn.q_proj.bias": "model-00028-of-00031.safetensors",
893
+ "model.layers.75.self_attn.q_proj.weight": "model-00028-of-00031.safetensors",
894
+ "model.layers.75.self_attn.v_proj.bias": "model-00028-of-00031.safetensors",
895
+ "model.layers.75.self_attn.v_proj.weight": "model-00028-of-00031.safetensors",
896
+ "model.layers.76.input_layernorm.weight": "model-00029-of-00031.safetensors",
897
+ "model.layers.76.mlp.down_proj.weight": "model-00029-of-00031.safetensors",
898
+ "model.layers.76.mlp.gate_proj.weight": "model-00029-of-00031.safetensors",
899
+ "model.layers.76.mlp.up_proj.weight": "model-00029-of-00031.safetensors",
900
+ "model.layers.76.post_attention_layernorm.weight": "model-00029-of-00031.safetensors",
901
+ "model.layers.76.self_attn.k_proj.bias": "model-00029-of-00031.safetensors",
902
+ "model.layers.76.self_attn.k_proj.weight": "model-00029-of-00031.safetensors",
903
+ "model.layers.76.self_attn.o_proj.weight": "model-00029-of-00031.safetensors",
904
+ "model.layers.76.self_attn.q_proj.bias": "model-00029-of-00031.safetensors",
905
+ "model.layers.76.self_attn.q_proj.weight": "model-00029-of-00031.safetensors",
906
+ "model.layers.76.self_attn.v_proj.bias": "model-00029-of-00031.safetensors",
907
+ "model.layers.76.self_attn.v_proj.weight": "model-00029-of-00031.safetensors",
908
+ "model.layers.77.input_layernorm.weight": "model-00029-of-00031.safetensors",
909
+ "model.layers.77.mlp.down_proj.weight": "model-00029-of-00031.safetensors",
910
+ "model.layers.77.mlp.gate_proj.weight": "model-00029-of-00031.safetensors",
911
+ "model.layers.77.mlp.up_proj.weight": "model-00029-of-00031.safetensors",
912
+ "model.layers.77.post_attention_layernorm.weight": "model-00029-of-00031.safetensors",
913
+ "model.layers.77.self_attn.k_proj.bias": "model-00029-of-00031.safetensors",
914
+ "model.layers.77.self_attn.k_proj.weight": "model-00029-of-00031.safetensors",
915
+ "model.layers.77.self_attn.o_proj.weight": "model-00029-of-00031.safetensors",
916
+ "model.layers.77.self_attn.q_proj.bias": "model-00029-of-00031.safetensors",
917
+ "model.layers.77.self_attn.q_proj.weight": "model-00029-of-00031.safetensors",
918
+ "model.layers.77.self_attn.v_proj.bias": "model-00029-of-00031.safetensors",
919
+ "model.layers.77.self_attn.v_proj.weight": "model-00029-of-00031.safetensors",
920
+ "model.layers.78.input_layernorm.weight": "model-00030-of-00031.safetensors",
921
+ "model.layers.78.mlp.down_proj.weight": "model-00030-of-00031.safetensors",
922
+ "model.layers.78.mlp.gate_proj.weight": "model-00030-of-00031.safetensors",
923
+ "model.layers.78.mlp.up_proj.weight": "model-00030-of-00031.safetensors",
924
+ "model.layers.78.post_attention_layernorm.weight": "model-00030-of-00031.safetensors",
925
+ "model.layers.78.self_attn.k_proj.bias": "model-00029-of-00031.safetensors",
926
+ "model.layers.78.self_attn.k_proj.weight": "model-00029-of-00031.safetensors",
927
+ "model.layers.78.self_attn.o_proj.weight": "model-00029-of-00031.safetensors",
928
+ "model.layers.78.self_attn.q_proj.bias": "model-00029-of-00031.safetensors",
929
+ "model.layers.78.self_attn.q_proj.weight": "model-00029-of-00031.safetensors",
930
+ "model.layers.78.self_attn.v_proj.bias": "model-00029-of-00031.safetensors",
931
+ "model.layers.78.self_attn.v_proj.weight": "model-00029-of-00031.safetensors",
932
+ "model.layers.79.input_layernorm.weight": "model-00030-of-00031.safetensors",
933
+ "model.layers.79.mlp.down_proj.weight": "model-00030-of-00031.safetensors",
934
+ "model.layers.79.mlp.gate_proj.weight": "model-00030-of-00031.safetensors",
935
+ "model.layers.79.mlp.up_proj.weight": "model-00030-of-00031.safetensors",
936
+ "model.layers.79.post_attention_layernorm.weight": "model-00030-of-00031.safetensors",
937
+ "model.layers.79.self_attn.k_proj.bias": "model-00030-of-00031.safetensors",
938
+ "model.layers.79.self_attn.k_proj.weight": "model-00030-of-00031.safetensors",
939
+ "model.layers.79.self_attn.o_proj.weight": "model-00030-of-00031.safetensors",
940
+ "model.layers.79.self_attn.q_proj.bias": "model-00030-of-00031.safetensors",
941
+ "model.layers.79.self_attn.q_proj.weight": "model-00030-of-00031.safetensors",
942
+ "model.layers.79.self_attn.v_proj.bias": "model-00030-of-00031.safetensors",
943
+ "model.layers.79.self_attn.v_proj.weight": "model-00030-of-00031.safetensors",
944
+ "model.layers.8.input_layernorm.weight": "model-00004-of-00031.safetensors",
945
+ "model.layers.8.mlp.down_proj.weight": "model-00004-of-00031.safetensors",
946
+ "model.layers.8.mlp.gate_proj.weight": "model-00004-of-00031.safetensors",
947
+ "model.layers.8.mlp.up_proj.weight": "model-00004-of-00031.safetensors",
948
+ "model.layers.8.post_attention_layernorm.weight": "model-00004-of-00031.safetensors",
949
+ "model.layers.8.self_attn.k_proj.bias": "model-00004-of-00031.safetensors",
950
+ "model.layers.8.self_attn.k_proj.weight": "model-00004-of-00031.safetensors",
951
+ "model.layers.8.self_attn.o_proj.weight": "model-00004-of-00031.safetensors",
952
+ "model.layers.8.self_attn.q_proj.bias": "model-00004-of-00031.safetensors",
953
+ "model.layers.8.self_attn.q_proj.weight": "model-00004-of-00031.safetensors",
954
+ "model.layers.8.self_attn.v_proj.bias": "model-00004-of-00031.safetensors",
955
+ "model.layers.8.self_attn.v_proj.weight": "model-00004-of-00031.safetensors",
956
+ "model.layers.9.input_layernorm.weight": "model-00005-of-00031.safetensors",
957
+ "model.layers.9.mlp.down_proj.weight": "model-00005-of-00031.safetensors",
958
+ "model.layers.9.mlp.gate_proj.weight": "model-00004-of-00031.safetensors",
959
+ "model.layers.9.mlp.up_proj.weight": "model-00005-of-00031.safetensors",
960
+ "model.layers.9.post_attention_layernorm.weight": "model-00005-of-00031.safetensors",
961
+ "model.layers.9.self_attn.k_proj.bias": "model-00004-of-00031.safetensors",
962
+ "model.layers.9.self_attn.k_proj.weight": "model-00004-of-00031.safetensors",
963
+ "model.layers.9.self_attn.o_proj.weight": "model-00004-of-00031.safetensors",
964
+ "model.layers.9.self_attn.q_proj.bias": "model-00004-of-00031.safetensors",
965
+ "model.layers.9.self_attn.q_proj.weight": "model-00004-of-00031.safetensors",
966
+ "model.layers.9.self_attn.v_proj.bias": "model-00004-of-00031.safetensors",
967
+ "model.layers.9.self_attn.v_proj.weight": "model-00004-of-00031.safetensors",
968
+ "model.norm.weight": "model-00030-of-00031.safetensors"
969
+ }
970
+ }
special_tokens_map.json ADDED
@@ -0,0 +1,20 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "additional_special_tokens": [
3
+ "<|im_start|>",
4
+ "<|im_end|>"
5
+ ],
6
+ "eos_token": {
7
+ "content": "<|im_end|>",
8
+ "lstrip": false,
9
+ "normalized": false,
10
+ "rstrip": false,
11
+ "single_word": false
12
+ },
13
+ "pad_token": {
14
+ "content": "<|endoftext|>",
15
+ "lstrip": false,
16
+ "normalized": false,
17
+ "rstrip": false,
18
+ "single_word": false
19
+ }
20
+ }
tokenizer_config.json ADDED
@@ -0,0 +1,44 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_prefix_space": false,
3
+ "added_tokens_decoder": {
4
+ "151643": {
5
+ "content": "<|endoftext|>",
6
+ "lstrip": false,
7
+ "normalized": false,
8
+ "rstrip": false,
9
+ "single_word": false,
10
+ "special": true
11
+ },
12
+ "151644": {
13
+ "content": "<|im_start|>",
14
+ "lstrip": false,
15
+ "normalized": false,
16
+ "rstrip": false,
17
+ "single_word": false,
18
+ "special": true
19
+ },
20
+ "151645": {
21
+ "content": "<|im_end|>",
22
+ "lstrip": false,
23
+ "normalized": false,
24
+ "rstrip": false,
25
+ "single_word": false,
26
+ "special": true
27
+ }
28
+ },
29
+ "additional_special_tokens": [
30
+ "<|im_start|>",
31
+ "<|im_end|>"
32
+ ],
33
+ "bos_token": null,
34
+ "chat_template": "{% set system_message = 'You are a helpful assistant.' %}{% if messages[0]['role'] == 'system' %}{% set system_message = messages[0]['content'] %}{% endif %}{% if system_message is defined %}{{ '<|im_start|>system\\n' + system_message + '<|im_end|>\\n' }}{% endif %}{% for message in messages %}{% set content = message['content'] %}{% if message['role'] == 'user' %}{{ '<|im_start|>user\\n' + content + '<|im_end|>\\n<|im_start|>assistant\\n' }}{% elif message['role'] == 'assistant' %}{{ content + '<|im_end|>' + '\\n' }}{% endif %}{% endfor %}",
35
+ "clean_up_tokenization_spaces": false,
36
+ "eos_token": "<|im_end|>",
37
+ "errors": "replace",
38
+ "model_max_length": 131072,
39
+ "pad_token": "<|endoftext|>",
40
+ "padding_side": "right",
41
+ "split_special_tokens": false,
42
+ "tokenizer_class": "Qwen2Tokenizer",
43
+ "unk_token": null
44
+ }
trainer_state.json ADDED
@@ -0,0 +1,2604 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 1.0,
5
+ "eval_steps": 500,
6
+ "global_step": 3694,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.0027070925825663237,
13
+ "grad_norm": 2.7909258086527373,
14
+ "learning_rate": 4.999977397559719e-05,
15
+ "loss": 1.4602,
16
+ "step": 10
17
+ },
18
+ {
19
+ "epoch": 0.005414185165132647,
20
+ "grad_norm": 2.9246419242894404,
21
+ "learning_rate": 4.9999095906475714e-05,
22
+ "loss": 1.4843,
23
+ "step": 20
24
+ },
25
+ {
26
+ "epoch": 0.008121277747698972,
27
+ "grad_norm": 2.60296538082041,
28
+ "learning_rate": 4.999796580489639e-05,
29
+ "loss": 1.3953,
30
+ "step": 30
31
+ },
32
+ {
33
+ "epoch": 0.010828370330265295,
34
+ "grad_norm": 2.427882849139683,
35
+ "learning_rate": 4.9996383691293656e-05,
36
+ "loss": 1.3886,
37
+ "step": 40
38
+ },
39
+ {
40
+ "epoch": 0.01353546291283162,
41
+ "grad_norm": 2.362478801666027,
42
+ "learning_rate": 4.999434959427522e-05,
43
+ "loss": 1.4087,
44
+ "step": 50
45
+ },
46
+ {
47
+ "epoch": 0.016242555495397944,
48
+ "grad_norm": 2.0562878551243253,
49
+ "learning_rate": 4.999186355062154e-05,
50
+ "loss": 1.3925,
51
+ "step": 60
52
+ },
53
+ {
54
+ "epoch": 0.018949648077964266,
55
+ "grad_norm": 1.4997595683614129,
56
+ "learning_rate": 4.998892560528511e-05,
57
+ "loss": 1.4251,
58
+ "step": 70
59
+ },
60
+ {
61
+ "epoch": 0.02165674066053059,
62
+ "grad_norm": 1.8396372795029057,
63
+ "learning_rate": 4.9985535811389726e-05,
64
+ "loss": 1.4299,
65
+ "step": 80
66
+ },
67
+ {
68
+ "epoch": 0.024363833243096916,
69
+ "grad_norm": 3.016445354360887,
70
+ "learning_rate": 4.99816942302295e-05,
71
+ "loss": 1.4179,
72
+ "step": 90
73
+ },
74
+ {
75
+ "epoch": 0.02707092582566324,
76
+ "grad_norm": 2.285685249444506,
77
+ "learning_rate": 4.997740093126769e-05,
78
+ "loss": 1.4453,
79
+ "step": 100
80
+ },
81
+ {
82
+ "epoch": 0.02977801840822956,
83
+ "grad_norm": 2.3733951587674054,
84
+ "learning_rate": 4.997265599213554e-05,
85
+ "loss": 1.3905,
86
+ "step": 110
87
+ },
88
+ {
89
+ "epoch": 0.03248511099079589,
90
+ "grad_norm": 3.8476837651749336,
91
+ "learning_rate": 4.996745949863082e-05,
92
+ "loss": 1.4418,
93
+ "step": 120
94
+ },
95
+ {
96
+ "epoch": 0.03519220357336221,
97
+ "grad_norm": 1.6916145193283187,
98
+ "learning_rate": 4.996181154471625e-05,
99
+ "loss": 1.4637,
100
+ "step": 130
101
+ },
102
+ {
103
+ "epoch": 0.03789929615592853,
104
+ "grad_norm": 1.6551273153572528,
105
+ "learning_rate": 4.9955712232517896e-05,
106
+ "loss": 1.422,
107
+ "step": 140
108
+ },
109
+ {
110
+ "epoch": 0.040606388738494856,
111
+ "grad_norm": 1.842180664471275,
112
+ "learning_rate": 4.99491616723232e-05,
113
+ "loss": 1.4768,
114
+ "step": 150
115
+ },
116
+ {
117
+ "epoch": 0.04331348132106118,
118
+ "grad_norm": 2.8401045455632503,
119
+ "learning_rate": 4.994215998257909e-05,
120
+ "loss": 1.4626,
121
+ "step": 160
122
+ },
123
+ {
124
+ "epoch": 0.0460205739036275,
125
+ "grad_norm": 2.137479556495621,
126
+ "learning_rate": 4.993470728988979e-05,
127
+ "loss": 1.4147,
128
+ "step": 170
129
+ },
130
+ {
131
+ "epoch": 0.04872766648619383,
132
+ "grad_norm": 1.9073264544958397,
133
+ "learning_rate": 4.992680372901454e-05,
134
+ "loss": 1.4119,
135
+ "step": 180
136
+ },
137
+ {
138
+ "epoch": 0.051434759068760154,
139
+ "grad_norm": 1.5685293387125376,
140
+ "learning_rate": 4.9918449442865126e-05,
141
+ "loss": 1.428,
142
+ "step": 190
143
+ },
144
+ {
145
+ "epoch": 0.05414185165132648,
146
+ "grad_norm": 2.128149762022408,
147
+ "learning_rate": 4.9909644582503366e-05,
148
+ "loss": 1.4497,
149
+ "step": 200
150
+ },
151
+ {
152
+ "epoch": 0.0568489442338928,
153
+ "grad_norm": 1.8550357912320867,
154
+ "learning_rate": 4.990038930713833e-05,
155
+ "loss": 1.4661,
156
+ "step": 210
157
+ },
158
+ {
159
+ "epoch": 0.05955603681645912,
160
+ "grad_norm": 1.811618523801653,
161
+ "learning_rate": 4.989068378412346e-05,
162
+ "loss": 1.418,
163
+ "step": 220
164
+ },
165
+ {
166
+ "epoch": 0.062263129399025445,
167
+ "grad_norm": 1.691670012971929,
168
+ "learning_rate": 4.9880528188953556e-05,
169
+ "loss": 1.4446,
170
+ "step": 230
171
+ },
172
+ {
173
+ "epoch": 0.06497022198159177,
174
+ "grad_norm": 1.7432757441956248,
175
+ "learning_rate": 4.986992270526161e-05,
176
+ "loss": 1.3911,
177
+ "step": 240
178
+ },
179
+ {
180
+ "epoch": 0.0676773145641581,
181
+ "grad_norm": 1.709102114001383,
182
+ "learning_rate": 4.9858867524815456e-05,
183
+ "loss": 1.4912,
184
+ "step": 250
185
+ },
186
+ {
187
+ "epoch": 0.07038440714672442,
188
+ "grad_norm": 1.7869823709380956,
189
+ "learning_rate": 4.984736284751436e-05,
190
+ "loss": 1.454,
191
+ "step": 260
192
+ },
193
+ {
194
+ "epoch": 0.07309149972929074,
195
+ "grad_norm": 1.6458598808862888,
196
+ "learning_rate": 4.983540888138534e-05,
197
+ "loss": 1.4278,
198
+ "step": 270
199
+ },
200
+ {
201
+ "epoch": 0.07579859231185707,
202
+ "grad_norm": 2.4415366747183054,
203
+ "learning_rate": 4.982300584257943e-05,
204
+ "loss": 1.4511,
205
+ "step": 280
206
+ },
207
+ {
208
+ "epoch": 0.07850568489442339,
209
+ "grad_norm": 1.7712176478354567,
210
+ "learning_rate": 4.98101539553678e-05,
211
+ "loss": 1.4221,
212
+ "step": 290
213
+ },
214
+ {
215
+ "epoch": 0.08121277747698971,
216
+ "grad_norm": 1.408329894177912,
217
+ "learning_rate": 4.979685345213766e-05,
218
+ "loss": 1.4153,
219
+ "step": 300
220
+ },
221
+ {
222
+ "epoch": 0.08391987005955603,
223
+ "grad_norm": 2.8322848403176106,
224
+ "learning_rate": 4.978310457338806e-05,
225
+ "loss": 1.4544,
226
+ "step": 310
227
+ },
228
+ {
229
+ "epoch": 0.08662696264212236,
230
+ "grad_norm": 1.349216831725713,
231
+ "learning_rate": 4.976890756772558e-05,
232
+ "loss": 1.4232,
233
+ "step": 320
234
+ },
235
+ {
236
+ "epoch": 0.08933405522468868,
237
+ "grad_norm": 3.302218333832659,
238
+ "learning_rate": 4.9754262691859786e-05,
239
+ "loss": 1.4716,
240
+ "step": 330
241
+ },
242
+ {
243
+ "epoch": 0.092041147807255,
244
+ "grad_norm": 2.049533306517282,
245
+ "learning_rate": 4.9739170210598646e-05,
246
+ "loss": 1.469,
247
+ "step": 340
248
+ },
249
+ {
250
+ "epoch": 0.09474824038982133,
251
+ "grad_norm": 2.329244123317743,
252
+ "learning_rate": 4.9723630396843665e-05,
253
+ "loss": 1.4268,
254
+ "step": 350
255
+ },
256
+ {
257
+ "epoch": 0.09745533297238766,
258
+ "grad_norm": 1.6230256143926636,
259
+ "learning_rate": 4.9707643531585026e-05,
260
+ "loss": 1.44,
261
+ "step": 360
262
+ },
263
+ {
264
+ "epoch": 0.10016242555495398,
265
+ "grad_norm": 2.048460766648452,
266
+ "learning_rate": 4.969120990389645e-05,
267
+ "loss": 1.4171,
268
+ "step": 370
269
+ },
270
+ {
271
+ "epoch": 0.10286951813752031,
272
+ "grad_norm": 1.8949361232151039,
273
+ "learning_rate": 4.967432981093002e-05,
274
+ "loss": 1.4555,
275
+ "step": 380
276
+ },
277
+ {
278
+ "epoch": 0.10557661072008663,
279
+ "grad_norm": 1.454874136363247,
280
+ "learning_rate": 4.965700355791075e-05,
281
+ "loss": 1.4637,
282
+ "step": 390
283
+ },
284
+ {
285
+ "epoch": 0.10828370330265295,
286
+ "grad_norm": 1.6866618414734464,
287
+ "learning_rate": 4.9639231458131155e-05,
288
+ "loss": 1.4882,
289
+ "step": 400
290
+ },
291
+ {
292
+ "epoch": 0.11099079588521928,
293
+ "grad_norm": 1.767137745890727,
294
+ "learning_rate": 4.962101383294546e-05,
295
+ "loss": 1.4643,
296
+ "step": 410
297
+ },
298
+ {
299
+ "epoch": 0.1136978884677856,
300
+ "grad_norm": 1.4955900585918869,
301
+ "learning_rate": 4.9602351011763916e-05,
302
+ "loss": 1.4916,
303
+ "step": 420
304
+ },
305
+ {
306
+ "epoch": 0.11640498105035192,
307
+ "grad_norm": 1.4089383124912003,
308
+ "learning_rate": 4.958324333204675e-05,
309
+ "loss": 1.4558,
310
+ "step": 430
311
+ },
312
+ {
313
+ "epoch": 0.11911207363291824,
314
+ "grad_norm": 1.794803201497381,
315
+ "learning_rate": 4.9563691139298115e-05,
316
+ "loss": 1.4685,
317
+ "step": 440
318
+ },
319
+ {
320
+ "epoch": 0.12181916621548457,
321
+ "grad_norm": 4.657348417565155,
322
+ "learning_rate": 4.954369478705984e-05,
323
+ "loss": 1.4517,
324
+ "step": 450
325
+ },
326
+ {
327
+ "epoch": 0.12452625879805089,
328
+ "grad_norm": 1.727020331207215,
329
+ "learning_rate": 4.952325463690499e-05,
330
+ "loss": 1.4905,
331
+ "step": 460
332
+ },
333
+ {
334
+ "epoch": 0.12723335138061723,
335
+ "grad_norm": 1.677461362340949,
336
+ "learning_rate": 4.95023710584314e-05,
337
+ "loss": 1.4786,
338
+ "step": 470
339
+ },
340
+ {
341
+ "epoch": 0.12994044396318355,
342
+ "grad_norm": 1.3718142212853295,
343
+ "learning_rate": 4.948104442925493e-05,
344
+ "loss": 1.5028,
345
+ "step": 480
346
+ },
347
+ {
348
+ "epoch": 0.13264753654574987,
349
+ "grad_norm": 1.762804612117205,
350
+ "learning_rate": 4.945927513500267e-05,
351
+ "loss": 1.53,
352
+ "step": 490
353
+ },
354
+ {
355
+ "epoch": 0.1353546291283162,
356
+ "grad_norm": 2.616132608044088,
357
+ "learning_rate": 4.943706356930596e-05,
358
+ "loss": 1.4513,
359
+ "step": 500
360
+ },
361
+ {
362
+ "epoch": 0.13806172171088252,
363
+ "grad_norm": 1.5396425075226725,
364
+ "learning_rate": 4.941441013379328e-05,
365
+ "loss": 1.4783,
366
+ "step": 510
367
+ },
368
+ {
369
+ "epoch": 0.14076881429344884,
370
+ "grad_norm": 1.5753718924201408,
371
+ "learning_rate": 4.939131523808295e-05,
372
+ "loss": 1.469,
373
+ "step": 520
374
+ },
375
+ {
376
+ "epoch": 0.14347590687601516,
377
+ "grad_norm": 1.990561438026612,
378
+ "learning_rate": 4.936777929977578e-05,
379
+ "loss": 1.5288,
380
+ "step": 530
381
+ },
382
+ {
383
+ "epoch": 0.1461829994585815,
384
+ "grad_norm": 1.535027221106918,
385
+ "learning_rate": 4.9343802744447485e-05,
386
+ "loss": 1.4864,
387
+ "step": 540
388
+ },
389
+ {
390
+ "epoch": 0.1488900920411478,
391
+ "grad_norm": 1.5120704229625987,
392
+ "learning_rate": 4.9319386005641e-05,
393
+ "loss": 1.5288,
394
+ "step": 550
395
+ },
396
+ {
397
+ "epoch": 0.15159718462371413,
398
+ "grad_norm": 1.3272029647679429,
399
+ "learning_rate": 4.9294529524858605e-05,
400
+ "loss": 1.4571,
401
+ "step": 560
402
+ },
403
+ {
404
+ "epoch": 0.15430427720628045,
405
+ "grad_norm": 1.3970253346744186,
406
+ "learning_rate": 4.926923375155402e-05,
407
+ "loss": 1.4252,
408
+ "step": 570
409
+ },
410
+ {
411
+ "epoch": 0.15701136978884678,
412
+ "grad_norm": 1.8691453221841268,
413
+ "learning_rate": 4.924349914312421e-05,
414
+ "loss": 1.4326,
415
+ "step": 580
416
+ },
417
+ {
418
+ "epoch": 0.1597184623714131,
419
+ "grad_norm": 1.949334637283444,
420
+ "learning_rate": 4.921732616490111e-05,
421
+ "loss": 1.4056,
422
+ "step": 590
423
+ },
424
+ {
425
+ "epoch": 0.16242555495397942,
426
+ "grad_norm": 1.5924101773217003,
427
+ "learning_rate": 4.919071529014329e-05,
428
+ "loss": 1.4681,
429
+ "step": 600
430
+ },
431
+ {
432
+ "epoch": 0.16513264753654575,
433
+ "grad_norm": 1.6221588411171692,
434
+ "learning_rate": 4.91636670000273e-05,
435
+ "loss": 1.509,
436
+ "step": 610
437
+ },
438
+ {
439
+ "epoch": 0.16783974011911207,
440
+ "grad_norm": 1.5149301503130714,
441
+ "learning_rate": 4.913618178363904e-05,
442
+ "loss": 1.5005,
443
+ "step": 620
444
+ },
445
+ {
446
+ "epoch": 0.1705468327016784,
447
+ "grad_norm": 1.6644176831544628,
448
+ "learning_rate": 4.9108260137964865e-05,
449
+ "loss": 1.4638,
450
+ "step": 630
451
+ },
452
+ {
453
+ "epoch": 0.17325392528424471,
454
+ "grad_norm": 1.569403594423818,
455
+ "learning_rate": 4.9079902567882665e-05,
456
+ "loss": 1.4826,
457
+ "step": 640
458
+ },
459
+ {
460
+ "epoch": 0.17596101786681104,
461
+ "grad_norm": 1.3674634000662147,
462
+ "learning_rate": 4.9051109586152634e-05,
463
+ "loss": 1.4344,
464
+ "step": 650
465
+ },
466
+ {
467
+ "epoch": 0.17866811044937736,
468
+ "grad_norm": 1.1930764967763101,
469
+ "learning_rate": 4.902188171340811e-05,
470
+ "loss": 1.4396,
471
+ "step": 660
472
+ },
473
+ {
474
+ "epoch": 0.18137520303194368,
475
+ "grad_norm": 1.89545649756294,
476
+ "learning_rate": 4.8992219478146094e-05,
477
+ "loss": 1.4858,
478
+ "step": 670
479
+ },
480
+ {
481
+ "epoch": 0.18408229561451,
482
+ "grad_norm": 4.121840228113138,
483
+ "learning_rate": 4.8962123416717706e-05,
484
+ "loss": 1.4912,
485
+ "step": 680
486
+ },
487
+ {
488
+ "epoch": 0.18678938819707633,
489
+ "grad_norm": 1.292830871259127,
490
+ "learning_rate": 4.893159407331849e-05,
491
+ "loss": 1.5105,
492
+ "step": 690
493
+ },
494
+ {
495
+ "epoch": 0.18949648077964265,
496
+ "grad_norm": 2.0378465705988797,
497
+ "learning_rate": 4.8900631999978565e-05,
498
+ "loss": 1.4478,
499
+ "step": 700
500
+ },
501
+ {
502
+ "epoch": 0.19220357336220897,
503
+ "grad_norm": 1.4086096037731015,
504
+ "learning_rate": 4.886923775655267e-05,
505
+ "loss": 1.4748,
506
+ "step": 710
507
+ },
508
+ {
509
+ "epoch": 0.19491066594477532,
510
+ "grad_norm": 1.6471666031866425,
511
+ "learning_rate": 4.883741191071001e-05,
512
+ "loss": 1.4616,
513
+ "step": 720
514
+ },
515
+ {
516
+ "epoch": 0.19761775852734165,
517
+ "grad_norm": 1.3016146758942742,
518
+ "learning_rate": 4.880515503792402e-05,
519
+ "loss": 1.4502,
520
+ "step": 730
521
+ },
522
+ {
523
+ "epoch": 0.20032485110990797,
524
+ "grad_norm": 1.7755038959773146,
525
+ "learning_rate": 4.8772467721461916e-05,
526
+ "loss": 1.4929,
527
+ "step": 740
528
+ },
529
+ {
530
+ "epoch": 0.2030319436924743,
531
+ "grad_norm": 1.8173068291163943,
532
+ "learning_rate": 4.8739350552374206e-05,
533
+ "loss": 1.4514,
534
+ "step": 750
535
+ },
536
+ {
537
+ "epoch": 0.20573903627504062,
538
+ "grad_norm": 1.3957806556643733,
539
+ "learning_rate": 4.870580412948397e-05,
540
+ "loss": 1.4901,
541
+ "step": 760
542
+ },
543
+ {
544
+ "epoch": 0.20844612885760694,
545
+ "grad_norm": 1.7574648378127529,
546
+ "learning_rate": 4.8671829059375996e-05,
547
+ "loss": 1.4239,
548
+ "step": 770
549
+ },
550
+ {
551
+ "epoch": 0.21115322144017326,
552
+ "grad_norm": 1.988134441146335,
553
+ "learning_rate": 4.863742595638589e-05,
554
+ "loss": 1.4929,
555
+ "step": 780
556
+ },
557
+ {
558
+ "epoch": 0.21386031402273958,
559
+ "grad_norm": 1.6248880174612887,
560
+ "learning_rate": 4.8602595442588926e-05,
561
+ "loss": 1.5049,
562
+ "step": 790
563
+ },
564
+ {
565
+ "epoch": 0.2165674066053059,
566
+ "grad_norm": 1.4759465960672522,
567
+ "learning_rate": 4.856733814778878e-05,
568
+ "loss": 1.4617,
569
+ "step": 800
570
+ },
571
+ {
572
+ "epoch": 0.21927449918787223,
573
+ "grad_norm": 1.6391142091975623,
574
+ "learning_rate": 4.853165470950617e-05,
575
+ "loss": 1.496,
576
+ "step": 810
577
+ },
578
+ {
579
+ "epoch": 0.22198159177043855,
580
+ "grad_norm": 1.5875776432159403,
581
+ "learning_rate": 4.8495545772967336e-05,
582
+ "loss": 1.4696,
583
+ "step": 820
584
+ },
585
+ {
586
+ "epoch": 0.22468868435300487,
587
+ "grad_norm": 1.3488488116129105,
588
+ "learning_rate": 4.8459011991092335e-05,
589
+ "loss": 1.4865,
590
+ "step": 830
591
+ },
592
+ {
593
+ "epoch": 0.2273957769355712,
594
+ "grad_norm": 1.2491528956283275,
595
+ "learning_rate": 4.842205402448326e-05,
596
+ "loss": 1.4818,
597
+ "step": 840
598
+ },
599
+ {
600
+ "epoch": 0.23010286951813752,
601
+ "grad_norm": 1.4146633075540191,
602
+ "learning_rate": 4.8384672541412314e-05,
603
+ "loss": 1.4461,
604
+ "step": 850
605
+ },
606
+ {
607
+ "epoch": 0.23280996210070384,
608
+ "grad_norm": 2.8216763132669405,
609
+ "learning_rate": 4.834686821780966e-05,
610
+ "loss": 1.4905,
611
+ "step": 860
612
+ },
613
+ {
614
+ "epoch": 0.23551705468327017,
615
+ "grad_norm": 1.3681470128440785,
616
+ "learning_rate": 4.83086417372513e-05,
617
+ "loss": 1.4762,
618
+ "step": 870
619
+ },
620
+ {
621
+ "epoch": 0.2382241472658365,
622
+ "grad_norm": 1.6949909678615045,
623
+ "learning_rate": 4.826999379094662e-05,
624
+ "loss": 1.4548,
625
+ "step": 880
626
+ },
627
+ {
628
+ "epoch": 0.2409312398484028,
629
+ "grad_norm": 1.7298574479777804,
630
+ "learning_rate": 4.823092507772593e-05,
631
+ "loss": 1.4457,
632
+ "step": 890
633
+ },
634
+ {
635
+ "epoch": 0.24363833243096913,
636
+ "grad_norm": 1.3751587504284029,
637
+ "learning_rate": 4.8191436304027846e-05,
638
+ "loss": 1.4485,
639
+ "step": 900
640
+ },
641
+ {
642
+ "epoch": 0.24634542501353546,
643
+ "grad_norm": 1.98272939662037,
644
+ "learning_rate": 4.815152818388647e-05,
645
+ "loss": 1.4827,
646
+ "step": 910
647
+ },
648
+ {
649
+ "epoch": 0.24905251759610178,
650
+ "grad_norm": 1.48867070663095,
651
+ "learning_rate": 4.811120143891855e-05,
652
+ "loss": 1.4961,
653
+ "step": 920
654
+ },
655
+ {
656
+ "epoch": 0.2517596101786681,
657
+ "grad_norm": 1.6582119972239975,
658
+ "learning_rate": 4.807045679831035e-05,
659
+ "loss": 1.5125,
660
+ "step": 930
661
+ },
662
+ {
663
+ "epoch": 0.25446670276123445,
664
+ "grad_norm": 2.2411742740260143,
665
+ "learning_rate": 4.8029294998804515e-05,
666
+ "loss": 1.481,
667
+ "step": 940
668
+ },
669
+ {
670
+ "epoch": 0.25717379534380075,
671
+ "grad_norm": 3.1196395093384757,
672
+ "learning_rate": 4.798771678468673e-05,
673
+ "loss": 1.4766,
674
+ "step": 950
675
+ },
676
+ {
677
+ "epoch": 0.2598808879263671,
678
+ "grad_norm": 1.747388174474549,
679
+ "learning_rate": 4.794572290777228e-05,
680
+ "loss": 1.4664,
681
+ "step": 960
682
+ },
683
+ {
684
+ "epoch": 0.2625879805089334,
685
+ "grad_norm": 1.244225229661525,
686
+ "learning_rate": 4.790331412739246e-05,
687
+ "loss": 1.4881,
688
+ "step": 970
689
+ },
690
+ {
691
+ "epoch": 0.26529507309149974,
692
+ "grad_norm": 1.5361914036706255,
693
+ "learning_rate": 4.786049121038078e-05,
694
+ "loss": 1.4481,
695
+ "step": 980
696
+ },
697
+ {
698
+ "epoch": 0.26800216567406604,
699
+ "grad_norm": 2.004415613855818,
700
+ "learning_rate": 4.78172549310592e-05,
701
+ "loss": 1.4654,
702
+ "step": 990
703
+ },
704
+ {
705
+ "epoch": 0.2707092582566324,
706
+ "grad_norm": 1.6485133719660112,
707
+ "learning_rate": 4.777360607122405e-05,
708
+ "loss": 1.4569,
709
+ "step": 1000
710
+ },
711
+ {
712
+ "epoch": 0.2734163508391987,
713
+ "grad_norm": 1.3237479316055099,
714
+ "learning_rate": 4.772954542013193e-05,
715
+ "loss": 1.4616,
716
+ "step": 1010
717
+ },
718
+ {
719
+ "epoch": 0.27612344342176504,
720
+ "grad_norm": 1.4682831465817014,
721
+ "learning_rate": 4.7685073774485424e-05,
722
+ "loss": 1.4312,
723
+ "step": 1020
724
+ },
725
+ {
726
+ "epoch": 0.27883053600433133,
727
+ "grad_norm": 2.66972568412979,
728
+ "learning_rate": 4.764019193841871e-05,
729
+ "loss": 1.4638,
730
+ "step": 1030
731
+ },
732
+ {
733
+ "epoch": 0.2815376285868977,
734
+ "grad_norm": 1.8874806530406238,
735
+ "learning_rate": 4.7594900723482994e-05,
736
+ "loss": 1.4342,
737
+ "step": 1040
738
+ },
739
+ {
740
+ "epoch": 0.284244721169464,
741
+ "grad_norm": 1.7012798675260252,
742
+ "learning_rate": 4.7549200948631875e-05,
743
+ "loss": 1.4808,
744
+ "step": 1050
745
+ },
746
+ {
747
+ "epoch": 0.2869518137520303,
748
+ "grad_norm": 1.5041238189640276,
749
+ "learning_rate": 4.750309344020649e-05,
750
+ "loss": 1.477,
751
+ "step": 1060
752
+ },
753
+ {
754
+ "epoch": 0.2896589063345966,
755
+ "grad_norm": 1.445182392004961,
756
+ "learning_rate": 4.74565790319206e-05,
757
+ "loss": 1.4684,
758
+ "step": 1070
759
+ },
760
+ {
761
+ "epoch": 0.292365998917163,
762
+ "grad_norm": 1.5350175117888023,
763
+ "learning_rate": 4.7409658564845526e-05,
764
+ "loss": 1.4786,
765
+ "step": 1080
766
+ },
767
+ {
768
+ "epoch": 0.29507309149972927,
769
+ "grad_norm": 1.6518274612672268,
770
+ "learning_rate": 4.73623328873949e-05,
771
+ "loss": 1.4819,
772
+ "step": 1090
773
+ },
774
+ {
775
+ "epoch": 0.2977801840822956,
776
+ "grad_norm": 1.7095589003674432,
777
+ "learning_rate": 4.731460285530936e-05,
778
+ "loss": 1.4976,
779
+ "step": 1100
780
+ },
781
+ {
782
+ "epoch": 0.3004872766648619,
783
+ "grad_norm": 1.723367767756465,
784
+ "learning_rate": 4.726646933164108e-05,
785
+ "loss": 1.4504,
786
+ "step": 1110
787
+ },
788
+ {
789
+ "epoch": 0.30319436924742826,
790
+ "grad_norm": 1.9040390036164423,
791
+ "learning_rate": 4.7217933186738114e-05,
792
+ "loss": 1.4579,
793
+ "step": 1120
794
+ },
795
+ {
796
+ "epoch": 0.30590146182999456,
797
+ "grad_norm": 1.2294419736565405,
798
+ "learning_rate": 4.716899529822874e-05,
799
+ "loss": 1.5031,
800
+ "step": 1130
801
+ },
802
+ {
803
+ "epoch": 0.3086085544125609,
804
+ "grad_norm": 2.0118944901286997,
805
+ "learning_rate": 4.7119656551005505e-05,
806
+ "loss": 1.4532,
807
+ "step": 1140
808
+ },
809
+ {
810
+ "epoch": 0.31131564699512726,
811
+ "grad_norm": 1.6217590845134082,
812
+ "learning_rate": 4.706991783720928e-05,
813
+ "loss": 1.4871,
814
+ "step": 1150
815
+ },
816
+ {
817
+ "epoch": 0.31402273957769355,
818
+ "grad_norm": 1.509260025154433,
819
+ "learning_rate": 4.701978005621311e-05,
820
+ "loss": 1.4464,
821
+ "step": 1160
822
+ },
823
+ {
824
+ "epoch": 0.3167298321602599,
825
+ "grad_norm": 1.5171464183166403,
826
+ "learning_rate": 4.696924411460596e-05,
827
+ "loss": 1.4486,
828
+ "step": 1170
829
+ },
830
+ {
831
+ "epoch": 0.3194369247428262,
832
+ "grad_norm": 1.2572808716835593,
833
+ "learning_rate": 4.6918310926176306e-05,
834
+ "loss": 1.4458,
835
+ "step": 1180
836
+ },
837
+ {
838
+ "epoch": 0.32214401732539255,
839
+ "grad_norm": 1.7849046370327453,
840
+ "learning_rate": 4.686698141189564e-05,
841
+ "loss": 1.4404,
842
+ "step": 1190
843
+ },
844
+ {
845
+ "epoch": 0.32485110990795885,
846
+ "grad_norm": 1.914525739200728,
847
+ "learning_rate": 4.6815256499901775e-05,
848
+ "loss": 1.4898,
849
+ "step": 1200
850
+ },
851
+ {
852
+ "epoch": 0.3275582024905252,
853
+ "grad_norm": 1.611833069659245,
854
+ "learning_rate": 4.6763137125482104e-05,
855
+ "loss": 1.4847,
856
+ "step": 1210
857
+ },
858
+ {
859
+ "epoch": 0.3302652950730915,
860
+ "grad_norm": 1.753641811163595,
861
+ "learning_rate": 4.671062423105667e-05,
862
+ "loss": 1.433,
863
+ "step": 1220
864
+ },
865
+ {
866
+ "epoch": 0.33297238765565784,
867
+ "grad_norm": 1.1996803367158615,
868
+ "learning_rate": 4.66577187661611e-05,
869
+ "loss": 1.4533,
870
+ "step": 1230
871
+ },
872
+ {
873
+ "epoch": 0.33567948023822414,
874
+ "grad_norm": 1.554114829537543,
875
+ "learning_rate": 4.660442168742951e-05,
876
+ "loss": 1.4708,
877
+ "step": 1240
878
+ },
879
+ {
880
+ "epoch": 0.3383865728207905,
881
+ "grad_norm": 6.405248874344029,
882
+ "learning_rate": 4.6550733958577115e-05,
883
+ "loss": 1.4721,
884
+ "step": 1250
885
+ },
886
+ {
887
+ "epoch": 0.3410936654033568,
888
+ "grad_norm": 1.5464469028099914,
889
+ "learning_rate": 4.6496656550382864e-05,
890
+ "loss": 1.4524,
891
+ "step": 1260
892
+ },
893
+ {
894
+ "epoch": 0.34380075798592313,
895
+ "grad_norm": 1.5438246148321,
896
+ "learning_rate": 4.6442190440671886e-05,
897
+ "loss": 1.5217,
898
+ "step": 1270
899
+ },
900
+ {
901
+ "epoch": 0.34650785056848943,
902
+ "grad_norm": 1.6459917338592895,
903
+ "learning_rate": 4.638733661429775e-05,
904
+ "loss": 1.431,
905
+ "step": 1280
906
+ },
907
+ {
908
+ "epoch": 0.3492149431510558,
909
+ "grad_norm": 1.7332054776628096,
910
+ "learning_rate": 4.633209606312473e-05,
911
+ "loss": 1.4896,
912
+ "step": 1290
913
+ },
914
+ {
915
+ "epoch": 0.3519220357336221,
916
+ "grad_norm": 1.7861896691639296,
917
+ "learning_rate": 4.6276469786009846e-05,
918
+ "loss": 1.4587,
919
+ "step": 1300
920
+ },
921
+ {
922
+ "epoch": 0.3546291283161884,
923
+ "grad_norm": 1.380983484664312,
924
+ "learning_rate": 4.622045878878477e-05,
925
+ "loss": 1.4928,
926
+ "step": 1310
927
+ },
928
+ {
929
+ "epoch": 0.3573362208987547,
930
+ "grad_norm": 1.6722892346510752,
931
+ "learning_rate": 4.616406408423769e-05,
932
+ "loss": 1.3885,
933
+ "step": 1320
934
+ },
935
+ {
936
+ "epoch": 0.36004331348132107,
937
+ "grad_norm": 1.302543220749753,
938
+ "learning_rate": 4.610728669209494e-05,
939
+ "loss": 1.4281,
940
+ "step": 1330
941
+ },
942
+ {
943
+ "epoch": 0.36275040606388737,
944
+ "grad_norm": 1.3027170210142245,
945
+ "learning_rate": 4.605012763900264e-05,
946
+ "loss": 1.4695,
947
+ "step": 1340
948
+ },
949
+ {
950
+ "epoch": 0.3654574986464537,
951
+ "grad_norm": 1.604586924613264,
952
+ "learning_rate": 4.599258795850803e-05,
953
+ "loss": 1.449,
954
+ "step": 1350
955
+ },
956
+ {
957
+ "epoch": 0.36816459122902,
958
+ "grad_norm": 1.2848250120256988,
959
+ "learning_rate": 4.593466869104087e-05,
960
+ "loss": 1.41,
961
+ "step": 1360
962
+ },
963
+ {
964
+ "epoch": 0.37087168381158636,
965
+ "grad_norm": 1.5746776369025801,
966
+ "learning_rate": 4.5876370883894606e-05,
967
+ "loss": 1.5164,
968
+ "step": 1370
969
+ },
970
+ {
971
+ "epoch": 0.37357877639415266,
972
+ "grad_norm": 1.8869095435477306,
973
+ "learning_rate": 4.581769559120738e-05,
974
+ "loss": 1.4649,
975
+ "step": 1380
976
+ },
977
+ {
978
+ "epoch": 0.376285868976719,
979
+ "grad_norm": 1.717413158097039,
980
+ "learning_rate": 4.575864387394304e-05,
981
+ "loss": 1.4801,
982
+ "step": 1390
983
+ },
984
+ {
985
+ "epoch": 0.3789929615592853,
986
+ "grad_norm": 2.539587222307521,
987
+ "learning_rate": 4.569921679987192e-05,
988
+ "loss": 1.4385,
989
+ "step": 1400
990
+ },
991
+ {
992
+ "epoch": 0.38170005414185165,
993
+ "grad_norm": 1.2980114991302223,
994
+ "learning_rate": 4.563941544355152e-05,
995
+ "loss": 1.4934,
996
+ "step": 1410
997
+ },
998
+ {
999
+ "epoch": 0.38440714672441795,
1000
+ "grad_norm": 1.6715775537496502,
1001
+ "learning_rate": 4.5579240886307136e-05,
1002
+ "loss": 1.4538,
1003
+ "step": 1420
1004
+ },
1005
+ {
1006
+ "epoch": 0.3871142393069843,
1007
+ "grad_norm": 2.709253749500051,
1008
+ "learning_rate": 4.551869421621221e-05,
1009
+ "loss": 1.4554,
1010
+ "step": 1430
1011
+ },
1012
+ {
1013
+ "epoch": 0.38982133188955065,
1014
+ "grad_norm": 1.5064962735011314,
1015
+ "learning_rate": 4.5457776528068743e-05,
1016
+ "loss": 1.5152,
1017
+ "step": 1440
1018
+ },
1019
+ {
1020
+ "epoch": 0.39252842447211694,
1021
+ "grad_norm": 1.4866135650495016,
1022
+ "learning_rate": 4.539648892338747e-05,
1023
+ "loss": 1.4701,
1024
+ "step": 1450
1025
+ },
1026
+ {
1027
+ "epoch": 0.3952355170546833,
1028
+ "grad_norm": 1.425267538841413,
1029
+ "learning_rate": 4.533483251036792e-05,
1030
+ "loss": 1.4794,
1031
+ "step": 1460
1032
+ },
1033
+ {
1034
+ "epoch": 0.3979426096372496,
1035
+ "grad_norm": 1.4920137928311075,
1036
+ "learning_rate": 4.527280840387842e-05,
1037
+ "loss": 1.4884,
1038
+ "step": 1470
1039
+ },
1040
+ {
1041
+ "epoch": 0.40064970221981594,
1042
+ "grad_norm": 1.4659663002416385,
1043
+ "learning_rate": 4.5210417725435896e-05,
1044
+ "loss": 1.4436,
1045
+ "step": 1480
1046
+ },
1047
+ {
1048
+ "epoch": 0.40335679480238223,
1049
+ "grad_norm": 1.4406570198350586,
1050
+ "learning_rate": 4.514766160318561e-05,
1051
+ "loss": 1.473,
1052
+ "step": 1490
1053
+ },
1054
+ {
1055
+ "epoch": 0.4060638873849486,
1056
+ "grad_norm": 1.4236232583730681,
1057
+ "learning_rate": 4.508454117188077e-05,
1058
+ "loss": 1.441,
1059
+ "step": 1500
1060
+ },
1061
+ {
1062
+ "epoch": 0.4087709799675149,
1063
+ "grad_norm": 1.658008187554401,
1064
+ "learning_rate": 4.5021057572862e-05,
1065
+ "loss": 1.4656,
1066
+ "step": 1510
1067
+ },
1068
+ {
1069
+ "epoch": 0.41147807255008123,
1070
+ "grad_norm": 1.5768571336837218,
1071
+ "learning_rate": 4.495721195403671e-05,
1072
+ "loss": 1.4681,
1073
+ "step": 1520
1074
+ },
1075
+ {
1076
+ "epoch": 0.4141851651326475,
1077
+ "grad_norm": 1.770673573444082,
1078
+ "learning_rate": 4.4893005469858316e-05,
1079
+ "loss": 1.42,
1080
+ "step": 1530
1081
+ },
1082
+ {
1083
+ "epoch": 0.4168922577152139,
1084
+ "grad_norm": 1.3645638511742517,
1085
+ "learning_rate": 4.482843928130541e-05,
1086
+ "loss": 1.4911,
1087
+ "step": 1540
1088
+ },
1089
+ {
1090
+ "epoch": 0.41959935029778017,
1091
+ "grad_norm": 1.6871925830755095,
1092
+ "learning_rate": 4.476351455586072e-05,
1093
+ "loss": 1.5269,
1094
+ "step": 1550
1095
+ },
1096
+ {
1097
+ "epoch": 0.4223064428803465,
1098
+ "grad_norm": 1.4062614507732203,
1099
+ "learning_rate": 4.469823246749003e-05,
1100
+ "loss": 1.445,
1101
+ "step": 1560
1102
+ },
1103
+ {
1104
+ "epoch": 0.4250135354629128,
1105
+ "grad_norm": 1.732815716768682,
1106
+ "learning_rate": 4.4632594196620945e-05,
1107
+ "loss": 1.5041,
1108
+ "step": 1570
1109
+ },
1110
+ {
1111
+ "epoch": 0.42772062804547917,
1112
+ "grad_norm": 1.2938163974151866,
1113
+ "learning_rate": 4.456660093012155e-05,
1114
+ "loss": 1.4565,
1115
+ "step": 1580
1116
+ },
1117
+ {
1118
+ "epoch": 0.43042772062804546,
1119
+ "grad_norm": 1.4591593663238542,
1120
+ "learning_rate": 4.4500253861278926e-05,
1121
+ "loss": 1.457,
1122
+ "step": 1590
1123
+ },
1124
+ {
1125
+ "epoch": 0.4331348132106118,
1126
+ "grad_norm": 1.619966018635216,
1127
+ "learning_rate": 4.443355418977761e-05,
1128
+ "loss": 1.4848,
1129
+ "step": 1600
1130
+ },
1131
+ {
1132
+ "epoch": 0.4358419057931781,
1133
+ "grad_norm": 1.5713970862575257,
1134
+ "learning_rate": 4.436650312167787e-05,
1135
+ "loss": 1.4311,
1136
+ "step": 1610
1137
+ },
1138
+ {
1139
+ "epoch": 0.43854899837574446,
1140
+ "grad_norm": 4.836628503687221,
1141
+ "learning_rate": 4.429910186939392e-05,
1142
+ "loss": 1.5039,
1143
+ "step": 1620
1144
+ },
1145
+ {
1146
+ "epoch": 0.44125609095831075,
1147
+ "grad_norm": 1.5617479359446638,
1148
+ "learning_rate": 4.4231351651671985e-05,
1149
+ "loss": 1.4387,
1150
+ "step": 1630
1151
+ },
1152
+ {
1153
+ "epoch": 0.4439631835408771,
1154
+ "grad_norm": 4.127569209273957,
1155
+ "learning_rate": 4.4163253693568266e-05,
1156
+ "loss": 1.4654,
1157
+ "step": 1640
1158
+ },
1159
+ {
1160
+ "epoch": 0.4466702761234434,
1161
+ "grad_norm": 1.9045020579112555,
1162
+ "learning_rate": 4.409480922642677e-05,
1163
+ "loss": 1.403,
1164
+ "step": 1650
1165
+ },
1166
+ {
1167
+ "epoch": 0.44937736870600975,
1168
+ "grad_norm": 1.444246465499322,
1169
+ "learning_rate": 4.40260194878571e-05,
1170
+ "loss": 1.4568,
1171
+ "step": 1660
1172
+ },
1173
+ {
1174
+ "epoch": 0.45208446128857604,
1175
+ "grad_norm": 1.3435184719673208,
1176
+ "learning_rate": 4.395688572171203e-05,
1177
+ "loss": 1.5321,
1178
+ "step": 1670
1179
+ },
1180
+ {
1181
+ "epoch": 0.4547915538711424,
1182
+ "grad_norm": 1.8192617831889093,
1183
+ "learning_rate": 4.3887409178064995e-05,
1184
+ "loss": 1.4693,
1185
+ "step": 1680
1186
+ },
1187
+ {
1188
+ "epoch": 0.4574986464537087,
1189
+ "grad_norm": 1.3323020396326049,
1190
+ "learning_rate": 4.3817591113187563e-05,
1191
+ "loss": 1.4333,
1192
+ "step": 1690
1193
+ },
1194
+ {
1195
+ "epoch": 0.46020573903627504,
1196
+ "grad_norm": 1.2996234776965605,
1197
+ "learning_rate": 4.374743278952662e-05,
1198
+ "loss": 1.4034,
1199
+ "step": 1700
1200
+ },
1201
+ {
1202
+ "epoch": 0.4629128316188414,
1203
+ "grad_norm": 1.4381391819706373,
1204
+ "learning_rate": 4.367693547568163e-05,
1205
+ "loss": 1.4924,
1206
+ "step": 1710
1207
+ },
1208
+ {
1209
+ "epoch": 0.4656199242014077,
1210
+ "grad_norm": 1.8146691731599425,
1211
+ "learning_rate": 4.360610044638167e-05,
1212
+ "loss": 1.4284,
1213
+ "step": 1720
1214
+ },
1215
+ {
1216
+ "epoch": 0.46832701678397404,
1217
+ "grad_norm": 2.968939444897473,
1218
+ "learning_rate": 4.353492898246234e-05,
1219
+ "loss": 1.4586,
1220
+ "step": 1730
1221
+ },
1222
+ {
1223
+ "epoch": 0.47103410936654033,
1224
+ "grad_norm": 2.510758529664001,
1225
+ "learning_rate": 4.346342237084267e-05,
1226
+ "loss": 1.4913,
1227
+ "step": 1740
1228
+ },
1229
+ {
1230
+ "epoch": 0.4737412019491067,
1231
+ "grad_norm": 1.6561919074934996,
1232
+ "learning_rate": 4.339158190450176e-05,
1233
+ "loss": 1.474,
1234
+ "step": 1750
1235
+ },
1236
+ {
1237
+ "epoch": 0.476448294531673,
1238
+ "grad_norm": 1.6475904490305402,
1239
+ "learning_rate": 4.331940888245552e-05,
1240
+ "loss": 1.4331,
1241
+ "step": 1760
1242
+ },
1243
+ {
1244
+ "epoch": 0.47915538711423933,
1245
+ "grad_norm": 1.5072770205173374,
1246
+ "learning_rate": 4.3246904609733074e-05,
1247
+ "loss": 1.4949,
1248
+ "step": 1770
1249
+ },
1250
+ {
1251
+ "epoch": 0.4818624796968056,
1252
+ "grad_norm": 1.3220720078913577,
1253
+ "learning_rate": 4.317407039735323e-05,
1254
+ "loss": 1.4037,
1255
+ "step": 1780
1256
+ },
1257
+ {
1258
+ "epoch": 0.484569572279372,
1259
+ "grad_norm": 2.630264685866369,
1260
+ "learning_rate": 4.310090756230072e-05,
1261
+ "loss": 1.4366,
1262
+ "step": 1790
1263
+ },
1264
+ {
1265
+ "epoch": 0.48727666486193827,
1266
+ "grad_norm": 1.8523936235855012,
1267
+ "learning_rate": 4.3027417427502435e-05,
1268
+ "loss": 1.4757,
1269
+ "step": 1800
1270
+ },
1271
+ {
1272
+ "epoch": 0.4899837574445046,
1273
+ "grad_norm": 1.6259780791806493,
1274
+ "learning_rate": 4.2953601321803494e-05,
1275
+ "loss": 1.4554,
1276
+ "step": 1810
1277
+ },
1278
+ {
1279
+ "epoch": 0.4926908500270709,
1280
+ "grad_norm": 1.3829948059505168,
1281
+ "learning_rate": 4.287946057994318e-05,
1282
+ "loss": 1.444,
1283
+ "step": 1820
1284
+ },
1285
+ {
1286
+ "epoch": 0.49539794260963727,
1287
+ "grad_norm": 1.472490908240983,
1288
+ "learning_rate": 4.2804996542530854e-05,
1289
+ "loss": 1.4799,
1290
+ "step": 1830
1291
+ },
1292
+ {
1293
+ "epoch": 0.49810503519220356,
1294
+ "grad_norm": 1.4418412770727913,
1295
+ "learning_rate": 4.273021055602168e-05,
1296
+ "loss": 1.4539,
1297
+ "step": 1840
1298
+ },
1299
+ {
1300
+ "epoch": 0.5008121277747699,
1301
+ "grad_norm": 1.3579847745967637,
1302
+ "learning_rate": 4.265510397269229e-05,
1303
+ "loss": 1.4866,
1304
+ "step": 1850
1305
+ },
1306
+ {
1307
+ "epoch": 0.5035192203573362,
1308
+ "grad_norm": 1.2861587741021074,
1309
+ "learning_rate": 4.2579678150616344e-05,
1310
+ "loss": 1.4509,
1311
+ "step": 1860
1312
+ },
1313
+ {
1314
+ "epoch": 0.5062263129399025,
1315
+ "grad_norm": 1.69680502271894,
1316
+ "learning_rate": 4.2503934453639945e-05,
1317
+ "loss": 1.4417,
1318
+ "step": 1870
1319
+ },
1320
+ {
1321
+ "epoch": 0.5089334055224689,
1322
+ "grad_norm": 1.374930502329713,
1323
+ "learning_rate": 4.2427874251357e-05,
1324
+ "loss": 1.4459,
1325
+ "step": 1880
1326
+ },
1327
+ {
1328
+ "epoch": 0.5116404981050352,
1329
+ "grad_norm": 1.6333223217571196,
1330
+ "learning_rate": 4.235149891908447e-05,
1331
+ "loss": 1.4187,
1332
+ "step": 1890
1333
+ },
1334
+ {
1335
+ "epoch": 0.5143475906876015,
1336
+ "grad_norm": 1.2550594285330077,
1337
+ "learning_rate": 4.2274809837837444e-05,
1338
+ "loss": 1.4777,
1339
+ "step": 1900
1340
+ },
1341
+ {
1342
+ "epoch": 0.5170546832701678,
1343
+ "grad_norm": 1.7713884342868287,
1344
+ "learning_rate": 4.219780839430424e-05,
1345
+ "loss": 1.4742,
1346
+ "step": 1910
1347
+ },
1348
+ {
1349
+ "epoch": 0.5197617758527342,
1350
+ "grad_norm": 1.8526614078728505,
1351
+ "learning_rate": 4.212049598082126e-05,
1352
+ "loss": 1.4639,
1353
+ "step": 1920
1354
+ },
1355
+ {
1356
+ "epoch": 0.5224688684353005,
1357
+ "grad_norm": 1.7077160912171838,
1358
+ "learning_rate": 4.204287399534791e-05,
1359
+ "loss": 1.4678,
1360
+ "step": 1930
1361
+ },
1362
+ {
1363
+ "epoch": 0.5251759610178668,
1364
+ "grad_norm": 1.7342679697841041,
1365
+ "learning_rate": 4.196494384144119e-05,
1366
+ "loss": 1.4587,
1367
+ "step": 1940
1368
+ },
1369
+ {
1370
+ "epoch": 0.5278830536004331,
1371
+ "grad_norm": 1.3760999544704575,
1372
+ "learning_rate": 4.188670692823042e-05,
1373
+ "loss": 1.4364,
1374
+ "step": 1950
1375
+ },
1376
+ {
1377
+ "epoch": 0.5305901461829995,
1378
+ "grad_norm": 1.4927106037863223,
1379
+ "learning_rate": 4.1808164670391747e-05,
1380
+ "loss": 1.4636,
1381
+ "step": 1960
1382
+ },
1383
+ {
1384
+ "epoch": 0.5332972387655658,
1385
+ "grad_norm": 1.3148570909335389,
1386
+ "learning_rate": 4.172931848812251e-05,
1387
+ "loss": 1.4781,
1388
+ "step": 1970
1389
+ },
1390
+ {
1391
+ "epoch": 0.5360043313481321,
1392
+ "grad_norm": 1.3941339635980279,
1393
+ "learning_rate": 4.1650169807115623e-05,
1394
+ "loss": 1.4214,
1395
+ "step": 1980
1396
+ },
1397
+ {
1398
+ "epoch": 0.5387114239306985,
1399
+ "grad_norm": 1.1834123922199669,
1400
+ "learning_rate": 4.157072005853373e-05,
1401
+ "loss": 1.4226,
1402
+ "step": 1990
1403
+ },
1404
+ {
1405
+ "epoch": 0.5414185165132648,
1406
+ "grad_norm": 1.564745160522187,
1407
+ "learning_rate": 4.149097067898342e-05,
1408
+ "loss": 1.4631,
1409
+ "step": 2000
1410
+ },
1411
+ {
1412
+ "epoch": 0.5441256090958311,
1413
+ "grad_norm": 1.5351238663722413,
1414
+ "learning_rate": 4.141092311048914e-05,
1415
+ "loss": 1.4231,
1416
+ "step": 2010
1417
+ },
1418
+ {
1419
+ "epoch": 0.5468327016783974,
1420
+ "grad_norm": 1.4673872230064178,
1421
+ "learning_rate": 4.133057880046721e-05,
1422
+ "loss": 1.4308,
1423
+ "step": 2020
1424
+ },
1425
+ {
1426
+ "epoch": 0.5495397942609638,
1427
+ "grad_norm": 2.0555674903653576,
1428
+ "learning_rate": 4.1249939201699605e-05,
1429
+ "loss": 1.4054,
1430
+ "step": 2030
1431
+ },
1432
+ {
1433
+ "epoch": 0.5522468868435301,
1434
+ "grad_norm": 1.4275260085968484,
1435
+ "learning_rate": 4.116900577230769e-05,
1436
+ "loss": 1.4171,
1437
+ "step": 2040
1438
+ },
1439
+ {
1440
+ "epoch": 0.5549539794260964,
1441
+ "grad_norm": 1.2596642799013085,
1442
+ "learning_rate": 4.108777997572588e-05,
1443
+ "loss": 1.4064,
1444
+ "step": 2050
1445
+ },
1446
+ {
1447
+ "epoch": 0.5576610720086627,
1448
+ "grad_norm": 1.475728319316803,
1449
+ "learning_rate": 4.100626328067515e-05,
1450
+ "loss": 1.451,
1451
+ "step": 2060
1452
+ },
1453
+ {
1454
+ "epoch": 0.5603681645912291,
1455
+ "grad_norm": 1.4445720383429121,
1456
+ "learning_rate": 4.0924457161136465e-05,
1457
+ "loss": 1.4357,
1458
+ "step": 2070
1459
+ },
1460
+ {
1461
+ "epoch": 0.5630752571737954,
1462
+ "grad_norm": 1.277001561017148,
1463
+ "learning_rate": 4.0842363096324186e-05,
1464
+ "loss": 1.4152,
1465
+ "step": 2080
1466
+ },
1467
+ {
1468
+ "epoch": 0.5657823497563617,
1469
+ "grad_norm": 4.077914450657582,
1470
+ "learning_rate": 4.075998257065927e-05,
1471
+ "loss": 1.4219,
1472
+ "step": 2090
1473
+ },
1474
+ {
1475
+ "epoch": 0.568489442338928,
1476
+ "grad_norm": 1.4881560211628038,
1477
+ "learning_rate": 4.067731707374245e-05,
1478
+ "loss": 1.4558,
1479
+ "step": 2100
1480
+ },
1481
+ {
1482
+ "epoch": 0.5711965349214944,
1483
+ "grad_norm": 1.8102222242604538,
1484
+ "learning_rate": 4.0594368100327276e-05,
1485
+ "loss": 1.4418,
1486
+ "step": 2110
1487
+ },
1488
+ {
1489
+ "epoch": 0.5739036275040607,
1490
+ "grad_norm": 1.5342688729092908,
1491
+ "learning_rate": 4.051113715029314e-05,
1492
+ "loss": 1.4186,
1493
+ "step": 2120
1494
+ },
1495
+ {
1496
+ "epoch": 0.576610720086627,
1497
+ "grad_norm": 1.4776107653456874,
1498
+ "learning_rate": 4.0427625728618094e-05,
1499
+ "loss": 1.4241,
1500
+ "step": 2130
1501
+ },
1502
+ {
1503
+ "epoch": 0.5793178126691932,
1504
+ "grad_norm": 1.8005823424782537,
1505
+ "learning_rate": 4.034383534535169e-05,
1506
+ "loss": 1.4604,
1507
+ "step": 2140
1508
+ },
1509
+ {
1510
+ "epoch": 0.5820249052517596,
1511
+ "grad_norm": 1.4602346950120104,
1512
+ "learning_rate": 4.0259767515587607e-05,
1513
+ "loss": 1.4164,
1514
+ "step": 2150
1515
+ },
1516
+ {
1517
+ "epoch": 0.584731997834326,
1518
+ "grad_norm": 2.2063404920756025,
1519
+ "learning_rate": 4.017542375943635e-05,
1520
+ "loss": 1.4546,
1521
+ "step": 2160
1522
+ },
1523
+ {
1524
+ "epoch": 0.5874390904168922,
1525
+ "grad_norm": 1.3913645582438086,
1526
+ "learning_rate": 4.009080560199767e-05,
1527
+ "loss": 1.4499,
1528
+ "step": 2170
1529
+ },
1530
+ {
1531
+ "epoch": 0.5901461829994585,
1532
+ "grad_norm": 1.4511449754518972,
1533
+ "learning_rate": 4.0005914573333076e-05,
1534
+ "loss": 1.3874,
1535
+ "step": 2180
1536
+ },
1537
+ {
1538
+ "epoch": 0.5928532755820249,
1539
+ "grad_norm": 1.374243913815498,
1540
+ "learning_rate": 3.992075220843806e-05,
1541
+ "loss": 1.3881,
1542
+ "step": 2190
1543
+ },
1544
+ {
1545
+ "epoch": 0.5955603681645912,
1546
+ "grad_norm": 1.4550895039236171,
1547
+ "learning_rate": 3.983532004721446e-05,
1548
+ "loss": 1.4356,
1549
+ "step": 2200
1550
+ },
1551
+ {
1552
+ "epoch": 0.5982674607471575,
1553
+ "grad_norm": 2.8598201176092597,
1554
+ "learning_rate": 3.974961963444252e-05,
1555
+ "loss": 1.4194,
1556
+ "step": 2210
1557
+ },
1558
+ {
1559
+ "epoch": 0.6009745533297238,
1560
+ "grad_norm": 1.2900018758134322,
1561
+ "learning_rate": 3.9663652519753016e-05,
1562
+ "loss": 1.4254,
1563
+ "step": 2220
1564
+ },
1565
+ {
1566
+ "epoch": 0.6036816459122902,
1567
+ "grad_norm": 1.3550238391234248,
1568
+ "learning_rate": 3.95774202575992e-05,
1569
+ "loss": 1.4265,
1570
+ "step": 2230
1571
+ },
1572
+ {
1573
+ "epoch": 0.6063887384948565,
1574
+ "grad_norm": 1.2357167108126608,
1575
+ "learning_rate": 3.9490924407228725e-05,
1576
+ "loss": 1.439,
1577
+ "step": 2240
1578
+ },
1579
+ {
1580
+ "epoch": 0.6090958310774228,
1581
+ "grad_norm": 1.7838096381387492,
1582
+ "learning_rate": 3.940416653265542e-05,
1583
+ "loss": 1.3754,
1584
+ "step": 2250
1585
+ },
1586
+ {
1587
+ "epoch": 0.6118029236599891,
1588
+ "grad_norm": 1.5918950572416704,
1589
+ "learning_rate": 3.931714820263104e-05,
1590
+ "loss": 1.4193,
1591
+ "step": 2260
1592
+ },
1593
+ {
1594
+ "epoch": 0.6145100162425555,
1595
+ "grad_norm": 1.9008726278943935,
1596
+ "learning_rate": 3.922987099061685e-05,
1597
+ "loss": 1.4014,
1598
+ "step": 2270
1599
+ },
1600
+ {
1601
+ "epoch": 0.6172171088251218,
1602
+ "grad_norm": 1.5355669123167321,
1603
+ "learning_rate": 3.914233647475525e-05,
1604
+ "loss": 1.4242,
1605
+ "step": 2280
1606
+ },
1607
+ {
1608
+ "epoch": 0.6199242014076881,
1609
+ "grad_norm": 8.645566079065198,
1610
+ "learning_rate": 3.9054546237841156e-05,
1611
+ "loss": 1.3978,
1612
+ "step": 2290
1613
+ },
1614
+ {
1615
+ "epoch": 0.6226312939902545,
1616
+ "grad_norm": 1.4299000700088789,
1617
+ "learning_rate": 3.896650186729345e-05,
1618
+ "loss": 1.4444,
1619
+ "step": 2300
1620
+ },
1621
+ {
1622
+ "epoch": 0.6253383865728208,
1623
+ "grad_norm": 1.5568306708216317,
1624
+ "learning_rate": 3.8878204955126224e-05,
1625
+ "loss": 1.4535,
1626
+ "step": 2310
1627
+ },
1628
+ {
1629
+ "epoch": 0.6280454791553871,
1630
+ "grad_norm": 1.4298727206313604,
1631
+ "learning_rate": 3.878965709792003e-05,
1632
+ "loss": 1.3964,
1633
+ "step": 2320
1634
+ },
1635
+ {
1636
+ "epoch": 0.6307525717379534,
1637
+ "grad_norm": 1.9081562507364116,
1638
+ "learning_rate": 3.8700859896793004e-05,
1639
+ "loss": 1.4637,
1640
+ "step": 2330
1641
+ },
1642
+ {
1643
+ "epoch": 0.6334596643205198,
1644
+ "grad_norm": 1.7135745696260376,
1645
+ "learning_rate": 3.861181495737187e-05,
1646
+ "loss": 1.4429,
1647
+ "step": 2340
1648
+ },
1649
+ {
1650
+ "epoch": 0.6361667569030861,
1651
+ "grad_norm": 1.5360272794198329,
1652
+ "learning_rate": 3.852252388976299e-05,
1653
+ "loss": 1.4268,
1654
+ "step": 2350
1655
+ },
1656
+ {
1657
+ "epoch": 0.6388738494856524,
1658
+ "grad_norm": 1.2103422788055682,
1659
+ "learning_rate": 3.843298830852317e-05,
1660
+ "loss": 1.4469,
1661
+ "step": 2360
1662
+ },
1663
+ {
1664
+ "epoch": 0.6415809420682187,
1665
+ "grad_norm": 1.2503359557443374,
1666
+ "learning_rate": 3.834320983263052e-05,
1667
+ "loss": 1.4208,
1668
+ "step": 2370
1669
+ },
1670
+ {
1671
+ "epoch": 0.6442880346507851,
1672
+ "grad_norm": 1.704839306648677,
1673
+ "learning_rate": 3.8253190085455136e-05,
1674
+ "loss": 1.4205,
1675
+ "step": 2380
1676
+ },
1677
+ {
1678
+ "epoch": 0.6469951272333514,
1679
+ "grad_norm": 1.7016731029603525,
1680
+ "learning_rate": 3.816293069472981e-05,
1681
+ "loss": 1.4311,
1682
+ "step": 2390
1683
+ },
1684
+ {
1685
+ "epoch": 0.6497022198159177,
1686
+ "grad_norm": 1.4362792263340989,
1687
+ "learning_rate": 3.8072433292520526e-05,
1688
+ "loss": 1.38,
1689
+ "step": 2400
1690
+ },
1691
+ {
1692
+ "epoch": 0.652409312398484,
1693
+ "grad_norm": 1.5654681480694452,
1694
+ "learning_rate": 3.7981699515196975e-05,
1695
+ "loss": 1.4094,
1696
+ "step": 2410
1697
+ },
1698
+ {
1699
+ "epoch": 0.6551164049810504,
1700
+ "grad_norm": 1.4659564972472934,
1701
+ "learning_rate": 3.789073100340298e-05,
1702
+ "loss": 1.409,
1703
+ "step": 2420
1704
+ },
1705
+ {
1706
+ "epoch": 0.6578234975636167,
1707
+ "grad_norm": 7.417972847111027,
1708
+ "learning_rate": 3.779952940202686e-05,
1709
+ "loss": 1.4294,
1710
+ "step": 2430
1711
+ },
1712
+ {
1713
+ "epoch": 0.660530590146183,
1714
+ "grad_norm": 1.3497894617366657,
1715
+ "learning_rate": 3.770809636017158e-05,
1716
+ "loss": 1.407,
1717
+ "step": 2440
1718
+ },
1719
+ {
1720
+ "epoch": 0.6632376827287493,
1721
+ "grad_norm": 1.3081753915783962,
1722
+ "learning_rate": 3.7616433531125035e-05,
1723
+ "loss": 1.3756,
1724
+ "step": 2450
1725
+ },
1726
+ {
1727
+ "epoch": 0.6659447753113157,
1728
+ "grad_norm": 1.527466315347287,
1729
+ "learning_rate": 3.752454257233013e-05,
1730
+ "loss": 1.4118,
1731
+ "step": 2460
1732
+ },
1733
+ {
1734
+ "epoch": 0.668651867893882,
1735
+ "grad_norm": 1.6627023176259257,
1736
+ "learning_rate": 3.7432425145354804e-05,
1737
+ "loss": 1.4329,
1738
+ "step": 2470
1739
+ },
1740
+ {
1741
+ "epoch": 0.6713589604764483,
1742
+ "grad_norm": 1.309294160787172,
1743
+ "learning_rate": 3.734008291586195e-05,
1744
+ "loss": 1.3788,
1745
+ "step": 2480
1746
+ },
1747
+ {
1748
+ "epoch": 0.6740660530590146,
1749
+ "grad_norm": 1.4372591663740029,
1750
+ "learning_rate": 3.724751755357937e-05,
1751
+ "loss": 1.3044,
1752
+ "step": 2490
1753
+ },
1754
+ {
1755
+ "epoch": 0.676773145641581,
1756
+ "grad_norm": 1.8886812778978754,
1757
+ "learning_rate": 3.71547307322695e-05,
1758
+ "loss": 1.3798,
1759
+ "step": 2500
1760
+ },
1761
+ {
1762
+ "epoch": 0.6794802382241473,
1763
+ "grad_norm": 1.9753319009276904,
1764
+ "learning_rate": 3.7061724129699233e-05,
1765
+ "loss": 1.4362,
1766
+ "step": 2510
1767
+ },
1768
+ {
1769
+ "epoch": 0.6821873308067136,
1770
+ "grad_norm": 1.5870592037312612,
1771
+ "learning_rate": 3.6968499427609493e-05,
1772
+ "loss": 1.434,
1773
+ "step": 2520
1774
+ },
1775
+ {
1776
+ "epoch": 0.6848944233892799,
1777
+ "grad_norm": 1.2460580041038836,
1778
+ "learning_rate": 3.6875058311684896e-05,
1779
+ "loss": 1.3991,
1780
+ "step": 2530
1781
+ },
1782
+ {
1783
+ "epoch": 0.6876015159718463,
1784
+ "grad_norm": 1.984255011434617,
1785
+ "learning_rate": 3.6781402471523244e-05,
1786
+ "loss": 1.426,
1787
+ "step": 2540
1788
+ },
1789
+ {
1790
+ "epoch": 0.6903086085544126,
1791
+ "grad_norm": 1.7368812988388407,
1792
+ "learning_rate": 3.6687533600604955e-05,
1793
+ "loss": 1.3894,
1794
+ "step": 2550
1795
+ },
1796
+ {
1797
+ "epoch": 0.6930157011369789,
1798
+ "grad_norm": 1.4619686462990809,
1799
+ "learning_rate": 3.659345339626247e-05,
1800
+ "loss": 1.4415,
1801
+ "step": 2560
1802
+ },
1803
+ {
1804
+ "epoch": 0.6957227937195453,
1805
+ "grad_norm": 1.2916377265272263,
1806
+ "learning_rate": 3.649916355964956e-05,
1807
+ "loss": 1.4081,
1808
+ "step": 2570
1809
+ },
1810
+ {
1811
+ "epoch": 0.6984298863021116,
1812
+ "grad_norm": 1.4083989609601546,
1813
+ "learning_rate": 3.640466579571052e-05,
1814
+ "loss": 1.3883,
1815
+ "step": 2580
1816
+ },
1817
+ {
1818
+ "epoch": 0.7011369788846779,
1819
+ "grad_norm": 1.7091470425749693,
1820
+ "learning_rate": 3.630996181314944e-05,
1821
+ "loss": 1.4253,
1822
+ "step": 2590
1823
+ },
1824
+ {
1825
+ "epoch": 0.7038440714672441,
1826
+ "grad_norm": 1.3001138215713586,
1827
+ "learning_rate": 3.621505332439918e-05,
1828
+ "loss": 1.4529,
1829
+ "step": 2600
1830
+ },
1831
+ {
1832
+ "epoch": 0.7065511640498106,
1833
+ "grad_norm": 1.7216936795211795,
1834
+ "learning_rate": 3.611994204559051e-05,
1835
+ "loss": 1.4673,
1836
+ "step": 2610
1837
+ },
1838
+ {
1839
+ "epoch": 0.7092582566323768,
1840
+ "grad_norm": 1.781463187900956,
1841
+ "learning_rate": 3.602462969652102e-05,
1842
+ "loss": 1.4071,
1843
+ "step": 2620
1844
+ },
1845
+ {
1846
+ "epoch": 0.7119653492149431,
1847
+ "grad_norm": 1.3210728327580619,
1848
+ "learning_rate": 3.5929118000624064e-05,
1849
+ "loss": 1.394,
1850
+ "step": 2630
1851
+ },
1852
+ {
1853
+ "epoch": 0.7146724417975094,
1854
+ "grad_norm": 1.7986279315081266,
1855
+ "learning_rate": 3.5833408684937563e-05,
1856
+ "loss": 1.4134,
1857
+ "step": 2640
1858
+ },
1859
+ {
1860
+ "epoch": 0.7173795343800758,
1861
+ "grad_norm": 1.8124093789053668,
1862
+ "learning_rate": 3.573750348007278e-05,
1863
+ "loss": 1.395,
1864
+ "step": 2650
1865
+ },
1866
+ {
1867
+ "epoch": 0.7200866269626421,
1868
+ "grad_norm": 1.3733544619692244,
1869
+ "learning_rate": 3.564140412018306e-05,
1870
+ "loss": 1.4952,
1871
+ "step": 2660
1872
+ },
1873
+ {
1874
+ "epoch": 0.7227937195452084,
1875
+ "grad_norm": 1.1625577227023642,
1876
+ "learning_rate": 3.554511234293243e-05,
1877
+ "loss": 1.3995,
1878
+ "step": 2670
1879
+ },
1880
+ {
1881
+ "epoch": 0.7255008121277747,
1882
+ "grad_norm": 1.1967978911566575,
1883
+ "learning_rate": 3.54486298894642e-05,
1884
+ "loss": 1.3751,
1885
+ "step": 2680
1886
+ },
1887
+ {
1888
+ "epoch": 0.7282079047103411,
1889
+ "grad_norm": 1.5901060329868977,
1890
+ "learning_rate": 3.5351958504369504e-05,
1891
+ "loss": 1.3714,
1892
+ "step": 2690
1893
+ },
1894
+ {
1895
+ "epoch": 0.7309149972929074,
1896
+ "grad_norm": 1.4331430962829053,
1897
+ "learning_rate": 3.525509993565569e-05,
1898
+ "loss": 1.4538,
1899
+ "step": 2700
1900
+ },
1901
+ {
1902
+ "epoch": 0.7336220898754737,
1903
+ "grad_norm": 1.578052835087518,
1904
+ "learning_rate": 3.5158055934714784e-05,
1905
+ "loss": 1.4321,
1906
+ "step": 2710
1907
+ },
1908
+ {
1909
+ "epoch": 0.73632918245804,
1910
+ "grad_norm": 1.4900021275863136,
1911
+ "learning_rate": 3.506082825629176e-05,
1912
+ "loss": 1.4224,
1913
+ "step": 2720
1914
+ },
1915
+ {
1916
+ "epoch": 0.7390362750406064,
1917
+ "grad_norm": 1.803432159658249,
1918
+ "learning_rate": 3.496341865845286e-05,
1919
+ "loss": 1.4066,
1920
+ "step": 2730
1921
+ },
1922
+ {
1923
+ "epoch": 0.7417433676231727,
1924
+ "grad_norm": 1.8906057321665828,
1925
+ "learning_rate": 3.4865828902553786e-05,
1926
+ "loss": 1.3412,
1927
+ "step": 2740
1928
+ },
1929
+ {
1930
+ "epoch": 0.744450460205739,
1931
+ "grad_norm": 1.5022839535897015,
1932
+ "learning_rate": 3.476806075320784e-05,
1933
+ "loss": 1.4204,
1934
+ "step": 2750
1935
+ },
1936
+ {
1937
+ "epoch": 0.7471575527883053,
1938
+ "grad_norm": 1.2487314478311355,
1939
+ "learning_rate": 3.467011597825402e-05,
1940
+ "loss": 1.3934,
1941
+ "step": 2760
1942
+ },
1943
+ {
1944
+ "epoch": 0.7498646453708717,
1945
+ "grad_norm": 1.546439548647951,
1946
+ "learning_rate": 3.4571996348725066e-05,
1947
+ "loss": 1.3971,
1948
+ "step": 2770
1949
+ },
1950
+ {
1951
+ "epoch": 0.752571737953438,
1952
+ "grad_norm": 1.2557741585940914,
1953
+ "learning_rate": 3.447370363881543e-05,
1954
+ "loss": 1.4221,
1955
+ "step": 2780
1956
+ },
1957
+ {
1958
+ "epoch": 0.7552788305360043,
1959
+ "grad_norm": 1.3960693603390584,
1960
+ "learning_rate": 3.437523962584923e-05,
1961
+ "loss": 1.4134,
1962
+ "step": 2790
1963
+ },
1964
+ {
1965
+ "epoch": 0.7579859231185706,
1966
+ "grad_norm": 1.5039396517235502,
1967
+ "learning_rate": 3.427660609024799e-05,
1968
+ "loss": 1.3927,
1969
+ "step": 2800
1970
+ },
1971
+ {
1972
+ "epoch": 0.760693015701137,
1973
+ "grad_norm": 1.0815563013041047,
1974
+ "learning_rate": 3.417780481549863e-05,
1975
+ "loss": 1.3605,
1976
+ "step": 2810
1977
+ },
1978
+ {
1979
+ "epoch": 0.7634001082837033,
1980
+ "grad_norm": 1.3586955443715294,
1981
+ "learning_rate": 3.407883758812106e-05,
1982
+ "loss": 1.4041,
1983
+ "step": 2820
1984
+ },
1985
+ {
1986
+ "epoch": 0.7661072008662696,
1987
+ "grad_norm": 1.0920930772742923,
1988
+ "learning_rate": 3.397970619763597e-05,
1989
+ "loss": 1.428,
1990
+ "step": 2830
1991
+ },
1992
+ {
1993
+ "epoch": 0.7688142934488359,
1994
+ "grad_norm": 1.3785331176013547,
1995
+ "learning_rate": 3.388041243653242e-05,
1996
+ "loss": 1.4384,
1997
+ "step": 2840
1998
+ },
1999
+ {
2000
+ "epoch": 0.7715213860314023,
2001
+ "grad_norm": 1.3052185999792532,
2002
+ "learning_rate": 3.3780958100235445e-05,
2003
+ "loss": 1.3429,
2004
+ "step": 2850
2005
+ },
2006
+ {
2007
+ "epoch": 0.7742284786139686,
2008
+ "grad_norm": 2.042603724777467,
2009
+ "learning_rate": 3.3681344987073615e-05,
2010
+ "loss": 1.3328,
2011
+ "step": 2860
2012
+ },
2013
+ {
2014
+ "epoch": 0.7769355711965349,
2015
+ "grad_norm": 1.441878338125671,
2016
+ "learning_rate": 3.358157489824648e-05,
2017
+ "loss": 1.4065,
2018
+ "step": 2870
2019
+ },
2020
+ {
2021
+ "epoch": 0.7796426637791013,
2022
+ "grad_norm": 1.35241437286716,
2023
+ "learning_rate": 3.348164963779202e-05,
2024
+ "loss": 1.4298,
2025
+ "step": 2880
2026
+ },
2027
+ {
2028
+ "epoch": 0.7823497563616676,
2029
+ "grad_norm": 1.6143507738425422,
2030
+ "learning_rate": 3.3381571012554024e-05,
2031
+ "loss": 1.3649,
2032
+ "step": 2890
2033
+ },
2034
+ {
2035
+ "epoch": 0.7850568489442339,
2036
+ "grad_norm": 1.4514578355155894,
2037
+ "learning_rate": 3.328134083214941e-05,
2038
+ "loss": 1.3838,
2039
+ "step": 2900
2040
+ },
2041
+ {
2042
+ "epoch": 0.7877639415268002,
2043
+ "grad_norm": 1.515687318078472,
2044
+ "learning_rate": 3.318096090893551e-05,
2045
+ "loss": 1.4035,
2046
+ "step": 2910
2047
+ },
2048
+ {
2049
+ "epoch": 0.7904710341093666,
2050
+ "grad_norm": 1.2884021166011788,
2051
+ "learning_rate": 3.308043305797729e-05,
2052
+ "loss": 1.3559,
2053
+ "step": 2920
2054
+ },
2055
+ {
2056
+ "epoch": 0.7931781266919329,
2057
+ "grad_norm": 1.4337183263775037,
2058
+ "learning_rate": 3.297975909701457e-05,
2059
+ "loss": 1.4059,
2060
+ "step": 2930
2061
+ },
2062
+ {
2063
+ "epoch": 0.7958852192744992,
2064
+ "grad_norm": 1.5371766509521272,
2065
+ "learning_rate": 3.2878940846429085e-05,
2066
+ "loss": 1.3774,
2067
+ "step": 2940
2068
+ },
2069
+ {
2070
+ "epoch": 0.7985923118570655,
2071
+ "grad_norm": 1.3405567316289324,
2072
+ "learning_rate": 3.277798012921163e-05,
2073
+ "loss": 1.3904,
2074
+ "step": 2950
2075
+ },
2076
+ {
2077
+ "epoch": 0.8012994044396319,
2078
+ "grad_norm": 1.4555652196907567,
2079
+ "learning_rate": 3.2676878770929075e-05,
2080
+ "loss": 1.4146,
2081
+ "step": 2960
2082
+ },
2083
+ {
2084
+ "epoch": 0.8040064970221982,
2085
+ "grad_norm": 1.4152140878656931,
2086
+ "learning_rate": 3.257563859969134e-05,
2087
+ "loss": 1.4307,
2088
+ "step": 2970
2089
+ },
2090
+ {
2091
+ "epoch": 0.8067135896047645,
2092
+ "grad_norm": 1.4286723786741538,
2093
+ "learning_rate": 3.2474261446118384e-05,
2094
+ "loss": 1.3802,
2095
+ "step": 2980
2096
+ },
2097
+ {
2098
+ "epoch": 0.8094206821873308,
2099
+ "grad_norm": 1.7253495020201928,
2100
+ "learning_rate": 3.237274914330704e-05,
2101
+ "loss": 1.3832,
2102
+ "step": 2990
2103
+ },
2104
+ {
2105
+ "epoch": 0.8121277747698972,
2106
+ "grad_norm": 1.42537117722243,
2107
+ "learning_rate": 3.227110352679791e-05,
2108
+ "loss": 1.4196,
2109
+ "step": 3000
2110
+ },
2111
+ {
2112
+ "epoch": 0.8148348673524635,
2113
+ "grad_norm": 1.6839610738971422,
2114
+ "learning_rate": 3.216932643454219e-05,
2115
+ "loss": 1.3631,
2116
+ "step": 3010
2117
+ },
2118
+ {
2119
+ "epoch": 0.8175419599350298,
2120
+ "grad_norm": 1.971059453029315,
2121
+ "learning_rate": 3.2067419706868404e-05,
2122
+ "loss": 1.3943,
2123
+ "step": 3020
2124
+ },
2125
+ {
2126
+ "epoch": 0.8202490525175961,
2127
+ "grad_norm": 1.4912077914008548,
2128
+ "learning_rate": 3.196538518644911e-05,
2129
+ "loss": 1.3607,
2130
+ "step": 3030
2131
+ },
2132
+ {
2133
+ "epoch": 0.8229561451001625,
2134
+ "grad_norm": 2.642611457217887,
2135
+ "learning_rate": 3.186322471826765e-05,
2136
+ "loss": 1.3885,
2137
+ "step": 3040
2138
+ },
2139
+ {
2140
+ "epoch": 0.8256632376827288,
2141
+ "grad_norm": 2.0970669091185052,
2142
+ "learning_rate": 3.176094014958473e-05,
2143
+ "loss": 1.4174,
2144
+ "step": 3050
2145
+ },
2146
+ {
2147
+ "epoch": 0.828370330265295,
2148
+ "grad_norm": 1.6411214953570172,
2149
+ "learning_rate": 3.1658533329905034e-05,
2150
+ "loss": 1.3908,
2151
+ "step": 3060
2152
+ },
2153
+ {
2154
+ "epoch": 0.8310774228478613,
2155
+ "grad_norm": 1.4488420914686069,
2156
+ "learning_rate": 3.155600611094377e-05,
2157
+ "loss": 1.3525,
2158
+ "step": 3070
2159
+ },
2160
+ {
2161
+ "epoch": 0.8337845154304278,
2162
+ "grad_norm": 1.3921616791086928,
2163
+ "learning_rate": 3.145336034659322e-05,
2164
+ "loss": 1.3815,
2165
+ "step": 3080
2166
+ },
2167
+ {
2168
+ "epoch": 0.836491608012994,
2169
+ "grad_norm": 1.3485960002369286,
2170
+ "learning_rate": 3.135059789288919e-05,
2171
+ "loss": 1.3718,
2172
+ "step": 3090
2173
+ },
2174
+ {
2175
+ "epoch": 0.8391987005955603,
2176
+ "grad_norm": 3.0257814626066333,
2177
+ "learning_rate": 3.1247720607977465e-05,
2178
+ "loss": 1.3773,
2179
+ "step": 3100
2180
+ },
2181
+ {
2182
+ "epoch": 0.8419057931781266,
2183
+ "grad_norm": 1.607580731614463,
2184
+ "learning_rate": 3.11447303520802e-05,
2185
+ "loss": 1.3527,
2186
+ "step": 3110
2187
+ },
2188
+ {
2189
+ "epoch": 0.844612885760693,
2190
+ "grad_norm": 1.370395503929159,
2191
+ "learning_rate": 3.1041628987462256e-05,
2192
+ "loss": 1.3978,
2193
+ "step": 3120
2194
+ },
2195
+ {
2196
+ "epoch": 0.8473199783432593,
2197
+ "grad_norm": 1.6192384678646978,
2198
+ "learning_rate": 3.0938418378397615e-05,
2199
+ "loss": 1.3757,
2200
+ "step": 3130
2201
+ },
2202
+ {
2203
+ "epoch": 0.8500270709258256,
2204
+ "grad_norm": 1.4273589649993463,
2205
+ "learning_rate": 3.0835100391135555e-05,
2206
+ "loss": 1.3831,
2207
+ "step": 3140
2208
+ },
2209
+ {
2210
+ "epoch": 0.852734163508392,
2211
+ "grad_norm": 1.35446316177469,
2212
+ "learning_rate": 3.0731676893866995e-05,
2213
+ "loss": 1.3877,
2214
+ "step": 3150
2215
+ },
2216
+ {
2217
+ "epoch": 0.8554412560909583,
2218
+ "grad_norm": 2.163498528349308,
2219
+ "learning_rate": 3.062814975669067e-05,
2220
+ "loss": 1.3685,
2221
+ "step": 3160
2222
+ },
2223
+ {
2224
+ "epoch": 0.8581483486735246,
2225
+ "grad_norm": 1.3900238288203155,
2226
+ "learning_rate": 3.052452085157933e-05,
2227
+ "loss": 1.358,
2228
+ "step": 3170
2229
+ },
2230
+ {
2231
+ "epoch": 0.8608554412560909,
2232
+ "grad_norm": 1.216239709612049,
2233
+ "learning_rate": 3.042079205234589e-05,
2234
+ "loss": 1.3548,
2235
+ "step": 3180
2236
+ },
2237
+ {
2238
+ "epoch": 0.8635625338386573,
2239
+ "grad_norm": 1.349220629569809,
2240
+ "learning_rate": 3.0316965234609535e-05,
2241
+ "loss": 1.3874,
2242
+ "step": 3190
2243
+ },
2244
+ {
2245
+ "epoch": 0.8662696264212236,
2246
+ "grad_norm": 1.1250681117090537,
2247
+ "learning_rate": 3.021304227576182e-05,
2248
+ "loss": 1.3876,
2249
+ "step": 3200
2250
+ },
2251
+ {
2252
+ "epoch": 0.8689767190037899,
2253
+ "grad_norm": 1.3913714515749762,
2254
+ "learning_rate": 3.0109025054932744e-05,
2255
+ "loss": 1.347,
2256
+ "step": 3210
2257
+ },
2258
+ {
2259
+ "epoch": 0.8716838115863562,
2260
+ "grad_norm": 1.0609773784043655,
2261
+ "learning_rate": 3.0004915452956695e-05,
2262
+ "loss": 1.3667,
2263
+ "step": 3220
2264
+ },
2265
+ {
2266
+ "epoch": 0.8743909041689226,
2267
+ "grad_norm": 1.4538812767833513,
2268
+ "learning_rate": 2.9900715352338538e-05,
2269
+ "loss": 1.3696,
2270
+ "step": 3230
2271
+ },
2272
+ {
2273
+ "epoch": 0.8770979967514889,
2274
+ "grad_norm": 1.7108604924584878,
2275
+ "learning_rate": 2.979642663721951e-05,
2276
+ "loss": 1.3795,
2277
+ "step": 3240
2278
+ },
2279
+ {
2280
+ "epoch": 0.8798050893340552,
2281
+ "grad_norm": 2.5139594534456546,
2282
+ "learning_rate": 2.9692051193343184e-05,
2283
+ "loss": 1.3734,
2284
+ "step": 3250
2285
+ },
2286
+ {
2287
+ "epoch": 0.8825121819166215,
2288
+ "grad_norm": 2.5623897245796736,
2289
+ "learning_rate": 2.958759090802134e-05,
2290
+ "loss": 1.372,
2291
+ "step": 3260
2292
+ },
2293
+ {
2294
+ "epoch": 0.8852192744991879,
2295
+ "grad_norm": 1.7887583210364995,
2296
+ "learning_rate": 2.948304767009986e-05,
2297
+ "loss": 1.3814,
2298
+ "step": 3270
2299
+ },
2300
+ {
2301
+ "epoch": 0.8879263670817542,
2302
+ "grad_norm": 1.0710727849835535,
2303
+ "learning_rate": 2.93784233699246e-05,
2304
+ "loss": 1.4349,
2305
+ "step": 3280
2306
+ },
2307
+ {
2308
+ "epoch": 0.8906334596643205,
2309
+ "grad_norm": 2.563978550658464,
2310
+ "learning_rate": 2.927371989930714e-05,
2311
+ "loss": 1.3437,
2312
+ "step": 3290
2313
+ },
2314
+ {
2315
+ "epoch": 0.8933405522468868,
2316
+ "grad_norm": 2.145609531812958,
2317
+ "learning_rate": 2.9168939151490638e-05,
2318
+ "loss": 1.4138,
2319
+ "step": 3300
2320
+ },
2321
+ {
2322
+ "epoch": 0.8960476448294532,
2323
+ "grad_norm": 1.4751882863646832,
2324
+ "learning_rate": 2.906408302111556e-05,
2325
+ "loss": 1.3525,
2326
+ "step": 3310
2327
+ },
2328
+ {
2329
+ "epoch": 0.8987547374120195,
2330
+ "grad_norm": 1.499040784897915,
2331
+ "learning_rate": 2.8959153404185468e-05,
2332
+ "loss": 1.3717,
2333
+ "step": 3320
2334
+ },
2335
+ {
2336
+ "epoch": 0.9014618299945858,
2337
+ "grad_norm": 1.3438190150311098,
2338
+ "learning_rate": 2.885415219803267e-05,
2339
+ "loss": 1.3545,
2340
+ "step": 3330
2341
+ },
2342
+ {
2343
+ "epoch": 0.9041689225771521,
2344
+ "grad_norm": 1.4506182307044833,
2345
+ "learning_rate": 2.8749081301283955e-05,
2346
+ "loss": 1.3841,
2347
+ "step": 3340
2348
+ },
2349
+ {
2350
+ "epoch": 0.9068760151597185,
2351
+ "grad_norm": 1.5878698130305084,
2352
+ "learning_rate": 2.8643942613826264e-05,
2353
+ "loss": 1.3263,
2354
+ "step": 3350
2355
+ },
2356
+ {
2357
+ "epoch": 0.9095831077422848,
2358
+ "grad_norm": 1.3742244372155115,
2359
+ "learning_rate": 2.8538738036772318e-05,
2360
+ "loss": 1.3403,
2361
+ "step": 3360
2362
+ },
2363
+ {
2364
+ "epoch": 0.9122902003248511,
2365
+ "grad_norm": 1.2893381483305373,
2366
+ "learning_rate": 2.8433469472426257e-05,
2367
+ "loss": 1.3252,
2368
+ "step": 3370
2369
+ },
2370
+ {
2371
+ "epoch": 0.9149972929074174,
2372
+ "grad_norm": 1.7256165310292995,
2373
+ "learning_rate": 2.832813882424923e-05,
2374
+ "loss": 1.3595,
2375
+ "step": 3380
2376
+ },
2377
+ {
2378
+ "epoch": 0.9177043854899838,
2379
+ "grad_norm": 1.574410258409617,
2380
+ "learning_rate": 2.8222747996824973e-05,
2381
+ "loss": 1.3498,
2382
+ "step": 3390
2383
+ },
2384
+ {
2385
+ "epoch": 0.9204114780725501,
2386
+ "grad_norm": 1.4304720741459473,
2387
+ "learning_rate": 2.8117298895825417e-05,
2388
+ "loss": 1.3613,
2389
+ "step": 3400
2390
+ },
2391
+ {
2392
+ "epoch": 0.9231185706551164,
2393
+ "grad_norm": 1.8155878547283097,
2394
+ "learning_rate": 2.8011793427976152e-05,
2395
+ "loss": 1.3675,
2396
+ "step": 3410
2397
+ },
2398
+ {
2399
+ "epoch": 0.9258256632376828,
2400
+ "grad_norm": 1.568315689782668,
2401
+ "learning_rate": 2.7906233501022005e-05,
2402
+ "loss": 1.3878,
2403
+ "step": 3420
2404
+ },
2405
+ {
2406
+ "epoch": 0.9285327558202491,
2407
+ "grad_norm": 1.3449271497175002,
2408
+ "learning_rate": 2.780062102369255e-05,
2409
+ "loss": 1.3451,
2410
+ "step": 3430
2411
+ },
2412
+ {
2413
+ "epoch": 0.9312398484028154,
2414
+ "grad_norm": 1.8796135561662028,
2415
+ "learning_rate": 2.7694957905667533e-05,
2416
+ "loss": 1.3541,
2417
+ "step": 3440
2418
+ },
2419
+ {
2420
+ "epoch": 0.9339469409853817,
2421
+ "grad_norm": 1.5881147795174977,
2422
+ "learning_rate": 2.7589246057542424e-05,
2423
+ "loss": 1.3481,
2424
+ "step": 3450
2425
+ },
2426
+ {
2427
+ "epoch": 0.9366540335679481,
2428
+ "grad_norm": 1.5669861875980806,
2429
+ "learning_rate": 2.74834873907938e-05,
2430
+ "loss": 1.3049,
2431
+ "step": 3460
2432
+ },
2433
+ {
2434
+ "epoch": 0.9393611261505144,
2435
+ "grad_norm": 1.3159700318764556,
2436
+ "learning_rate": 2.7377683817744825e-05,
2437
+ "loss": 1.3724,
2438
+ "step": 3470
2439
+ },
2440
+ {
2441
+ "epoch": 0.9420682187330807,
2442
+ "grad_norm": 1.1861068665908672,
2443
+ "learning_rate": 2.727183725153066e-05,
2444
+ "loss": 1.3996,
2445
+ "step": 3480
2446
+ },
2447
+ {
2448
+ "epoch": 0.944775311315647,
2449
+ "grad_norm": 1.588796314194285,
2450
+ "learning_rate": 2.7165949606063838e-05,
2451
+ "loss": 1.3868,
2452
+ "step": 3490
2453
+ },
2454
+ {
2455
+ "epoch": 0.9474824038982134,
2456
+ "grad_norm": 1.8405833406640584,
2457
+ "learning_rate": 2.7060022795999722e-05,
2458
+ "loss": 1.3248,
2459
+ "step": 3500
2460
+ },
2461
+ {
2462
+ "epoch": 0.9501894964807797,
2463
+ "grad_norm": 1.4845932153084007,
2464
+ "learning_rate": 2.6954058736701825e-05,
2465
+ "loss": 1.3113,
2466
+ "step": 3510
2467
+ },
2468
+ {
2469
+ "epoch": 0.952896589063346,
2470
+ "grad_norm": 2.2421879491898347,
2471
+ "learning_rate": 2.684805934420721e-05,
2472
+ "loss": 1.3569,
2473
+ "step": 3520
2474
+ },
2475
+ {
2476
+ "epoch": 0.9556036816459123,
2477
+ "grad_norm": 1.8263161858131314,
2478
+ "learning_rate": 2.674202653519183e-05,
2479
+ "loss": 1.353,
2480
+ "step": 3530
2481
+ },
2482
+ {
2483
+ "epoch": 0.9583107742284787,
2484
+ "grad_norm": 1.6830556154140768,
2485
+ "learning_rate": 2.6635962226935844e-05,
2486
+ "loss": 1.3581,
2487
+ "step": 3540
2488
+ },
2489
+ {
2490
+ "epoch": 0.961017866811045,
2491
+ "grad_norm": 1.1962057897073894,
2492
+ "learning_rate": 2.652986833728905e-05,
2493
+ "loss": 1.3377,
2494
+ "step": 3550
2495
+ },
2496
+ {
2497
+ "epoch": 0.9637249593936112,
2498
+ "grad_norm": 1.1605034457482457,
2499
+ "learning_rate": 2.642374678463606e-05,
2500
+ "loss": 1.3142,
2501
+ "step": 3560
2502
+ },
2503
+ {
2504
+ "epoch": 0.9664320519761775,
2505
+ "grad_norm": 1.6116484287446706,
2506
+ "learning_rate": 2.6317599487861732e-05,
2507
+ "loss": 1.3704,
2508
+ "step": 3570
2509
+ },
2510
+ {
2511
+ "epoch": 0.969139144558744,
2512
+ "grad_norm": 2.022789120877381,
2513
+ "learning_rate": 2.6211428366316414e-05,
2514
+ "loss": 1.3188,
2515
+ "step": 3580
2516
+ },
2517
+ {
2518
+ "epoch": 0.9718462371413102,
2519
+ "grad_norm": 2.0186900985710245,
2520
+ "learning_rate": 2.6105235339781254e-05,
2521
+ "loss": 1.3168,
2522
+ "step": 3590
2523
+ },
2524
+ {
2525
+ "epoch": 0.9745533297238765,
2526
+ "grad_norm": 2.040600986429446,
2527
+ "learning_rate": 2.599902232843348e-05,
2528
+ "loss": 1.361,
2529
+ "step": 3600
2530
+ },
2531
+ {
2532
+ "epoch": 0.9772604223064428,
2533
+ "grad_norm": 1.2311848491778494,
2534
+ "learning_rate": 2.5892791252811694e-05,
2535
+ "loss": 1.3658,
2536
+ "step": 3610
2537
+ },
2538
+ {
2539
+ "epoch": 0.9799675148890092,
2540
+ "grad_norm": 1.8540679978665582,
2541
+ "learning_rate": 2.578654403378112e-05,
2542
+ "loss": 1.373,
2543
+ "step": 3620
2544
+ },
2545
+ {
2546
+ "epoch": 0.9826746074715755,
2547
+ "grad_norm": 1.3073823909407571,
2548
+ "learning_rate": 2.5680282592498913e-05,
2549
+ "loss": 1.3845,
2550
+ "step": 3630
2551
+ },
2552
+ {
2553
+ "epoch": 0.9853817000541418,
2554
+ "grad_norm": 1.6720630501211762,
2555
+ "learning_rate": 2.5574008850379366e-05,
2556
+ "loss": 1.3438,
2557
+ "step": 3640
2558
+ },
2559
+ {
2560
+ "epoch": 0.9880887926367081,
2561
+ "grad_norm": 1.4546581832324275,
2562
+ "learning_rate": 2.546772472905921e-05,
2563
+ "loss": 1.2815,
2564
+ "step": 3650
2565
+ },
2566
+ {
2567
+ "epoch": 0.9907958852192745,
2568
+ "grad_norm": 1.4509964601293808,
2569
+ "learning_rate": 2.5361432150362848e-05,
2570
+ "loss": 1.3469,
2571
+ "step": 3660
2572
+ },
2573
+ {
2574
+ "epoch": 0.9935029778018408,
2575
+ "grad_norm": 1.4316417505770653,
2576
+ "learning_rate": 2.5255133036267608e-05,
2577
+ "loss": 1.3315,
2578
+ "step": 3670
2579
+ },
2580
+ {
2581
+ "epoch": 0.9962100703844071,
2582
+ "grad_norm": 1.7706024949129144,
2583
+ "learning_rate": 2.5148829308869004e-05,
2584
+ "loss": 1.3607,
2585
+ "step": 3680
2586
+ },
2587
+ {
2588
+ "epoch": 0.9989171629669734,
2589
+ "grad_norm": 1.5222816405267459,
2590
+ "learning_rate": 2.5042522890345936e-05,
2591
+ "loss": 1.2815,
2592
+ "step": 3690
2593
+ }
2594
+ ],
2595
+ "logging_steps": 10,
2596
+ "max_steps": 7388,
2597
+ "num_input_tokens_seen": 0,
2598
+ "num_train_epochs": 2,
2599
+ "save_steps": 500,
2600
+ "total_flos": 1754891515854848.0,
2601
+ "train_batch_size": 1,
2602
+ "trial_name": null,
2603
+ "trial_params": null
2604
+ }
training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6fe531666a82ca88ea44768b422f5e09b21034d1b63638087ab6cd9d654b3fab
3
+ size 6459
vocab.json ADDED
The diff for this file is too large to render. See raw diff