Delta-Vector commited on
Commit
26cf120
·
verified ·
1 Parent(s): 4876489

Upload folder using huggingface_hub

Browse files
.gitattributes CHANGED
@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
 
 
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ tokenizer.json filter=lfs diff=lfs merge=lfs -text
config.json ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "MistralForCausalLM"
4
+ ],
5
+ "attention_dropout": 0.0,
6
+ "bos_token_id": 1,
7
+ "eos_token_id": 2,
8
+ "head_dim": 128,
9
+ "hidden_act": "silu",
10
+ "hidden_size": 5120,
11
+ "initializer_range": 0.02,
12
+ "intermediate_size": 32768,
13
+ "max_position_embeddings": 131072,
14
+ "model_type": "mistral",
15
+ "num_attention_heads": 32,
16
+ "num_hidden_layers": 40,
17
+ "num_key_value_heads": 8,
18
+ "rms_norm_eps": 1e-05,
19
+ "rope_theta": 1000000000.0,
20
+ "sliding_window": null,
21
+ "tie_word_embeddings": false,
22
+ "torch_dtype": "bfloat16",
23
+ "transformers_version": "4.51.3",
24
+ "use_cache": false,
25
+ "vocab_size": 131072
26
+ }
generation_config.json ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 1,
4
+ "do_sample": true,
5
+ "eos_token_id": 2,
6
+ "transformers_version": "4.51.3"
7
+ }
latest ADDED
@@ -0,0 +1 @@
 
 
1
+ global_step255
model-00001-of-00010.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:32433c65657ef5ea09a4f0974ec22ee1521961a293720125bcd7d5707aacdce2
3
+ size 4781571736
model-00002-of-00010.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:360176f6f97a32856bb7634f393c0d65768ea699496c96b9fd5fa7b0ff622743
3
+ size 4781592784
model-00003-of-00010.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:316252e89b322fac157cf360301955e298037aa931b2ef0440747f8dca6c54a3
3
+ size 4781592800
model-00004-of-00010.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0e6f9bf29fe4060e7fae78e60f7b7b419db8149a5b88d76685df80a45ab95cb9
3
+ size 4886471600
model-00005-of-00010.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b89556a9fc6e80238aa9ca2631cdba80a29e49b7462d1596bc714c7773e70cf2
3
+ size 4781592824
model-00006-of-00010.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c67e6390615f3fb3d8a30c1f9a0c8d2b38572735c07f85346166446f682f4907
3
+ size 4781592816
model-00007-of-00010.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:515a86e0eb9037ee3a775fee9f736e1c17262fd4c336fb11e91ffdf71d0de437
3
+ size 4886471600
model-00008-of-00010.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1b3f3f40a754dd6687bd25b40dd47604b686409b87d5bfb6ccf667728d8bcf34
3
+ size 4781592824
model-00009-of-00010.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:81bb6549dfd807ad33bb6eaa8e75042df7b82ef36d3dbe8e88d1e50c45bd1c49
3
+ size 4781592816
model-00010-of-00010.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ae14fa2f6624c128142f2ca3f6f33aa683e7d38b4ad46845277dffe5a1593b77
3
+ size 3900777072
model.safetensors.index.json ADDED
@@ -0,0 +1,370 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "metadata": {
3
+ "total_size": 47144806400
4
+ },
5
+ "weight_map": {
6
+ "lm_head.weight": "model-00010-of-00010.safetensors",
7
+ "model.embed_tokens.weight": "model-00001-of-00010.safetensors",
8
+ "model.layers.0.input_layernorm.weight": "model-00001-of-00010.safetensors",
9
+ "model.layers.0.mlp.down_proj.weight": "model-00001-of-00010.safetensors",
10
+ "model.layers.0.mlp.gate_proj.weight": "model-00001-of-00010.safetensors",
11
+ "model.layers.0.mlp.up_proj.weight": "model-00001-of-00010.safetensors",
12
+ "model.layers.0.post_attention_layernorm.weight": "model-00001-of-00010.safetensors",
13
+ "model.layers.0.self_attn.k_proj.weight": "model-00001-of-00010.safetensors",
14
+ "model.layers.0.self_attn.o_proj.weight": "model-00001-of-00010.safetensors",
15
+ "model.layers.0.self_attn.q_proj.weight": "model-00001-of-00010.safetensors",
16
+ "model.layers.0.self_attn.v_proj.weight": "model-00001-of-00010.safetensors",
17
+ "model.layers.1.input_layernorm.weight": "model-00001-of-00010.safetensors",
18
+ "model.layers.1.mlp.down_proj.weight": "model-00001-of-00010.safetensors",
19
+ "model.layers.1.mlp.gate_proj.weight": "model-00001-of-00010.safetensors",
20
+ "model.layers.1.mlp.up_proj.weight": "model-00001-of-00010.safetensors",
21
+ "model.layers.1.post_attention_layernorm.weight": "model-00001-of-00010.safetensors",
22
+ "model.layers.1.self_attn.k_proj.weight": "model-00001-of-00010.safetensors",
23
+ "model.layers.1.self_attn.o_proj.weight": "model-00001-of-00010.safetensors",
24
+ "model.layers.1.self_attn.q_proj.weight": "model-00001-of-00010.safetensors",
25
+ "model.layers.1.self_attn.v_proj.weight": "model-00001-of-00010.safetensors",
26
+ "model.layers.10.input_layernorm.weight": "model-00003-of-00010.safetensors",
27
+ "model.layers.10.mlp.down_proj.weight": "model-00003-of-00010.safetensors",
28
+ "model.layers.10.mlp.gate_proj.weight": "model-00003-of-00010.safetensors",
29
+ "model.layers.10.mlp.up_proj.weight": "model-00003-of-00010.safetensors",
30
+ "model.layers.10.post_attention_layernorm.weight": "model-00003-of-00010.safetensors",
31
+ "model.layers.10.self_attn.k_proj.weight": "model-00003-of-00010.safetensors",
32
+ "model.layers.10.self_attn.o_proj.weight": "model-00003-of-00010.safetensors",
33
+ "model.layers.10.self_attn.q_proj.weight": "model-00003-of-00010.safetensors",
34
+ "model.layers.10.self_attn.v_proj.weight": "model-00003-of-00010.safetensors",
35
+ "model.layers.11.input_layernorm.weight": "model-00004-of-00010.safetensors",
36
+ "model.layers.11.mlp.down_proj.weight": "model-00004-of-00010.safetensors",
37
+ "model.layers.11.mlp.gate_proj.weight": "model-00003-of-00010.safetensors",
38
+ "model.layers.11.mlp.up_proj.weight": "model-00003-of-00010.safetensors",
39
+ "model.layers.11.post_attention_layernorm.weight": "model-00004-of-00010.safetensors",
40
+ "model.layers.11.self_attn.k_proj.weight": "model-00003-of-00010.safetensors",
41
+ "model.layers.11.self_attn.o_proj.weight": "model-00003-of-00010.safetensors",
42
+ "model.layers.11.self_attn.q_proj.weight": "model-00003-of-00010.safetensors",
43
+ "model.layers.11.self_attn.v_proj.weight": "model-00003-of-00010.safetensors",
44
+ "model.layers.12.input_layernorm.weight": "model-00004-of-00010.safetensors",
45
+ "model.layers.12.mlp.down_proj.weight": "model-00004-of-00010.safetensors",
46
+ "model.layers.12.mlp.gate_proj.weight": "model-00004-of-00010.safetensors",
47
+ "model.layers.12.mlp.up_proj.weight": "model-00004-of-00010.safetensors",
48
+ "model.layers.12.post_attention_layernorm.weight": "model-00004-of-00010.safetensors",
49
+ "model.layers.12.self_attn.k_proj.weight": "model-00004-of-00010.safetensors",
50
+ "model.layers.12.self_attn.o_proj.weight": "model-00004-of-00010.safetensors",
51
+ "model.layers.12.self_attn.q_proj.weight": "model-00004-of-00010.safetensors",
52
+ "model.layers.12.self_attn.v_proj.weight": "model-00004-of-00010.safetensors",
53
+ "model.layers.13.input_layernorm.weight": "model-00004-of-00010.safetensors",
54
+ "model.layers.13.mlp.down_proj.weight": "model-00004-of-00010.safetensors",
55
+ "model.layers.13.mlp.gate_proj.weight": "model-00004-of-00010.safetensors",
56
+ "model.layers.13.mlp.up_proj.weight": "model-00004-of-00010.safetensors",
57
+ "model.layers.13.post_attention_layernorm.weight": "model-00004-of-00010.safetensors",
58
+ "model.layers.13.self_attn.k_proj.weight": "model-00004-of-00010.safetensors",
59
+ "model.layers.13.self_attn.o_proj.weight": "model-00004-of-00010.safetensors",
60
+ "model.layers.13.self_attn.q_proj.weight": "model-00004-of-00010.safetensors",
61
+ "model.layers.13.self_attn.v_proj.weight": "model-00004-of-00010.safetensors",
62
+ "model.layers.14.input_layernorm.weight": "model-00004-of-00010.safetensors",
63
+ "model.layers.14.mlp.down_proj.weight": "model-00004-of-00010.safetensors",
64
+ "model.layers.14.mlp.gate_proj.weight": "model-00004-of-00010.safetensors",
65
+ "model.layers.14.mlp.up_proj.weight": "model-00004-of-00010.safetensors",
66
+ "model.layers.14.post_attention_layernorm.weight": "model-00004-of-00010.safetensors",
67
+ "model.layers.14.self_attn.k_proj.weight": "model-00004-of-00010.safetensors",
68
+ "model.layers.14.self_attn.o_proj.weight": "model-00004-of-00010.safetensors",
69
+ "model.layers.14.self_attn.q_proj.weight": "model-00004-of-00010.safetensors",
70
+ "model.layers.14.self_attn.v_proj.weight": "model-00004-of-00010.safetensors",
71
+ "model.layers.15.input_layernorm.weight": "model-00004-of-00010.safetensors",
72
+ "model.layers.15.mlp.down_proj.weight": "model-00004-of-00010.safetensors",
73
+ "model.layers.15.mlp.gate_proj.weight": "model-00004-of-00010.safetensors",
74
+ "model.layers.15.mlp.up_proj.weight": "model-00004-of-00010.safetensors",
75
+ "model.layers.15.post_attention_layernorm.weight": "model-00004-of-00010.safetensors",
76
+ "model.layers.15.self_attn.k_proj.weight": "model-00004-of-00010.safetensors",
77
+ "model.layers.15.self_attn.o_proj.weight": "model-00004-of-00010.safetensors",
78
+ "model.layers.15.self_attn.q_proj.weight": "model-00004-of-00010.safetensors",
79
+ "model.layers.15.self_attn.v_proj.weight": "model-00004-of-00010.safetensors",
80
+ "model.layers.16.input_layernorm.weight": "model-00005-of-00010.safetensors",
81
+ "model.layers.16.mlp.down_proj.weight": "model-00005-of-00010.safetensors",
82
+ "model.layers.16.mlp.gate_proj.weight": "model-00005-of-00010.safetensors",
83
+ "model.layers.16.mlp.up_proj.weight": "model-00005-of-00010.safetensors",
84
+ "model.layers.16.post_attention_layernorm.weight": "model-00005-of-00010.safetensors",
85
+ "model.layers.16.self_attn.k_proj.weight": "model-00004-of-00010.safetensors",
86
+ "model.layers.16.self_attn.o_proj.weight": "model-00004-of-00010.safetensors",
87
+ "model.layers.16.self_attn.q_proj.weight": "model-00004-of-00010.safetensors",
88
+ "model.layers.16.self_attn.v_proj.weight": "model-00004-of-00010.safetensors",
89
+ "model.layers.17.input_layernorm.weight": "model-00005-of-00010.safetensors",
90
+ "model.layers.17.mlp.down_proj.weight": "model-00005-of-00010.safetensors",
91
+ "model.layers.17.mlp.gate_proj.weight": "model-00005-of-00010.safetensors",
92
+ "model.layers.17.mlp.up_proj.weight": "model-00005-of-00010.safetensors",
93
+ "model.layers.17.post_attention_layernorm.weight": "model-00005-of-00010.safetensors",
94
+ "model.layers.17.self_attn.k_proj.weight": "model-00005-of-00010.safetensors",
95
+ "model.layers.17.self_attn.o_proj.weight": "model-00005-of-00010.safetensors",
96
+ "model.layers.17.self_attn.q_proj.weight": "model-00005-of-00010.safetensors",
97
+ "model.layers.17.self_attn.v_proj.weight": "model-00005-of-00010.safetensors",
98
+ "model.layers.18.input_layernorm.weight": "model-00005-of-00010.safetensors",
99
+ "model.layers.18.mlp.down_proj.weight": "model-00005-of-00010.safetensors",
100
+ "model.layers.18.mlp.gate_proj.weight": "model-00005-of-00010.safetensors",
101
+ "model.layers.18.mlp.up_proj.weight": "model-00005-of-00010.safetensors",
102
+ "model.layers.18.post_attention_layernorm.weight": "model-00005-of-00010.safetensors",
103
+ "model.layers.18.self_attn.k_proj.weight": "model-00005-of-00010.safetensors",
104
+ "model.layers.18.self_attn.o_proj.weight": "model-00005-of-00010.safetensors",
105
+ "model.layers.18.self_attn.q_proj.weight": "model-00005-of-00010.safetensors",
106
+ "model.layers.18.self_attn.v_proj.weight": "model-00005-of-00010.safetensors",
107
+ "model.layers.19.input_layernorm.weight": "model-00005-of-00010.safetensors",
108
+ "model.layers.19.mlp.down_proj.weight": "model-00005-of-00010.safetensors",
109
+ "model.layers.19.mlp.gate_proj.weight": "model-00005-of-00010.safetensors",
110
+ "model.layers.19.mlp.up_proj.weight": "model-00005-of-00010.safetensors",
111
+ "model.layers.19.post_attention_layernorm.weight": "model-00005-of-00010.safetensors",
112
+ "model.layers.19.self_attn.k_proj.weight": "model-00005-of-00010.safetensors",
113
+ "model.layers.19.self_attn.o_proj.weight": "model-00005-of-00010.safetensors",
114
+ "model.layers.19.self_attn.q_proj.weight": "model-00005-of-00010.safetensors",
115
+ "model.layers.19.self_attn.v_proj.weight": "model-00005-of-00010.safetensors",
116
+ "model.layers.2.input_layernorm.weight": "model-00001-of-00010.safetensors",
117
+ "model.layers.2.mlp.down_proj.weight": "model-00001-of-00010.safetensors",
118
+ "model.layers.2.mlp.gate_proj.weight": "model-00001-of-00010.safetensors",
119
+ "model.layers.2.mlp.up_proj.weight": "model-00001-of-00010.safetensors",
120
+ "model.layers.2.post_attention_layernorm.weight": "model-00001-of-00010.safetensors",
121
+ "model.layers.2.self_attn.k_proj.weight": "model-00001-of-00010.safetensors",
122
+ "model.layers.2.self_attn.o_proj.weight": "model-00001-of-00010.safetensors",
123
+ "model.layers.2.self_attn.q_proj.weight": "model-00001-of-00010.safetensors",
124
+ "model.layers.2.self_attn.v_proj.weight": "model-00001-of-00010.safetensors",
125
+ "model.layers.20.input_layernorm.weight": "model-00006-of-00010.safetensors",
126
+ "model.layers.20.mlp.down_proj.weight": "model-00006-of-00010.safetensors",
127
+ "model.layers.20.mlp.gate_proj.weight": "model-00005-of-00010.safetensors",
128
+ "model.layers.20.mlp.up_proj.weight": "model-00006-of-00010.safetensors",
129
+ "model.layers.20.post_attention_layernorm.weight": "model-00006-of-00010.safetensors",
130
+ "model.layers.20.self_attn.k_proj.weight": "model-00005-of-00010.safetensors",
131
+ "model.layers.20.self_attn.o_proj.weight": "model-00005-of-00010.safetensors",
132
+ "model.layers.20.self_attn.q_proj.weight": "model-00005-of-00010.safetensors",
133
+ "model.layers.20.self_attn.v_proj.weight": "model-00005-of-00010.safetensors",
134
+ "model.layers.21.input_layernorm.weight": "model-00006-of-00010.safetensors",
135
+ "model.layers.21.mlp.down_proj.weight": "model-00006-of-00010.safetensors",
136
+ "model.layers.21.mlp.gate_proj.weight": "model-00006-of-00010.safetensors",
137
+ "model.layers.21.mlp.up_proj.weight": "model-00006-of-00010.safetensors",
138
+ "model.layers.21.post_attention_layernorm.weight": "model-00006-of-00010.safetensors",
139
+ "model.layers.21.self_attn.k_proj.weight": "model-00006-of-00010.safetensors",
140
+ "model.layers.21.self_attn.o_proj.weight": "model-00006-of-00010.safetensors",
141
+ "model.layers.21.self_attn.q_proj.weight": "model-00006-of-00010.safetensors",
142
+ "model.layers.21.self_attn.v_proj.weight": "model-00006-of-00010.safetensors",
143
+ "model.layers.22.input_layernorm.weight": "model-00006-of-00010.safetensors",
144
+ "model.layers.22.mlp.down_proj.weight": "model-00006-of-00010.safetensors",
145
+ "model.layers.22.mlp.gate_proj.weight": "model-00006-of-00010.safetensors",
146
+ "model.layers.22.mlp.up_proj.weight": "model-00006-of-00010.safetensors",
147
+ "model.layers.22.post_attention_layernorm.weight": "model-00006-of-00010.safetensors",
148
+ "model.layers.22.self_attn.k_proj.weight": "model-00006-of-00010.safetensors",
149
+ "model.layers.22.self_attn.o_proj.weight": "model-00006-of-00010.safetensors",
150
+ "model.layers.22.self_attn.q_proj.weight": "model-00006-of-00010.safetensors",
151
+ "model.layers.22.self_attn.v_proj.weight": "model-00006-of-00010.safetensors",
152
+ "model.layers.23.input_layernorm.weight": "model-00006-of-00010.safetensors",
153
+ "model.layers.23.mlp.down_proj.weight": "model-00006-of-00010.safetensors",
154
+ "model.layers.23.mlp.gate_proj.weight": "model-00006-of-00010.safetensors",
155
+ "model.layers.23.mlp.up_proj.weight": "model-00006-of-00010.safetensors",
156
+ "model.layers.23.post_attention_layernorm.weight": "model-00006-of-00010.safetensors",
157
+ "model.layers.23.self_attn.k_proj.weight": "model-00006-of-00010.safetensors",
158
+ "model.layers.23.self_attn.o_proj.weight": "model-00006-of-00010.safetensors",
159
+ "model.layers.23.self_attn.q_proj.weight": "model-00006-of-00010.safetensors",
160
+ "model.layers.23.self_attn.v_proj.weight": "model-00006-of-00010.safetensors",
161
+ "model.layers.24.input_layernorm.weight": "model-00007-of-00010.safetensors",
162
+ "model.layers.24.mlp.down_proj.weight": "model-00007-of-00010.safetensors",
163
+ "model.layers.24.mlp.gate_proj.weight": "model-00006-of-00010.safetensors",
164
+ "model.layers.24.mlp.up_proj.weight": "model-00006-of-00010.safetensors",
165
+ "model.layers.24.post_attention_layernorm.weight": "model-00007-of-00010.safetensors",
166
+ "model.layers.24.self_attn.k_proj.weight": "model-00006-of-00010.safetensors",
167
+ "model.layers.24.self_attn.o_proj.weight": "model-00006-of-00010.safetensors",
168
+ "model.layers.24.self_attn.q_proj.weight": "model-00006-of-00010.safetensors",
169
+ "model.layers.24.self_attn.v_proj.weight": "model-00006-of-00010.safetensors",
170
+ "model.layers.25.input_layernorm.weight": "model-00007-of-00010.safetensors",
171
+ "model.layers.25.mlp.down_proj.weight": "model-00007-of-00010.safetensors",
172
+ "model.layers.25.mlp.gate_proj.weight": "model-00007-of-00010.safetensors",
173
+ "model.layers.25.mlp.up_proj.weight": "model-00007-of-00010.safetensors",
174
+ "model.layers.25.post_attention_layernorm.weight": "model-00007-of-00010.safetensors",
175
+ "model.layers.25.self_attn.k_proj.weight": "model-00007-of-00010.safetensors",
176
+ "model.layers.25.self_attn.o_proj.weight": "model-00007-of-00010.safetensors",
177
+ "model.layers.25.self_attn.q_proj.weight": "model-00007-of-00010.safetensors",
178
+ "model.layers.25.self_attn.v_proj.weight": "model-00007-of-00010.safetensors",
179
+ "model.layers.26.input_layernorm.weight": "model-00007-of-00010.safetensors",
180
+ "model.layers.26.mlp.down_proj.weight": "model-00007-of-00010.safetensors",
181
+ "model.layers.26.mlp.gate_proj.weight": "model-00007-of-00010.safetensors",
182
+ "model.layers.26.mlp.up_proj.weight": "model-00007-of-00010.safetensors",
183
+ "model.layers.26.post_attention_layernorm.weight": "model-00007-of-00010.safetensors",
184
+ "model.layers.26.self_attn.k_proj.weight": "model-00007-of-00010.safetensors",
185
+ "model.layers.26.self_attn.o_proj.weight": "model-00007-of-00010.safetensors",
186
+ "model.layers.26.self_attn.q_proj.weight": "model-00007-of-00010.safetensors",
187
+ "model.layers.26.self_attn.v_proj.weight": "model-00007-of-00010.safetensors",
188
+ "model.layers.27.input_layernorm.weight": "model-00007-of-00010.safetensors",
189
+ "model.layers.27.mlp.down_proj.weight": "model-00007-of-00010.safetensors",
190
+ "model.layers.27.mlp.gate_proj.weight": "model-00007-of-00010.safetensors",
191
+ "model.layers.27.mlp.up_proj.weight": "model-00007-of-00010.safetensors",
192
+ "model.layers.27.post_attention_layernorm.weight": "model-00007-of-00010.safetensors",
193
+ "model.layers.27.self_attn.k_proj.weight": "model-00007-of-00010.safetensors",
194
+ "model.layers.27.self_attn.o_proj.weight": "model-00007-of-00010.safetensors",
195
+ "model.layers.27.self_attn.q_proj.weight": "model-00007-of-00010.safetensors",
196
+ "model.layers.27.self_attn.v_proj.weight": "model-00007-of-00010.safetensors",
197
+ "model.layers.28.input_layernorm.weight": "model-00007-of-00010.safetensors",
198
+ "model.layers.28.mlp.down_proj.weight": "model-00007-of-00010.safetensors",
199
+ "model.layers.28.mlp.gate_proj.weight": "model-00007-of-00010.safetensors",
200
+ "model.layers.28.mlp.up_proj.weight": "model-00007-of-00010.safetensors",
201
+ "model.layers.28.post_attention_layernorm.weight": "model-00007-of-00010.safetensors",
202
+ "model.layers.28.self_attn.k_proj.weight": "model-00007-of-00010.safetensors",
203
+ "model.layers.28.self_attn.o_proj.weight": "model-00007-of-00010.safetensors",
204
+ "model.layers.28.self_attn.q_proj.weight": "model-00007-of-00010.safetensors",
205
+ "model.layers.28.self_attn.v_proj.weight": "model-00007-of-00010.safetensors",
206
+ "model.layers.29.input_layernorm.weight": "model-00008-of-00010.safetensors",
207
+ "model.layers.29.mlp.down_proj.weight": "model-00008-of-00010.safetensors",
208
+ "model.layers.29.mlp.gate_proj.weight": "model-00008-of-00010.safetensors",
209
+ "model.layers.29.mlp.up_proj.weight": "model-00008-of-00010.safetensors",
210
+ "model.layers.29.post_attention_layernorm.weight": "model-00008-of-00010.safetensors",
211
+ "model.layers.29.self_attn.k_proj.weight": "model-00007-of-00010.safetensors",
212
+ "model.layers.29.self_attn.o_proj.weight": "model-00007-of-00010.safetensors",
213
+ "model.layers.29.self_attn.q_proj.weight": "model-00007-of-00010.safetensors",
214
+ "model.layers.29.self_attn.v_proj.weight": "model-00007-of-00010.safetensors",
215
+ "model.layers.3.input_layernorm.weight": "model-00002-of-00010.safetensors",
216
+ "model.layers.3.mlp.down_proj.weight": "model-00002-of-00010.safetensors",
217
+ "model.layers.3.mlp.gate_proj.weight": "model-00002-of-00010.safetensors",
218
+ "model.layers.3.mlp.up_proj.weight": "model-00002-of-00010.safetensors",
219
+ "model.layers.3.post_attention_layernorm.weight": "model-00002-of-00010.safetensors",
220
+ "model.layers.3.self_attn.k_proj.weight": "model-00001-of-00010.safetensors",
221
+ "model.layers.3.self_attn.o_proj.weight": "model-00001-of-00010.safetensors",
222
+ "model.layers.3.self_attn.q_proj.weight": "model-00001-of-00010.safetensors",
223
+ "model.layers.3.self_attn.v_proj.weight": "model-00001-of-00010.safetensors",
224
+ "model.layers.30.input_layernorm.weight": "model-00008-of-00010.safetensors",
225
+ "model.layers.30.mlp.down_proj.weight": "model-00008-of-00010.safetensors",
226
+ "model.layers.30.mlp.gate_proj.weight": "model-00008-of-00010.safetensors",
227
+ "model.layers.30.mlp.up_proj.weight": "model-00008-of-00010.safetensors",
228
+ "model.layers.30.post_attention_layernorm.weight": "model-00008-of-00010.safetensors",
229
+ "model.layers.30.self_attn.k_proj.weight": "model-00008-of-00010.safetensors",
230
+ "model.layers.30.self_attn.o_proj.weight": "model-00008-of-00010.safetensors",
231
+ "model.layers.30.self_attn.q_proj.weight": "model-00008-of-00010.safetensors",
232
+ "model.layers.30.self_attn.v_proj.weight": "model-00008-of-00010.safetensors",
233
+ "model.layers.31.input_layernorm.weight": "model-00008-of-00010.safetensors",
234
+ "model.layers.31.mlp.down_proj.weight": "model-00008-of-00010.safetensors",
235
+ "model.layers.31.mlp.gate_proj.weight": "model-00008-of-00010.safetensors",
236
+ "model.layers.31.mlp.up_proj.weight": "model-00008-of-00010.safetensors",
237
+ "model.layers.31.post_attention_layernorm.weight": "model-00008-of-00010.safetensors",
238
+ "model.layers.31.self_attn.k_proj.weight": "model-00008-of-00010.safetensors",
239
+ "model.layers.31.self_attn.o_proj.weight": "model-00008-of-00010.safetensors",
240
+ "model.layers.31.self_attn.q_proj.weight": "model-00008-of-00010.safetensors",
241
+ "model.layers.31.self_attn.v_proj.weight": "model-00008-of-00010.safetensors",
242
+ "model.layers.32.input_layernorm.weight": "model-00008-of-00010.safetensors",
243
+ "model.layers.32.mlp.down_proj.weight": "model-00008-of-00010.safetensors",
244
+ "model.layers.32.mlp.gate_proj.weight": "model-00008-of-00010.safetensors",
245
+ "model.layers.32.mlp.up_proj.weight": "model-00008-of-00010.safetensors",
246
+ "model.layers.32.post_attention_layernorm.weight": "model-00008-of-00010.safetensors",
247
+ "model.layers.32.self_attn.k_proj.weight": "model-00008-of-00010.safetensors",
248
+ "model.layers.32.self_attn.o_proj.weight": "model-00008-of-00010.safetensors",
249
+ "model.layers.32.self_attn.q_proj.weight": "model-00008-of-00010.safetensors",
250
+ "model.layers.32.self_attn.v_proj.weight": "model-00008-of-00010.safetensors",
251
+ "model.layers.33.input_layernorm.weight": "model-00009-of-00010.safetensors",
252
+ "model.layers.33.mlp.down_proj.weight": "model-00009-of-00010.safetensors",
253
+ "model.layers.33.mlp.gate_proj.weight": "model-00008-of-00010.safetensors",
254
+ "model.layers.33.mlp.up_proj.weight": "model-00009-of-00010.safetensors",
255
+ "model.layers.33.post_attention_layernorm.weight": "model-00009-of-00010.safetensors",
256
+ "model.layers.33.self_attn.k_proj.weight": "model-00008-of-00010.safetensors",
257
+ "model.layers.33.self_attn.o_proj.weight": "model-00008-of-00010.safetensors",
258
+ "model.layers.33.self_attn.q_proj.weight": "model-00008-of-00010.safetensors",
259
+ "model.layers.33.self_attn.v_proj.weight": "model-00008-of-00010.safetensors",
260
+ "model.layers.34.input_layernorm.weight": "model-00009-of-00010.safetensors",
261
+ "model.layers.34.mlp.down_proj.weight": "model-00009-of-00010.safetensors",
262
+ "model.layers.34.mlp.gate_proj.weight": "model-00009-of-00010.safetensors",
263
+ "model.layers.34.mlp.up_proj.weight": "model-00009-of-00010.safetensors",
264
+ "model.layers.34.post_attention_layernorm.weight": "model-00009-of-00010.safetensors",
265
+ "model.layers.34.self_attn.k_proj.weight": "model-00009-of-00010.safetensors",
266
+ "model.layers.34.self_attn.o_proj.weight": "model-00009-of-00010.safetensors",
267
+ "model.layers.34.self_attn.q_proj.weight": "model-00009-of-00010.safetensors",
268
+ "model.layers.34.self_attn.v_proj.weight": "model-00009-of-00010.safetensors",
269
+ "model.layers.35.input_layernorm.weight": "model-00009-of-00010.safetensors",
270
+ "model.layers.35.mlp.down_proj.weight": "model-00009-of-00010.safetensors",
271
+ "model.layers.35.mlp.gate_proj.weight": "model-00009-of-00010.safetensors",
272
+ "model.layers.35.mlp.up_proj.weight": "model-00009-of-00010.safetensors",
273
+ "model.layers.35.post_attention_layernorm.weight": "model-00009-of-00010.safetensors",
274
+ "model.layers.35.self_attn.k_proj.weight": "model-00009-of-00010.safetensors",
275
+ "model.layers.35.self_attn.o_proj.weight": "model-00009-of-00010.safetensors",
276
+ "model.layers.35.self_attn.q_proj.weight": "model-00009-of-00010.safetensors",
277
+ "model.layers.35.self_attn.v_proj.weight": "model-00009-of-00010.safetensors",
278
+ "model.layers.36.input_layernorm.weight": "model-00009-of-00010.safetensors",
279
+ "model.layers.36.mlp.down_proj.weight": "model-00009-of-00010.safetensors",
280
+ "model.layers.36.mlp.gate_proj.weight": "model-00009-of-00010.safetensors",
281
+ "model.layers.36.mlp.up_proj.weight": "model-00009-of-00010.safetensors",
282
+ "model.layers.36.post_attention_layernorm.weight": "model-00009-of-00010.safetensors",
283
+ "model.layers.36.self_attn.k_proj.weight": "model-00009-of-00010.safetensors",
284
+ "model.layers.36.self_attn.o_proj.weight": "model-00009-of-00010.safetensors",
285
+ "model.layers.36.self_attn.q_proj.weight": "model-00009-of-00010.safetensors",
286
+ "model.layers.36.self_attn.v_proj.weight": "model-00009-of-00010.safetensors",
287
+ "model.layers.37.input_layernorm.weight": "model-00010-of-00010.safetensors",
288
+ "model.layers.37.mlp.down_proj.weight": "model-00010-of-00010.safetensors",
289
+ "model.layers.37.mlp.gate_proj.weight": "model-00009-of-00010.safetensors",
290
+ "model.layers.37.mlp.up_proj.weight": "model-00009-of-00010.safetensors",
291
+ "model.layers.37.post_attention_layernorm.weight": "model-00010-of-00010.safetensors",
292
+ "model.layers.37.self_attn.k_proj.weight": "model-00009-of-00010.safetensors",
293
+ "model.layers.37.self_attn.o_proj.weight": "model-00009-of-00010.safetensors",
294
+ "model.layers.37.self_attn.q_proj.weight": "model-00009-of-00010.safetensors",
295
+ "model.layers.37.self_attn.v_proj.weight": "model-00009-of-00010.safetensors",
296
+ "model.layers.38.input_layernorm.weight": "model-00010-of-00010.safetensors",
297
+ "model.layers.38.mlp.down_proj.weight": "model-00010-of-00010.safetensors",
298
+ "model.layers.38.mlp.gate_proj.weight": "model-00010-of-00010.safetensors",
299
+ "model.layers.38.mlp.up_proj.weight": "model-00010-of-00010.safetensors",
300
+ "model.layers.38.post_attention_layernorm.weight": "model-00010-of-00010.safetensors",
301
+ "model.layers.38.self_attn.k_proj.weight": "model-00010-of-00010.safetensors",
302
+ "model.layers.38.self_attn.o_proj.weight": "model-00010-of-00010.safetensors",
303
+ "model.layers.38.self_attn.q_proj.weight": "model-00010-of-00010.safetensors",
304
+ "model.layers.38.self_attn.v_proj.weight": "model-00010-of-00010.safetensors",
305
+ "model.layers.39.input_layernorm.weight": "model-00010-of-00010.safetensors",
306
+ "model.layers.39.mlp.down_proj.weight": "model-00010-of-00010.safetensors",
307
+ "model.layers.39.mlp.gate_proj.weight": "model-00010-of-00010.safetensors",
308
+ "model.layers.39.mlp.up_proj.weight": "model-00010-of-00010.safetensors",
309
+ "model.layers.39.post_attention_layernorm.weight": "model-00010-of-00010.safetensors",
310
+ "model.layers.39.self_attn.k_proj.weight": "model-00010-of-00010.safetensors",
311
+ "model.layers.39.self_attn.o_proj.weight": "model-00010-of-00010.safetensors",
312
+ "model.layers.39.self_attn.q_proj.weight": "model-00010-of-00010.safetensors",
313
+ "model.layers.39.self_attn.v_proj.weight": "model-00010-of-00010.safetensors",
314
+ "model.layers.4.input_layernorm.weight": "model-00002-of-00010.safetensors",
315
+ "model.layers.4.mlp.down_proj.weight": "model-00002-of-00010.safetensors",
316
+ "model.layers.4.mlp.gate_proj.weight": "model-00002-of-00010.safetensors",
317
+ "model.layers.4.mlp.up_proj.weight": "model-00002-of-00010.safetensors",
318
+ "model.layers.4.post_attention_layernorm.weight": "model-00002-of-00010.safetensors",
319
+ "model.layers.4.self_attn.k_proj.weight": "model-00002-of-00010.safetensors",
320
+ "model.layers.4.self_attn.o_proj.weight": "model-00002-of-00010.safetensors",
321
+ "model.layers.4.self_attn.q_proj.weight": "model-00002-of-00010.safetensors",
322
+ "model.layers.4.self_attn.v_proj.weight": "model-00002-of-00010.safetensors",
323
+ "model.layers.5.input_layernorm.weight": "model-00002-of-00010.safetensors",
324
+ "model.layers.5.mlp.down_proj.weight": "model-00002-of-00010.safetensors",
325
+ "model.layers.5.mlp.gate_proj.weight": "model-00002-of-00010.safetensors",
326
+ "model.layers.5.mlp.up_proj.weight": "model-00002-of-00010.safetensors",
327
+ "model.layers.5.post_attention_layernorm.weight": "model-00002-of-00010.safetensors",
328
+ "model.layers.5.self_attn.k_proj.weight": "model-00002-of-00010.safetensors",
329
+ "model.layers.5.self_attn.o_proj.weight": "model-00002-of-00010.safetensors",
330
+ "model.layers.5.self_attn.q_proj.weight": "model-00002-of-00010.safetensors",
331
+ "model.layers.5.self_attn.v_proj.weight": "model-00002-of-00010.safetensors",
332
+ "model.layers.6.input_layernorm.weight": "model-00002-of-00010.safetensors",
333
+ "model.layers.6.mlp.down_proj.weight": "model-00002-of-00010.safetensors",
334
+ "model.layers.6.mlp.gate_proj.weight": "model-00002-of-00010.safetensors",
335
+ "model.layers.6.mlp.up_proj.weight": "model-00002-of-00010.safetensors",
336
+ "model.layers.6.post_attention_layernorm.weight": "model-00002-of-00010.safetensors",
337
+ "model.layers.6.self_attn.k_proj.weight": "model-00002-of-00010.safetensors",
338
+ "model.layers.6.self_attn.o_proj.weight": "model-00002-of-00010.safetensors",
339
+ "model.layers.6.self_attn.q_proj.weight": "model-00002-of-00010.safetensors",
340
+ "model.layers.6.self_attn.v_proj.weight": "model-00002-of-00010.safetensors",
341
+ "model.layers.7.input_layernorm.weight": "model-00003-of-00010.safetensors",
342
+ "model.layers.7.mlp.down_proj.weight": "model-00003-of-00010.safetensors",
343
+ "model.layers.7.mlp.gate_proj.weight": "model-00002-of-00010.safetensors",
344
+ "model.layers.7.mlp.up_proj.weight": "model-00003-of-00010.safetensors",
345
+ "model.layers.7.post_attention_layernorm.weight": "model-00003-of-00010.safetensors",
346
+ "model.layers.7.self_attn.k_proj.weight": "model-00002-of-00010.safetensors",
347
+ "model.layers.7.self_attn.o_proj.weight": "model-00002-of-00010.safetensors",
348
+ "model.layers.7.self_attn.q_proj.weight": "model-00002-of-00010.safetensors",
349
+ "model.layers.7.self_attn.v_proj.weight": "model-00002-of-00010.safetensors",
350
+ "model.layers.8.input_layernorm.weight": "model-00003-of-00010.safetensors",
351
+ "model.layers.8.mlp.down_proj.weight": "model-00003-of-00010.safetensors",
352
+ "model.layers.8.mlp.gate_proj.weight": "model-00003-of-00010.safetensors",
353
+ "model.layers.8.mlp.up_proj.weight": "model-00003-of-00010.safetensors",
354
+ "model.layers.8.post_attention_layernorm.weight": "model-00003-of-00010.safetensors",
355
+ "model.layers.8.self_attn.k_proj.weight": "model-00003-of-00010.safetensors",
356
+ "model.layers.8.self_attn.o_proj.weight": "model-00003-of-00010.safetensors",
357
+ "model.layers.8.self_attn.q_proj.weight": "model-00003-of-00010.safetensors",
358
+ "model.layers.8.self_attn.v_proj.weight": "model-00003-of-00010.safetensors",
359
+ "model.layers.9.input_layernorm.weight": "model-00003-of-00010.safetensors",
360
+ "model.layers.9.mlp.down_proj.weight": "model-00003-of-00010.safetensors",
361
+ "model.layers.9.mlp.gate_proj.weight": "model-00003-of-00010.safetensors",
362
+ "model.layers.9.mlp.up_proj.weight": "model-00003-of-00010.safetensors",
363
+ "model.layers.9.post_attention_layernorm.weight": "model-00003-of-00010.safetensors",
364
+ "model.layers.9.self_attn.k_proj.weight": "model-00003-of-00010.safetensors",
365
+ "model.layers.9.self_attn.o_proj.weight": "model-00003-of-00010.safetensors",
366
+ "model.layers.9.self_attn.q_proj.weight": "model-00003-of-00010.safetensors",
367
+ "model.layers.9.self_attn.v_proj.weight": "model-00003-of-00010.safetensors",
368
+ "model.norm.weight": "model-00010-of-00010.safetensors"
369
+ }
370
+ }
rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b75930684c955ecf0c0d4663e934224dd9427dadc59769259f2549965d357d51
3
+ size 16389
rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8e0fc296e6f82fe9a1f2b1274ad316ee068039fb68d980e22ef7add272b3df2c
3
+ size 16389
rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c6399da60ecd02ef2a24796b1a0b8d7be3d9569d07a5cca1dae98bb711d07adc
3
+ size 16389
rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7e1f1975564e49e4b373c7185880a8efd678d42e7f4c9f4eb2ca3822853f4c41
3
+ size 16389
rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f85c9d057fba0d1cd5d7ef8ef2d2c0bcab17f8bcefd896e1954ae652e7ea9742
3
+ size 16389
rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fae7634eccbdfd7191d6e3316d715529147ef6845a5db00336d6d8894115f6b9
3
+ size 16389
rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fc671e9ecff8caa15e423e1d0ff23ea3b53712f483676c07e0841767cc8899f5
3
+ size 16389
rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:89ae733cbfe3211b283a70ccd5dc72b6a7638e14344d7116a42b9efff9d6c977
3
+ size 16389
scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a3bf2153a5f048b14eb14b950e980eca144ab12899ee02a89a2faa101ffd127b
3
+ size 1465
special_tokens_map.json ADDED
@@ -0,0 +1,1032 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "additional_special_tokens": [
3
+ "<unk>",
4
+ "<s>",
5
+ "</s>",
6
+ "[INST]",
7
+ "[/INST]",
8
+ "[AVAILABLE_TOOLS]",
9
+ "[/AVAILABLE_TOOLS]",
10
+ "[TOOL_RESULTS]",
11
+ "[/TOOL_RESULTS]",
12
+ "[TOOL_CALLS]",
13
+ "[IMG]",
14
+ "<pad>",
15
+ "[IMG_BREAK]",
16
+ "[IMG_END]",
17
+ "[PREFIX]",
18
+ "[MIDDLE]",
19
+ "[SUFFIX]",
20
+ "[SYSTEM_PROMPT]",
21
+ "[/SYSTEM_PROMPT]",
22
+ "[TOOL_CONTENT]",
23
+ "<SPECIAL_20>",
24
+ "<SPECIAL_21>",
25
+ "<SPECIAL_22>",
26
+ "<SPECIAL_23>",
27
+ "<SPECIAL_24>",
28
+ "<SPECIAL_25>",
29
+ "<SPECIAL_26>",
30
+ "<SPECIAL_27>",
31
+ "<SPECIAL_28>",
32
+ "<SPECIAL_29>",
33
+ "<SPECIAL_30>",
34
+ "<SPECIAL_31>",
35
+ "<SPECIAL_32>",
36
+ "<SPECIAL_33>",
37
+ "<SPECIAL_34>",
38
+ "<SPECIAL_35>",
39
+ "<SPECIAL_36>",
40
+ "<SPECIAL_37>",
41
+ "<SPECIAL_38>",
42
+ "<SPECIAL_39>",
43
+ "<SPECIAL_40>",
44
+ "<SPECIAL_41>",
45
+ "<SPECIAL_42>",
46
+ "<SPECIAL_43>",
47
+ "<SPECIAL_44>",
48
+ "<SPECIAL_45>",
49
+ "<SPECIAL_46>",
50
+ "<SPECIAL_47>",
51
+ "<SPECIAL_48>",
52
+ "<SPECIAL_49>",
53
+ "<SPECIAL_50>",
54
+ "<SPECIAL_51>",
55
+ "<SPECIAL_52>",
56
+ "<SPECIAL_53>",
57
+ "<SPECIAL_54>",
58
+ "<SPECIAL_55>",
59
+ "<SPECIAL_56>",
60
+ "<SPECIAL_57>",
61
+ "<SPECIAL_58>",
62
+ "<SPECIAL_59>",
63
+ "<SPECIAL_60>",
64
+ "<SPECIAL_61>",
65
+ "<SPECIAL_62>",
66
+ "<SPECIAL_63>",
67
+ "<SPECIAL_64>",
68
+ "<SPECIAL_65>",
69
+ "<SPECIAL_66>",
70
+ "<SPECIAL_67>",
71
+ "<SPECIAL_68>",
72
+ "<SPECIAL_69>",
73
+ "<SPECIAL_70>",
74
+ "<SPECIAL_71>",
75
+ "<SPECIAL_72>",
76
+ "<SPECIAL_73>",
77
+ "<SPECIAL_74>",
78
+ "<SPECIAL_75>",
79
+ "<SPECIAL_76>",
80
+ "<SPECIAL_77>",
81
+ "<SPECIAL_78>",
82
+ "<SPECIAL_79>",
83
+ "<SPECIAL_80>",
84
+ "<SPECIAL_81>",
85
+ "<SPECIAL_82>",
86
+ "<SPECIAL_83>",
87
+ "<SPECIAL_84>",
88
+ "<SPECIAL_85>",
89
+ "<SPECIAL_86>",
90
+ "<SPECIAL_87>",
91
+ "<SPECIAL_88>",
92
+ "<SPECIAL_89>",
93
+ "<SPECIAL_90>",
94
+ "<SPECIAL_91>",
95
+ "<SPECIAL_92>",
96
+ "<SPECIAL_93>",
97
+ "<SPECIAL_94>",
98
+ "<SPECIAL_95>",
99
+ "<SPECIAL_96>",
100
+ "<SPECIAL_97>",
101
+ "<SPECIAL_98>",
102
+ "<SPECIAL_99>",
103
+ "<SPECIAL_100>",
104
+ "<SPECIAL_101>",
105
+ "<SPECIAL_102>",
106
+ "<SPECIAL_103>",
107
+ "<SPECIAL_104>",
108
+ "<SPECIAL_105>",
109
+ "<SPECIAL_106>",
110
+ "<SPECIAL_107>",
111
+ "<SPECIAL_108>",
112
+ "<SPECIAL_109>",
113
+ "<SPECIAL_110>",
114
+ "<SPECIAL_111>",
115
+ "<SPECIAL_112>",
116
+ "<SPECIAL_113>",
117
+ "<SPECIAL_114>",
118
+ "<SPECIAL_115>",
119
+ "<SPECIAL_116>",
120
+ "<SPECIAL_117>",
121
+ "<SPECIAL_118>",
122
+ "<SPECIAL_119>",
123
+ "<SPECIAL_120>",
124
+ "<SPECIAL_121>",
125
+ "<SPECIAL_122>",
126
+ "<SPECIAL_123>",
127
+ "<SPECIAL_124>",
128
+ "<SPECIAL_125>",
129
+ "<SPECIAL_126>",
130
+ "<SPECIAL_127>",
131
+ "<SPECIAL_128>",
132
+ "<SPECIAL_129>",
133
+ "<SPECIAL_130>",
134
+ "<SPECIAL_131>",
135
+ "<SPECIAL_132>",
136
+ "<SPECIAL_133>",
137
+ "<SPECIAL_134>",
138
+ "<SPECIAL_135>",
139
+ "<SPECIAL_136>",
140
+ "<SPECIAL_137>",
141
+ "<SPECIAL_138>",
142
+ "<SPECIAL_139>",
143
+ "<SPECIAL_140>",
144
+ "<SPECIAL_141>",
145
+ "<SPECIAL_142>",
146
+ "<SPECIAL_143>",
147
+ "<SPECIAL_144>",
148
+ "<SPECIAL_145>",
149
+ "<SPECIAL_146>",
150
+ "<SPECIAL_147>",
151
+ "<SPECIAL_148>",
152
+ "<SPECIAL_149>",
153
+ "<SPECIAL_150>",
154
+ "<SPECIAL_151>",
155
+ "<SPECIAL_152>",
156
+ "<SPECIAL_153>",
157
+ "<SPECIAL_154>",
158
+ "<SPECIAL_155>",
159
+ "<SPECIAL_156>",
160
+ "<SPECIAL_157>",
161
+ "<SPECIAL_158>",
162
+ "<SPECIAL_159>",
163
+ "<SPECIAL_160>",
164
+ "<SPECIAL_161>",
165
+ "<SPECIAL_162>",
166
+ "<SPECIAL_163>",
167
+ "<SPECIAL_164>",
168
+ "<SPECIAL_165>",
169
+ "<SPECIAL_166>",
170
+ "<SPECIAL_167>",
171
+ "<SPECIAL_168>",
172
+ "<SPECIAL_169>",
173
+ "<SPECIAL_170>",
174
+ "<SPECIAL_171>",
175
+ "<SPECIAL_172>",
176
+ "<SPECIAL_173>",
177
+ "<SPECIAL_174>",
178
+ "<SPECIAL_175>",
179
+ "<SPECIAL_176>",
180
+ "<SPECIAL_177>",
181
+ "<SPECIAL_178>",
182
+ "<SPECIAL_179>",
183
+ "<SPECIAL_180>",
184
+ "<SPECIAL_181>",
185
+ "<SPECIAL_182>",
186
+ "<SPECIAL_183>",
187
+ "<SPECIAL_184>",
188
+ "<SPECIAL_185>",
189
+ "<SPECIAL_186>",
190
+ "<SPECIAL_187>",
191
+ "<SPECIAL_188>",
192
+ "<SPECIAL_189>",
193
+ "<SPECIAL_190>",
194
+ "<SPECIAL_191>",
195
+ "<SPECIAL_192>",
196
+ "<SPECIAL_193>",
197
+ "<SPECIAL_194>",
198
+ "<SPECIAL_195>",
199
+ "<SPECIAL_196>",
200
+ "<SPECIAL_197>",
201
+ "<SPECIAL_198>",
202
+ "<SPECIAL_199>",
203
+ "<SPECIAL_200>",
204
+ "<SPECIAL_201>",
205
+ "<SPECIAL_202>",
206
+ "<SPECIAL_203>",
207
+ "<SPECIAL_204>",
208
+ "<SPECIAL_205>",
209
+ "<SPECIAL_206>",
210
+ "<SPECIAL_207>",
211
+ "<SPECIAL_208>",
212
+ "<SPECIAL_209>",
213
+ "<SPECIAL_210>",
214
+ "<SPECIAL_211>",
215
+ "<SPECIAL_212>",
216
+ "<SPECIAL_213>",
217
+ "<SPECIAL_214>",
218
+ "<SPECIAL_215>",
219
+ "<SPECIAL_216>",
220
+ "<SPECIAL_217>",
221
+ "<SPECIAL_218>",
222
+ "<SPECIAL_219>",
223
+ "<SPECIAL_220>",
224
+ "<SPECIAL_221>",
225
+ "<SPECIAL_222>",
226
+ "<SPECIAL_223>",
227
+ "<SPECIAL_224>",
228
+ "<SPECIAL_225>",
229
+ "<SPECIAL_226>",
230
+ "<SPECIAL_227>",
231
+ "<SPECIAL_228>",
232
+ "<SPECIAL_229>",
233
+ "<SPECIAL_230>",
234
+ "<SPECIAL_231>",
235
+ "<SPECIAL_232>",
236
+ "<SPECIAL_233>",
237
+ "<SPECIAL_234>",
238
+ "<SPECIAL_235>",
239
+ "<SPECIAL_236>",
240
+ "<SPECIAL_237>",
241
+ "<SPECIAL_238>",
242
+ "<SPECIAL_239>",
243
+ "<SPECIAL_240>",
244
+ "<SPECIAL_241>",
245
+ "<SPECIAL_242>",
246
+ "<SPECIAL_243>",
247
+ "<SPECIAL_244>",
248
+ "<SPECIAL_245>",
249
+ "<SPECIAL_246>",
250
+ "<SPECIAL_247>",
251
+ "<SPECIAL_248>",
252
+ "<SPECIAL_249>",
253
+ "<SPECIAL_250>",
254
+ "<SPECIAL_251>",
255
+ "<SPECIAL_252>",
256
+ "<SPECIAL_253>",
257
+ "<SPECIAL_254>",
258
+ "<SPECIAL_255>",
259
+ "<SPECIAL_256>",
260
+ "<SPECIAL_257>",
261
+ "<SPECIAL_258>",
262
+ "<SPECIAL_259>",
263
+ "<SPECIAL_260>",
264
+ "<SPECIAL_261>",
265
+ "<SPECIAL_262>",
266
+ "<SPECIAL_263>",
267
+ "<SPECIAL_264>",
268
+ "<SPECIAL_265>",
269
+ "<SPECIAL_266>",
270
+ "<SPECIAL_267>",
271
+ "<SPECIAL_268>",
272
+ "<SPECIAL_269>",
273
+ "<SPECIAL_270>",
274
+ "<SPECIAL_271>",
275
+ "<SPECIAL_272>",
276
+ "<SPECIAL_273>",
277
+ "<SPECIAL_274>",
278
+ "<SPECIAL_275>",
279
+ "<SPECIAL_276>",
280
+ "<SPECIAL_277>",
281
+ "<SPECIAL_278>",
282
+ "<SPECIAL_279>",
283
+ "<SPECIAL_280>",
284
+ "<SPECIAL_281>",
285
+ "<SPECIAL_282>",
286
+ "<SPECIAL_283>",
287
+ "<SPECIAL_284>",
288
+ "<SPECIAL_285>",
289
+ "<SPECIAL_286>",
290
+ "<SPECIAL_287>",
291
+ "<SPECIAL_288>",
292
+ "<SPECIAL_289>",
293
+ "<SPECIAL_290>",
294
+ "<SPECIAL_291>",
295
+ "<SPECIAL_292>",
296
+ "<SPECIAL_293>",
297
+ "<SPECIAL_294>",
298
+ "<SPECIAL_295>",
299
+ "<SPECIAL_296>",
300
+ "<SPECIAL_297>",
301
+ "<SPECIAL_298>",
302
+ "<SPECIAL_299>",
303
+ "<SPECIAL_300>",
304
+ "<SPECIAL_301>",
305
+ "<SPECIAL_302>",
306
+ "<SPECIAL_303>",
307
+ "<SPECIAL_304>",
308
+ "<SPECIAL_305>",
309
+ "<SPECIAL_306>",
310
+ "<SPECIAL_307>",
311
+ "<SPECIAL_308>",
312
+ "<SPECIAL_309>",
313
+ "<SPECIAL_310>",
314
+ "<SPECIAL_311>",
315
+ "<SPECIAL_312>",
316
+ "<SPECIAL_313>",
317
+ "<SPECIAL_314>",
318
+ "<SPECIAL_315>",
319
+ "<SPECIAL_316>",
320
+ "<SPECIAL_317>",
321
+ "<SPECIAL_318>",
322
+ "<SPECIAL_319>",
323
+ "<SPECIAL_320>",
324
+ "<SPECIAL_321>",
325
+ "<SPECIAL_322>",
326
+ "<SPECIAL_323>",
327
+ "<SPECIAL_324>",
328
+ "<SPECIAL_325>",
329
+ "<SPECIAL_326>",
330
+ "<SPECIAL_327>",
331
+ "<SPECIAL_328>",
332
+ "<SPECIAL_329>",
333
+ "<SPECIAL_330>",
334
+ "<SPECIAL_331>",
335
+ "<SPECIAL_332>",
336
+ "<SPECIAL_333>",
337
+ "<SPECIAL_334>",
338
+ "<SPECIAL_335>",
339
+ "<SPECIAL_336>",
340
+ "<SPECIAL_337>",
341
+ "<SPECIAL_338>",
342
+ "<SPECIAL_339>",
343
+ "<SPECIAL_340>",
344
+ "<SPECIAL_341>",
345
+ "<SPECIAL_342>",
346
+ "<SPECIAL_343>",
347
+ "<SPECIAL_344>",
348
+ "<SPECIAL_345>",
349
+ "<SPECIAL_346>",
350
+ "<SPECIAL_347>",
351
+ "<SPECIAL_348>",
352
+ "<SPECIAL_349>",
353
+ "<SPECIAL_350>",
354
+ "<SPECIAL_351>",
355
+ "<SPECIAL_352>",
356
+ "<SPECIAL_353>",
357
+ "<SPECIAL_354>",
358
+ "<SPECIAL_355>",
359
+ "<SPECIAL_356>",
360
+ "<SPECIAL_357>",
361
+ "<SPECIAL_358>",
362
+ "<SPECIAL_359>",
363
+ "<SPECIAL_360>",
364
+ "<SPECIAL_361>",
365
+ "<SPECIAL_362>",
366
+ "<SPECIAL_363>",
367
+ "<SPECIAL_364>",
368
+ "<SPECIAL_365>",
369
+ "<SPECIAL_366>",
370
+ "<SPECIAL_367>",
371
+ "<SPECIAL_368>",
372
+ "<SPECIAL_369>",
373
+ "<SPECIAL_370>",
374
+ "<SPECIAL_371>",
375
+ "<SPECIAL_372>",
376
+ "<SPECIAL_373>",
377
+ "<SPECIAL_374>",
378
+ "<SPECIAL_375>",
379
+ "<SPECIAL_376>",
380
+ "<SPECIAL_377>",
381
+ "<SPECIAL_378>",
382
+ "<SPECIAL_379>",
383
+ "<SPECIAL_380>",
384
+ "<SPECIAL_381>",
385
+ "<SPECIAL_382>",
386
+ "<SPECIAL_383>",
387
+ "<SPECIAL_384>",
388
+ "<SPECIAL_385>",
389
+ "<SPECIAL_386>",
390
+ "<SPECIAL_387>",
391
+ "<SPECIAL_388>",
392
+ "<SPECIAL_389>",
393
+ "<SPECIAL_390>",
394
+ "<SPECIAL_391>",
395
+ "<SPECIAL_392>",
396
+ "<SPECIAL_393>",
397
+ "<SPECIAL_394>",
398
+ "<SPECIAL_395>",
399
+ "<SPECIAL_396>",
400
+ "<SPECIAL_397>",
401
+ "<SPECIAL_398>",
402
+ "<SPECIAL_399>",
403
+ "<SPECIAL_400>",
404
+ "<SPECIAL_401>",
405
+ "<SPECIAL_402>",
406
+ "<SPECIAL_403>",
407
+ "<SPECIAL_404>",
408
+ "<SPECIAL_405>",
409
+ "<SPECIAL_406>",
410
+ "<SPECIAL_407>",
411
+ "<SPECIAL_408>",
412
+ "<SPECIAL_409>",
413
+ "<SPECIAL_410>",
414
+ "<SPECIAL_411>",
415
+ "<SPECIAL_412>",
416
+ "<SPECIAL_413>",
417
+ "<SPECIAL_414>",
418
+ "<SPECIAL_415>",
419
+ "<SPECIAL_416>",
420
+ "<SPECIAL_417>",
421
+ "<SPECIAL_418>",
422
+ "<SPECIAL_419>",
423
+ "<SPECIAL_420>",
424
+ "<SPECIAL_421>",
425
+ "<SPECIAL_422>",
426
+ "<SPECIAL_423>",
427
+ "<SPECIAL_424>",
428
+ "<SPECIAL_425>",
429
+ "<SPECIAL_426>",
430
+ "<SPECIAL_427>",
431
+ "<SPECIAL_428>",
432
+ "<SPECIAL_429>",
433
+ "<SPECIAL_430>",
434
+ "<SPECIAL_431>",
435
+ "<SPECIAL_432>",
436
+ "<SPECIAL_433>",
437
+ "<SPECIAL_434>",
438
+ "<SPECIAL_435>",
439
+ "<SPECIAL_436>",
440
+ "<SPECIAL_437>",
441
+ "<SPECIAL_438>",
442
+ "<SPECIAL_439>",
443
+ "<SPECIAL_440>",
444
+ "<SPECIAL_441>",
445
+ "<SPECIAL_442>",
446
+ "<SPECIAL_443>",
447
+ "<SPECIAL_444>",
448
+ "<SPECIAL_445>",
449
+ "<SPECIAL_446>",
450
+ "<SPECIAL_447>",
451
+ "<SPECIAL_448>",
452
+ "<SPECIAL_449>",
453
+ "<SPECIAL_450>",
454
+ "<SPECIAL_451>",
455
+ "<SPECIAL_452>",
456
+ "<SPECIAL_453>",
457
+ "<SPECIAL_454>",
458
+ "<SPECIAL_455>",
459
+ "<SPECIAL_456>",
460
+ "<SPECIAL_457>",
461
+ "<SPECIAL_458>",
462
+ "<SPECIAL_459>",
463
+ "<SPECIAL_460>",
464
+ "<SPECIAL_461>",
465
+ "<SPECIAL_462>",
466
+ "<SPECIAL_463>",
467
+ "<SPECIAL_464>",
468
+ "<SPECIAL_465>",
469
+ "<SPECIAL_466>",
470
+ "<SPECIAL_467>",
471
+ "<SPECIAL_468>",
472
+ "<SPECIAL_469>",
473
+ "<SPECIAL_470>",
474
+ "<SPECIAL_471>",
475
+ "<SPECIAL_472>",
476
+ "<SPECIAL_473>",
477
+ "<SPECIAL_474>",
478
+ "<SPECIAL_475>",
479
+ "<SPECIAL_476>",
480
+ "<SPECIAL_477>",
481
+ "<SPECIAL_478>",
482
+ "<SPECIAL_479>",
483
+ "<SPECIAL_480>",
484
+ "<SPECIAL_481>",
485
+ "<SPECIAL_482>",
486
+ "<SPECIAL_483>",
487
+ "<SPECIAL_484>",
488
+ "<SPECIAL_485>",
489
+ "<SPECIAL_486>",
490
+ "<SPECIAL_487>",
491
+ "<SPECIAL_488>",
492
+ "<SPECIAL_489>",
493
+ "<SPECIAL_490>",
494
+ "<SPECIAL_491>",
495
+ "<SPECIAL_492>",
496
+ "<SPECIAL_493>",
497
+ "<SPECIAL_494>",
498
+ "<SPECIAL_495>",
499
+ "<SPECIAL_496>",
500
+ "<SPECIAL_497>",
501
+ "<SPECIAL_498>",
502
+ "<SPECIAL_499>",
503
+ "<SPECIAL_500>",
504
+ "<SPECIAL_501>",
505
+ "<SPECIAL_502>",
506
+ "<SPECIAL_503>",
507
+ "<SPECIAL_504>",
508
+ "<SPECIAL_505>",
509
+ "<SPECIAL_506>",
510
+ "<SPECIAL_507>",
511
+ "<SPECIAL_508>",
512
+ "<SPECIAL_509>",
513
+ "<SPECIAL_510>",
514
+ "<SPECIAL_511>",
515
+ "<SPECIAL_512>",
516
+ "<SPECIAL_513>",
517
+ "<SPECIAL_514>",
518
+ "<SPECIAL_515>",
519
+ "<SPECIAL_516>",
520
+ "<SPECIAL_517>",
521
+ "<SPECIAL_518>",
522
+ "<SPECIAL_519>",
523
+ "<SPECIAL_520>",
524
+ "<SPECIAL_521>",
525
+ "<SPECIAL_522>",
526
+ "<SPECIAL_523>",
527
+ "<SPECIAL_524>",
528
+ "<SPECIAL_525>",
529
+ "<SPECIAL_526>",
530
+ "<SPECIAL_527>",
531
+ "<SPECIAL_528>",
532
+ "<SPECIAL_529>",
533
+ "<SPECIAL_530>",
534
+ "<SPECIAL_531>",
535
+ "<SPECIAL_532>",
536
+ "<SPECIAL_533>",
537
+ "<SPECIAL_534>",
538
+ "<SPECIAL_535>",
539
+ "<SPECIAL_536>",
540
+ "<SPECIAL_537>",
541
+ "<SPECIAL_538>",
542
+ "<SPECIAL_539>",
543
+ "<SPECIAL_540>",
544
+ "<SPECIAL_541>",
545
+ "<SPECIAL_542>",
546
+ "<SPECIAL_543>",
547
+ "<SPECIAL_544>",
548
+ "<SPECIAL_545>",
549
+ "<SPECIAL_546>",
550
+ "<SPECIAL_547>",
551
+ "<SPECIAL_548>",
552
+ "<SPECIAL_549>",
553
+ "<SPECIAL_550>",
554
+ "<SPECIAL_551>",
555
+ "<SPECIAL_552>",
556
+ "<SPECIAL_553>",
557
+ "<SPECIAL_554>",
558
+ "<SPECIAL_555>",
559
+ "<SPECIAL_556>",
560
+ "<SPECIAL_557>",
561
+ "<SPECIAL_558>",
562
+ "<SPECIAL_559>",
563
+ "<SPECIAL_560>",
564
+ "<SPECIAL_561>",
565
+ "<SPECIAL_562>",
566
+ "<SPECIAL_563>",
567
+ "<SPECIAL_564>",
568
+ "<SPECIAL_565>",
569
+ "<SPECIAL_566>",
570
+ "<SPECIAL_567>",
571
+ "<SPECIAL_568>",
572
+ "<SPECIAL_569>",
573
+ "<SPECIAL_570>",
574
+ "<SPECIAL_571>",
575
+ "<SPECIAL_572>",
576
+ "<SPECIAL_573>",
577
+ "<SPECIAL_574>",
578
+ "<SPECIAL_575>",
579
+ "<SPECIAL_576>",
580
+ "<SPECIAL_577>",
581
+ "<SPECIAL_578>",
582
+ "<SPECIAL_579>",
583
+ "<SPECIAL_580>",
584
+ "<SPECIAL_581>",
585
+ "<SPECIAL_582>",
586
+ "<SPECIAL_583>",
587
+ "<SPECIAL_584>",
588
+ "<SPECIAL_585>",
589
+ "<SPECIAL_586>",
590
+ "<SPECIAL_587>",
591
+ "<SPECIAL_588>",
592
+ "<SPECIAL_589>",
593
+ "<SPECIAL_590>",
594
+ "<SPECIAL_591>",
595
+ "<SPECIAL_592>",
596
+ "<SPECIAL_593>",
597
+ "<SPECIAL_594>",
598
+ "<SPECIAL_595>",
599
+ "<SPECIAL_596>",
600
+ "<SPECIAL_597>",
601
+ "<SPECIAL_598>",
602
+ "<SPECIAL_599>",
603
+ "<SPECIAL_600>",
604
+ "<SPECIAL_601>",
605
+ "<SPECIAL_602>",
606
+ "<SPECIAL_603>",
607
+ "<SPECIAL_604>",
608
+ "<SPECIAL_605>",
609
+ "<SPECIAL_606>",
610
+ "<SPECIAL_607>",
611
+ "<SPECIAL_608>",
612
+ "<SPECIAL_609>",
613
+ "<SPECIAL_610>",
614
+ "<SPECIAL_611>",
615
+ "<SPECIAL_612>",
616
+ "<SPECIAL_613>",
617
+ "<SPECIAL_614>",
618
+ "<SPECIAL_615>",
619
+ "<SPECIAL_616>",
620
+ "<SPECIAL_617>",
621
+ "<SPECIAL_618>",
622
+ "<SPECIAL_619>",
623
+ "<SPECIAL_620>",
624
+ "<SPECIAL_621>",
625
+ "<SPECIAL_622>",
626
+ "<SPECIAL_623>",
627
+ "<SPECIAL_624>",
628
+ "<SPECIAL_625>",
629
+ "<SPECIAL_626>",
630
+ "<SPECIAL_627>",
631
+ "<SPECIAL_628>",
632
+ "<SPECIAL_629>",
633
+ "<SPECIAL_630>",
634
+ "<SPECIAL_631>",
635
+ "<SPECIAL_632>",
636
+ "<SPECIAL_633>",
637
+ "<SPECIAL_634>",
638
+ "<SPECIAL_635>",
639
+ "<SPECIAL_636>",
640
+ "<SPECIAL_637>",
641
+ "<SPECIAL_638>",
642
+ "<SPECIAL_639>",
643
+ "<SPECIAL_640>",
644
+ "<SPECIAL_641>",
645
+ "<SPECIAL_642>",
646
+ "<SPECIAL_643>",
647
+ "<SPECIAL_644>",
648
+ "<SPECIAL_645>",
649
+ "<SPECIAL_646>",
650
+ "<SPECIAL_647>",
651
+ "<SPECIAL_648>",
652
+ "<SPECIAL_649>",
653
+ "<SPECIAL_650>",
654
+ "<SPECIAL_651>",
655
+ "<SPECIAL_652>",
656
+ "<SPECIAL_653>",
657
+ "<SPECIAL_654>",
658
+ "<SPECIAL_655>",
659
+ "<SPECIAL_656>",
660
+ "<SPECIAL_657>",
661
+ "<SPECIAL_658>",
662
+ "<SPECIAL_659>",
663
+ "<SPECIAL_660>",
664
+ "<SPECIAL_661>",
665
+ "<SPECIAL_662>",
666
+ "<SPECIAL_663>",
667
+ "<SPECIAL_664>",
668
+ "<SPECIAL_665>",
669
+ "<SPECIAL_666>",
670
+ "<SPECIAL_667>",
671
+ "<SPECIAL_668>",
672
+ "<SPECIAL_669>",
673
+ "<SPECIAL_670>",
674
+ "<SPECIAL_671>",
675
+ "<SPECIAL_672>",
676
+ "<SPECIAL_673>",
677
+ "<SPECIAL_674>",
678
+ "<SPECIAL_675>",
679
+ "<SPECIAL_676>",
680
+ "<SPECIAL_677>",
681
+ "<SPECIAL_678>",
682
+ "<SPECIAL_679>",
683
+ "<SPECIAL_680>",
684
+ "<SPECIAL_681>",
685
+ "<SPECIAL_682>",
686
+ "<SPECIAL_683>",
687
+ "<SPECIAL_684>",
688
+ "<SPECIAL_685>",
689
+ "<SPECIAL_686>",
690
+ "<SPECIAL_687>",
691
+ "<SPECIAL_688>",
692
+ "<SPECIAL_689>",
693
+ "<SPECIAL_690>",
694
+ "<SPECIAL_691>",
695
+ "<SPECIAL_692>",
696
+ "<SPECIAL_693>",
697
+ "<SPECIAL_694>",
698
+ "<SPECIAL_695>",
699
+ "<SPECIAL_696>",
700
+ "<SPECIAL_697>",
701
+ "<SPECIAL_698>",
702
+ "<SPECIAL_699>",
703
+ "<SPECIAL_700>",
704
+ "<SPECIAL_701>",
705
+ "<SPECIAL_702>",
706
+ "<SPECIAL_703>",
707
+ "<SPECIAL_704>",
708
+ "<SPECIAL_705>",
709
+ "<SPECIAL_706>",
710
+ "<SPECIAL_707>",
711
+ "<SPECIAL_708>",
712
+ "<SPECIAL_709>",
713
+ "<SPECIAL_710>",
714
+ "<SPECIAL_711>",
715
+ "<SPECIAL_712>",
716
+ "<SPECIAL_713>",
717
+ "<SPECIAL_714>",
718
+ "<SPECIAL_715>",
719
+ "<SPECIAL_716>",
720
+ "<SPECIAL_717>",
721
+ "<SPECIAL_718>",
722
+ "<SPECIAL_719>",
723
+ "<SPECIAL_720>",
724
+ "<SPECIAL_721>",
725
+ "<SPECIAL_722>",
726
+ "<SPECIAL_723>",
727
+ "<SPECIAL_724>",
728
+ "<SPECIAL_725>",
729
+ "<SPECIAL_726>",
730
+ "<SPECIAL_727>",
731
+ "<SPECIAL_728>",
732
+ "<SPECIAL_729>",
733
+ "<SPECIAL_730>",
734
+ "<SPECIAL_731>",
735
+ "<SPECIAL_732>",
736
+ "<SPECIAL_733>",
737
+ "<SPECIAL_734>",
738
+ "<SPECIAL_735>",
739
+ "<SPECIAL_736>",
740
+ "<SPECIAL_737>",
741
+ "<SPECIAL_738>",
742
+ "<SPECIAL_739>",
743
+ "<SPECIAL_740>",
744
+ "<SPECIAL_741>",
745
+ "<SPECIAL_742>",
746
+ "<SPECIAL_743>",
747
+ "<SPECIAL_744>",
748
+ "<SPECIAL_745>",
749
+ "<SPECIAL_746>",
750
+ "<SPECIAL_747>",
751
+ "<SPECIAL_748>",
752
+ "<SPECIAL_749>",
753
+ "<SPECIAL_750>",
754
+ "<SPECIAL_751>",
755
+ "<SPECIAL_752>",
756
+ "<SPECIAL_753>",
757
+ "<SPECIAL_754>",
758
+ "<SPECIAL_755>",
759
+ "<SPECIAL_756>",
760
+ "<SPECIAL_757>",
761
+ "<SPECIAL_758>",
762
+ "<SPECIAL_759>",
763
+ "<SPECIAL_760>",
764
+ "<SPECIAL_761>",
765
+ "<SPECIAL_762>",
766
+ "<SPECIAL_763>",
767
+ "<SPECIAL_764>",
768
+ "<SPECIAL_765>",
769
+ "<SPECIAL_766>",
770
+ "<SPECIAL_767>",
771
+ "<SPECIAL_768>",
772
+ "<SPECIAL_769>",
773
+ "<SPECIAL_770>",
774
+ "<SPECIAL_771>",
775
+ "<SPECIAL_772>",
776
+ "<SPECIAL_773>",
777
+ "<SPECIAL_774>",
778
+ "<SPECIAL_775>",
779
+ "<SPECIAL_776>",
780
+ "<SPECIAL_777>",
781
+ "<SPECIAL_778>",
782
+ "<SPECIAL_779>",
783
+ "<SPECIAL_780>",
784
+ "<SPECIAL_781>",
785
+ "<SPECIAL_782>",
786
+ "<SPECIAL_783>",
787
+ "<SPECIAL_784>",
788
+ "<SPECIAL_785>",
789
+ "<SPECIAL_786>",
790
+ "<SPECIAL_787>",
791
+ "<SPECIAL_788>",
792
+ "<SPECIAL_789>",
793
+ "<SPECIAL_790>",
794
+ "<SPECIAL_791>",
795
+ "<SPECIAL_792>",
796
+ "<SPECIAL_793>",
797
+ "<SPECIAL_794>",
798
+ "<SPECIAL_795>",
799
+ "<SPECIAL_796>",
800
+ "<SPECIAL_797>",
801
+ "<SPECIAL_798>",
802
+ "<SPECIAL_799>",
803
+ "<SPECIAL_800>",
804
+ "<SPECIAL_801>",
805
+ "<SPECIAL_802>",
806
+ "<SPECIAL_803>",
807
+ "<SPECIAL_804>",
808
+ "<SPECIAL_805>",
809
+ "<SPECIAL_806>",
810
+ "<SPECIAL_807>",
811
+ "<SPECIAL_808>",
812
+ "<SPECIAL_809>",
813
+ "<SPECIAL_810>",
814
+ "<SPECIAL_811>",
815
+ "<SPECIAL_812>",
816
+ "<SPECIAL_813>",
817
+ "<SPECIAL_814>",
818
+ "<SPECIAL_815>",
819
+ "<SPECIAL_816>",
820
+ "<SPECIAL_817>",
821
+ "<SPECIAL_818>",
822
+ "<SPECIAL_819>",
823
+ "<SPECIAL_820>",
824
+ "<SPECIAL_821>",
825
+ "<SPECIAL_822>",
826
+ "<SPECIAL_823>",
827
+ "<SPECIAL_824>",
828
+ "<SPECIAL_825>",
829
+ "<SPECIAL_826>",
830
+ "<SPECIAL_827>",
831
+ "<SPECIAL_828>",
832
+ "<SPECIAL_829>",
833
+ "<SPECIAL_830>",
834
+ "<SPECIAL_831>",
835
+ "<SPECIAL_832>",
836
+ "<SPECIAL_833>",
837
+ "<SPECIAL_834>",
838
+ "<SPECIAL_835>",
839
+ "<SPECIAL_836>",
840
+ "<SPECIAL_837>",
841
+ "<SPECIAL_838>",
842
+ "<SPECIAL_839>",
843
+ "<SPECIAL_840>",
844
+ "<SPECIAL_841>",
845
+ "<SPECIAL_842>",
846
+ "<SPECIAL_843>",
847
+ "<SPECIAL_844>",
848
+ "<SPECIAL_845>",
849
+ "<SPECIAL_846>",
850
+ "<SPECIAL_847>",
851
+ "<SPECIAL_848>",
852
+ "<SPECIAL_849>",
853
+ "<SPECIAL_850>",
854
+ "<SPECIAL_851>",
855
+ "<SPECIAL_852>",
856
+ "<SPECIAL_853>",
857
+ "<SPECIAL_854>",
858
+ "<SPECIAL_855>",
859
+ "<SPECIAL_856>",
860
+ "<SPECIAL_857>",
861
+ "<SPECIAL_858>",
862
+ "<SPECIAL_859>",
863
+ "<SPECIAL_860>",
864
+ "<SPECIAL_861>",
865
+ "<SPECIAL_862>",
866
+ "<SPECIAL_863>",
867
+ "<SPECIAL_864>",
868
+ "<SPECIAL_865>",
869
+ "<SPECIAL_866>",
870
+ "<SPECIAL_867>",
871
+ "<SPECIAL_868>",
872
+ "<SPECIAL_869>",
873
+ "<SPECIAL_870>",
874
+ "<SPECIAL_871>",
875
+ "<SPECIAL_872>",
876
+ "<SPECIAL_873>",
877
+ "<SPECIAL_874>",
878
+ "<SPECIAL_875>",
879
+ "<SPECIAL_876>",
880
+ "<SPECIAL_877>",
881
+ "<SPECIAL_878>",
882
+ "<SPECIAL_879>",
883
+ "<SPECIAL_880>",
884
+ "<SPECIAL_881>",
885
+ "<SPECIAL_882>",
886
+ "<SPECIAL_883>",
887
+ "<SPECIAL_884>",
888
+ "<SPECIAL_885>",
889
+ "<SPECIAL_886>",
890
+ "<SPECIAL_887>",
891
+ "<SPECIAL_888>",
892
+ "<SPECIAL_889>",
893
+ "<SPECIAL_890>",
894
+ "<SPECIAL_891>",
895
+ "<SPECIAL_892>",
896
+ "<SPECIAL_893>",
897
+ "<SPECIAL_894>",
898
+ "<SPECIAL_895>",
899
+ "<SPECIAL_896>",
900
+ "<SPECIAL_897>",
901
+ "<SPECIAL_898>",
902
+ "<SPECIAL_899>",
903
+ "<SPECIAL_900>",
904
+ "<SPECIAL_901>",
905
+ "<SPECIAL_902>",
906
+ "<SPECIAL_903>",
907
+ "<SPECIAL_904>",
908
+ "<SPECIAL_905>",
909
+ "<SPECIAL_906>",
910
+ "<SPECIAL_907>",
911
+ "<SPECIAL_908>",
912
+ "<SPECIAL_909>",
913
+ "<SPECIAL_910>",
914
+ "<SPECIAL_911>",
915
+ "<SPECIAL_912>",
916
+ "<SPECIAL_913>",
917
+ "<SPECIAL_914>",
918
+ "<SPECIAL_915>",
919
+ "<SPECIAL_916>",
920
+ "<SPECIAL_917>",
921
+ "<SPECIAL_918>",
922
+ "<SPECIAL_919>",
923
+ "<SPECIAL_920>",
924
+ "<SPECIAL_921>",
925
+ "<SPECIAL_922>",
926
+ "<SPECIAL_923>",
927
+ "<SPECIAL_924>",
928
+ "<SPECIAL_925>",
929
+ "<SPECIAL_926>",
930
+ "<SPECIAL_927>",
931
+ "<SPECIAL_928>",
932
+ "<SPECIAL_929>",
933
+ "<SPECIAL_930>",
934
+ "<SPECIAL_931>",
935
+ "<SPECIAL_932>",
936
+ "<SPECIAL_933>",
937
+ "<SPECIAL_934>",
938
+ "<SPECIAL_935>",
939
+ "<SPECIAL_936>",
940
+ "<SPECIAL_937>",
941
+ "<SPECIAL_938>",
942
+ "<SPECIAL_939>",
943
+ "<SPECIAL_940>",
944
+ "<SPECIAL_941>",
945
+ "<SPECIAL_942>",
946
+ "<SPECIAL_943>",
947
+ "<SPECIAL_944>",
948
+ "<SPECIAL_945>",
949
+ "<SPECIAL_946>",
950
+ "<SPECIAL_947>",
951
+ "<SPECIAL_948>",
952
+ "<SPECIAL_949>",
953
+ "<SPECIAL_950>",
954
+ "<SPECIAL_951>",
955
+ "<SPECIAL_952>",
956
+ "<SPECIAL_953>",
957
+ "<SPECIAL_954>",
958
+ "<SPECIAL_955>",
959
+ "<SPECIAL_956>",
960
+ "<SPECIAL_957>",
961
+ "<SPECIAL_958>",
962
+ "<SPECIAL_959>",
963
+ "<SPECIAL_960>",
964
+ "<SPECIAL_961>",
965
+ "<SPECIAL_962>",
966
+ "<SPECIAL_963>",
967
+ "<SPECIAL_964>",
968
+ "<SPECIAL_965>",
969
+ "<SPECIAL_966>",
970
+ "<SPECIAL_967>",
971
+ "<SPECIAL_968>",
972
+ "<SPECIAL_969>",
973
+ "<SPECIAL_970>",
974
+ "<SPECIAL_971>",
975
+ "<SPECIAL_972>",
976
+ "<SPECIAL_973>",
977
+ "<SPECIAL_974>",
978
+ "<SPECIAL_975>",
979
+ "<SPECIAL_976>",
980
+ "<SPECIAL_977>",
981
+ "<SPECIAL_978>",
982
+ "<SPECIAL_979>",
983
+ "<SPECIAL_980>",
984
+ "<SPECIAL_981>",
985
+ "<SPECIAL_982>",
986
+ "<SPECIAL_983>",
987
+ "<SPECIAL_984>",
988
+ "<SPECIAL_985>",
989
+ "<SPECIAL_986>",
990
+ "<SPECIAL_987>",
991
+ "<SPECIAL_988>",
992
+ "<SPECIAL_989>",
993
+ "<SPECIAL_990>",
994
+ "<SPECIAL_991>",
995
+ "<SPECIAL_992>",
996
+ "<SPECIAL_993>",
997
+ "<SPECIAL_994>",
998
+ "<SPECIAL_995>",
999
+ "<SPECIAL_996>",
1000
+ "<SPECIAL_997>",
1001
+ "<SPECIAL_998>",
1002
+ "<SPECIAL_999>"
1003
+ ],
1004
+ "bos_token": {
1005
+ "content": "<s>",
1006
+ "lstrip": false,
1007
+ "normalized": false,
1008
+ "rstrip": false,
1009
+ "single_word": false
1010
+ },
1011
+ "eos_token": {
1012
+ "content": "</s>",
1013
+ "lstrip": false,
1014
+ "normalized": false,
1015
+ "rstrip": false,
1016
+ "single_word": false
1017
+ },
1018
+ "pad_token": {
1019
+ "content": "<pad>",
1020
+ "lstrip": false,
1021
+ "normalized": false,
1022
+ "rstrip": false,
1023
+ "single_word": false
1024
+ },
1025
+ "unk_token": {
1026
+ "content": "<unk>",
1027
+ "lstrip": false,
1028
+ "normalized": false,
1029
+ "rstrip": false,
1030
+ "single_word": false
1031
+ }
1032
+ }
tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b76085f9923309d873994d444989f7eb6ec074b06f25b58f1e8d7b7741070949
3
+ size 17078037
tokenizer_config.json ADDED
The diff for this file is too large to render. See raw diff
 
trainer_state.json ADDED
@@ -0,0 +1,1898 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 1.9883268482490273,
6
+ "eval_steps": 32,
7
+ "global_step": 256,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.007782101167315175,
14
+ "grad_norm": 13.2358487482735,
15
+ "learning_rate": 4e-08,
16
+ "loss": 1.3028,
17
+ "step": 1
18
+ },
19
+ {
20
+ "epoch": 0.007782101167315175,
21
+ "eval_loss": 1.1581796407699585,
22
+ "eval_runtime": 193.4094,
23
+ "eval_samples_per_second": 13.939,
24
+ "eval_steps_per_second": 0.222,
25
+ "step": 1
26
+ },
27
+ {
28
+ "epoch": 0.01556420233463035,
29
+ "grad_norm": 13.517130491540176,
30
+ "learning_rate": 8e-08,
31
+ "loss": 1.2437,
32
+ "step": 2
33
+ },
34
+ {
35
+ "epoch": 0.023346303501945526,
36
+ "grad_norm": 13.028970161750262,
37
+ "learning_rate": 1.2000000000000002e-07,
38
+ "loss": 1.3071,
39
+ "step": 3
40
+ },
41
+ {
42
+ "epoch": 0.0311284046692607,
43
+ "grad_norm": 13.538257754320586,
44
+ "learning_rate": 1.6e-07,
45
+ "loss": 1.3013,
46
+ "step": 4
47
+ },
48
+ {
49
+ "epoch": 0.038910505836575876,
50
+ "grad_norm": 13.666771312447045,
51
+ "learning_rate": 2e-07,
52
+ "loss": 1.2823,
53
+ "step": 5
54
+ },
55
+ {
56
+ "epoch": 0.04669260700389105,
57
+ "grad_norm": 12.758203259942219,
58
+ "learning_rate": 2.4000000000000003e-07,
59
+ "loss": 1.2447,
60
+ "step": 6
61
+ },
62
+ {
63
+ "epoch": 0.054474708171206226,
64
+ "grad_norm": 12.521308995567729,
65
+ "learning_rate": 2.8e-07,
66
+ "loss": 1.261,
67
+ "step": 7
68
+ },
69
+ {
70
+ "epoch": 0.0622568093385214,
71
+ "grad_norm": 11.561801223157566,
72
+ "learning_rate": 3.2e-07,
73
+ "loss": 1.2665,
74
+ "step": 8
75
+ },
76
+ {
77
+ "epoch": 0.07003891050583658,
78
+ "grad_norm": 11.156739432208562,
79
+ "learning_rate": 3.6e-07,
80
+ "loss": 1.2488,
81
+ "step": 9
82
+ },
83
+ {
84
+ "epoch": 0.07782101167315175,
85
+ "grad_norm": 10.490731019339595,
86
+ "learning_rate": 4e-07,
87
+ "loss": 1.2695,
88
+ "step": 10
89
+ },
90
+ {
91
+ "epoch": 0.08560311284046693,
92
+ "grad_norm": 8.664509193480505,
93
+ "learning_rate": 4.3999999999999997e-07,
94
+ "loss": 1.2306,
95
+ "step": 11
96
+ },
97
+ {
98
+ "epoch": 0.0933852140077821,
99
+ "grad_norm": 7.187121902510894,
100
+ "learning_rate": 4.800000000000001e-07,
101
+ "loss": 1.2614,
102
+ "step": 12
103
+ },
104
+ {
105
+ "epoch": 0.10116731517509728,
106
+ "grad_norm": 6.109576574934582,
107
+ "learning_rate": 5.2e-07,
108
+ "loss": 1.227,
109
+ "step": 13
110
+ },
111
+ {
112
+ "epoch": 0.10894941634241245,
113
+ "grad_norm": 5.033716100027243,
114
+ "learning_rate": 5.6e-07,
115
+ "loss": 1.2676,
116
+ "step": 14
117
+ },
118
+ {
119
+ "epoch": 0.11673151750972763,
120
+ "grad_norm": 3.4452961961682815,
121
+ "learning_rate": 6e-07,
122
+ "loss": 1.177,
123
+ "step": 15
124
+ },
125
+ {
126
+ "epoch": 0.1245136186770428,
127
+ "grad_norm": 3.1164741035862455,
128
+ "learning_rate": 6.4e-07,
129
+ "loss": 1.229,
130
+ "step": 16
131
+ },
132
+ {
133
+ "epoch": 0.13229571984435798,
134
+ "grad_norm": 3.1002564736005267,
135
+ "learning_rate": 6.8e-07,
136
+ "loss": 1.2488,
137
+ "step": 17
138
+ },
139
+ {
140
+ "epoch": 0.14007782101167315,
141
+ "grad_norm": 3.1257169665944833,
142
+ "learning_rate": 7.2e-07,
143
+ "loss": 1.2324,
144
+ "step": 18
145
+ },
146
+ {
147
+ "epoch": 0.14785992217898833,
148
+ "grad_norm": 3.0581482597501832,
149
+ "learning_rate": 7.599999999999999e-07,
150
+ "loss": 1.1873,
151
+ "step": 19
152
+ },
153
+ {
154
+ "epoch": 0.1556420233463035,
155
+ "grad_norm": 3.043731725003644,
156
+ "learning_rate": 8e-07,
157
+ "loss": 1.1759,
158
+ "step": 20
159
+ },
160
+ {
161
+ "epoch": 0.16342412451361868,
162
+ "grad_norm": 2.734793910693522,
163
+ "learning_rate": 8.4e-07,
164
+ "loss": 1.165,
165
+ "step": 21
166
+ },
167
+ {
168
+ "epoch": 0.17120622568093385,
169
+ "grad_norm": 2.5471637005098233,
170
+ "learning_rate": 8.799999999999999e-07,
171
+ "loss": 1.2258,
172
+ "step": 22
173
+ },
174
+ {
175
+ "epoch": 0.17898832684824903,
176
+ "grad_norm": 2.133879636511503,
177
+ "learning_rate": 9.2e-07,
178
+ "loss": 1.1924,
179
+ "step": 23
180
+ },
181
+ {
182
+ "epoch": 0.1867704280155642,
183
+ "grad_norm": 1.809481995522435,
184
+ "learning_rate": 9.600000000000001e-07,
185
+ "loss": 1.1285,
186
+ "step": 24
187
+ },
188
+ {
189
+ "epoch": 0.19455252918287938,
190
+ "grad_norm": 1.5258999116809353,
191
+ "learning_rate": 1e-06,
192
+ "loss": 1.1613,
193
+ "step": 25
194
+ },
195
+ {
196
+ "epoch": 0.20233463035019456,
197
+ "grad_norm": 1.363595982806325,
198
+ "learning_rate": 9.995654063450673e-07,
199
+ "loss": 1.1791,
200
+ "step": 26
201
+ },
202
+ {
203
+ "epoch": 0.21011673151750973,
204
+ "grad_norm": 1.4125313988908728,
205
+ "learning_rate": 9.9912739965096e-07,
206
+ "loss": 1.2031,
207
+ "step": 27
208
+ },
209
+ {
210
+ "epoch": 0.2178988326848249,
211
+ "grad_norm": 1.446135093305486,
212
+ "learning_rate": 9.986859395532194e-07,
213
+ "loss": 1.1514,
214
+ "step": 28
215
+ },
216
+ {
217
+ "epoch": 0.22568093385214008,
218
+ "grad_norm": 1.342713959761476,
219
+ "learning_rate": 9.98240985048373e-07,
220
+ "loss": 1.1741,
221
+ "step": 29
222
+ },
223
+ {
224
+ "epoch": 0.23346303501945526,
225
+ "grad_norm": 1.2716097075011323,
226
+ "learning_rate": 9.977924944812361e-07,
227
+ "loss": 1.1696,
228
+ "step": 30
229
+ },
230
+ {
231
+ "epoch": 0.24124513618677043,
232
+ "grad_norm": 1.1283654885451528,
233
+ "learning_rate": 9.973404255319148e-07,
234
+ "loss": 1.1564,
235
+ "step": 31
236
+ },
237
+ {
238
+ "epoch": 0.2490272373540856,
239
+ "grad_norm": 1.0833672902427003,
240
+ "learning_rate": 9.968847352024923e-07,
241
+ "loss": 1.1924,
242
+ "step": 32
243
+ },
244
+ {
245
+ "epoch": 0.2490272373540856,
246
+ "eval_loss": 1.017075777053833,
247
+ "eval_runtime": 192.9829,
248
+ "eval_samples_per_second": 13.97,
249
+ "eval_steps_per_second": 0.223,
250
+ "step": 32
251
+ },
252
+ {
253
+ "epoch": 0.25680933852140075,
254
+ "grad_norm": 1.1221714895781139,
255
+ "learning_rate": 9.964253798033959e-07,
256
+ "loss": 1.1953,
257
+ "step": 33
258
+ },
259
+ {
260
+ "epoch": 0.26459143968871596,
261
+ "grad_norm": 1.1667935320099643,
262
+ "learning_rate": 9.959623149394347e-07,
263
+ "loss": 1.1631,
264
+ "step": 34
265
+ },
266
+ {
267
+ "epoch": 0.2723735408560311,
268
+ "grad_norm": 1.1728769110368067,
269
+ "learning_rate": 9.954954954954955e-07,
270
+ "loss": 1.1979,
271
+ "step": 35
272
+ },
273
+ {
274
+ "epoch": 0.2801556420233463,
275
+ "grad_norm": 1.1879928099076524,
276
+ "learning_rate": 9.950248756218905e-07,
277
+ "loss": 1.2109,
278
+ "step": 36
279
+ },
280
+ {
281
+ "epoch": 0.28793774319066145,
282
+ "grad_norm": 1.0793090685595952,
283
+ "learning_rate": 9.94550408719346e-07,
284
+ "loss": 1.1517,
285
+ "step": 37
286
+ },
287
+ {
288
+ "epoch": 0.29571984435797666,
289
+ "grad_norm": 1.2162353553544407,
290
+ "learning_rate": 9.940720474236205e-07,
291
+ "loss": 1.1667,
292
+ "step": 38
293
+ },
294
+ {
295
+ "epoch": 0.3035019455252918,
296
+ "grad_norm": 1.1553136659353096,
297
+ "learning_rate": 9.935897435897434e-07,
298
+ "loss": 1.1564,
299
+ "step": 39
300
+ },
301
+ {
302
+ "epoch": 0.311284046692607,
303
+ "grad_norm": 0.8975925277420491,
304
+ "learning_rate": 9.931034482758622e-07,
305
+ "loss": 1.1409,
306
+ "step": 40
307
+ },
308
+ {
309
+ "epoch": 0.31906614785992216,
310
+ "grad_norm": 0.848010373104077,
311
+ "learning_rate": 9.92613111726685e-07,
312
+ "loss": 1.1489,
313
+ "step": 41
314
+ },
315
+ {
316
+ "epoch": 0.32684824902723736,
317
+ "grad_norm": 0.8445987480588121,
318
+ "learning_rate": 9.921186833565135e-07,
319
+ "loss": 1.1518,
320
+ "step": 42
321
+ },
322
+ {
323
+ "epoch": 0.3346303501945525,
324
+ "grad_norm": 0.8910485581415633,
325
+ "learning_rate": 9.916201117318436e-07,
326
+ "loss": 1.1663,
327
+ "step": 43
328
+ },
329
+ {
330
+ "epoch": 0.3424124513618677,
331
+ "grad_norm": 0.8226662462385251,
332
+ "learning_rate": 9.911173445535296e-07,
333
+ "loss": 1.1975,
334
+ "step": 44
335
+ },
336
+ {
337
+ "epoch": 0.35019455252918286,
338
+ "grad_norm": 1.266452159867259,
339
+ "learning_rate": 9.906103286384975e-07,
340
+ "loss": 1.19,
341
+ "step": 45
342
+ },
343
+ {
344
+ "epoch": 0.35797665369649806,
345
+ "grad_norm": 0.8228834741800612,
346
+ "learning_rate": 9.900990099009902e-07,
347
+ "loss": 1.1672,
348
+ "step": 46
349
+ },
350
+ {
351
+ "epoch": 0.3657587548638132,
352
+ "grad_norm": 0.7967316496651314,
353
+ "learning_rate": 9.895833333333333e-07,
354
+ "loss": 1.1653,
355
+ "step": 47
356
+ },
357
+ {
358
+ "epoch": 0.3735408560311284,
359
+ "grad_norm": 0.8385685679438281,
360
+ "learning_rate": 9.8906324298621e-07,
361
+ "loss": 1.2161,
362
+ "step": 48
363
+ },
364
+ {
365
+ "epoch": 0.38132295719844356,
366
+ "grad_norm": 0.7832731689890803,
367
+ "learning_rate": 9.88538681948424e-07,
368
+ "loss": 1.1493,
369
+ "step": 49
370
+ },
371
+ {
372
+ "epoch": 0.38910505836575876,
373
+ "grad_norm": 0.7470950974115376,
374
+ "learning_rate": 9.88009592326139e-07,
375
+ "loss": 1.1444,
376
+ "step": 50
377
+ },
378
+ {
379
+ "epoch": 0.3968871595330739,
380
+ "grad_norm": 0.9407081004608256,
381
+ "learning_rate": 9.874759152215799e-07,
382
+ "loss": 1.2262,
383
+ "step": 51
384
+ },
385
+ {
386
+ "epoch": 0.4046692607003891,
387
+ "grad_norm": 0.7807613751449256,
388
+ "learning_rate": 9.869375907111757e-07,
389
+ "loss": 1.1761,
390
+ "step": 52
391
+ },
392
+ {
393
+ "epoch": 0.41245136186770426,
394
+ "grad_norm": 0.7814091677207984,
395
+ "learning_rate": 9.86394557823129e-07,
396
+ "loss": 1.1684,
397
+ "step": 53
398
+ },
399
+ {
400
+ "epoch": 0.42023346303501946,
401
+ "grad_norm": 0.7209307303105739,
402
+ "learning_rate": 9.858467545143973e-07,
403
+ "loss": 1.1429,
404
+ "step": 54
405
+ },
406
+ {
407
+ "epoch": 0.4280155642023346,
408
+ "grad_norm": 0.6839759799403674,
409
+ "learning_rate": 9.852941176470587e-07,
410
+ "loss": 1.1591,
411
+ "step": 55
412
+ },
413
+ {
414
+ "epoch": 0.4357976653696498,
415
+ "grad_norm": 0.7399500476567901,
416
+ "learning_rate": 9.847365829640571e-07,
417
+ "loss": 1.1264,
418
+ "step": 56
419
+ },
420
+ {
421
+ "epoch": 0.44357976653696496,
422
+ "grad_norm": 0.7250794425256079,
423
+ "learning_rate": 9.841740850642927e-07,
424
+ "loss": 1.1764,
425
+ "step": 57
426
+ },
427
+ {
428
+ "epoch": 0.45136186770428016,
429
+ "grad_norm": 0.6729705016341,
430
+ "learning_rate": 9.836065573770493e-07,
431
+ "loss": 1.1557,
432
+ "step": 58
433
+ },
434
+ {
435
+ "epoch": 0.4591439688715953,
436
+ "grad_norm": 0.6814295448695143,
437
+ "learning_rate": 9.830339321357286e-07,
438
+ "loss": 1.1674,
439
+ "step": 59
440
+ },
441
+ {
442
+ "epoch": 0.4669260700389105,
443
+ "grad_norm": 0.6657154699765654,
444
+ "learning_rate": 9.824561403508773e-07,
445
+ "loss": 1.1441,
446
+ "step": 60
447
+ },
448
+ {
449
+ "epoch": 0.47470817120622566,
450
+ "grad_norm": 0.6256174713114324,
451
+ "learning_rate": 9.818731117824774e-07,
452
+ "loss": 1.1797,
453
+ "step": 61
454
+ },
455
+ {
456
+ "epoch": 0.48249027237354086,
457
+ "grad_norm": 0.6194904383458136,
458
+ "learning_rate": 9.81284774911482e-07,
459
+ "loss": 1.1476,
460
+ "step": 62
461
+ },
462
+ {
463
+ "epoch": 0.490272373540856,
464
+ "grad_norm": 0.7409013991341397,
465
+ "learning_rate": 9.80691056910569e-07,
466
+ "loss": 1.1563,
467
+ "step": 63
468
+ },
469
+ {
470
+ "epoch": 0.4980544747081712,
471
+ "grad_norm": 0.6529431777087924,
472
+ "learning_rate": 9.800918836140888e-07,
473
+ "loss": 1.13,
474
+ "step": 64
475
+ },
476
+ {
477
+ "epoch": 0.4980544747081712,
478
+ "eval_loss": 0.993022084236145,
479
+ "eval_runtime": 193.0792,
480
+ "eval_samples_per_second": 13.963,
481
+ "eval_steps_per_second": 0.223,
482
+ "step": 64
483
+ },
484
+ {
485
+ "epoch": 0.5058365758754864,
486
+ "grad_norm": 0.6249987093009363,
487
+ "learning_rate": 9.794871794871796e-07,
488
+ "loss": 1.1365,
489
+ "step": 65
490
+ },
491
+ {
492
+ "epoch": 0.5136186770428015,
493
+ "grad_norm": 0.5714052179303046,
494
+ "learning_rate": 9.788768675940237e-07,
495
+ "loss": 1.1486,
496
+ "step": 66
497
+ },
498
+ {
499
+ "epoch": 0.5214007782101168,
500
+ "grad_norm": 0.6266253830635832,
501
+ "learning_rate": 9.782608695652173e-07,
502
+ "loss": 1.1297,
503
+ "step": 67
504
+ },
505
+ {
506
+ "epoch": 0.5291828793774319,
507
+ "grad_norm": 0.6413361931814175,
508
+ "learning_rate": 9.776391055642225e-07,
509
+ "loss": 1.1346,
510
+ "step": 68
511
+ },
512
+ {
513
+ "epoch": 0.5369649805447471,
514
+ "grad_norm": 0.6401174657840715,
515
+ "learning_rate": 9.770114942528735e-07,
516
+ "loss": 1.1443,
517
+ "step": 69
518
+ },
519
+ {
520
+ "epoch": 0.5447470817120622,
521
+ "grad_norm": 0.6160976711769636,
522
+ "learning_rate": 9.763779527559055e-07,
523
+ "loss": 1.138,
524
+ "step": 70
525
+ },
526
+ {
527
+ "epoch": 0.5525291828793775,
528
+ "grad_norm": 0.6234000690415658,
529
+ "learning_rate": 9.757383966244725e-07,
530
+ "loss": 1.1018,
531
+ "step": 71
532
+ },
533
+ {
534
+ "epoch": 0.5603112840466926,
535
+ "grad_norm": 0.6528376037799454,
536
+ "learning_rate": 9.75092739798622e-07,
537
+ "loss": 1.1059,
538
+ "step": 72
539
+ },
540
+ {
541
+ "epoch": 0.5680933852140078,
542
+ "grad_norm": 0.5652314830088928,
543
+ "learning_rate": 9.744408945686901e-07,
544
+ "loss": 1.1222,
545
+ "step": 73
546
+ },
547
+ {
548
+ "epoch": 0.5758754863813229,
549
+ "grad_norm": 0.585673991341348,
550
+ "learning_rate": 9.737827715355804e-07,
551
+ "loss": 1.1067,
552
+ "step": 74
553
+ },
554
+ {
555
+ "epoch": 0.5836575875486382,
556
+ "grad_norm": 0.5888031772006102,
557
+ "learning_rate": 9.731182795698924e-07,
558
+ "loss": 1.1495,
559
+ "step": 75
560
+ },
561
+ {
562
+ "epoch": 0.5914396887159533,
563
+ "grad_norm": 0.6220423369607407,
564
+ "learning_rate": 9.72447325769854e-07,
565
+ "loss": 1.1197,
566
+ "step": 76
567
+ },
568
+ {
569
+ "epoch": 0.5992217898832685,
570
+ "grad_norm": 0.6112403683619732,
571
+ "learning_rate": 9.71769815418024e-07,
572
+ "loss": 1.1382,
573
+ "step": 77
574
+ },
575
+ {
576
+ "epoch": 0.6070038910505836,
577
+ "grad_norm": 0.6459405223806135,
578
+ "learning_rate": 9.710856519367158e-07,
579
+ "loss": 1.1518,
580
+ "step": 78
581
+ },
582
+ {
583
+ "epoch": 0.6147859922178989,
584
+ "grad_norm": 0.6209769863467987,
585
+ "learning_rate": 9.703947368421054e-07,
586
+ "loss": 1.1239,
587
+ "step": 79
588
+ },
589
+ {
590
+ "epoch": 0.622568093385214,
591
+ "grad_norm": 0.5843668163519747,
592
+ "learning_rate": 9.696969696969695e-07,
593
+ "loss": 1.1433,
594
+ "step": 80
595
+ },
596
+ {
597
+ "epoch": 0.6303501945525292,
598
+ "grad_norm": 0.5969841579813842,
599
+ "learning_rate": 9.689922480620153e-07,
600
+ "loss": 1.1312,
601
+ "step": 81
602
+ },
603
+ {
604
+ "epoch": 0.6381322957198443,
605
+ "grad_norm": 0.5567475366696957,
606
+ "learning_rate": 9.68280467445743e-07,
607
+ "loss": 1.0717,
608
+ "step": 82
609
+ },
610
+ {
611
+ "epoch": 0.6459143968871596,
612
+ "grad_norm": 0.8841164939701298,
613
+ "learning_rate": 9.675615212527965e-07,
614
+ "loss": 1.1229,
615
+ "step": 83
616
+ },
617
+ {
618
+ "epoch": 0.6536964980544747,
619
+ "grad_norm": 0.6021627871478876,
620
+ "learning_rate": 9.668353007307475e-07,
621
+ "loss": 1.119,
622
+ "step": 84
623
+ },
624
+ {
625
+ "epoch": 0.6614785992217899,
626
+ "grad_norm": 0.561817882301385,
627
+ "learning_rate": 9.661016949152542e-07,
628
+ "loss": 1.1298,
629
+ "step": 85
630
+ },
631
+ {
632
+ "epoch": 0.669260700389105,
633
+ "grad_norm": 0.6382338925455957,
634
+ "learning_rate": 9.653605905735377e-07,
635
+ "loss": 1.1168,
636
+ "step": 86
637
+ },
638
+ {
639
+ "epoch": 0.6770428015564203,
640
+ "grad_norm": 0.602322633909128,
641
+ "learning_rate": 9.646118721461186e-07,
642
+ "loss": 1.1291,
643
+ "step": 87
644
+ },
645
+ {
646
+ "epoch": 0.6848249027237354,
647
+ "grad_norm": 0.5990302538351283,
648
+ "learning_rate": 9.63855421686747e-07,
649
+ "loss": 1.1332,
650
+ "step": 88
651
+ },
652
+ {
653
+ "epoch": 0.6926070038910506,
654
+ "grad_norm": 0.5869777798515858,
655
+ "learning_rate": 9.630911188004613e-07,
656
+ "loss": 1.1404,
657
+ "step": 89
658
+ },
659
+ {
660
+ "epoch": 0.7003891050583657,
661
+ "grad_norm": 0.5933684142522954,
662
+ "learning_rate": 9.623188405797102e-07,
663
+ "loss": 1.1528,
664
+ "step": 90
665
+ },
666
+ {
667
+ "epoch": 0.708171206225681,
668
+ "grad_norm": 0.6118209958752469,
669
+ "learning_rate": 9.615384615384615e-07,
670
+ "loss": 1.1486,
671
+ "step": 91
672
+ },
673
+ {
674
+ "epoch": 0.7159533073929961,
675
+ "grad_norm": 0.626189163793506,
676
+ "learning_rate": 9.607498535442295e-07,
677
+ "loss": 1.1079,
678
+ "step": 92
679
+ },
680
+ {
681
+ "epoch": 0.7237354085603113,
682
+ "grad_norm": 0.5824843670586056,
683
+ "learning_rate": 9.599528857479386e-07,
684
+ "loss": 1.1471,
685
+ "step": 93
686
+ },
687
+ {
688
+ "epoch": 0.7315175097276264,
689
+ "grad_norm": 0.6045894845802466,
690
+ "learning_rate": 9.591474245115454e-07,
691
+ "loss": 1.0899,
692
+ "step": 94
693
+ },
694
+ {
695
+ "epoch": 0.7392996108949417,
696
+ "grad_norm": 0.5453528811988809,
697
+ "learning_rate": 9.583333333333334e-07,
698
+ "loss": 1.1232,
699
+ "step": 95
700
+ },
701
+ {
702
+ "epoch": 0.7470817120622568,
703
+ "grad_norm": 0.6430932607963977,
704
+ "learning_rate": 9.57510472770796e-07,
705
+ "loss": 1.0917,
706
+ "step": 96
707
+ },
708
+ {
709
+ "epoch": 0.7470817120622568,
710
+ "eval_loss": 0.9749420881271362,
711
+ "eval_runtime": 194.1545,
712
+ "eval_samples_per_second": 13.886,
713
+ "eval_steps_per_second": 0.221,
714
+ "step": 96
715
+ },
716
+ {
717
+ "epoch": 0.754863813229572,
718
+ "grad_norm": 0.5852779384368416,
719
+ "learning_rate": 9.566787003610106e-07,
720
+ "loss": 1.1517,
721
+ "step": 97
722
+ },
723
+ {
724
+ "epoch": 0.7626459143968871,
725
+ "grad_norm": 0.6409842274490922,
726
+ "learning_rate": 9.55837870538415e-07,
727
+ "loss": 1.1254,
728
+ "step": 98
729
+ },
730
+ {
731
+ "epoch": 0.7704280155642024,
732
+ "grad_norm": 0.5639528294869124,
733
+ "learning_rate": 9.549878345498782e-07,
734
+ "loss": 1.1121,
735
+ "step": 99
736
+ },
737
+ {
738
+ "epoch": 0.7782101167315175,
739
+ "grad_norm": 0.6213244609041039,
740
+ "learning_rate": 9.541284403669725e-07,
741
+ "loss": 1.0915,
742
+ "step": 100
743
+ },
744
+ {
745
+ "epoch": 0.7859922178988327,
746
+ "grad_norm": 0.6551114483360119,
747
+ "learning_rate": 9.53259532595326e-07,
748
+ "loss": 1.1534,
749
+ "step": 101
750
+ },
751
+ {
752
+ "epoch": 0.7937743190661478,
753
+ "grad_norm": 0.6413194021766461,
754
+ "learning_rate": 9.523809523809523e-07,
755
+ "loss": 1.1432,
756
+ "step": 102
757
+ },
758
+ {
759
+ "epoch": 0.8015564202334631,
760
+ "grad_norm": 0.5912246112879433,
761
+ "learning_rate": 9.514925373134328e-07,
762
+ "loss": 1.1856,
763
+ "step": 103
764
+ },
765
+ {
766
+ "epoch": 0.8093385214007782,
767
+ "grad_norm": 0.5929209222521526,
768
+ "learning_rate": 9.505941213258286e-07,
769
+ "loss": 1.1097,
770
+ "step": 104
771
+ },
772
+ {
773
+ "epoch": 0.8171206225680934,
774
+ "grad_norm": 0.5781185505602303,
775
+ "learning_rate": 9.496855345911948e-07,
776
+ "loss": 1.093,
777
+ "step": 105
778
+ },
779
+ {
780
+ "epoch": 0.8249027237354085,
781
+ "grad_norm": 0.5850063235452896,
782
+ "learning_rate": 9.487666034155598e-07,
783
+ "loss": 1.1053,
784
+ "step": 106
785
+ },
786
+ {
787
+ "epoch": 0.8326848249027238,
788
+ "grad_norm": 0.5557542208140502,
789
+ "learning_rate": 9.478371501272264e-07,
790
+ "loss": 1.1072,
791
+ "step": 107
792
+ },
793
+ {
794
+ "epoch": 0.8404669260700389,
795
+ "grad_norm": 0.6939378070209432,
796
+ "learning_rate": 9.46896992962252e-07,
797
+ "loss": 1.1557,
798
+ "step": 108
799
+ },
800
+ {
801
+ "epoch": 0.8482490272373541,
802
+ "grad_norm": 0.5821602482631876,
803
+ "learning_rate": 9.459459459459459e-07,
804
+ "loss": 1.1008,
805
+ "step": 109
806
+ },
807
+ {
808
+ "epoch": 0.8560311284046692,
809
+ "grad_norm": 0.5641222765431608,
810
+ "learning_rate": 9.449838187702264e-07,
811
+ "loss": 1.1066,
812
+ "step": 110
813
+ },
814
+ {
815
+ "epoch": 0.8638132295719845,
816
+ "grad_norm": 0.6073853509255491,
817
+ "learning_rate": 9.440104166666666e-07,
818
+ "loss": 1.1347,
819
+ "step": 111
820
+ },
821
+ {
822
+ "epoch": 0.8715953307392996,
823
+ "grad_norm": 0.6652073708936578,
824
+ "learning_rate": 9.430255402750491e-07,
825
+ "loss": 1.1539,
826
+ "step": 112
827
+ },
828
+ {
829
+ "epoch": 0.8793774319066148,
830
+ "grad_norm": 0.5779065641712691,
831
+ "learning_rate": 9.420289855072464e-07,
832
+ "loss": 1.1286,
833
+ "step": 113
834
+ },
835
+ {
836
+ "epoch": 0.8871595330739299,
837
+ "grad_norm": 1.0893325101395284,
838
+ "learning_rate": 9.410205434062292e-07,
839
+ "loss": 1.1327,
840
+ "step": 114
841
+ },
842
+ {
843
+ "epoch": 0.8949416342412452,
844
+ "grad_norm": 0.5897769203979899,
845
+ "learning_rate": 9.399999999999999e-07,
846
+ "loss": 1.1083,
847
+ "step": 115
848
+ },
849
+ {
850
+ "epoch": 0.9027237354085603,
851
+ "grad_norm": 0.5962317172776513,
852
+ "learning_rate": 9.389671361502347e-07,
853
+ "loss": 1.0842,
854
+ "step": 116
855
+ },
856
+ {
857
+ "epoch": 0.9105058365758755,
858
+ "grad_norm": 0.5955932544091112,
859
+ "learning_rate": 9.379217273954116e-07,
860
+ "loss": 1.1366,
861
+ "step": 117
862
+ },
863
+ {
864
+ "epoch": 0.9182879377431906,
865
+ "grad_norm": 0.5949693376391756,
866
+ "learning_rate": 9.368635437881874e-07,
867
+ "loss": 1.1094,
868
+ "step": 118
869
+ },
870
+ {
871
+ "epoch": 0.9260700389105059,
872
+ "grad_norm": 0.6044514449577904,
873
+ "learning_rate": 9.357923497267759e-07,
874
+ "loss": 1.1639,
875
+ "step": 119
876
+ },
877
+ {
878
+ "epoch": 0.933852140077821,
879
+ "grad_norm": 0.625428671903296,
880
+ "learning_rate": 9.347079037800687e-07,
881
+ "loss": 1.1054,
882
+ "step": 120
883
+ },
884
+ {
885
+ "epoch": 0.9416342412451362,
886
+ "grad_norm": 0.5297520430435139,
887
+ "learning_rate": 9.33609958506224e-07,
888
+ "loss": 1.1396,
889
+ "step": 121
890
+ },
891
+ {
892
+ "epoch": 0.9494163424124513,
893
+ "grad_norm": 0.604395748227856,
894
+ "learning_rate": 9.324982602644397e-07,
895
+ "loss": 1.1297,
896
+ "step": 122
897
+ },
898
+ {
899
+ "epoch": 0.9571984435797666,
900
+ "grad_norm": 0.5935060150814615,
901
+ "learning_rate": 9.313725490196079e-07,
902
+ "loss": 1.1327,
903
+ "step": 123
904
+ },
905
+ {
906
+ "epoch": 0.9649805447470817,
907
+ "grad_norm": 0.6213246371577336,
908
+ "learning_rate": 9.30232558139535e-07,
909
+ "loss": 1.1655,
910
+ "step": 124
911
+ },
912
+ {
913
+ "epoch": 0.9727626459143969,
914
+ "grad_norm": 0.5459987203799985,
915
+ "learning_rate": 9.290780141843972e-07,
916
+ "loss": 1.1119,
917
+ "step": 125
918
+ },
919
+ {
920
+ "epoch": 0.980544747081712,
921
+ "grad_norm": 0.5862633284111678,
922
+ "learning_rate": 9.279086366880799e-07,
923
+ "loss": 1.1102,
924
+ "step": 126
925
+ },
926
+ {
927
+ "epoch": 0.9883268482490273,
928
+ "grad_norm": 0.5749701605073742,
929
+ "learning_rate": 9.267241379310344e-07,
930
+ "loss": 1.1215,
931
+ "step": 127
932
+ },
933
+ {
934
+ "epoch": 0.9961089494163424,
935
+ "grad_norm": 0.5608948491944076,
936
+ "learning_rate": 9.25524222704266e-07,
937
+ "loss": 1.1171,
938
+ "step": 128
939
+ },
940
+ {
941
+ "epoch": 0.9961089494163424,
942
+ "eval_loss": 0.9638092517852783,
943
+ "eval_runtime": 193.8483,
944
+ "eval_samples_per_second": 13.908,
945
+ "eval_steps_per_second": 0.222,
946
+ "step": 128
947
+ },
948
+ {
949
+ "epoch": 1.0,
950
+ "grad_norm": 0.5608948491944076,
951
+ "learning_rate": 9.243085880640464e-07,
952
+ "loss": 1.1281,
953
+ "step": 129
954
+ },
955
+ {
956
+ "epoch": 1.0077821011673151,
957
+ "grad_norm": 0.8826714731817153,
958
+ "learning_rate": 9.230769230769229e-07,
959
+ "loss": 1.1385,
960
+ "step": 130
961
+ },
962
+ {
963
+ "epoch": 1.0155642023346303,
964
+ "grad_norm": 0.585537303127348,
965
+ "learning_rate": 9.218289085545723e-07,
966
+ "loss": 1.0753,
967
+ "step": 131
968
+ },
969
+ {
970
+ "epoch": 1.0233463035019454,
971
+ "grad_norm": 0.574845853397158,
972
+ "learning_rate": 9.205642167780252e-07,
973
+ "loss": 1.1466,
974
+ "step": 132
975
+ },
976
+ {
977
+ "epoch": 1.0311284046692606,
978
+ "grad_norm": 0.573244467973897,
979
+ "learning_rate": 9.192825112107622e-07,
980
+ "loss": 1.1359,
981
+ "step": 133
982
+ },
983
+ {
984
+ "epoch": 1.038910505836576,
985
+ "grad_norm": 0.6964714993014424,
986
+ "learning_rate": 9.179834462001504e-07,
987
+ "loss": 1.1122,
988
+ "step": 134
989
+ },
990
+ {
991
+ "epoch": 1.046692607003891,
992
+ "grad_norm": 0.641322046313728,
993
+ "learning_rate": 9.166666666666665e-07,
994
+ "loss": 1.0825,
995
+ "step": 135
996
+ },
997
+ {
998
+ "epoch": 1.0544747081712063,
999
+ "grad_norm": 0.5360716115631284,
1000
+ "learning_rate": 9.153318077803201e-07,
1001
+ "loss": 1.1006,
1002
+ "step": 136
1003
+ },
1004
+ {
1005
+ "epoch": 1.0622568093385214,
1006
+ "grad_norm": 0.5816374932590664,
1007
+ "learning_rate": 9.139784946236559e-07,
1008
+ "loss": 1.1116,
1009
+ "step": 137
1010
+ },
1011
+ {
1012
+ "epoch": 1.0700389105058365,
1013
+ "grad_norm": 0.6372126076146449,
1014
+ "learning_rate": 9.126063418406805e-07,
1015
+ "loss": 1.0979,
1016
+ "step": 138
1017
+ },
1018
+ {
1019
+ "epoch": 1.0778210116731517,
1020
+ "grad_norm": 0.5578000693621826,
1021
+ "learning_rate": 9.11214953271028e-07,
1022
+ "loss": 1.1263,
1023
+ "step": 139
1024
+ },
1025
+ {
1026
+ "epoch": 1.0856031128404668,
1027
+ "grad_norm": 0.5923651609485125,
1028
+ "learning_rate": 9.098039215686274e-07,
1029
+ "loss": 1.0966,
1030
+ "step": 140
1031
+ },
1032
+ {
1033
+ "epoch": 1.0933852140077822,
1034
+ "grad_norm": 0.533195556847129,
1035
+ "learning_rate": 9.083728278041073e-07,
1036
+ "loss": 1.1349,
1037
+ "step": 141
1038
+ },
1039
+ {
1040
+ "epoch": 1.1011673151750974,
1041
+ "grad_norm": 0.6329974401580136,
1042
+ "learning_rate": 9.069212410501193e-07,
1043
+ "loss": 1.1072,
1044
+ "step": 142
1045
+ },
1046
+ {
1047
+ "epoch": 1.1089494163424125,
1048
+ "grad_norm": 0.5546033043372056,
1049
+ "learning_rate": 9.054487179487179e-07,
1050
+ "loss": 1.1565,
1051
+ "step": 143
1052
+ },
1053
+ {
1054
+ "epoch": 1.1167315175097277,
1055
+ "grad_norm": 0.5975088088440533,
1056
+ "learning_rate": 9.03954802259887e-07,
1057
+ "loss": 1.0724,
1058
+ "step": 144
1059
+ },
1060
+ {
1061
+ "epoch": 1.1245136186770428,
1062
+ "grad_norm": 0.57801276427658,
1063
+ "learning_rate": 9.024390243902439e-07,
1064
+ "loss": 1.1346,
1065
+ "step": 145
1066
+ },
1067
+ {
1068
+ "epoch": 1.132295719844358,
1069
+ "grad_norm": 0.5879398504293435,
1070
+ "learning_rate": 9.009009009009008e-07,
1071
+ "loss": 1.159,
1072
+ "step": 146
1073
+ },
1074
+ {
1075
+ "epoch": 1.140077821011673,
1076
+ "grad_norm": 0.593628488224272,
1077
+ "learning_rate": 8.993399339933992e-07,
1078
+ "loss": 1.1464,
1079
+ "step": 147
1080
+ },
1081
+ {
1082
+ "epoch": 1.1478599221789882,
1083
+ "grad_norm": 0.5593315767964653,
1084
+ "learning_rate": 8.977556109725684e-07,
1085
+ "loss": 1.1054,
1086
+ "step": 148
1087
+ },
1088
+ {
1089
+ "epoch": 1.1556420233463034,
1090
+ "grad_norm": 0.5787699146029086,
1091
+ "learning_rate": 8.96147403685092e-07,
1092
+ "loss": 1.096,
1093
+ "step": 149
1094
+ },
1095
+ {
1096
+ "epoch": 1.1634241245136188,
1097
+ "grad_norm": 0.5877657497768165,
1098
+ "learning_rate": 8.945147679324893e-07,
1099
+ "loss": 1.0887,
1100
+ "step": 150
1101
+ },
1102
+ {
1103
+ "epoch": 1.171206225680934,
1104
+ "grad_norm": 0.6197879903738175,
1105
+ "learning_rate": 8.928571428571428e-07,
1106
+ "loss": 1.1514,
1107
+ "step": 151
1108
+ },
1109
+ {
1110
+ "epoch": 1.178988326848249,
1111
+ "grad_norm": 0.5393516857845513,
1112
+ "learning_rate": 8.911739502999142e-07,
1113
+ "loss": 1.1215,
1114
+ "step": 152
1115
+ },
1116
+ {
1117
+ "epoch": 1.1867704280155642,
1118
+ "grad_norm": 0.5758698948439026,
1119
+ "learning_rate": 8.894645941278064e-07,
1120
+ "loss": 1.0557,
1121
+ "step": 153
1122
+ },
1123
+ {
1124
+ "epoch": 1.1945525291828794,
1125
+ "grad_norm": 0.5870780530298951,
1126
+ "learning_rate": 8.877284595300261e-07,
1127
+ "loss": 1.0892,
1128
+ "step": 154
1129
+ },
1130
+ {
1131
+ "epoch": 1.2023346303501945,
1132
+ "grad_norm": 0.5933780513917563,
1133
+ "learning_rate": 8.859649122807017e-07,
1134
+ "loss": 1.1125,
1135
+ "step": 155
1136
+ },
1137
+ {
1138
+ "epoch": 1.2101167315175096,
1139
+ "grad_norm": 0.5615488624761017,
1140
+ "learning_rate": 8.841732979664014e-07,
1141
+ "loss": 1.1394,
1142
+ "step": 156
1143
+ },
1144
+ {
1145
+ "epoch": 1.217898832684825,
1146
+ "grad_norm": 0.5960612312151845,
1147
+ "learning_rate": 8.823529411764706e-07,
1148
+ "loss": 1.0859,
1149
+ "step": 157
1150
+ },
1151
+ {
1152
+ "epoch": 1.2256809338521402,
1153
+ "grad_norm": 0.6325808125841859,
1154
+ "learning_rate": 8.80503144654088e-07,
1155
+ "loss": 1.1128,
1156
+ "step": 158
1157
+ },
1158
+ {
1159
+ "epoch": 1.2334630350194553,
1160
+ "grad_norm": 0.6076576722454798,
1161
+ "learning_rate": 8.78623188405797e-07,
1162
+ "loss": 1.1092,
1163
+ "step": 159
1164
+ },
1165
+ {
1166
+ "epoch": 1.2412451361867705,
1167
+ "grad_norm": 0.5821560609671981,
1168
+ "learning_rate": 8.767123287671233e-07,
1169
+ "loss": 1.0963,
1170
+ "step": 160
1171
+ },
1172
+ {
1173
+ "epoch": 1.2412451361867705,
1174
+ "eval_loss": 0.9557842016220093,
1175
+ "eval_runtime": 193.7527,
1176
+ "eval_samples_per_second": 13.915,
1177
+ "eval_steps_per_second": 0.222,
1178
+ "step": 160
1179
+ },
1180
+ {
1181
+ "epoch": 1.2490272373540856,
1182
+ "grad_norm": 0.5621223931485095,
1183
+ "learning_rate": 8.747697974217309e-07,
1184
+ "loss": 1.1338,
1185
+ "step": 161
1186
+ },
1187
+ {
1188
+ "epoch": 1.2568093385214008,
1189
+ "grad_norm": 0.5962800493628467,
1190
+ "learning_rate": 8.727948003714019e-07,
1191
+ "loss": 1.1388,
1192
+ "step": 162
1193
+ },
1194
+ {
1195
+ "epoch": 1.264591439688716,
1196
+ "grad_norm": 0.5814349430158473,
1197
+ "learning_rate": 8.707865168539326e-07,
1198
+ "loss": 1.1038,
1199
+ "step": 163
1200
+ },
1201
+ {
1202
+ "epoch": 1.272373540856031,
1203
+ "grad_norm": 0.6004347939909851,
1204
+ "learning_rate": 8.687440982058545e-07,
1205
+ "loss": 1.143,
1206
+ "step": 164
1207
+ },
1208
+ {
1209
+ "epoch": 1.2801556420233462,
1210
+ "grad_norm": 0.5948532065914537,
1211
+ "learning_rate": 8.666666666666666e-07,
1212
+ "loss": 1.155,
1213
+ "step": 165
1214
+ },
1215
+ {
1216
+ "epoch": 1.2879377431906613,
1217
+ "grad_norm": 0.5570114843646816,
1218
+ "learning_rate": 8.645533141210375e-07,
1219
+ "loss": 1.095,
1220
+ "step": 166
1221
+ },
1222
+ {
1223
+ "epoch": 1.2957198443579767,
1224
+ "grad_norm": 0.6592254796476543,
1225
+ "learning_rate": 8.624031007751938e-07,
1226
+ "loss": 1.1114,
1227
+ "step": 167
1228
+ },
1229
+ {
1230
+ "epoch": 1.3035019455252919,
1231
+ "grad_norm": 0.5375793451912569,
1232
+ "learning_rate": 8.602150537634409e-07,
1233
+ "loss": 1.1018,
1234
+ "step": 168
1235
+ },
1236
+ {
1237
+ "epoch": 1.311284046692607,
1238
+ "grad_norm": 0.5774524273066066,
1239
+ "learning_rate": 8.579881656804733e-07,
1240
+ "loss": 1.0866,
1241
+ "step": 169
1242
+ },
1243
+ {
1244
+ "epoch": 1.3190661478599222,
1245
+ "grad_norm": 0.5767675794480738,
1246
+ "learning_rate": 8.557213930348258e-07,
1247
+ "loss": 1.0919,
1248
+ "step": 170
1249
+ },
1250
+ {
1251
+ "epoch": 1.3268482490272373,
1252
+ "grad_norm": 0.5424563326577124,
1253
+ "learning_rate": 8.534136546184737e-07,
1254
+ "loss": 1.097,
1255
+ "step": 171
1256
+ },
1257
+ {
1258
+ "epoch": 1.3346303501945525,
1259
+ "grad_norm": 0.5504324773913235,
1260
+ "learning_rate": 8.510638297872341e-07,
1261
+ "loss": 1.1088,
1262
+ "step": 172
1263
+ },
1264
+ {
1265
+ "epoch": 1.3424124513618678,
1266
+ "grad_norm": 0.5631592132878254,
1267
+ "learning_rate": 8.486707566462167e-07,
1268
+ "loss": 1.141,
1269
+ "step": 173
1270
+ },
1271
+ {
1272
+ "epoch": 1.350194552529183,
1273
+ "grad_norm": 0.5987091201900138,
1274
+ "learning_rate": 8.462332301341588e-07,
1275
+ "loss": 1.1352,
1276
+ "step": 174
1277
+ },
1278
+ {
1279
+ "epoch": 1.3579766536964981,
1280
+ "grad_norm": 0.5644686177164949,
1281
+ "learning_rate": 8.4375e-07,
1282
+ "loss": 1.1086,
1283
+ "step": 175
1284
+ },
1285
+ {
1286
+ "epoch": 1.3657587548638133,
1287
+ "grad_norm": 0.5603933667416149,
1288
+ "learning_rate": 8.412197686645635e-07,
1289
+ "loss": 1.1066,
1290
+ "step": 176
1291
+ },
1292
+ {
1293
+ "epoch": 1.3735408560311284,
1294
+ "grad_norm": 1.1467757879130558,
1295
+ "learning_rate": 8.386411889596603e-07,
1296
+ "loss": 1.1587,
1297
+ "step": 177
1298
+ },
1299
+ {
1300
+ "epoch": 1.3813229571984436,
1301
+ "grad_norm": 0.7887321457214312,
1302
+ "learning_rate": 8.360128617363344e-07,
1303
+ "loss": 1.088,
1304
+ "step": 178
1305
+ },
1306
+ {
1307
+ "epoch": 1.3891050583657587,
1308
+ "grad_norm": 1.3469817773930703,
1309
+ "learning_rate": 8.333333333333332e-07,
1310
+ "loss": 1.0843,
1311
+ "step": 179
1312
+ },
1313
+ {
1314
+ "epoch": 1.3968871595330739,
1315
+ "grad_norm": 0.5467899896948426,
1316
+ "learning_rate": 8.306010928961747e-07,
1317
+ "loss": 1.1669,
1318
+ "step": 180
1319
+ },
1320
+ {
1321
+ "epoch": 1.404669260700389,
1322
+ "grad_norm": 0.5788709158633432,
1323
+ "learning_rate": 8.278145695364237e-07,
1324
+ "loss": 1.1163,
1325
+ "step": 181
1326
+ },
1327
+ {
1328
+ "epoch": 1.4124513618677041,
1329
+ "grad_norm": 0.5711519564174999,
1330
+ "learning_rate": 8.249721293199554e-07,
1331
+ "loss": 1.1085,
1332
+ "step": 182
1333
+ },
1334
+ {
1335
+ "epoch": 1.4202334630350195,
1336
+ "grad_norm": 0.5648595416368181,
1337
+ "learning_rate": 8.220720720720721e-07,
1338
+ "loss": 1.0845,
1339
+ "step": 183
1340
+ },
1341
+ {
1342
+ "epoch": 1.4280155642023347,
1343
+ "grad_norm": 0.525420807610935,
1344
+ "learning_rate": 8.191126279863481e-07,
1345
+ "loss": 1.1037,
1346
+ "step": 184
1347
+ },
1348
+ {
1349
+ "epoch": 1.4357976653696498,
1350
+ "grad_norm": 0.6990978764976382,
1351
+ "learning_rate": 8.160919540229884e-07,
1352
+ "loss": 1.0672,
1353
+ "step": 185
1354
+ },
1355
+ {
1356
+ "epoch": 1.443579766536965,
1357
+ "grad_norm": 0.565722083537725,
1358
+ "learning_rate": 8.130081300813006e-07,
1359
+ "loss": 1.1214,
1360
+ "step": 186
1361
+ },
1362
+ {
1363
+ "epoch": 1.45136186770428,
1364
+ "grad_norm": 0.6283006226852244,
1365
+ "learning_rate": 8.098591549295774e-07,
1366
+ "loss": 1.0992,
1367
+ "step": 187
1368
+ },
1369
+ {
1370
+ "epoch": 1.4591439688715953,
1371
+ "grad_norm": 0.5550197869791744,
1372
+ "learning_rate": 8.066429418742585e-07,
1373
+ "loss": 1.1136,
1374
+ "step": 188
1375
+ },
1376
+ {
1377
+ "epoch": 1.4669260700389106,
1378
+ "grad_norm": 0.5659938483591545,
1379
+ "learning_rate": 8.03357314148681e-07,
1380
+ "loss": 1.0884,
1381
+ "step": 189
1382
+ },
1383
+ {
1384
+ "epoch": 1.4747081712062258,
1385
+ "grad_norm": 0.5872567653519262,
1386
+ "learning_rate": 8e-07,
1387
+ "loss": 1.1283,
1388
+ "step": 190
1389
+ },
1390
+ {
1391
+ "epoch": 1.482490272373541,
1392
+ "grad_norm": 0.6010641188429918,
1393
+ "learning_rate": 7.965686274509804e-07,
1394
+ "loss": 1.0959,
1395
+ "step": 191
1396
+ },
1397
+ {
1398
+ "epoch": 1.490272373540856,
1399
+ "grad_norm": 0.559777631958916,
1400
+ "learning_rate": 7.930607187112763e-07,
1401
+ "loss": 1.1047,
1402
+ "step": 192
1403
+ },
1404
+ {
1405
+ "epoch": 1.490272373540856,
1406
+ "eval_loss": 0.949611485004425,
1407
+ "eval_runtime": 193.3019,
1408
+ "eval_samples_per_second": 13.947,
1409
+ "eval_steps_per_second": 0.222,
1410
+ "step": 192
1411
+ },
1412
+ {
1413
+ "epoch": 1.4980544747081712,
1414
+ "grad_norm": 0.5608029800864122,
1415
+ "learning_rate": 7.894736842105262e-07,
1416
+ "loss": 1.0798,
1417
+ "step": 193
1418
+ },
1419
+ {
1420
+ "epoch": 1.5058365758754864,
1421
+ "grad_norm": 0.5765576143527928,
1422
+ "learning_rate": 7.85804816223067e-07,
1423
+ "loss": 1.0887,
1424
+ "step": 194
1425
+ },
1426
+ {
1427
+ "epoch": 1.5136186770428015,
1428
+ "grad_norm": 0.5783903086304087,
1429
+ "learning_rate": 7.82051282051282e-07,
1430
+ "loss": 1.1015,
1431
+ "step": 195
1432
+ },
1433
+ {
1434
+ "epoch": 1.5214007782101167,
1435
+ "grad_norm": 0.5252594928277367,
1436
+ "learning_rate": 7.782101167315173e-07,
1437
+ "loss": 1.0802,
1438
+ "step": 196
1439
+ },
1440
+ {
1441
+ "epoch": 1.5291828793774318,
1442
+ "grad_norm": 0.5435740719293682,
1443
+ "learning_rate": 7.742782152230972e-07,
1444
+ "loss": 1.0874,
1445
+ "step": 197
1446
+ },
1447
+ {
1448
+ "epoch": 1.536964980544747,
1449
+ "grad_norm": 0.5369255828107669,
1450
+ "learning_rate": 7.702523240371846e-07,
1451
+ "loss": 1.095,
1452
+ "step": 198
1453
+ },
1454
+ {
1455
+ "epoch": 1.544747081712062,
1456
+ "grad_norm": 0.6850191931654889,
1457
+ "learning_rate": 7.661290322580645e-07,
1458
+ "loss": 1.0912,
1459
+ "step": 199
1460
+ },
1461
+ {
1462
+ "epoch": 1.5525291828793775,
1463
+ "grad_norm": 0.5688231956299873,
1464
+ "learning_rate": 7.619047619047617e-07,
1465
+ "loss": 1.0552,
1466
+ "step": 200
1467
+ },
1468
+ {
1469
+ "epoch": 1.5603112840466926,
1470
+ "grad_norm": 0.5928997625619618,
1471
+ "learning_rate": 7.575757575757575e-07,
1472
+ "loss": 1.0621,
1473
+ "step": 201
1474
+ },
1475
+ {
1476
+ "epoch": 1.5680933852140078,
1477
+ "grad_norm": 0.7716125599428396,
1478
+ "learning_rate": 7.531380753138075e-07,
1479
+ "loss": 1.0758,
1480
+ "step": 202
1481
+ },
1482
+ {
1483
+ "epoch": 1.575875486381323,
1484
+ "grad_norm": 0.5629855625488145,
1485
+ "learning_rate": 7.48587570621469e-07,
1486
+ "loss": 1.0606,
1487
+ "step": 203
1488
+ },
1489
+ {
1490
+ "epoch": 1.5836575875486383,
1491
+ "grad_norm": 0.5635371246521241,
1492
+ "learning_rate": 7.439198855507868e-07,
1493
+ "loss": 1.1029,
1494
+ "step": 204
1495
+ },
1496
+ {
1497
+ "epoch": 1.5914396887159534,
1498
+ "grad_norm": 0.6498217881509594,
1499
+ "learning_rate": 7.391304347826086e-07,
1500
+ "loss": 1.0737,
1501
+ "step": 205
1502
+ },
1503
+ {
1504
+ "epoch": 1.5992217898832686,
1505
+ "grad_norm": 0.5578772825539674,
1506
+ "learning_rate": 7.342143906020557e-07,
1507
+ "loss": 1.0938,
1508
+ "step": 206
1509
+ },
1510
+ {
1511
+ "epoch": 1.6070038910505837,
1512
+ "grad_norm": 0.6895801155757268,
1513
+ "learning_rate": 7.291666666666667e-07,
1514
+ "loss": 1.108,
1515
+ "step": 207
1516
+ },
1517
+ {
1518
+ "epoch": 1.6147859922178989,
1519
+ "grad_norm": 0.5472071760512172,
1520
+ "learning_rate": 7.239819004524887e-07,
1521
+ "loss": 1.0799,
1522
+ "step": 208
1523
+ },
1524
+ {
1525
+ "epoch": 1.622568093385214,
1526
+ "grad_norm": 0.5551855311336273,
1527
+ "learning_rate": 7.186544342507645e-07,
1528
+ "loss": 1.1011,
1529
+ "step": 209
1530
+ },
1531
+ {
1532
+ "epoch": 1.6303501945525292,
1533
+ "grad_norm": 0.5922020624280246,
1534
+ "learning_rate": 7.131782945736434e-07,
1535
+ "loss": 1.0883,
1536
+ "step": 210
1537
+ },
1538
+ {
1539
+ "epoch": 1.6381322957198443,
1540
+ "grad_norm": 0.5855500909084618,
1541
+ "learning_rate": 7.075471698113208e-07,
1542
+ "loss": 1.0277,
1543
+ "step": 211
1544
+ },
1545
+ {
1546
+ "epoch": 1.6459143968871595,
1547
+ "grad_norm": 0.5436280978302126,
1548
+ "learning_rate": 7.017543859649121e-07,
1549
+ "loss": 1.0829,
1550
+ "step": 212
1551
+ },
1552
+ {
1553
+ "epoch": 1.6536964980544746,
1554
+ "grad_norm": 0.568803772556928,
1555
+ "learning_rate": 6.957928802588997e-07,
1556
+ "loss": 1.0778,
1557
+ "step": 213
1558
+ },
1559
+ {
1560
+ "epoch": 1.6614785992217898,
1561
+ "grad_norm": 0.5954446162983249,
1562
+ "learning_rate": 6.89655172413793e-07,
1563
+ "loss": 1.0885,
1564
+ "step": 214
1565
+ },
1566
+ {
1567
+ "epoch": 1.669260700389105,
1568
+ "grad_norm": 0.6318388963180205,
1569
+ "learning_rate": 6.833333333333333e-07,
1570
+ "loss": 1.0761,
1571
+ "step": 215
1572
+ },
1573
+ {
1574
+ "epoch": 1.6770428015564203,
1575
+ "grad_norm": 1.0424216403770146,
1576
+ "learning_rate": 6.768189509306259e-07,
1577
+ "loss": 1.0882,
1578
+ "step": 216
1579
+ },
1580
+ {
1581
+ "epoch": 1.6848249027237354,
1582
+ "grad_norm": 0.5848369566451875,
1583
+ "learning_rate": 6.701030927835052e-07,
1584
+ "loss": 1.0933,
1585
+ "step": 217
1586
+ },
1587
+ {
1588
+ "epoch": 1.6926070038910506,
1589
+ "grad_norm": 0.6540113258331898,
1590
+ "learning_rate": 6.63176265270506e-07,
1591
+ "loss": 1.0997,
1592
+ "step": 218
1593
+ },
1594
+ {
1595
+ "epoch": 1.7003891050583657,
1596
+ "grad_norm": 0.5945247307598128,
1597
+ "learning_rate": 6.560283687943263e-07,
1598
+ "loss": 1.1121,
1599
+ "step": 219
1600
+ },
1601
+ {
1602
+ "epoch": 1.708171206225681,
1603
+ "grad_norm": 0.5812576783107466,
1604
+ "learning_rate": 6.486486486486486e-07,
1605
+ "loss": 1.1089,
1606
+ "step": 220
1607
+ },
1608
+ {
1609
+ "epoch": 1.7159533073929962,
1610
+ "grad_norm": 0.5422687962188623,
1611
+ "learning_rate": 6.41025641025641e-07,
1612
+ "loss": 1.0689,
1613
+ "step": 221
1614
+ },
1615
+ {
1616
+ "epoch": 1.7237354085603114,
1617
+ "grad_norm": 0.5655113478395188,
1618
+ "learning_rate": 6.33147113594041e-07,
1619
+ "loss": 1.1083,
1620
+ "step": 222
1621
+ },
1622
+ {
1623
+ "epoch": 1.7315175097276265,
1624
+ "grad_norm": 0.5479997273898842,
1625
+ "learning_rate": 6.249999999999999e-07,
1626
+ "loss": 1.0495,
1627
+ "step": 223
1628
+ },
1629
+ {
1630
+ "epoch": 1.7392996108949417,
1631
+ "grad_norm": 0.6082276343842562,
1632
+ "learning_rate": 6.165703275529864e-07,
1633
+ "loss": 1.0847,
1634
+ "step": 224
1635
+ },
1636
+ {
1637
+ "epoch": 1.7392996108949417,
1638
+ "eval_loss": 0.9448344707489014,
1639
+ "eval_runtime": 193.8348,
1640
+ "eval_samples_per_second": 13.909,
1641
+ "eval_steps_per_second": 0.222,
1642
+ "step": 224
1643
+ },
1644
+ {
1645
+ "epoch": 1.7470817120622568,
1646
+ "grad_norm": 0.5456728126246078,
1647
+ "learning_rate": 6.078431372549019e-07,
1648
+ "loss": 1.0543,
1649
+ "step": 225
1650
+ },
1651
+ {
1652
+ "epoch": 1.754863813229572,
1653
+ "grad_norm": 0.8395615938618125,
1654
+ "learning_rate": 5.988023952095807e-07,
1655
+ "loss": 1.1133,
1656
+ "step": 226
1657
+ },
1658
+ {
1659
+ "epoch": 1.7626459143968871,
1660
+ "grad_norm": 0.5699433030581755,
1661
+ "learning_rate": 5.89430894308943e-07,
1662
+ "loss": 1.0878,
1663
+ "step": 227
1664
+ },
1665
+ {
1666
+ "epoch": 1.7704280155642023,
1667
+ "grad_norm": 0.5372956205386954,
1668
+ "learning_rate": 5.797101449275362e-07,
1669
+ "loss": 1.0749,
1670
+ "step": 228
1671
+ },
1672
+ {
1673
+ "epoch": 1.7782101167315174,
1674
+ "grad_norm": 0.5485335276065648,
1675
+ "learning_rate": 5.696202531645569e-07,
1676
+ "loss": 1.056,
1677
+ "step": 229
1678
+ },
1679
+ {
1680
+ "epoch": 1.7859922178988326,
1681
+ "grad_norm": 0.5739239028380502,
1682
+ "learning_rate": 5.591397849462365e-07,
1683
+ "loss": 1.1175,
1684
+ "step": 230
1685
+ },
1686
+ {
1687
+ "epoch": 1.7937743190661477,
1688
+ "grad_norm": 0.5674136555108149,
1689
+ "learning_rate": 5.482456140350876e-07,
1690
+ "loss": 1.1061,
1691
+ "step": 231
1692
+ },
1693
+ {
1694
+ "epoch": 1.801556420233463,
1695
+ "grad_norm": 0.6414824329123738,
1696
+ "learning_rate": 5.369127516778523e-07,
1697
+ "loss": 1.1502,
1698
+ "step": 232
1699
+ },
1700
+ {
1701
+ "epoch": 1.8093385214007782,
1702
+ "grad_norm": 0.5567692960590215,
1703
+ "learning_rate": 5.251141552511415e-07,
1704
+ "loss": 1.074,
1705
+ "step": 233
1706
+ },
1707
+ {
1708
+ "epoch": 1.8171206225680934,
1709
+ "grad_norm": 0.5203307823955872,
1710
+ "learning_rate": 5.128205128205127e-07,
1711
+ "loss": 1.0576,
1712
+ "step": 234
1713
+ },
1714
+ {
1715
+ "epoch": 1.8249027237354085,
1716
+ "grad_norm": 0.5480039922419879,
1717
+ "learning_rate": 5e-07,
1718
+ "loss": 1.0692,
1719
+ "step": 235
1720
+ },
1721
+ {
1722
+ "epoch": 1.8326848249027239,
1723
+ "grad_norm": 0.6602968949965664,
1724
+ "learning_rate": 4.8661800486618e-07,
1725
+ "loss": 1.0716,
1726
+ "step": 236
1727
+ },
1728
+ {
1729
+ "epoch": 1.840466926070039,
1730
+ "grad_norm": 0.5621257570421472,
1731
+ "learning_rate": 4.72636815920398e-07,
1732
+ "loss": 1.1222,
1733
+ "step": 237
1734
+ },
1735
+ {
1736
+ "epoch": 1.8482490272373542,
1737
+ "grad_norm": 0.5778853314353605,
1738
+ "learning_rate": 4.5801526717557246e-07,
1739
+ "loss": 1.0651,
1740
+ "step": 238
1741
+ },
1742
+ {
1743
+ "epoch": 1.8560311284046693,
1744
+ "grad_norm": 0.5284341924121511,
1745
+ "learning_rate": 4.4270833333333337e-07,
1746
+ "loss": 1.0708,
1747
+ "step": 239
1748
+ },
1749
+ {
1750
+ "epoch": 1.8638132295719845,
1751
+ "grad_norm": 0.5965113916130589,
1752
+ "learning_rate": 4.266666666666667e-07,
1753
+ "loss": 1.0981,
1754
+ "step": 240
1755
+ },
1756
+ {
1757
+ "epoch": 1.8715953307392996,
1758
+ "grad_norm": 0.5597034969209712,
1759
+ "learning_rate": 4.098360655737704e-07,
1760
+ "loss": 1.1179,
1761
+ "step": 241
1762
+ },
1763
+ {
1764
+ "epoch": 1.8793774319066148,
1765
+ "grad_norm": 0.5536241929831267,
1766
+ "learning_rate": 3.92156862745098e-07,
1767
+ "loss": 1.0935,
1768
+ "step": 242
1769
+ },
1770
+ {
1771
+ "epoch": 1.88715953307393,
1772
+ "grad_norm": 0.5584595178829068,
1773
+ "learning_rate": 3.7356321839080463e-07,
1774
+ "loss": 1.1041,
1775
+ "step": 243
1776
+ },
1777
+ {
1778
+ "epoch": 1.894941634241245,
1779
+ "grad_norm": 0.5427158370327506,
1780
+ "learning_rate": 3.539823008849558e-07,
1781
+ "loss": 1.0756,
1782
+ "step": 244
1783
+ },
1784
+ {
1785
+ "epoch": 1.9027237354085602,
1786
+ "grad_norm": 0.5336572931407305,
1787
+ "learning_rate": 3.333333333333333e-07,
1788
+ "loss": 1.0514,
1789
+ "step": 245
1790
+ },
1791
+ {
1792
+ "epoch": 1.9105058365758754,
1793
+ "grad_norm": 0.5369948018962315,
1794
+ "learning_rate": 3.1152647975077885e-07,
1795
+ "loss": 1.1033,
1796
+ "step": 246
1797
+ },
1798
+ {
1799
+ "epoch": 1.9182879377431905,
1800
+ "grad_norm": 0.5717043508183633,
1801
+ "learning_rate": 2.8846153846153846e-07,
1802
+ "loss": 1.0766,
1803
+ "step": 247
1804
+ },
1805
+ {
1806
+ "epoch": 1.9260700389105059,
1807
+ "grad_norm": 0.5722908601877614,
1808
+ "learning_rate": 2.64026402640264e-07,
1809
+ "loss": 1.13,
1810
+ "step": 248
1811
+ },
1812
+ {
1813
+ "epoch": 1.933852140077821,
1814
+ "grad_norm": 0.5419564077709611,
1815
+ "learning_rate": 2.3809523809523806e-07,
1816
+ "loss": 1.0729,
1817
+ "step": 249
1818
+ },
1819
+ {
1820
+ "epoch": 1.9416342412451362,
1821
+ "grad_norm": 0.5886685089221407,
1822
+ "learning_rate": 2.1052631578947366e-07,
1823
+ "loss": 1.1071,
1824
+ "step": 250
1825
+ },
1826
+ {
1827
+ "epoch": 1.9494163424124513,
1828
+ "grad_norm": 0.5403936814687964,
1829
+ "learning_rate": 1.8115942028985505e-07,
1830
+ "loss": 1.0983,
1831
+ "step": 251
1832
+ },
1833
+ {
1834
+ "epoch": 1.9571984435797667,
1835
+ "grad_norm": 0.528388946685679,
1836
+ "learning_rate": 1.4981273408239696e-07,
1837
+ "loss": 1.1008,
1838
+ "step": 252
1839
+ },
1840
+ {
1841
+ "epoch": 1.9649805447470818,
1842
+ "grad_norm": 0.5354770235277335,
1843
+ "learning_rate": 1.1627906976744186e-07,
1844
+ "loss": 1.1341,
1845
+ "step": 253
1846
+ },
1847
+ {
1848
+ "epoch": 1.972762645914397,
1849
+ "grad_norm": 0.5252514289927873,
1850
+ "learning_rate": 8.032128514056224e-08,
1851
+ "loss": 1.0818,
1852
+ "step": 254
1853
+ },
1854
+ {
1855
+ "epoch": 1.9805447470817121,
1856
+ "grad_norm": 0.5858247065123809,
1857
+ "learning_rate": 4.166666666666666e-08,
1858
+ "loss": 1.0804,
1859
+ "step": 255
1860
+ },
1861
+ {
1862
+ "epoch": 1.9883268482490273,
1863
+ "grad_norm": 0.5257477128880242,
1864
+ "learning_rate": 0,
1865
+ "loss": 1.0901,
1866
+ "step": 256
1867
+ },
1868
+ {
1869
+ "epoch": 1.9883268482490273,
1870
+ "eval_loss": 0.9421924352645874,
1871
+ "eval_runtime": 193.7536,
1872
+ "eval_samples_per_second": 13.915,
1873
+ "eval_steps_per_second": 0.222,
1874
+ "step": 256
1875
+ }
1876
+ ],
1877
+ "logging_steps": 1,
1878
+ "max_steps": 256,
1879
+ "num_input_tokens_seen": 0,
1880
+ "num_train_epochs": 2,
1881
+ "save_steps": 64,
1882
+ "stateful_callbacks": {
1883
+ "TrainerControl": {
1884
+ "args": {
1885
+ "should_epoch_stop": false,
1886
+ "should_evaluate": false,
1887
+ "should_log": false,
1888
+ "should_save": true,
1889
+ "should_training_stop": true
1890
+ },
1891
+ "attributes": {}
1892
+ }
1893
+ },
1894
+ "total_flos": 7.554618867587219e+18,
1895
+ "train_batch_size": 8,
1896
+ "trial_name": null,
1897
+ "trial_params": null
1898
+ }
training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:17496cd4d83504a4a3783884f366a4a092e8bd034b660325033ae7304af0dede
3
+ size 8977
zero_to_fp32.py ADDED
@@ -0,0 +1,674 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env python
2
+
3
+ # Copyright (c) Microsoft Corporation.
4
+ # SPDX-License-Identifier: Apache-2.0
5
+
6
+ # DeepSpeed Team
7
+
8
+ # This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
9
+ # copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
10
+ # the future. Once extracted, the weights don't require DeepSpeed and can be used in any
11
+ # application.
12
+ #
13
+ # example:
14
+ # python zero_to_fp32.py . output_dir/
15
+ # or
16
+ # python zero_to_fp32.py . output_dir/ --safe_serialization
17
+
18
+ import argparse
19
+ import torch
20
+ import glob
21
+ import math
22
+ import os
23
+ import re
24
+ import json
25
+ from tqdm import tqdm
26
+ from collections import OrderedDict
27
+ from dataclasses import dataclass
28
+
29
+ # while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
30
+ # DeepSpeed data structures it has to be available in the current python environment.
31
+ from deepspeed.utils import logger
32
+ from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
33
+ FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
34
+ FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
35
+
36
+
37
+ @dataclass
38
+ class zero_model_state:
39
+ buffers: dict()
40
+ param_shapes: dict()
41
+ shared_params: list
42
+ ds_version: int
43
+ frozen_param_shapes: dict()
44
+ frozen_param_fragments: dict()
45
+
46
+
47
+ debug = 0
48
+
49
+ # load to cpu
50
+ device = torch.device('cpu')
51
+
52
+
53
+ def atoi(text):
54
+ return int(text) if text.isdigit() else text
55
+
56
+
57
+ def natural_keys(text):
58
+ '''
59
+ alist.sort(key=natural_keys) sorts in human order
60
+ http://nedbatchelder.com/blog/200712/human_sorting.html
61
+ (See Toothy's implementation in the comments)
62
+ '''
63
+ return [atoi(c) for c in re.split(r'(\d+)', text)]
64
+
65
+
66
+ def get_model_state_file(checkpoint_dir, zero_stage):
67
+ if not os.path.isdir(checkpoint_dir):
68
+ raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
69
+
70
+ # there should be only one file
71
+ if zero_stage <= 2:
72
+ file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
73
+ elif zero_stage == 3:
74
+ file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
75
+
76
+ if not os.path.exists(file):
77
+ raise FileNotFoundError(f"can't find model states file at '{file}'")
78
+
79
+ return file
80
+
81
+
82
+ def get_checkpoint_files(checkpoint_dir, glob_pattern):
83
+ # XXX: need to test that this simple glob rule works for multi-node setup too
84
+ ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
85
+
86
+ if len(ckpt_files) == 0:
87
+ raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
88
+
89
+ return ckpt_files
90
+
91
+
92
+ def get_optim_files(checkpoint_dir):
93
+ return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
94
+
95
+
96
+ def get_model_state_files(checkpoint_dir):
97
+ return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
98
+
99
+
100
+ def parse_model_states(files):
101
+ zero_model_states = []
102
+ for file in files:
103
+ state_dict = torch.load(file, map_location=device)
104
+
105
+ if BUFFER_NAMES not in state_dict:
106
+ raise ValueError(f"{file} is not a model state checkpoint")
107
+ buffer_names = state_dict[BUFFER_NAMES]
108
+ if debug:
109
+ print("Found buffers:", buffer_names)
110
+
111
+ # recover just the buffers while restoring them to fp32 if they were saved in fp16
112
+ buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
113
+ param_shapes = state_dict[PARAM_SHAPES]
114
+
115
+ # collect parameters that are included in param_shapes
116
+ param_names = []
117
+ for s in param_shapes:
118
+ for name in s.keys():
119
+ param_names.append(name)
120
+
121
+ # update with frozen parameters
122
+ frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
123
+ if frozen_param_shapes is not None:
124
+ if debug:
125
+ print(f"Found frozen_param_shapes: {frozen_param_shapes}")
126
+ param_names += list(frozen_param_shapes.keys())
127
+
128
+ # handle shared params
129
+ shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
130
+
131
+ ds_version = state_dict.get(DS_VERSION, None)
132
+
133
+ frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
134
+
135
+ z_model_state = zero_model_state(buffers=buffers,
136
+ param_shapes=param_shapes,
137
+ shared_params=shared_params,
138
+ ds_version=ds_version,
139
+ frozen_param_shapes=frozen_param_shapes,
140
+ frozen_param_fragments=frozen_param_fragments)
141
+ zero_model_states.append(z_model_state)
142
+
143
+ return zero_model_states
144
+
145
+
146
+ def parse_optim_states(files, ds_checkpoint_dir):
147
+ total_files = len(files)
148
+ state_dicts = []
149
+ for f in files:
150
+ state_dict = torch.load(f, map_location=device)
151
+ # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
152
+ # and also handle the case where it was already removed by another helper script
153
+ state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
154
+ state_dicts.append(state_dict)
155
+
156
+ if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
157
+ raise ValueError(f"{files[0]} is not a zero checkpoint")
158
+ zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
159
+ world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
160
+
161
+ # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
162
+ # parameters can be different from data parallelism for non-expert parameters. So we can just
163
+ # use the max of the partition_count to get the dp world_size.
164
+
165
+ if type(world_size) is list:
166
+ world_size = max(world_size)
167
+
168
+ if world_size != total_files:
169
+ raise ValueError(
170
+ f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
171
+ "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
172
+ )
173
+
174
+ # the groups are named differently in each stage
175
+ if zero_stage <= 2:
176
+ fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
177
+ elif zero_stage == 3:
178
+ fp32_groups_key = FP32_FLAT_GROUPS
179
+ else:
180
+ raise ValueError(f"unknown zero stage {zero_stage}")
181
+
182
+ if zero_stage <= 2:
183
+ fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
184
+ elif zero_stage == 3:
185
+ # if there is more than one param group, there will be multiple flattened tensors - one
186
+ # flattened tensor per group - for simplicity merge them into a single tensor
187
+ #
188
+ # XXX: could make the script more memory efficient for when there are multiple groups - it
189
+ # will require matching the sub-lists of param_shapes for each param group flattened tensor
190
+
191
+ fp32_flat_groups = [
192
+ torch.cat(state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key], 0) for i in range(len(state_dicts))
193
+ ]
194
+
195
+ return zero_stage, world_size, fp32_flat_groups
196
+
197
+
198
+ def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters):
199
+ """
200
+ Returns fp32 state_dict reconstructed from ds checkpoint
201
+
202
+ Args:
203
+ - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
204
+
205
+ """
206
+ print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
207
+
208
+ optim_files = get_optim_files(ds_checkpoint_dir)
209
+ zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
210
+ print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
211
+
212
+ model_files = get_model_state_files(ds_checkpoint_dir)
213
+
214
+ zero_model_states = parse_model_states(model_files)
215
+ print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
216
+
217
+ if zero_stage <= 2:
218
+ return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
219
+ exclude_frozen_parameters)
220
+ elif zero_stage == 3:
221
+ return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
222
+ exclude_frozen_parameters)
223
+
224
+
225
+ def _zero2_merge_frozen_params(state_dict, zero_model_states):
226
+ if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
227
+ return
228
+
229
+ frozen_param_shapes = zero_model_states[0].frozen_param_shapes
230
+ frozen_param_fragments = zero_model_states[0].frozen_param_fragments
231
+
232
+ if debug:
233
+ num_elem = sum(s.numel() for s in frozen_param_shapes.values())
234
+ print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
235
+
236
+ wanted_params = len(frozen_param_shapes)
237
+ wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
238
+ avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
239
+ print(f'Frozen params: Have {avail_numel} numels to process.')
240
+ print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
241
+
242
+ total_params = 0
243
+ total_numel = 0
244
+ for name, shape in frozen_param_shapes.items():
245
+ total_params += 1
246
+ unpartitioned_numel = shape.numel()
247
+ total_numel += unpartitioned_numel
248
+
249
+ state_dict[name] = frozen_param_fragments[name]
250
+
251
+ if debug:
252
+ print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
253
+
254
+ print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
255
+
256
+
257
+ def _has_callable(obj, fn):
258
+ attr = getattr(obj, fn, None)
259
+ return callable(attr)
260
+
261
+
262
+ def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
263
+ param_shapes = zero_model_states[0].param_shapes
264
+
265
+ # Reconstruction protocol:
266
+ #
267
+ # XXX: document this
268
+
269
+ if debug:
270
+ for i in range(world_size):
271
+ for j in range(len(fp32_flat_groups[0])):
272
+ print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
273
+
274
+ # XXX: memory usage doubles here (zero2)
275
+ num_param_groups = len(fp32_flat_groups[0])
276
+ merged_single_partition_of_fp32_groups = []
277
+ for i in range(num_param_groups):
278
+ merged_partitions = [sd[i] for sd in fp32_flat_groups]
279
+ full_single_fp32_vector = torch.cat(merged_partitions, 0)
280
+ merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
281
+ avail_numel = sum(
282
+ [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
283
+
284
+ if debug:
285
+ wanted_params = sum([len(shapes) for shapes in param_shapes])
286
+ wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
287
+ # not asserting if there is a mismatch due to possible padding
288
+ print(f"Have {avail_numel} numels to process.")
289
+ print(f"Need {wanted_numel} numels in {wanted_params} params.")
290
+
291
+ # params
292
+ # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
293
+ # out-of-core computing solution
294
+ total_numel = 0
295
+ total_params = 0
296
+ for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
297
+ offset = 0
298
+ avail_numel = full_single_fp32_vector.numel()
299
+ for name, shape in shapes.items():
300
+
301
+ unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
302
+ total_numel += unpartitioned_numel
303
+ total_params += 1
304
+
305
+ if debug:
306
+ print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
307
+ state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
308
+ offset += unpartitioned_numel
309
+
310
+ # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
311
+ # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
312
+ # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
313
+ # live optimizer object, so we are checking that the numbers are within the right range
314
+ align_to = 2 * world_size
315
+
316
+ def zero2_align(x):
317
+ return align_to * math.ceil(x / align_to)
318
+
319
+ if debug:
320
+ print(f"original offset={offset}, avail_numel={avail_numel}")
321
+
322
+ offset = zero2_align(offset)
323
+ avail_numel = zero2_align(avail_numel)
324
+
325
+ if debug:
326
+ print(f"aligned offset={offset}, avail_numel={avail_numel}")
327
+
328
+ # Sanity check
329
+ if offset != avail_numel:
330
+ raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
331
+
332
+ print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
333
+
334
+
335
+ def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
336
+ exclude_frozen_parameters):
337
+ state_dict = OrderedDict()
338
+
339
+ # buffers
340
+ buffers = zero_model_states[0].buffers
341
+ state_dict.update(buffers)
342
+ if debug:
343
+ print(f"added {len(buffers)} buffers")
344
+
345
+ if not exclude_frozen_parameters:
346
+ _zero2_merge_frozen_params(state_dict, zero_model_states)
347
+
348
+ _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
349
+
350
+ # recover shared parameters
351
+ for pair in zero_model_states[0].shared_params:
352
+ if pair[1] in state_dict:
353
+ state_dict[pair[0]] = state_dict[pair[1]]
354
+
355
+ return state_dict
356
+
357
+
358
+ def zero3_partitioned_param_info(unpartitioned_numel, world_size):
359
+ remainder = unpartitioned_numel % world_size
360
+ padding_numel = (world_size - remainder) if remainder else 0
361
+ partitioned_numel = math.ceil(unpartitioned_numel / world_size)
362
+ return partitioned_numel, padding_numel
363
+
364
+
365
+ def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
366
+ if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
367
+ return
368
+
369
+ if debug:
370
+ for i in range(world_size):
371
+ num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
372
+ print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
373
+
374
+ frozen_param_shapes = zero_model_states[0].frozen_param_shapes
375
+ wanted_params = len(frozen_param_shapes)
376
+ wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
377
+ avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
378
+ print(f'Frozen params: Have {avail_numel} numels to process.')
379
+ print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
380
+
381
+ total_params = 0
382
+ total_numel = 0
383
+ for name, shape in zero_model_states[0].frozen_param_shapes.items():
384
+ total_params += 1
385
+ unpartitioned_numel = shape.numel()
386
+ total_numel += unpartitioned_numel
387
+
388
+ param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
389
+ state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
390
+
391
+ partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
392
+
393
+ if debug:
394
+ print(
395
+ f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
396
+ )
397
+
398
+ print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
399
+
400
+
401
+ def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
402
+ param_shapes = zero_model_states[0].param_shapes
403
+ avail_numel = fp32_flat_groups[0].numel() * world_size
404
+ # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
405
+ # param, re-consolidating each param, while dealing with padding if any
406
+
407
+ # merge list of dicts, preserving order
408
+ param_shapes = {k: v for d in param_shapes for k, v in d.items()}
409
+
410
+ if debug:
411
+ for i in range(world_size):
412
+ print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
413
+
414
+ wanted_params = len(param_shapes)
415
+ wanted_numel = sum(shape.numel() for shape in param_shapes.values())
416
+ # not asserting if there is a mismatch due to possible padding
417
+ avail_numel = fp32_flat_groups[0].numel() * world_size
418
+ print(f"Trainable params: Have {avail_numel} numels to process.")
419
+ print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
420
+
421
+ # params
422
+ # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
423
+ # out-of-core computing solution
424
+ offset = 0
425
+ total_numel = 0
426
+ total_params = 0
427
+ for name, shape in tqdm(param_shapes.items(), desc='Gathering Sharded Weights'):
428
+ unpartitioned_numel = shape.numel()
429
+ total_numel += unpartitioned_numel
430
+ total_params += 1
431
+ partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
432
+
433
+ if debug:
434
+ print(
435
+ f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
436
+ )
437
+
438
+ # XXX: memory usage doubles here
439
+ state_dict[name] = torch.cat(
440
+ tuple(fp32_flat_groups[i].narrow(0, offset, partitioned_numel) for i in range(world_size)),
441
+ 0).narrow(0, 0, unpartitioned_numel).view(shape)
442
+ offset += partitioned_numel
443
+
444
+ offset *= world_size
445
+
446
+ # Sanity check
447
+ if offset != avail_numel:
448
+ raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
449
+
450
+ print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
451
+
452
+
453
+ def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
454
+ exclude_frozen_parameters):
455
+ state_dict = OrderedDict()
456
+
457
+ # buffers
458
+ buffers = zero_model_states[0].buffers
459
+ state_dict.update(buffers)
460
+ if debug:
461
+ print(f"added {len(buffers)} buffers")
462
+
463
+ if not exclude_frozen_parameters:
464
+ _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
465
+
466
+ _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
467
+
468
+ # recover shared parameters
469
+ for pair in zero_model_states[0].shared_params:
470
+ if pair[1] in state_dict:
471
+ state_dict[pair[0]] = state_dict[pair[1]]
472
+
473
+ return state_dict
474
+
475
+
476
+ def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None, exclude_frozen_parameters=False):
477
+ """
478
+ Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
479
+ ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
480
+ via a model hub.
481
+
482
+ Args:
483
+ - ``checkpoint_dir``: path to the desired checkpoint folder
484
+ - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
485
+ - ``exclude_frozen_parameters``: exclude frozen parameters
486
+
487
+ Returns:
488
+ - pytorch ``state_dict``
489
+
490
+ Note: this approach may not work if your application doesn't have sufficient free CPU memory and
491
+ you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
492
+ the checkpoint.
493
+
494
+ A typical usage might be ::
495
+
496
+ from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
497
+ # do the training and checkpoint saving
498
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
499
+ model = model.cpu() # move to cpu
500
+ model.load_state_dict(state_dict)
501
+ # submit to model hub or save the model to share with others
502
+
503
+ In this example the ``model`` will no longer be usable in the deepspeed context of the same
504
+ application. i.e. you will need to re-initialize the deepspeed engine, since
505
+ ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
506
+
507
+ If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
508
+
509
+ """
510
+ if tag is None:
511
+ latest_path = os.path.join(checkpoint_dir, 'latest')
512
+ if os.path.isfile(latest_path):
513
+ with open(latest_path, 'r') as fd:
514
+ tag = fd.read().strip()
515
+ else:
516
+ raise ValueError(f"Unable to find 'latest' file at {latest_path}")
517
+
518
+ ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
519
+
520
+ if not os.path.isdir(ds_checkpoint_dir):
521
+ raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
522
+
523
+ return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
524
+
525
+
526
+ def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir,
527
+ output_dir,
528
+ max_shard_size="5GB",
529
+ safe_serialization=False,
530
+ tag=None,
531
+ exclude_frozen_parameters=False):
532
+ """
533
+ Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
534
+ loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
535
+
536
+ Args:
537
+ - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
538
+ - ``output_dir``: directory to the pytorch fp32 state_dict output files
539
+ - ``max_shard_size``: the maximum size for a checkpoint before being sharded, default value is 5GB
540
+ - ``safe_serialization``: whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).
541
+ - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
542
+ - ``exclude_frozen_parameters``: exclude frozen parameters
543
+ """
544
+ # Dependency pre-check
545
+ if safe_serialization:
546
+ try:
547
+ from safetensors.torch import save_file
548
+ except ImportError:
549
+ print('If you want to use `safe_serialization`, please `pip install safetensors`')
550
+ raise
551
+ if max_shard_size is not None:
552
+ try:
553
+ from huggingface_hub import split_torch_state_dict_into_shards
554
+ except ImportError:
555
+ print('If you want to use `max_shard_size`, please `pip install huggingface_hub`')
556
+ raise
557
+
558
+ # Convert zero checkpoint to state_dict
559
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag, exclude_frozen_parameters)
560
+
561
+ # Shard the model if it is too big.
562
+ weights_name = "model.safetensors" if safe_serialization else "pytorch_model.bin"
563
+ if max_shard_size is not None:
564
+ filename_pattern = weights_name.replace(".bin", "{suffix}.bin").replace(".safetensors", "{suffix}.safetensors")
565
+ state_dict_split = split_torch_state_dict_into_shards(state_dict,
566
+ filename_pattern=filename_pattern,
567
+ max_shard_size=max_shard_size)
568
+ else:
569
+ from collections import namedtuple
570
+ StateDictSplit = namedtuple("StateDictSplit", ["is_sharded", "filename_to_tensors"])
571
+ state_dict_split = StateDictSplit(is_sharded=False,
572
+ filename_to_tensors={weights_name: list(state_dict.keys())})
573
+
574
+ # Save the model
575
+ filename_to_tensors = state_dict_split.filename_to_tensors.items()
576
+ for shard_file, tensors in tqdm(filename_to_tensors, desc="Saving checkpoint shards"):
577
+ shard = {tensor: state_dict[tensor].contiguous() for tensor in tensors}
578
+ output_path = os.path.join(output_dir, shard_file)
579
+ if safe_serialization:
580
+ save_file(shard, output_path, metadata={"format": "pt"})
581
+ else:
582
+ torch.save(shard, output_path)
583
+
584
+ # Save index if sharded
585
+ if state_dict_split.is_sharded:
586
+ index = {
587
+ "metadata": state_dict_split.metadata,
588
+ "weight_map": state_dict_split.tensor_to_filename,
589
+ }
590
+ save_index_file = "model.safetensors.index.json" if safe_serialization else "pytorch_model.bin.index.json"
591
+ save_index_file = os.path.join(output_dir, save_index_file)
592
+ with open(save_index_file, "w", encoding="utf-8") as f:
593
+ content = json.dumps(index, indent=2, sort_keys=True) + "\n"
594
+ f.write(content)
595
+
596
+
597
+ def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
598
+ """
599
+ 1. Put the provided model to cpu
600
+ 2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
601
+ 3. Load it into the provided model
602
+
603
+ Args:
604
+ - ``model``: the model object to update
605
+ - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
606
+ - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
607
+
608
+ Returns:
609
+ - ``model`: modified model
610
+
611
+ Make sure you have plenty of CPU memory available before you call this function. If you don't
612
+ have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
613
+ conveniently placed for you in the checkpoint folder.
614
+
615
+ A typical usage might be ::
616
+
617
+ from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
618
+ model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
619
+ # submit to model hub or save the model to share with others
620
+
621
+ Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
622
+ of the same application. i.e. you will need to re-initialize the deepspeed engine, since
623
+ ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
624
+
625
+ """
626
+ logger.info(f"Extracting fp32 weights")
627
+ state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
628
+
629
+ logger.info(f"Overwriting model with fp32 weights")
630
+ model = model.cpu()
631
+ model.load_state_dict(state_dict, strict=False)
632
+
633
+ return model
634
+
635
+
636
+ if __name__ == "__main__":
637
+ parser = argparse.ArgumentParser()
638
+ parser.add_argument("checkpoint_dir",
639
+ type=str,
640
+ help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
641
+ parser.add_argument("output_dir",
642
+ type=str,
643
+ help="directory to the pytorch fp32 state_dict output files"
644
+ "(e.g. path/checkpoint-12-output/)")
645
+ parser.add_argument(
646
+ "--max_shard_size",
647
+ type=str,
648
+ default="5GB",
649
+ help="The maximum size for a checkpoint before being sharded. Checkpoints shard will then be each of size"
650
+ "lower than this size. If expressed as a string, needs to be digits followed by a unit (like `5MB`"
651
+ "We default it to 5GB in order for models to be able to run easily on free-tier google colab instances"
652
+ "without CPU OOM issues.")
653
+ parser.add_argument(
654
+ "--safe_serialization",
655
+ default=False,
656
+ action='store_true',
657
+ help="Whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).")
658
+ parser.add_argument("-t",
659
+ "--tag",
660
+ type=str,
661
+ default=None,
662
+ help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
663
+ parser.add_argument("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
664
+ parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
665
+ args = parser.parse_args()
666
+
667
+ debug = args.debug
668
+
669
+ convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir,
670
+ args.output_dir,
671
+ max_shard_size=args.max_shard_size,
672
+ safe_serialization=args.safe_serialization,
673
+ tag=args.tag,
674
+ exclude_frozen_parameters=args.exclude_frozen_parameters)