sirui6011 commited on
Commit
90e7e22
·
verified ·
1 Parent(s): fe1c7a3

add checkpoints/codi_frozen_1.5b_logit

Browse files
Files changed (30) hide show
  1. .gitattributes +3 -0
  2. checkpoints/codi_frozen_1.5b_logit/checkpoint-1000/added_tokens.json +35 -0
  3. checkpoints/codi_frozen_1.5b_logit/checkpoint-1000/chat_template.jinja +54 -0
  4. checkpoints/codi_frozen_1.5b_logit/checkpoint-1000/config.json +58 -0
  5. checkpoints/codi_frozen_1.5b_logit/checkpoint-1000/merges.txt +0 -0
  6. checkpoints/codi_frozen_1.5b_logit/checkpoint-1000/pytorch_model.bin +3 -0
  7. checkpoints/codi_frozen_1.5b_logit/checkpoint-1000/special_tokens_map.json +31 -0
  8. checkpoints/codi_frozen_1.5b_logit/checkpoint-1000/thought_projector.pt +3 -0
  9. checkpoints/codi_frozen_1.5b_logit/checkpoint-1000/tokenizer.json +3 -0
  10. checkpoints/codi_frozen_1.5b_logit/checkpoint-1000/tokenizer_config.json +295 -0
  11. checkpoints/codi_frozen_1.5b_logit/checkpoint-1000/trainer_state.json +2034 -0
  12. checkpoints/codi_frozen_1.5b_logit/checkpoint-1000/vocab.json +0 -0
  13. checkpoints/codi_frozen_1.5b_logit/checkpoint-500/added_tokens.json +35 -0
  14. checkpoints/codi_frozen_1.5b_logit/checkpoint-500/chat_template.jinja +54 -0
  15. checkpoints/codi_frozen_1.5b_logit/checkpoint-500/config.json +58 -0
  16. checkpoints/codi_frozen_1.5b_logit/checkpoint-500/merges.txt +0 -0
  17. checkpoints/codi_frozen_1.5b_logit/checkpoint-500/pytorch_model.bin +3 -0
  18. checkpoints/codi_frozen_1.5b_logit/checkpoint-500/special_tokens_map.json +31 -0
  19. checkpoints/codi_frozen_1.5b_logit/checkpoint-500/thought_projector.pt +3 -0
  20. checkpoints/codi_frozen_1.5b_logit/checkpoint-500/tokenizer.json +3 -0
  21. checkpoints/codi_frozen_1.5b_logit/checkpoint-500/tokenizer_config.json +295 -0
  22. checkpoints/codi_frozen_1.5b_logit/checkpoint-500/trainer_state.json +1034 -0
  23. checkpoints/codi_frozen_1.5b_logit/checkpoint-500/vocab.json +0 -0
  24. checkpoints/codi_frozen_1.5b_logit/wandb/debug-internal.log +11 -0
  25. checkpoints/codi_frozen_1.5b_logit/wandb/debug.log +20 -0
  26. checkpoints/codi_frozen_1.5b_logit/wandb/offline-run-20260624_215746-xw4jltek/files/requirements.txt +229 -0
  27. checkpoints/codi_frozen_1.5b_logit/wandb/offline-run-20260624_215746-xw4jltek/logs/debug-core.log +15 -0
  28. checkpoints/codi_frozen_1.5b_logit/wandb/offline-run-20260624_215746-xw4jltek/logs/debug-internal.log +11 -0
  29. checkpoints/codi_frozen_1.5b_logit/wandb/offline-run-20260624_215746-xw4jltek/logs/debug.log +20 -0
  30. checkpoints/codi_frozen_1.5b_logit/wandb/offline-run-20260624_215746-xw4jltek/run-xw4jltek.wandb +3 -0
.gitattributes CHANGED
@@ -94,3 +94,6 @@ checkpoints/codi_frozen_1.5b_hidden/checkpoint-1000/tokenizer.json filter=lfs di
94
  checkpoints/codi_frozen_1.5b_hidden/checkpoint-1500/tokenizer.json filter=lfs diff=lfs merge=lfs -text
95
  checkpoints/codi_frozen_1.5b_hidden/checkpoint-500/tokenizer.json filter=lfs diff=lfs merge=lfs -text
96
  checkpoints/codi_frozen_1.5b_hidden/wandb/offline-run-20260624_182247-a09es240/run-a09es240.wandb filter=lfs diff=lfs merge=lfs -text
 
 
 
 
94
  checkpoints/codi_frozen_1.5b_hidden/checkpoint-1500/tokenizer.json filter=lfs diff=lfs merge=lfs -text
95
  checkpoints/codi_frozen_1.5b_hidden/checkpoint-500/tokenizer.json filter=lfs diff=lfs merge=lfs -text
96
  checkpoints/codi_frozen_1.5b_hidden/wandb/offline-run-20260624_182247-a09es240/run-a09es240.wandb filter=lfs diff=lfs merge=lfs -text
97
+ checkpoints/codi_frozen_1.5b_logit/checkpoint-1000/tokenizer.json filter=lfs diff=lfs merge=lfs -text
98
+ checkpoints/codi_frozen_1.5b_logit/checkpoint-500/tokenizer.json filter=lfs diff=lfs merge=lfs -text
99
+ checkpoints/codi_frozen_1.5b_logit/wandb/offline-run-20260624_215746-xw4jltek/run-xw4jltek.wandb filter=lfs diff=lfs merge=lfs -text
checkpoints/codi_frozen_1.5b_logit/checkpoint-1000/added_tokens.json ADDED
@@ -0,0 +1,35 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "</tool_call>": 151658,
3
+ "<tool_call>": 151657,
4
+ "<|action_sep|>": 151670,
5
+ "<|arg_sep|>": 151671,
6
+ "<|box_end|>": 151649,
7
+ "<|box_start|>": 151648,
8
+ "<|call_sep|>": 151666,
9
+ "<|end_of_text|>": 151673,
10
+ "<|endoftext|>": 151643,
11
+ "<|exception_sep|>": 151669,
12
+ "<|file_sep|>": 151664,
13
+ "<|fim_middle|>": 151660,
14
+ "<|fim_pad|>": 151662,
15
+ "<|fim_prefix|>": 151659,
16
+ "<|fim_suffix|>": 151661,
17
+ "<|frame_sep|>": 151672,
18
+ "<|im_end|>": 151645,
19
+ "<|im_start|>": 151644,
20
+ "<|image_pad|>": 151655,
21
+ "<|latent_end|>": 151675,
22
+ "<|latent_start|>": 151674,
23
+ "<|line_sep|>": 151667,
24
+ "<|object_ref_end|>": 151647,
25
+ "<|object_ref_start|>": 151646,
26
+ "<|quad_end|>": 151651,
27
+ "<|quad_start|>": 151650,
28
+ "<|repo_name|>": 151663,
29
+ "<|return_sep|>": 151668,
30
+ "<|trace_context_start|>": 151665,
31
+ "<|video_pad|>": 151656,
32
+ "<|vision_end|>": 151653,
33
+ "<|vision_pad|>": 151654,
34
+ "<|vision_start|>": 151652
35
+ }
checkpoints/codi_frozen_1.5b_logit/checkpoint-1000/chat_template.jinja ADDED
@@ -0,0 +1,54 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {%- if tools %}
2
+ {{- '<|im_start|>system\n' }}
3
+ {%- if messages[0]['role'] == 'system' %}
4
+ {{- messages[0]['content'] }}
5
+ {%- else %}
6
+ {{- 'You are Qwen, created by Alibaba Cloud. You are a helpful assistant.' }}
7
+ {%- endif %}
8
+ {{- "\n\n# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
9
+ {%- for tool in tools %}
10
+ {{- "\n" }}
11
+ {{- tool | tojson }}
12
+ {%- endfor %}
13
+ {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
14
+ {%- else %}
15
+ {%- if messages[0]['role'] == 'system' %}
16
+ {{- '<|im_start|>system\n' + messages[0]['content'] + '<|im_end|>\n' }}
17
+ {%- else %}
18
+ {{- '<|im_start|>system\nYou are Qwen, created by Alibaba Cloud. You are a helpful assistant.<|im_end|>\n' }}
19
+ {%- endif %}
20
+ {%- endif %}
21
+ {%- for message in messages %}
22
+ {%- if (message.role == "user") or (message.role == "system" and not loop.first) or (message.role == "assistant" and not message.tool_calls) %}
23
+ {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
24
+ {%- elif message.role == "assistant" %}
25
+ {{- '<|im_start|>' + message.role }}
26
+ {%- if message.content %}
27
+ {{- '\n' + message.content }}
28
+ {%- endif %}
29
+ {%- for tool_call in message.tool_calls %}
30
+ {%- if tool_call.function is defined %}
31
+ {%- set tool_call = tool_call.function %}
32
+ {%- endif %}
33
+ {{- '\n<tool_call>\n{"name": "' }}
34
+ {{- tool_call.name }}
35
+ {{- '", "arguments": ' }}
36
+ {{- tool_call.arguments | tojson }}
37
+ {{- '}\n</tool_call>' }}
38
+ {%- endfor %}
39
+ {{- '<|im_end|>\n' }}
40
+ {%- elif message.role == "tool" %}
41
+ {%- if (loop.index0 == 0) or (messages[loop.index0 - 1].role != "tool") %}
42
+ {{- '<|im_start|>user' }}
43
+ {%- endif %}
44
+ {{- '\n<tool_response>\n' }}
45
+ {{- message.content }}
46
+ {{- '\n</tool_response>' }}
47
+ {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
48
+ {{- '<|im_end|>\n' }}
49
+ {%- endif %}
50
+ {%- endif %}
51
+ {%- endfor %}
52
+ {%- if add_generation_prompt %}
53
+ {{- '<|im_start|>assistant\n' }}
54
+ {%- endif %}
checkpoints/codi_frozen_1.5b_logit/checkpoint-1000/config.json ADDED
@@ -0,0 +1,58 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Qwen2ForCausalLM"
4
+ ],
5
+ "attention_dropout": 0.0,
6
+ "dtype": "bfloat16",
7
+ "eos_token_id": 151643,
8
+ "hidden_act": "silu",
9
+ "hidden_size": 1536,
10
+ "initializer_range": 0.02,
11
+ "intermediate_size": 8960,
12
+ "layer_types": [
13
+ "full_attention",
14
+ "full_attention",
15
+ "full_attention",
16
+ "full_attention",
17
+ "full_attention",
18
+ "full_attention",
19
+ "full_attention",
20
+ "full_attention",
21
+ "full_attention",
22
+ "full_attention",
23
+ "full_attention",
24
+ "full_attention",
25
+ "full_attention",
26
+ "full_attention",
27
+ "full_attention",
28
+ "full_attention",
29
+ "full_attention",
30
+ "full_attention",
31
+ "full_attention",
32
+ "full_attention",
33
+ "full_attention",
34
+ "full_attention",
35
+ "full_attention",
36
+ "full_attention",
37
+ "full_attention",
38
+ "full_attention",
39
+ "full_attention",
40
+ "full_attention"
41
+ ],
42
+ "max_position_embeddings": 32768,
43
+ "max_window_layers": 28,
44
+ "model_type": "qwen2",
45
+ "num_attention_heads": 12,
46
+ "num_hidden_layers": 28,
47
+ "num_key_value_heads": 2,
48
+ "pad_token_id": 151643,
49
+ "rms_norm_eps": 1e-06,
50
+ "rope_scaling": null,
51
+ "rope_theta": 1000000.0,
52
+ "sliding_window": null,
53
+ "tie_word_embeddings": true,
54
+ "transformers_version": "4.57.6",
55
+ "use_cache": true,
56
+ "use_sliding_window": false,
57
+ "vocab_size": 151676
58
+ }
checkpoints/codi_frozen_1.5b_logit/checkpoint-1000/merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
checkpoints/codi_frozen_1.5b_logit/checkpoint-1000/pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:615e98bf511fcd45f002e38734edb61325911c17327e65becd01a9c0143e6b81
3
+ size 3096275515
checkpoints/codi_frozen_1.5b_logit/checkpoint-1000/special_tokens_map.json ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "additional_special_tokens": [
3
+ "<|im_start|>",
4
+ "<|im_end|>",
5
+ "<|object_ref_start|>",
6
+ "<|object_ref_end|>",
7
+ "<|box_start|>",
8
+ "<|box_end|>",
9
+ "<|quad_start|>",
10
+ "<|quad_end|>",
11
+ "<|vision_start|>",
12
+ "<|vision_end|>",
13
+ "<|vision_pad|>",
14
+ "<|image_pad|>",
15
+ "<|video_pad|>"
16
+ ],
17
+ "eos_token": {
18
+ "content": "<|endoftext|>",
19
+ "lstrip": false,
20
+ "normalized": false,
21
+ "rstrip": false,
22
+ "single_word": false
23
+ },
24
+ "pad_token": {
25
+ "content": "<|endoftext|>",
26
+ "lstrip": false,
27
+ "normalized": false,
28
+ "rstrip": false,
29
+ "single_word": false
30
+ }
31
+ }
checkpoints/codi_frozen_1.5b_logit/checkpoint-1000/thought_projector.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:013b03d8e979e63b55e5c468844b7c23374e54c3c4ca557a8f0990d7d290ed17
3
+ size 9445953
checkpoints/codi_frozen_1.5b_logit/checkpoint-1000/tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:83a790d654474f5dfe225f889afd0210313eb1083f942671f2c4b8e95a1c922b
3
+ size 11424004
checkpoints/codi_frozen_1.5b_logit/checkpoint-1000/tokenizer_config.json ADDED
@@ -0,0 +1,295 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_bos_token": false,
3
+ "add_prefix_space": false,
4
+ "added_tokens_decoder": {
5
+ "151643": {
6
+ "content": "<|endoftext|>",
7
+ "lstrip": false,
8
+ "normalized": false,
9
+ "rstrip": false,
10
+ "single_word": false,
11
+ "special": true
12
+ },
13
+ "151644": {
14
+ "content": "<|im_start|>",
15
+ "lstrip": false,
16
+ "normalized": false,
17
+ "rstrip": false,
18
+ "single_word": false,
19
+ "special": true
20
+ },
21
+ "151645": {
22
+ "content": "<|im_end|>",
23
+ "lstrip": false,
24
+ "normalized": false,
25
+ "rstrip": false,
26
+ "single_word": false,
27
+ "special": true
28
+ },
29
+ "151646": {
30
+ "content": "<|object_ref_start|>",
31
+ "lstrip": false,
32
+ "normalized": false,
33
+ "rstrip": false,
34
+ "single_word": false,
35
+ "special": true
36
+ },
37
+ "151647": {
38
+ "content": "<|object_ref_end|>",
39
+ "lstrip": false,
40
+ "normalized": false,
41
+ "rstrip": false,
42
+ "single_word": false,
43
+ "special": true
44
+ },
45
+ "151648": {
46
+ "content": "<|box_start|>",
47
+ "lstrip": false,
48
+ "normalized": false,
49
+ "rstrip": false,
50
+ "single_word": false,
51
+ "special": true
52
+ },
53
+ "151649": {
54
+ "content": "<|box_end|>",
55
+ "lstrip": false,
56
+ "normalized": false,
57
+ "rstrip": false,
58
+ "single_word": false,
59
+ "special": true
60
+ },
61
+ "151650": {
62
+ "content": "<|quad_start|>",
63
+ "lstrip": false,
64
+ "normalized": false,
65
+ "rstrip": false,
66
+ "single_word": false,
67
+ "special": true
68
+ },
69
+ "151651": {
70
+ "content": "<|quad_end|>",
71
+ "lstrip": false,
72
+ "normalized": false,
73
+ "rstrip": false,
74
+ "single_word": false,
75
+ "special": true
76
+ },
77
+ "151652": {
78
+ "content": "<|vision_start|>",
79
+ "lstrip": false,
80
+ "normalized": false,
81
+ "rstrip": false,
82
+ "single_word": false,
83
+ "special": true
84
+ },
85
+ "151653": {
86
+ "content": "<|vision_end|>",
87
+ "lstrip": false,
88
+ "normalized": false,
89
+ "rstrip": false,
90
+ "single_word": false,
91
+ "special": true
92
+ },
93
+ "151654": {
94
+ "content": "<|vision_pad|>",
95
+ "lstrip": false,
96
+ "normalized": false,
97
+ "rstrip": false,
98
+ "single_word": false,
99
+ "special": true
100
+ },
101
+ "151655": {
102
+ "content": "<|image_pad|>",
103
+ "lstrip": false,
104
+ "normalized": false,
105
+ "rstrip": false,
106
+ "single_word": false,
107
+ "special": true
108
+ },
109
+ "151656": {
110
+ "content": "<|video_pad|>",
111
+ "lstrip": false,
112
+ "normalized": false,
113
+ "rstrip": false,
114
+ "single_word": false,
115
+ "special": true
116
+ },
117
+ "151657": {
118
+ "content": "<tool_call>",
119
+ "lstrip": false,
120
+ "normalized": false,
121
+ "rstrip": false,
122
+ "single_word": false,
123
+ "special": false
124
+ },
125
+ "151658": {
126
+ "content": "</tool_call>",
127
+ "lstrip": false,
128
+ "normalized": false,
129
+ "rstrip": false,
130
+ "single_word": false,
131
+ "special": false
132
+ },
133
+ "151659": {
134
+ "content": "<|fim_prefix|>",
135
+ "lstrip": false,
136
+ "normalized": false,
137
+ "rstrip": false,
138
+ "single_word": false,
139
+ "special": false
140
+ },
141
+ "151660": {
142
+ "content": "<|fim_middle|>",
143
+ "lstrip": false,
144
+ "normalized": false,
145
+ "rstrip": false,
146
+ "single_word": false,
147
+ "special": false
148
+ },
149
+ "151661": {
150
+ "content": "<|fim_suffix|>",
151
+ "lstrip": false,
152
+ "normalized": false,
153
+ "rstrip": false,
154
+ "single_word": false,
155
+ "special": false
156
+ },
157
+ "151662": {
158
+ "content": "<|fim_pad|>",
159
+ "lstrip": false,
160
+ "normalized": false,
161
+ "rstrip": false,
162
+ "single_word": false,
163
+ "special": false
164
+ },
165
+ "151663": {
166
+ "content": "<|repo_name|>",
167
+ "lstrip": false,
168
+ "normalized": false,
169
+ "rstrip": false,
170
+ "single_word": false,
171
+ "special": false
172
+ },
173
+ "151664": {
174
+ "content": "<|file_sep|>",
175
+ "lstrip": false,
176
+ "normalized": false,
177
+ "rstrip": false,
178
+ "single_word": false,
179
+ "special": false
180
+ },
181
+ "151665": {
182
+ "content": "<|trace_context_start|>",
183
+ "lstrip": false,
184
+ "normalized": false,
185
+ "rstrip": false,
186
+ "single_word": false,
187
+ "special": true
188
+ },
189
+ "151666": {
190
+ "content": "<|call_sep|>",
191
+ "lstrip": false,
192
+ "normalized": false,
193
+ "rstrip": false,
194
+ "single_word": false,
195
+ "special": true
196
+ },
197
+ "151667": {
198
+ "content": "<|line_sep|>",
199
+ "lstrip": false,
200
+ "normalized": false,
201
+ "rstrip": false,
202
+ "single_word": false,
203
+ "special": true
204
+ },
205
+ "151668": {
206
+ "content": "<|return_sep|>",
207
+ "lstrip": false,
208
+ "normalized": false,
209
+ "rstrip": false,
210
+ "single_word": false,
211
+ "special": true
212
+ },
213
+ "151669": {
214
+ "content": "<|exception_sep|>",
215
+ "lstrip": false,
216
+ "normalized": false,
217
+ "rstrip": false,
218
+ "single_word": false,
219
+ "special": true
220
+ },
221
+ "151670": {
222
+ "content": "<|action_sep|>",
223
+ "lstrip": false,
224
+ "normalized": false,
225
+ "rstrip": false,
226
+ "single_word": false,
227
+ "special": true
228
+ },
229
+ "151671": {
230
+ "content": "<|arg_sep|>",
231
+ "lstrip": false,
232
+ "normalized": false,
233
+ "rstrip": false,
234
+ "single_word": false,
235
+ "special": true
236
+ },
237
+ "151672": {
238
+ "content": "<|frame_sep|>",
239
+ "lstrip": false,
240
+ "normalized": false,
241
+ "rstrip": false,
242
+ "single_word": false,
243
+ "special": true
244
+ },
245
+ "151673": {
246
+ "content": "<|end_of_text|>",
247
+ "lstrip": false,
248
+ "normalized": false,
249
+ "rstrip": false,
250
+ "single_word": false,
251
+ "special": true
252
+ },
253
+ "151674": {
254
+ "content": "<|latent_start|>",
255
+ "lstrip": false,
256
+ "normalized": false,
257
+ "rstrip": false,
258
+ "single_word": false,
259
+ "special": true
260
+ },
261
+ "151675": {
262
+ "content": "<|latent_end|>",
263
+ "lstrip": false,
264
+ "normalized": false,
265
+ "rstrip": false,
266
+ "single_word": false,
267
+ "special": true
268
+ }
269
+ },
270
+ "additional_special_tokens": [
271
+ "<|im_start|>",
272
+ "<|im_end|>",
273
+ "<|object_ref_start|>",
274
+ "<|object_ref_end|>",
275
+ "<|box_start|>",
276
+ "<|box_end|>",
277
+ "<|quad_start|>",
278
+ "<|quad_end|>",
279
+ "<|vision_start|>",
280
+ "<|vision_end|>",
281
+ "<|vision_pad|>",
282
+ "<|image_pad|>",
283
+ "<|video_pad|>"
284
+ ],
285
+ "bos_token": null,
286
+ "clean_up_tokenization_spaces": false,
287
+ "eos_token": "<|endoftext|>",
288
+ "errors": "replace",
289
+ "extra_special_tokens": {},
290
+ "model_max_length": 32768,
291
+ "pad_token": "<|endoftext|>",
292
+ "split_special_tokens": false,
293
+ "tokenizer_class": "Qwen2Tokenizer",
294
+ "unk_token": null
295
+ }
checkpoints/codi_frozen_1.5b_logit/checkpoint-1000/trainer_state.json ADDED
@@ -0,0 +1,2034 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 0.28835063437139563,
6
+ "eval_steps": 500,
7
+ "global_step": 1000,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.0014417531718569781,
14
+ "grad_norm": 1952.0,
15
+ "kd_loss": 1.354009985923767,
16
+ "learning_rate": 4.444444444444445e-07,
17
+ "loss": 4.4194,
18
+ "step": 5,
19
+ "student_loss": 0.10154028236865997,
20
+ "teacher_loss": 0.0
21
+ },
22
+ {
23
+ "epoch": 0.0028835063437139563,
24
+ "grad_norm": 768.0,
25
+ "kd_loss": 2.1859679222106934,
26
+ "learning_rate": 1.0000000000000002e-06,
27
+ "loss": 4.072,
28
+ "step": 10,
29
+ "student_loss": 0.23161625862121582,
30
+ "teacher_loss": 0.0
31
+ },
32
+ {
33
+ "epoch": 0.004325259515570935,
34
+ "grad_norm": 676.0,
35
+ "kd_loss": 0.8974658250808716,
36
+ "learning_rate": 1.5555555555555558e-06,
37
+ "loss": 3.0092,
38
+ "step": 15,
39
+ "student_loss": 0.07354968786239624,
40
+ "teacher_loss": 0.0
41
+ },
42
+ {
43
+ "epoch": 0.0057670126874279125,
44
+ "grad_norm": 784.0,
45
+ "kd_loss": 5.459206581115723,
46
+ "learning_rate": 2.1111111111111114e-06,
47
+ "loss": 2.0182,
48
+ "step": 20,
49
+ "student_loss": 0.24417075514793396,
50
+ "teacher_loss": 0.0
51
+ },
52
+ {
53
+ "epoch": 0.00720876585928489,
54
+ "grad_norm": 132.0,
55
+ "kd_loss": 0.039198435842990875,
56
+ "learning_rate": 2.666666666666667e-06,
57
+ "loss": 0.9093,
58
+ "step": 25,
59
+ "student_loss": 0.3775336444377899,
60
+ "teacher_loss": 0.0
61
+ },
62
+ {
63
+ "epoch": 0.00865051903114187,
64
+ "grad_norm": 82.0,
65
+ "kd_loss": 0.01129975076764822,
66
+ "learning_rate": 3.2222222222222227e-06,
67
+ "loss": 0.6102,
68
+ "step": 30,
69
+ "student_loss": 0.48634102940559387,
70
+ "teacher_loss": 0.0
71
+ },
72
+ {
73
+ "epoch": 0.010092272202998846,
74
+ "grad_norm": 67.5,
75
+ "kd_loss": 0.01766686514019966,
76
+ "learning_rate": 3.777777777777778e-06,
77
+ "loss": 0.5753,
78
+ "step": 35,
79
+ "student_loss": 0.16448530554771423,
80
+ "teacher_loss": 0.0
81
+ },
82
+ {
83
+ "epoch": 0.011534025374855825,
84
+ "grad_norm": 66.0,
85
+ "kd_loss": 0.02908974140882492,
86
+ "learning_rate": 4.333333333333334e-06,
87
+ "loss": 0.441,
88
+ "step": 40,
89
+ "student_loss": 0.11244617402553558,
90
+ "teacher_loss": 0.0
91
+ },
92
+ {
93
+ "epoch": 0.012975778546712802,
94
+ "grad_norm": 68.0,
95
+ "kd_loss": 0.2916097640991211,
96
+ "learning_rate": 4.888888888888889e-06,
97
+ "loss": 0.2934,
98
+ "step": 45,
99
+ "student_loss": 0.07623665779829025,
100
+ "teacher_loss": 0.0
101
+ },
102
+ {
103
+ "epoch": 0.01441753171856978,
104
+ "grad_norm": 88.5,
105
+ "kd_loss": 0.19340172410011292,
106
+ "learning_rate": 5.444444444444445e-06,
107
+ "loss": 0.3841,
108
+ "step": 50,
109
+ "student_loss": 0.009455127641558647,
110
+ "teacher_loss": 0.0
111
+ },
112
+ {
113
+ "epoch": 0.015859284890426758,
114
+ "grad_norm": 115.0,
115
+ "kd_loss": 0.01024007610976696,
116
+ "learning_rate": 6e-06,
117
+ "loss": 0.2854,
118
+ "step": 55,
119
+ "student_loss": 0.0043172105215489864,
120
+ "teacher_loss": 0.0
121
+ },
122
+ {
123
+ "epoch": 0.01730103806228374,
124
+ "grad_norm": 57.75,
125
+ "kd_loss": 0.0018813807982951403,
126
+ "learning_rate": 6.555555555555556e-06,
127
+ "loss": 0.2133,
128
+ "step": 60,
129
+ "student_loss": 0.012564695440232754,
130
+ "teacher_loss": 0.0
131
+ },
132
+ {
133
+ "epoch": 0.018742791234140715,
134
+ "grad_norm": 56.0,
135
+ "kd_loss": 0.0018871173961088061,
136
+ "learning_rate": 7.111111111111112e-06,
137
+ "loss": 0.2158,
138
+ "step": 65,
139
+ "student_loss": 0.050387192517519,
140
+ "teacher_loss": 0.0
141
+ },
142
+ {
143
+ "epoch": 0.020184544405997693,
144
+ "grad_norm": 28.625,
145
+ "kd_loss": 0.4643922448158264,
146
+ "learning_rate": 7.666666666666667e-06,
147
+ "loss": 0.1818,
148
+ "step": 70,
149
+ "student_loss": 0.023501692339777946,
150
+ "teacher_loss": 0.0
151
+ },
152
+ {
153
+ "epoch": 0.02162629757785467,
154
+ "grad_norm": 43.0,
155
+ "kd_loss": 0.010926949791610241,
156
+ "learning_rate": 8.222222222222222e-06,
157
+ "loss": 0.3426,
158
+ "step": 75,
159
+ "student_loss": 0.0022108303382992744,
160
+ "teacher_loss": 0.0
161
+ },
162
+ {
163
+ "epoch": 0.02306805074971165,
164
+ "grad_norm": 13.4375,
165
+ "kd_loss": 0.004479927476495504,
166
+ "learning_rate": 8.777777777777778e-06,
167
+ "loss": 0.3733,
168
+ "step": 80,
169
+ "student_loss": 0.0041625467129051685,
170
+ "teacher_loss": 0.0
171
+ },
172
+ {
173
+ "epoch": 0.024509803921568627,
174
+ "grad_norm": 19.5,
175
+ "kd_loss": 0.003717327257618308,
176
+ "learning_rate": 9.333333333333334e-06,
177
+ "loss": 0.2466,
178
+ "step": 85,
179
+ "student_loss": 0.0008013551123440266,
180
+ "teacher_loss": 0.0
181
+ },
182
+ {
183
+ "epoch": 0.025951557093425604,
184
+ "grad_norm": 8.4375,
185
+ "kd_loss": 0.3134710490703583,
186
+ "learning_rate": 9.88888888888889e-06,
187
+ "loss": 0.2059,
188
+ "step": 90,
189
+ "student_loss": 0.016512297093868256,
190
+ "teacher_loss": 0.0
191
+ },
192
+ {
193
+ "epoch": 0.027393310265282585,
194
+ "grad_norm": 42.0,
195
+ "kd_loss": 0.0037511922419071198,
196
+ "learning_rate": 9.99995337991255e-06,
197
+ "loss": 0.1824,
198
+ "step": 95,
199
+ "student_loss": 0.014032241888344288,
200
+ "teacher_loss": 0.0
201
+ },
202
+ {
203
+ "epoch": 0.02883506343713956,
204
+ "grad_norm": 42.75,
205
+ "kd_loss": 0.0027243352960795164,
206
+ "learning_rate": 9.999763987297266e-06,
207
+ "loss": 0.1423,
208
+ "step": 100,
209
+ "student_loss": 0.0006586164236068726,
210
+ "teacher_loss": 0.0
211
+ },
212
+ {
213
+ "epoch": 0.03027681660899654,
214
+ "grad_norm": 22.75,
215
+ "kd_loss": 0.007741209119558334,
216
+ "learning_rate": 9.999428913912876e-06,
217
+ "loss": 0.1432,
218
+ "step": 105,
219
+ "student_loss": 0.006049849092960358,
220
+ "teacher_loss": 0.0
221
+ },
222
+ {
223
+ "epoch": 0.031718569780853516,
224
+ "grad_norm": 8.0,
225
+ "kd_loss": 0.002043877262622118,
226
+ "learning_rate": 9.998948169522587e-06,
227
+ "loss": 0.2012,
228
+ "step": 110,
229
+ "student_loss": 0.03380074352025986,
230
+ "teacher_loss": 0.0
231
+ },
232
+ {
233
+ "epoch": 0.03316032295271049,
234
+ "grad_norm": 12.9375,
235
+ "kd_loss": 0.13418245315551758,
236
+ "learning_rate": 9.998321768134101e-06,
237
+ "loss": 0.1955,
238
+ "step": 115,
239
+ "student_loss": 0.0011221985332667828,
240
+ "teacher_loss": 0.0
241
+ },
242
+ {
243
+ "epoch": 0.03460207612456748,
244
+ "grad_norm": 5.5,
245
+ "kd_loss": 0.00385228986851871,
246
+ "learning_rate": 9.997549727999212e-06,
247
+ "loss": 0.1011,
248
+ "step": 120,
249
+ "student_loss": 0.0008886058931238949,
250
+ "teacher_loss": 0.0
251
+ },
252
+ {
253
+ "epoch": 0.036043829296424454,
254
+ "grad_norm": 21.75,
255
+ "kd_loss": 0.8833059072494507,
256
+ "learning_rate": 9.996632071613257e-06,
257
+ "loss": 0.1431,
258
+ "step": 125,
259
+ "student_loss": 0.037908557802438736,
260
+ "teacher_loss": 0.0
261
+ },
262
+ {
263
+ "epoch": 0.03748558246828143,
264
+ "grad_norm": 22.75,
265
+ "kd_loss": 0.0020795725286006927,
266
+ "learning_rate": 9.995568825714479e-06,
267
+ "loss": 0.2122,
268
+ "step": 130,
269
+ "student_loss": 0.0007846044027246535,
270
+ "teacher_loss": 0.0
271
+ },
272
+ {
273
+ "epoch": 0.03892733564013841,
274
+ "grad_norm": 27.5,
275
+ "kd_loss": 0.6685954928398132,
276
+ "learning_rate": 9.994360021283235e-06,
277
+ "loss": 0.1744,
278
+ "step": 135,
279
+ "student_loss": 0.03987047076225281,
280
+ "teacher_loss": 0.0
281
+ },
282
+ {
283
+ "epoch": 0.040369088811995385,
284
+ "grad_norm": 6.34375,
285
+ "kd_loss": 0.0027170090470463037,
286
+ "learning_rate": 9.993005693541098e-06,
287
+ "loss": 0.1217,
288
+ "step": 140,
289
+ "student_loss": 0.0007884247461333871,
290
+ "teacher_loss": 0.0
291
+ },
292
+ {
293
+ "epoch": 0.04181084198385236,
294
+ "grad_norm": 15.625,
295
+ "kd_loss": 0.7305672764778137,
296
+ "learning_rate": 9.991505881949837e-06,
297
+ "loss": 0.1177,
298
+ "step": 145,
299
+ "student_loss": 0.032465774565935135,
300
+ "teacher_loss": 0.0
301
+ },
302
+ {
303
+ "epoch": 0.04325259515570934,
304
+ "grad_norm": 11.9375,
305
+ "kd_loss": 0.0009196309256367385,
306
+ "learning_rate": 9.989860630210251e-06,
307
+ "loss": 0.1145,
308
+ "step": 150,
309
+ "student_loss": 0.0006800334667786956,
310
+ "teacher_loss": 0.0
311
+ },
312
+ {
313
+ "epoch": 0.04469434832756632,
314
+ "grad_norm": 3.75,
315
+ "kd_loss": 0.0008867154247127473,
316
+ "learning_rate": 9.98806998626091e-06,
317
+ "loss": 0.1703,
318
+ "step": 155,
319
+ "student_loss": 0.000513731618411839,
320
+ "teacher_loss": 0.0
321
+ },
322
+ {
323
+ "epoch": 0.0461361014994233,
324
+ "grad_norm": 13.5625,
325
+ "kd_loss": 0.0031801851000636816,
326
+ "learning_rate": 9.98613400227676e-06,
327
+ "loss": 0.1881,
328
+ "step": 160,
329
+ "student_loss": 0.095796138048172,
330
+ "teacher_loss": 0.0
331
+ },
332
+ {
333
+ "epoch": 0.04757785467128028,
334
+ "grad_norm": 8.75,
335
+ "kd_loss": 0.009066148661077023,
336
+ "learning_rate": 9.984052734667586e-06,
337
+ "loss": 0.147,
338
+ "step": 165,
339
+ "student_loss": 0.00036973567330278456,
340
+ "teacher_loss": 0.0
341
+ },
342
+ {
343
+ "epoch": 0.049019607843137254,
344
+ "grad_norm": 19.875,
345
+ "kd_loss": 0.0007732495432719588,
346
+ "learning_rate": 9.981826244076387e-06,
347
+ "loss": 0.1159,
348
+ "step": 170,
349
+ "student_loss": 0.0006115734577178955,
350
+ "teacher_loss": 0.0
351
+ },
352
+ {
353
+ "epoch": 0.05046136101499423,
354
+ "grad_norm": 17.75,
355
+ "kd_loss": 0.001440449501387775,
356
+ "learning_rate": 9.979454595377594e-06,
357
+ "loss": 0.1439,
358
+ "step": 175,
359
+ "student_loss": 0.0018107936484739184,
360
+ "teacher_loss": 0.0
361
+ },
362
+ {
363
+ "epoch": 0.05190311418685121,
364
+ "grad_norm": 12.6875,
365
+ "kd_loss": 0.003575589507818222,
366
+ "learning_rate": 9.97693785767519e-06,
367
+ "loss": 0.0823,
368
+ "step": 180,
369
+ "student_loss": 0.0005937832174822688,
370
+ "teacher_loss": 0.0
371
+ },
372
+ {
373
+ "epoch": 0.05334486735870819,
374
+ "grad_norm": 13.8125,
375
+ "kd_loss": 0.002183184027671814,
376
+ "learning_rate": 9.9742761043007e-06,
377
+ "loss": 0.2085,
378
+ "step": 185,
379
+ "student_loss": 0.058680690824985504,
380
+ "teacher_loss": 0.0
381
+ },
382
+ {
383
+ "epoch": 0.05478662053056517,
384
+ "grad_norm": 7.9375,
385
+ "kd_loss": 0.13934114575386047,
386
+ "learning_rate": 9.971469412811032e-06,
387
+ "loss": 0.1039,
388
+ "step": 190,
389
+ "student_loss": 0.003180636093020439,
390
+ "teacher_loss": 0.0
391
+ },
392
+ {
393
+ "epoch": 0.056228373702422146,
394
+ "grad_norm": 14.5625,
395
+ "kd_loss": 0.2642575800418854,
396
+ "learning_rate": 9.968517864986247e-06,
397
+ "loss": 0.0963,
398
+ "step": 195,
399
+ "student_loss": 0.023328492417931557,
400
+ "teacher_loss": 0.0
401
+ },
402
+ {
403
+ "epoch": 0.05767012687427912,
404
+ "grad_norm": 6.1875,
405
+ "kd_loss": 0.0009822066640481353,
406
+ "learning_rate": 9.965421546827153e-06,
407
+ "loss": 0.1046,
408
+ "step": 200,
409
+ "student_loss": 0.000694891030434519,
410
+ "teacher_loss": 0.0
411
+ },
412
+ {
413
+ "epoch": 0.0591118800461361,
414
+ "grad_norm": 13.5,
415
+ "kd_loss": 0.0020439866930246353,
416
+ "learning_rate": 9.962180548552812e-06,
417
+ "loss": 0.0897,
418
+ "step": 205,
419
+ "student_loss": 0.042425841093063354,
420
+ "teacher_loss": 0.0
421
+ },
422
+ {
423
+ "epoch": 0.06055363321799308,
424
+ "grad_norm": 5.6875,
425
+ "kd_loss": 0.06946802884340286,
426
+ "learning_rate": 9.958794964597906e-06,
427
+ "loss": 0.0994,
428
+ "step": 210,
429
+ "student_loss": 0.0007058735354803503,
430
+ "teacher_loss": 0.0
431
+ },
432
+ {
433
+ "epoch": 0.061995386389850055,
434
+ "grad_norm": 8.75,
435
+ "kd_loss": 0.0022037022281438112,
436
+ "learning_rate": 9.955264893609986e-06,
437
+ "loss": 0.1323,
438
+ "step": 215,
439
+ "student_loss": 0.0006759354146197438,
440
+ "teacher_loss": 0.0
441
+ },
442
+ {
443
+ "epoch": 0.06343713956170703,
444
+ "grad_norm": 23.5,
445
+ "kd_loss": 0.11546128988265991,
446
+ "learning_rate": 9.951590438446597e-06,
447
+ "loss": 0.108,
448
+ "step": 220,
449
+ "student_loss": 0.04613906517624855,
450
+ "teacher_loss": 0.0
451
+ },
452
+ {
453
+ "epoch": 0.06487889273356401,
454
+ "grad_norm": 11.0,
455
+ "kd_loss": 0.006272319238632917,
456
+ "learning_rate": 9.947771706172285e-06,
457
+ "loss": 0.1205,
458
+ "step": 225,
459
+ "student_loss": 0.03350772336125374,
460
+ "teacher_loss": 0.0
461
+ },
462
+ {
463
+ "epoch": 0.06632064590542099,
464
+ "grad_norm": 24.25,
465
+ "kd_loss": 0.26141810417175293,
466
+ "learning_rate": 9.943808808055472e-06,
467
+ "loss": 0.0788,
468
+ "step": 230,
469
+ "student_loss": 0.003604925936087966,
470
+ "teacher_loss": 0.0
471
+ },
472
+ {
473
+ "epoch": 0.06776239907727798,
474
+ "grad_norm": 15.6875,
475
+ "kd_loss": 0.6798978447914124,
476
+ "learning_rate": 9.93970185956522e-06,
477
+ "loss": 0.0869,
478
+ "step": 235,
479
+ "student_loss": 0.015447350218892097,
480
+ "teacher_loss": 0.0
481
+ },
482
+ {
483
+ "epoch": 0.06920415224913495,
484
+ "grad_norm": 1.71875,
485
+ "kd_loss": 0.003386611584573984,
486
+ "learning_rate": 9.93545098036786e-06,
487
+ "loss": 0.1246,
488
+ "step": 240,
489
+ "student_loss": 0.0006252449820749462,
490
+ "teacher_loss": 0.0
491
+ },
492
+ {
493
+ "epoch": 0.07064590542099193,
494
+ "grad_norm": 21.25,
495
+ "kd_loss": 0.2356187403202057,
496
+ "learning_rate": 9.931056294323507e-06,
497
+ "loss": 0.146,
498
+ "step": 245,
499
+ "student_loss": 0.0012863262090831995,
500
+ "teacher_loss": 0.0
501
+ },
502
+ {
503
+ "epoch": 0.07208765859284891,
504
+ "grad_norm": 12.9375,
505
+ "kd_loss": 0.0037490571849048138,
506
+ "learning_rate": 9.926517929482454e-06,
507
+ "loss": 0.1485,
508
+ "step": 250,
509
+ "student_loss": 0.00040524674113839865,
510
+ "teacher_loss": 0.0
511
+ },
512
+ {
513
+ "epoch": 0.07352941176470588,
514
+ "grad_norm": 21.125,
515
+ "kd_loss": 0.0018789542373269796,
516
+ "learning_rate": 9.921836018081438e-06,
517
+ "loss": 0.056,
518
+ "step": 255,
519
+ "student_loss": 0.0003012118686456233,
520
+ "teacher_loss": 0.0
521
+ },
522
+ {
523
+ "epoch": 0.07497116493656286,
524
+ "grad_norm": 10.25,
525
+ "kd_loss": 0.008267119526863098,
526
+ "learning_rate": 9.917010696539793e-06,
527
+ "loss": 0.2125,
528
+ "step": 260,
529
+ "student_loss": 0.04526061564683914,
530
+ "teacher_loss": 0.0
531
+ },
532
+ {
533
+ "epoch": 0.07641291810841984,
534
+ "grad_norm": 4.8125,
535
+ "kd_loss": 0.0011836340418085456,
536
+ "learning_rate": 9.912042105455462e-06,
537
+ "loss": 0.1144,
538
+ "step": 265,
539
+ "student_loss": 0.00038986533763818443,
540
+ "teacher_loss": 0.0
541
+ },
542
+ {
543
+ "epoch": 0.07785467128027682,
544
+ "grad_norm": 12.875,
545
+ "kd_loss": 0.0008742965874262154,
546
+ "learning_rate": 9.906930389600917e-06,
547
+ "loss": 0.1514,
548
+ "step": 270,
549
+ "student_loss": 0.0005801369552500546,
550
+ "teacher_loss": 0.0
551
+ },
552
+ {
553
+ "epoch": 0.07929642445213379,
554
+ "grad_norm": 17.75,
555
+ "kd_loss": 0.0008876683423295617,
556
+ "learning_rate": 9.901675697918925e-06,
557
+ "loss": 0.1775,
558
+ "step": 275,
559
+ "student_loss": 0.053293921053409576,
560
+ "teacher_loss": 0.0
561
+ },
562
+ {
563
+ "epoch": 0.08073817762399077,
564
+ "grad_norm": 26.75,
565
+ "kd_loss": 0.002074585296213627,
566
+ "learning_rate": 9.896278183518216e-06,
567
+ "loss": 0.1494,
568
+ "step": 280,
569
+ "student_loss": 0.004262140020728111,
570
+ "teacher_loss": 0.0
571
+ },
572
+ {
573
+ "epoch": 0.08217993079584775,
574
+ "grad_norm": 7.65625,
575
+ "kd_loss": 0.0025187216233462095,
576
+ "learning_rate": 9.890738003669029e-06,
577
+ "loss": 0.1171,
578
+ "step": 285,
579
+ "student_loss": 0.0007169336313381791,
580
+ "teacher_loss": 0.0
581
+ },
582
+ {
583
+ "epoch": 0.08362168396770472,
584
+ "grad_norm": 21.375,
585
+ "kd_loss": 0.007369345054030418,
586
+ "learning_rate": 9.885055319798514e-06,
587
+ "loss": 0.1821,
588
+ "step": 290,
589
+ "student_loss": 0.0005684102652594447,
590
+ "teacher_loss": 0.0
591
+ },
592
+ {
593
+ "epoch": 0.0850634371395617,
594
+ "grad_norm": 16.0,
595
+ "kd_loss": 0.006104703992605209,
596
+ "learning_rate": 9.879230297486034e-06,
597
+ "loss": 0.1704,
598
+ "step": 295,
599
+ "student_loss": 0.0010396749712526798,
600
+ "teacher_loss": 0.0
601
+ },
602
+ {
603
+ "epoch": 0.08650519031141868,
604
+ "grad_norm": 23.375,
605
+ "kd_loss": 0.08903778344392776,
606
+ "learning_rate": 9.873263106458353e-06,
607
+ "loss": 0.1398,
608
+ "step": 300,
609
+ "student_loss": 0.014311299659311771,
610
+ "teacher_loss": 0.0
611
+ },
612
+ {
613
+ "epoch": 0.08794694348327567,
614
+ "grad_norm": 21.125,
615
+ "kd_loss": 0.003498164936900139,
616
+ "learning_rate": 9.86715392058467e-06,
617
+ "loss": 0.1304,
618
+ "step": 305,
619
+ "student_loss": 0.018731793388724327,
620
+ "teacher_loss": 0.0
621
+ },
622
+ {
623
+ "epoch": 0.08938869665513265,
624
+ "grad_norm": 22.25,
625
+ "kd_loss": 0.0002319709019502625,
626
+ "learning_rate": 9.860902917871566e-06,
627
+ "loss": 0.1595,
628
+ "step": 310,
629
+ "student_loss": 0.0005537475226446986,
630
+ "teacher_loss": 0.0
631
+ },
632
+ {
633
+ "epoch": 0.09083044982698962,
634
+ "grad_norm": 28.125,
635
+ "kd_loss": 0.012743989005684853,
636
+ "learning_rate": 9.854510280457819e-06,
637
+ "loss": 0.1062,
638
+ "step": 315,
639
+ "student_loss": 0.012689951807260513,
640
+ "teacher_loss": 0.0
641
+ },
642
+ {
643
+ "epoch": 0.0922722029988466,
644
+ "grad_norm": 18.625,
645
+ "kd_loss": 0.0014882351970300078,
646
+ "learning_rate": 9.847976194609083e-06,
647
+ "loss": 0.1479,
648
+ "step": 320,
649
+ "student_loss": 0.0005163275054655969,
650
+ "teacher_loss": 0.0
651
+ },
652
+ {
653
+ "epoch": 0.09371395617070358,
654
+ "grad_norm": 22.625,
655
+ "kd_loss": 0.3226740062236786,
656
+ "learning_rate": 9.841300850712479e-06,
657
+ "loss": 0.1206,
658
+ "step": 325,
659
+ "student_loss": 0.005041609983891249,
660
+ "teacher_loss": 0.0
661
+ },
662
+ {
663
+ "epoch": 0.09515570934256055,
664
+ "grad_norm": 10.125,
665
+ "kd_loss": 0.4770796298980713,
666
+ "learning_rate": 9.834484443271035e-06,
667
+ "loss": 0.0984,
668
+ "step": 330,
669
+ "student_loss": 0.015074173919856548,
670
+ "teacher_loss": 0.0
671
+ },
672
+ {
673
+ "epoch": 0.09659746251441753,
674
+ "grad_norm": 13.25,
675
+ "kd_loss": 0.000536711944732815,
676
+ "learning_rate": 9.827527170898019e-06,
677
+ "loss": 0.0722,
678
+ "step": 335,
679
+ "student_loss": 0.0006227240664884448,
680
+ "teacher_loss": 0.0
681
+ },
682
+ {
683
+ "epoch": 0.09803921568627451,
684
+ "grad_norm": 26.0,
685
+ "kd_loss": 0.12674546241760254,
686
+ "learning_rate": 9.820429236311158e-06,
687
+ "loss": 0.1458,
688
+ "step": 340,
689
+ "student_loss": 0.012428171001374722,
690
+ "teacher_loss": 0.0
691
+ },
692
+ {
693
+ "epoch": 0.09948096885813149,
694
+ "grad_norm": 8.375,
695
+ "kd_loss": 0.0027160593308508396,
696
+ "learning_rate": 9.81319084632673e-06,
697
+ "loss": 0.1197,
698
+ "step": 345,
699
+ "student_loss": 0.01698848232626915,
700
+ "teacher_loss": 0.0
701
+ },
702
+ {
703
+ "epoch": 0.10092272202998846,
704
+ "grad_norm": 25.5,
705
+ "kd_loss": 0.2688156068325043,
706
+ "learning_rate": 9.805812211853534e-06,
707
+ "loss": 0.0684,
708
+ "step": 350,
709
+ "student_loss": 0.006060065235942602,
710
+ "teacher_loss": 0.0
711
+ },
712
+ {
713
+ "epoch": 0.10236447520184544,
714
+ "grad_norm": 26.0,
715
+ "kd_loss": 0.0023566237650811672,
716
+ "learning_rate": 9.798293547886748e-06,
717
+ "loss": 0.1159,
718
+ "step": 355,
719
+ "student_loss": 0.0004638656973838806,
720
+ "teacher_loss": 0.0
721
+ },
722
+ {
723
+ "epoch": 0.10380622837370242,
724
+ "grad_norm": 9.125,
725
+ "kd_loss": 0.004494955763220787,
726
+ "learning_rate": 9.790635073501662e-06,
727
+ "loss": 0.082,
728
+ "step": 360,
729
+ "student_loss": 0.0004987778957001865,
730
+ "teacher_loss": 0.0
731
+ },
732
+ {
733
+ "epoch": 0.1052479815455594,
734
+ "grad_norm": 10.375,
735
+ "kd_loss": 0.0029149462934583426,
736
+ "learning_rate": 9.782837011847294e-06,
737
+ "loss": 0.1926,
738
+ "step": 365,
739
+ "student_loss": 0.043784309178590775,
740
+ "teacher_loss": 0.0
741
+ },
742
+ {
743
+ "epoch": 0.10668973471741638,
744
+ "grad_norm": 24.625,
745
+ "kd_loss": 0.07756392657756805,
746
+ "learning_rate": 9.774899590139897e-06,
747
+ "loss": 0.0849,
748
+ "step": 370,
749
+ "student_loss": 0.01941766031086445,
750
+ "teacher_loss": 0.0
751
+ },
752
+ {
753
+ "epoch": 0.10813148788927336,
754
+ "grad_norm": 13.125,
755
+ "kd_loss": 0.00261987978592515,
756
+ "learning_rate": 9.766823039656328e-06,
757
+ "loss": 0.1517,
758
+ "step": 375,
759
+ "student_loss": 0.03552848473191261,
760
+ "teacher_loss": 0.0
761
+ },
762
+ {
763
+ "epoch": 0.10957324106113034,
764
+ "grad_norm": 6.28125,
765
+ "kd_loss": 0.0021297987550497055,
766
+ "learning_rate": 9.758607595727309e-06,
767
+ "loss": 0.1632,
768
+ "step": 380,
769
+ "student_loss": 0.0005093910149298608,
770
+ "teacher_loss": 0.0
771
+ },
772
+ {
773
+ "epoch": 0.11101499423298732,
774
+ "grad_norm": 28.125,
775
+ "kd_loss": 0.012473474256694317,
776
+ "learning_rate": 9.75025349773058e-06,
777
+ "loss": 0.1271,
778
+ "step": 385,
779
+ "student_loss": 0.00041385076474398375,
780
+ "teacher_loss": 0.0
781
+ },
782
+ {
783
+ "epoch": 0.11245674740484429,
784
+ "grad_norm": 24.375,
785
+ "kd_loss": 0.0008915087673813105,
786
+ "learning_rate": 9.741760989083914e-06,
787
+ "loss": 0.0936,
788
+ "step": 390,
789
+ "student_loss": 0.00028736997046507895,
790
+ "teacher_loss": 0.0
791
+ },
792
+ {
793
+ "epoch": 0.11389850057670127,
794
+ "grad_norm": 12.875,
795
+ "kd_loss": 0.007614473812282085,
796
+ "learning_rate": 9.73313031723803e-06,
797
+ "loss": 0.0901,
798
+ "step": 395,
799
+ "student_loss": 0.0005106693133711815,
800
+ "teacher_loss": 0.0
801
+ },
802
+ {
803
+ "epoch": 0.11534025374855825,
804
+ "grad_norm": 11.25,
805
+ "kd_loss": 0.00347925559617579,
806
+ "learning_rate": 9.724361733669383e-06,
807
+ "loss": 0.0546,
808
+ "step": 400,
809
+ "student_loss": 0.0008912020130082965,
810
+ "teacher_loss": 0.0
811
+ },
812
+ {
813
+ "epoch": 0.11678200692041522,
814
+ "grad_norm": 8.0625,
815
+ "kd_loss": 0.0011954058427363634,
816
+ "learning_rate": 9.715455493872827e-06,
817
+ "loss": 0.1324,
818
+ "step": 405,
819
+ "student_loss": 0.004991908557713032,
820
+ "teacher_loss": 0.0
821
+ },
822
+ {
823
+ "epoch": 0.1182237600922722,
824
+ "grad_norm": 9.1875,
825
+ "kd_loss": 0.0025417145807296038,
826
+ "learning_rate": 9.70641185735419e-06,
827
+ "loss": 0.1483,
828
+ "step": 410,
829
+ "student_loss": 0.01691470481455326,
830
+ "teacher_loss": 0.0
831
+ },
832
+ {
833
+ "epoch": 0.11966551326412918,
834
+ "grad_norm": 13.3125,
835
+ "kd_loss": 0.003752017393708229,
836
+ "learning_rate": 9.697231087622691e-06,
837
+ "loss": 0.1511,
838
+ "step": 415,
839
+ "student_loss": 0.0010631532641127706,
840
+ "teacher_loss": 0.0
841
+ },
842
+ {
843
+ "epoch": 0.12110726643598616,
844
+ "grad_norm": 10.9375,
845
+ "kd_loss": 0.2554698586463928,
846
+ "learning_rate": 9.687913452183277e-06,
847
+ "loss": 0.1095,
848
+ "step": 420,
849
+ "student_loss": 0.06524818390607834,
850
+ "teacher_loss": 0.0
851
+ },
852
+ {
853
+ "epoch": 0.12254901960784313,
854
+ "grad_norm": 9.125,
855
+ "kd_loss": 0.0005501045961864293,
856
+ "learning_rate": 9.678459222528825e-06,
857
+ "loss": 0.0577,
858
+ "step": 425,
859
+ "student_loss": 0.0013150365557521582,
860
+ "teacher_loss": 0.0
861
+ },
862
+ {
863
+ "epoch": 0.12399077277970011,
864
+ "grad_norm": 10.0625,
865
+ "kd_loss": 0.00420792680233717,
866
+ "learning_rate": 9.668868674132224e-06,
867
+ "loss": 0.1381,
868
+ "step": 430,
869
+ "student_loss": 0.05358680710196495,
870
+ "teacher_loss": 0.0
871
+ },
872
+ {
873
+ "epoch": 0.1254325259515571,
874
+ "grad_norm": 14.625,
875
+ "kd_loss": 0.0016143789980560541,
876
+ "learning_rate": 9.65914208643836e-06,
877
+ "loss": 0.0654,
878
+ "step": 435,
879
+ "student_loss": 0.000772066181525588,
880
+ "teacher_loss": 0.0
881
+ },
882
+ {
883
+ "epoch": 0.12687427912341406,
884
+ "grad_norm": 7.15625,
885
+ "kd_loss": 0.0023973328061401844,
886
+ "learning_rate": 9.649279742855964e-06,
887
+ "loss": 0.1371,
888
+ "step": 440,
889
+ "student_loss": 0.025605041533708572,
890
+ "teacher_loss": 0.0
891
+ },
892
+ {
893
+ "epoch": 0.12831603229527105,
894
+ "grad_norm": 23.0,
895
+ "kd_loss": 0.0013417273294180632,
896
+ "learning_rate": 9.639281930749363e-06,
897
+ "loss": 0.0814,
898
+ "step": 445,
899
+ "student_loss": 0.03157199174165726,
900
+ "teacher_loss": 0.0
901
+ },
902
+ {
903
+ "epoch": 0.12975778546712802,
904
+ "grad_norm": 13.25,
905
+ "kd_loss": 0.0016620807582512498,
906
+ "learning_rate": 9.629148941430096e-06,
907
+ "loss": 0.1442,
908
+ "step": 450,
909
+ "student_loss": 0.0006721625686623156,
910
+ "teacher_loss": 0.0
911
+ },
912
+ {
913
+ "epoch": 0.131199538638985,
914
+ "grad_norm": 28.0,
915
+ "kd_loss": 0.0016946454998105764,
916
+ "learning_rate": 9.618881070148436e-06,
917
+ "loss": 0.1405,
918
+ "step": 455,
919
+ "student_loss": 0.0009172292775474489,
920
+ "teacher_loss": 0.0
921
+ },
922
+ {
923
+ "epoch": 0.13264129181084197,
924
+ "grad_norm": 10.125,
925
+ "kd_loss": 0.027760690078139305,
926
+ "learning_rate": 9.608478616084784e-06,
927
+ "loss": 0.0505,
928
+ "step": 460,
929
+ "student_loss": 0.0005534313968382776,
930
+ "teacher_loss": 0.0
931
+ },
932
+ {
933
+ "epoch": 0.13408304498269896,
934
+ "grad_norm": 14.25,
935
+ "kd_loss": 0.008209657855331898,
936
+ "learning_rate": 9.597941882340947e-06,
937
+ "loss": 0.1803,
938
+ "step": 465,
939
+ "student_loss": 0.000469625141704455,
940
+ "teacher_loss": 0.0
941
+ },
942
+ {
943
+ "epoch": 0.13552479815455595,
944
+ "grad_norm": 13.4375,
945
+ "kd_loss": 0.0006104095955379307,
946
+ "learning_rate": 9.587271175931313e-06,
947
+ "loss": 0.1148,
948
+ "step": 470,
949
+ "student_loss": 0.00037698415690101683,
950
+ "teacher_loss": 0.0
951
+ },
952
+ {
953
+ "epoch": 0.13696655132641292,
954
+ "grad_norm": 6.5625,
955
+ "kd_loss": 0.012027841061353683,
956
+ "learning_rate": 9.5764668077739e-06,
957
+ "loss": 0.0626,
958
+ "step": 475,
959
+ "student_loss": 0.0005685262149199843,
960
+ "teacher_loss": 0.0
961
+ },
962
+ {
963
+ "epoch": 0.1384083044982699,
964
+ "grad_norm": 9.0625,
965
+ "kd_loss": 0.0022413916885852814,
966
+ "learning_rate": 9.5655290926813e-06,
967
+ "loss": 0.099,
968
+ "step": 480,
969
+ "student_loss": 0.012239435687661171,
970
+ "teacher_loss": 0.0
971
+ },
972
+ {
973
+ "epoch": 0.13985005767012687,
974
+ "grad_norm": 25.75,
975
+ "kd_loss": 0.0016755703836679459,
976
+ "learning_rate": 9.554458349351507e-06,
977
+ "loss": 0.0891,
978
+ "step": 485,
979
+ "student_loss": 0.027401834726333618,
980
+ "teacher_loss": 0.0
981
+ },
982
+ {
983
+ "epoch": 0.14129181084198386,
984
+ "grad_norm": 29.875,
985
+ "kd_loss": 0.46647658944129944,
986
+ "learning_rate": 9.54325490035863e-06,
987
+ "loss": 0.1581,
988
+ "step": 490,
989
+ "student_loss": 0.0034644226543605328,
990
+ "teacher_loss": 0.0
991
+ },
992
+ {
993
+ "epoch": 0.14273356401384082,
994
+ "grad_norm": 12.0,
995
+ "kd_loss": 0.10091336816549301,
996
+ "learning_rate": 9.531919072143488e-06,
997
+ "loss": 0.1283,
998
+ "step": 495,
999
+ "student_loss": 0.0010429032845422626,
1000
+ "teacher_loss": 0.0
1001
+ },
1002
+ {
1003
+ "epoch": 0.14417531718569782,
1004
+ "grad_norm": 18.125,
1005
+ "kd_loss": 0.25293171405792236,
1006
+ "learning_rate": 9.520451195004107e-06,
1007
+ "loss": 0.0874,
1008
+ "step": 500,
1009
+ "student_loss": 0.0008772966684773564,
1010
+ "teacher_loss": 0.0
1011
+ },
1012
+ {
1013
+ "epoch": 0.14561707035755478,
1014
+ "grad_norm": 1.453125,
1015
+ "kd_loss": 0.012721131555736065,
1016
+ "learning_rate": 9.508851603086094e-06,
1017
+ "loss": 0.124,
1018
+ "step": 505,
1019
+ "student_loss": 0.0005022344412282109,
1020
+ "teacher_loss": 0.0
1021
+ },
1022
+ {
1023
+ "epoch": 0.14705882352941177,
1024
+ "grad_norm": 14.375,
1025
+ "kd_loss": 0.000858868530485779,
1026
+ "learning_rate": 9.497120634372894e-06,
1027
+ "loss": 0.1335,
1028
+ "step": 510,
1029
+ "student_loss": 0.00032862432999536395,
1030
+ "teacher_loss": 0.0
1031
+ },
1032
+ {
1033
+ "epoch": 0.14850057670126873,
1034
+ "grad_norm": 9.6875,
1035
+ "kd_loss": 0.0011465316638350487,
1036
+ "learning_rate": 9.48525863067595e-06,
1037
+ "loss": 0.0708,
1038
+ "step": 515,
1039
+ "student_loss": 0.008267870172858238,
1040
+ "teacher_loss": 0.0
1041
+ },
1042
+ {
1043
+ "epoch": 0.14994232987312572,
1044
+ "grad_norm": 10.625,
1045
+ "kd_loss": 0.002247657161206007,
1046
+ "learning_rate": 9.473265937624748e-06,
1047
+ "loss": 0.0503,
1048
+ "step": 520,
1049
+ "student_loss": 0.041273463517427444,
1050
+ "teacher_loss": 0.0
1051
+ },
1052
+ {
1053
+ "epoch": 0.1513840830449827,
1054
+ "grad_norm": 29.25,
1055
+ "kd_loss": 0.0018336102366447449,
1056
+ "learning_rate": 9.461142904656722e-06,
1057
+ "loss": 0.1143,
1058
+ "step": 525,
1059
+ "student_loss": 0.00042416868382133543,
1060
+ "teacher_loss": 0.0
1061
+ },
1062
+ {
1063
+ "epoch": 0.15282583621683968,
1064
+ "grad_norm": 1.765625,
1065
+ "kd_loss": 0.0032041582744568586,
1066
+ "learning_rate": 9.448889885007105e-06,
1067
+ "loss": 0.0643,
1068
+ "step": 530,
1069
+ "student_loss": 0.00041470438009127975,
1070
+ "teacher_loss": 0.0
1071
+ },
1072
+ {
1073
+ "epoch": 0.15426758938869667,
1074
+ "grad_norm": 6.125,
1075
+ "kd_loss": 0.0082230931147933,
1076
+ "learning_rate": 9.436507235698613e-06,
1077
+ "loss": 0.0916,
1078
+ "step": 535,
1079
+ "student_loss": 0.0004584086127579212,
1080
+ "teacher_loss": 0.0
1081
+ },
1082
+ {
1083
+ "epoch": 0.15570934256055363,
1084
+ "grad_norm": 8.6875,
1085
+ "kd_loss": 0.0010036047315225005,
1086
+ "learning_rate": 9.423995317531049e-06,
1087
+ "loss": 0.128,
1088
+ "step": 540,
1089
+ "student_loss": 0.0008725057705305517,
1090
+ "teacher_loss": 0.0
1091
+ },
1092
+ {
1093
+ "epoch": 0.15715109573241062,
1094
+ "grad_norm": 4.6875,
1095
+ "kd_loss": 0.004125351086258888,
1096
+ "learning_rate": 9.4113544950708e-06,
1097
+ "loss": 0.0659,
1098
+ "step": 545,
1099
+ "student_loss": 0.0022284940350800753,
1100
+ "teacher_loss": 0.0
1101
+ },
1102
+ {
1103
+ "epoch": 0.15859284890426759,
1104
+ "grad_norm": 5.78125,
1105
+ "kd_loss": 0.0035471171140670776,
1106
+ "learning_rate": 9.398585136640195e-06,
1107
+ "loss": 0.0526,
1108
+ "step": 550,
1109
+ "student_loss": 0.0001974322076421231,
1110
+ "teacher_loss": 0.0
1111
+ },
1112
+ {
1113
+ "epoch": 0.16003460207612458,
1114
+ "grad_norm": 15.0,
1115
+ "kd_loss": 0.6907945871353149,
1116
+ "learning_rate": 9.38568761430679e-06,
1117
+ "loss": 0.1251,
1118
+ "step": 555,
1119
+ "student_loss": 0.025625621899962425,
1120
+ "teacher_loss": 0.0
1121
+ },
1122
+ {
1123
+ "epoch": 0.16147635524798154,
1124
+ "grad_norm": 14.5625,
1125
+ "kd_loss": 0.9356266856193542,
1126
+ "learning_rate": 9.372662303872514e-06,
1127
+ "loss": 0.0997,
1128
+ "step": 560,
1129
+ "student_loss": 0.0429786816239357,
1130
+ "teacher_loss": 0.0
1131
+ },
1132
+ {
1133
+ "epoch": 0.16291810841983853,
1134
+ "grad_norm": 7.59375,
1135
+ "kd_loss": 0.0022694910876452923,
1136
+ "learning_rate": 9.359509584862735e-06,
1137
+ "loss": 0.1102,
1138
+ "step": 565,
1139
+ "student_loss": 0.038102827966213226,
1140
+ "teacher_loss": 0.0
1141
+ },
1142
+ {
1143
+ "epoch": 0.1643598615916955,
1144
+ "grad_norm": 24.5,
1145
+ "kd_loss": 0.002567964605987072,
1146
+ "learning_rate": 9.346229840515186e-06,
1147
+ "loss": 0.1791,
1148
+ "step": 570,
1149
+ "student_loss": 0.010701650753617287,
1150
+ "teacher_loss": 0.0
1151
+ },
1152
+ {
1153
+ "epoch": 0.16580161476355249,
1154
+ "grad_norm": 47.75,
1155
+ "kd_loss": 0.0009870093781501055,
1156
+ "learning_rate": 9.332823457768805e-06,
1157
+ "loss": 0.0795,
1158
+ "step": 575,
1159
+ "student_loss": 0.0004428588435985148,
1160
+ "teacher_loss": 0.0
1161
+ },
1162
+ {
1163
+ "epoch": 0.16724336793540945,
1164
+ "grad_norm": 5.625,
1165
+ "kd_loss": 0.018184596672654152,
1166
+ "learning_rate": 9.31929082725246e-06,
1167
+ "loss": 0.1173,
1168
+ "step": 580,
1169
+ "student_loss": 0.0014106971211731434,
1170
+ "teacher_loss": 0.0
1171
+ },
1172
+ {
1173
+ "epoch": 0.16868512110726644,
1174
+ "grad_norm": 5.46875,
1175
+ "kd_loss": 0.10436643660068512,
1176
+ "learning_rate": 9.305632343273567e-06,
1177
+ "loss": 0.0741,
1178
+ "step": 585,
1179
+ "student_loss": 0.0011741393245756626,
1180
+ "teacher_loss": 0.0
1181
+ },
1182
+ {
1183
+ "epoch": 0.1701268742791234,
1184
+ "grad_norm": 21.375,
1185
+ "kd_loss": 0.008095277473330498,
1186
+ "learning_rate": 9.291848403806604e-06,
1187
+ "loss": 0.1016,
1188
+ "step": 590,
1189
+ "student_loss": 0.00045060788397677243,
1190
+ "teacher_loss": 0.0
1191
+ },
1192
+ {
1193
+ "epoch": 0.1715686274509804,
1194
+ "grad_norm": 8.0625,
1195
+ "kd_loss": 0.004174170084297657,
1196
+ "learning_rate": 9.277939410481507e-06,
1197
+ "loss": 0.0639,
1198
+ "step": 595,
1199
+ "student_loss": 0.00036720611387863755,
1200
+ "teacher_loss": 0.0
1201
+ },
1202
+ {
1203
+ "epoch": 0.17301038062283736,
1204
+ "grad_norm": 31.375,
1205
+ "kd_loss": 0.0980350524187088,
1206
+ "learning_rate": 9.263905768571977e-06,
1207
+ "loss": 0.1305,
1208
+ "step": 600,
1209
+ "student_loss": 0.039228975772857666,
1210
+ "teacher_loss": 0.0
1211
+ },
1212
+ {
1213
+ "epoch": 0.17445213379469435,
1214
+ "grad_norm": 7.84375,
1215
+ "kd_loss": 0.0019628205336630344,
1216
+ "learning_rate": 9.249747886983664e-06,
1217
+ "loss": 0.1027,
1218
+ "step": 605,
1219
+ "student_loss": 0.10567551851272583,
1220
+ "teacher_loss": 0.0
1221
+ },
1222
+ {
1223
+ "epoch": 0.17589388696655134,
1224
+ "grad_norm": 10.8125,
1225
+ "kd_loss": 0.15832534432411194,
1226
+ "learning_rate": 9.235466178242255e-06,
1227
+ "loss": 0.0776,
1228
+ "step": 610,
1229
+ "student_loss": 0.008307134732604027,
1230
+ "teacher_loss": 0.0
1231
+ },
1232
+ {
1233
+ "epoch": 0.1773356401384083,
1234
+ "grad_norm": 7.4375,
1235
+ "kd_loss": 0.0004231311031617224,
1236
+ "learning_rate": 9.221061058481456e-06,
1237
+ "loss": 0.0726,
1238
+ "step": 615,
1239
+ "student_loss": 0.00020960887195542455,
1240
+ "teacher_loss": 0.0
1241
+ },
1242
+ {
1243
+ "epoch": 0.1787773933102653,
1244
+ "grad_norm": 2.9375,
1245
+ "kd_loss": 0.000930448470171541,
1246
+ "learning_rate": 9.206532947430865e-06,
1247
+ "loss": 0.0953,
1248
+ "step": 620,
1249
+ "student_loss": 0.012740620411932468,
1250
+ "teacher_loss": 0.0
1251
+ },
1252
+ {
1253
+ "epoch": 0.18021914648212226,
1254
+ "grad_norm": 10.9375,
1255
+ "kd_loss": 0.00044887661351822317,
1256
+ "learning_rate": 9.191882268403743e-06,
1257
+ "loss": 0.0502,
1258
+ "step": 625,
1259
+ "student_loss": 0.006197873968631029,
1260
+ "teacher_loss": 0.0
1261
+ },
1262
+ {
1263
+ "epoch": 0.18166089965397925,
1264
+ "grad_norm": 39.75,
1265
+ "kd_loss": 2.8654651641845703,
1266
+ "learning_rate": 9.177109448284673e-06,
1267
+ "loss": 0.1582,
1268
+ "step": 630,
1269
+ "student_loss": 0.05481014773249626,
1270
+ "teacher_loss": 0.0
1271
+ },
1272
+ {
1273
+ "epoch": 0.1831026528258362,
1274
+ "grad_norm": 5.15625,
1275
+ "kd_loss": 0.0009848499903455377,
1276
+ "learning_rate": 9.162214917517139e-06,
1277
+ "loss": 0.1122,
1278
+ "step": 635,
1279
+ "student_loss": 0.0012388643808662891,
1280
+ "teacher_loss": 0.0
1281
+ },
1282
+ {
1283
+ "epoch": 0.1845444059976932,
1284
+ "grad_norm": 8.8125,
1285
+ "kd_loss": 0.0020470728632062674,
1286
+ "learning_rate": 9.14719911009096e-06,
1287
+ "loss": 0.0638,
1288
+ "step": 640,
1289
+ "student_loss": 0.00041219103150069714,
1290
+ "teacher_loss": 0.0
1291
+ },
1292
+ {
1293
+ "epoch": 0.18598615916955016,
1294
+ "grad_norm": 8.8125,
1295
+ "kd_loss": 0.003420033724978566,
1296
+ "learning_rate": 9.132062463529666e-06,
1297
+ "loss": 0.0673,
1298
+ "step": 645,
1299
+ "student_loss": 0.0004483861557673663,
1300
+ "teacher_loss": 0.0
1301
+ },
1302
+ {
1303
+ "epoch": 0.18742791234140715,
1304
+ "grad_norm": 16.375,
1305
+ "kd_loss": 0.0017254488775506616,
1306
+ "learning_rate": 9.11680541887774e-06,
1307
+ "loss": 0.088,
1308
+ "step": 650,
1309
+ "student_loss": 0.0007219853578135371,
1310
+ "teacher_loss": 0.0
1311
+ },
1312
+ {
1313
+ "epoch": 0.18886966551326412,
1314
+ "grad_norm": 12.4375,
1315
+ "kd_loss": 0.004693194758147001,
1316
+ "learning_rate": 9.101428420687759e-06,
1317
+ "loss": 0.0916,
1318
+ "step": 655,
1319
+ "student_loss": 0.0004059349885210395,
1320
+ "teacher_loss": 0.0
1321
+ },
1322
+ {
1323
+ "epoch": 0.1903114186851211,
1324
+ "grad_norm": 17.125,
1325
+ "kd_loss": 0.0011971323983743787,
1326
+ "learning_rate": 9.085931917007464e-06,
1327
+ "loss": 0.088,
1328
+ "step": 660,
1329
+ "student_loss": 0.0004724525206256658,
1330
+ "teacher_loss": 0.0
1331
+ },
1332
+ {
1333
+ "epoch": 0.19175317185697807,
1334
+ "grad_norm": 7.90625,
1335
+ "kd_loss": 0.01377791166305542,
1336
+ "learning_rate": 9.070316359366681e-06,
1337
+ "loss": 0.0904,
1338
+ "step": 665,
1339
+ "student_loss": 0.00045861524995416403,
1340
+ "teacher_loss": 0.0
1341
+ },
1342
+ {
1343
+ "epoch": 0.19319492502883506,
1344
+ "grad_norm": 36.5,
1345
+ "kd_loss": 1.8900201320648193,
1346
+ "learning_rate": 9.054582202764175e-06,
1347
+ "loss": 0.0963,
1348
+ "step": 670,
1349
+ "student_loss": 0.07230479270219803,
1350
+ "teacher_loss": 0.0
1351
+ },
1352
+ {
1353
+ "epoch": 0.19463667820069205,
1354
+ "grad_norm": 4.0,
1355
+ "kd_loss": 0.0026326398365199566,
1356
+ "learning_rate": 9.038729905654393e-06,
1357
+ "loss": 0.1095,
1358
+ "step": 675,
1359
+ "student_loss": 0.0004619761893991381,
1360
+ "teacher_loss": 0.0
1361
+ },
1362
+ {
1363
+ "epoch": 0.19607843137254902,
1364
+ "grad_norm": 10.5,
1365
+ "kd_loss": 0.18342895805835724,
1366
+ "learning_rate": 9.022759929934109e-06,
1367
+ "loss": 0.0867,
1368
+ "step": 680,
1369
+ "student_loss": 0.07520744204521179,
1370
+ "teacher_loss": 0.0
1371
+ },
1372
+ {
1373
+ "epoch": 0.197520184544406,
1374
+ "grad_norm": 18.5,
1375
+ "kd_loss": 0.0016399289015680552,
1376
+ "learning_rate": 9.006672740928952e-06,
1377
+ "loss": 0.1277,
1378
+ "step": 685,
1379
+ "student_loss": 0.041107166558504105,
1380
+ "teacher_loss": 0.0
1381
+ },
1382
+ {
1383
+ "epoch": 0.19896193771626297,
1384
+ "grad_norm": 22.5,
1385
+ "kd_loss": 0.8545010089874268,
1386
+ "learning_rate": 8.990468807379861e-06,
1387
+ "loss": 0.0946,
1388
+ "step": 690,
1389
+ "student_loss": 0.014823048375546932,
1390
+ "teacher_loss": 0.0
1391
+ },
1392
+ {
1393
+ "epoch": 0.20040369088811996,
1394
+ "grad_norm": 6.1875,
1395
+ "kd_loss": 0.0027170374523848295,
1396
+ "learning_rate": 8.974148601429426e-06,
1397
+ "loss": 0.1075,
1398
+ "step": 695,
1399
+ "student_loss": 0.005867379251867533,
1400
+ "teacher_loss": 0.0
1401
+ },
1402
+ {
1403
+ "epoch": 0.20184544405997693,
1404
+ "grad_norm": 13.6875,
1405
+ "kd_loss": 0.005027920473366976,
1406
+ "learning_rate": 8.957712598608123e-06,
1407
+ "loss": 0.1523,
1408
+ "step": 700,
1409
+ "student_loss": 0.0006474690162576735,
1410
+ "teacher_loss": 0.0
1411
+ },
1412
+ {
1413
+ "epoch": 0.20328719723183392,
1414
+ "grad_norm": 10.5625,
1415
+ "kd_loss": 0.018207697197794914,
1416
+ "learning_rate": 8.941161277820462e-06,
1417
+ "loss": 0.0628,
1418
+ "step": 705,
1419
+ "student_loss": 0.0015416023088619113,
1420
+ "teacher_loss": 0.0
1421
+ },
1422
+ {
1423
+ "epoch": 0.20472895040369088,
1424
+ "grad_norm": 5.15625,
1425
+ "kd_loss": 0.31215962767601013,
1426
+ "learning_rate": 8.924495121331038e-06,
1427
+ "loss": 0.099,
1428
+ "step": 710,
1429
+ "student_loss": 0.007331720553338528,
1430
+ "teacher_loss": 0.0
1431
+ },
1432
+ {
1433
+ "epoch": 0.20617070357554787,
1434
+ "grad_norm": 15.875,
1435
+ "kd_loss": 0.0036377429496496916,
1436
+ "learning_rate": 8.907714614750473e-06,
1437
+ "loss": 0.0935,
1438
+ "step": 715,
1439
+ "student_loss": 0.008250439539551735,
1440
+ "teacher_loss": 0.0
1441
+ },
1442
+ {
1443
+ "epoch": 0.20761245674740483,
1444
+ "grad_norm": 3.171875,
1445
+ "kd_loss": 0.04166782274842262,
1446
+ "learning_rate": 8.890820247021263e-06,
1447
+ "loss": 0.1256,
1448
+ "step": 720,
1449
+ "student_loss": 0.0010231786873191595,
1450
+ "teacher_loss": 0.0
1451
+ },
1452
+ {
1453
+ "epoch": 0.20905420991926182,
1454
+ "grad_norm": 10.8125,
1455
+ "kd_loss": 0.002339277183637023,
1456
+ "learning_rate": 8.873812510403545e-06,
1457
+ "loss": 0.0612,
1458
+ "step": 725,
1459
+ "student_loss": 0.00030273603624664247,
1460
+ "teacher_loss": 0.0
1461
+ },
1462
+ {
1463
+ "epoch": 0.2104959630911188,
1464
+ "grad_norm": 20.75,
1465
+ "kd_loss": 0.07314825803041458,
1466
+ "learning_rate": 8.85669190046074e-06,
1467
+ "loss": 0.1104,
1468
+ "step": 730,
1469
+ "student_loss": 0.021018339321017265,
1470
+ "teacher_loss": 0.0
1471
+ },
1472
+ {
1473
+ "epoch": 0.21193771626297578,
1474
+ "grad_norm": 5.53125,
1475
+ "kd_loss": 0.0017747314414009452,
1476
+ "learning_rate": 8.839458916045117e-06,
1477
+ "loss": 0.0881,
1478
+ "step": 735,
1479
+ "student_loss": 0.00029675825498998165,
1480
+ "teacher_loss": 0.0
1481
+ },
1482
+ {
1483
+ "epoch": 0.21337946943483277,
1484
+ "grad_norm": 4.6875,
1485
+ "kd_loss": 0.001755792647600174,
1486
+ "learning_rate": 8.822114059283267e-06,
1487
+ "loss": 0.0907,
1488
+ "step": 740,
1489
+ "student_loss": 0.004126196261495352,
1490
+ "teacher_loss": 0.0
1491
+ },
1492
+ {
1493
+ "epoch": 0.21482122260668973,
1494
+ "grad_norm": 20.625,
1495
+ "kd_loss": 0.001788739929907024,
1496
+ "learning_rate": 8.804657835561456e-06,
1497
+ "loss": 0.1042,
1498
+ "step": 745,
1499
+ "student_loss": 0.0402885265648365,
1500
+ "teacher_loss": 0.0
1501
+ },
1502
+ {
1503
+ "epoch": 0.21626297577854672,
1504
+ "grad_norm": 7.28125,
1505
+ "kd_loss": 0.7170178294181824,
1506
+ "learning_rate": 8.787090753510917e-06,
1507
+ "loss": 0.0756,
1508
+ "step": 750,
1509
+ "student_loss": 0.022320456802845,
1510
+ "teacher_loss": 0.0
1511
+ },
1512
+ {
1513
+ "epoch": 0.2177047289504037,
1514
+ "grad_norm": 12.875,
1515
+ "kd_loss": 0.0030846295412629843,
1516
+ "learning_rate": 8.76941332499301e-06,
1517
+ "loss": 0.1132,
1518
+ "step": 755,
1519
+ "student_loss": 0.03327662870287895,
1520
+ "teacher_loss": 0.0
1521
+ },
1522
+ {
1523
+ "epoch": 0.21914648212226068,
1524
+ "grad_norm": 12.4375,
1525
+ "kd_loss": 0.0014672816032543778,
1526
+ "learning_rate": 8.751626065084328e-06,
1527
+ "loss": 0.1147,
1528
+ "step": 760,
1529
+ "student_loss": 0.00921363290399313,
1530
+ "teacher_loss": 0.0
1531
+ },
1532
+ {
1533
+ "epoch": 0.22058823529411764,
1534
+ "grad_norm": 14.125,
1535
+ "kd_loss": 0.005435326602309942,
1536
+ "learning_rate": 8.733729492061675e-06,
1537
+ "loss": 0.0476,
1538
+ "step": 765,
1539
+ "student_loss": 0.004831620957702398,
1540
+ "teacher_loss": 0.0
1541
+ },
1542
+ {
1543
+ "epoch": 0.22202998846597463,
1544
+ "grad_norm": 4.0,
1545
+ "kd_loss": 0.03225019574165344,
1546
+ "learning_rate": 8.715724127386971e-06,
1547
+ "loss": 0.103,
1548
+ "step": 770,
1549
+ "student_loss": 0.0014119932893663645,
1550
+ "teacher_loss": 0.0
1551
+ },
1552
+ {
1553
+ "epoch": 0.2234717416378316,
1554
+ "grad_norm": 3.390625,
1555
+ "kd_loss": 0.10136047750711441,
1556
+ "learning_rate": 8.697610495692055e-06,
1557
+ "loss": 0.0591,
1558
+ "step": 775,
1559
+ "student_loss": 0.0035214233212172985,
1560
+ "teacher_loss": 0.0
1561
+ },
1562
+ {
1563
+ "epoch": 0.22491349480968859,
1564
+ "grad_norm": 21.25,
1565
+ "kd_loss": 0.0011202533496543765,
1566
+ "learning_rate": 8.679389124763396e-06,
1567
+ "loss": 0.1158,
1568
+ "step": 780,
1569
+ "student_loss": 0.0014409847790375352,
1570
+ "teacher_loss": 0.0
1571
+ },
1572
+ {
1573
+ "epoch": 0.22635524798154555,
1574
+ "grad_norm": 9.0,
1575
+ "kd_loss": 0.000793208135291934,
1576
+ "learning_rate": 8.661060545526724e-06,
1577
+ "loss": 0.0603,
1578
+ "step": 785,
1579
+ "student_loss": 0.0002954969822894782,
1580
+ "teacher_loss": 0.0
1581
+ },
1582
+ {
1583
+ "epoch": 0.22779700115340254,
1584
+ "grad_norm": 7.5,
1585
+ "kd_loss": 0.0026095786597579718,
1586
+ "learning_rate": 8.64262529203155e-06,
1587
+ "loss": 0.0422,
1588
+ "step": 790,
1589
+ "student_loss": 0.002638905541971326,
1590
+ "teacher_loss": 0.0
1591
+ },
1592
+ {
1593
+ "epoch": 0.2292387543252595,
1594
+ "grad_norm": 3.28125,
1595
+ "kd_loss": 0.002780693117529154,
1596
+ "learning_rate": 8.624083901435613e-06,
1597
+ "loss": 0.0766,
1598
+ "step": 795,
1599
+ "student_loss": 0.000536055420525372,
1600
+ "teacher_loss": 0.0
1601
+ },
1602
+ {
1603
+ "epoch": 0.2306805074971165,
1604
+ "grad_norm": 50.0,
1605
+ "kd_loss": 0.13559553027153015,
1606
+ "learning_rate": 8.60543691398922e-06,
1607
+ "loss": 0.1005,
1608
+ "step": 800,
1609
+ "student_loss": 0.06979604810476303,
1610
+ "teacher_loss": 0.0
1611
+ },
1612
+ {
1613
+ "epoch": 0.23212226066897348,
1614
+ "grad_norm": 35.75,
1615
+ "kd_loss": 0.00036787468707188964,
1616
+ "learning_rate": 8.586684873019513e-06,
1617
+ "loss": 0.1055,
1618
+ "step": 805,
1619
+ "student_loss": 0.0003365253796800971,
1620
+ "teacher_loss": 0.0
1621
+ },
1622
+ {
1623
+ "epoch": 0.23356401384083045,
1624
+ "grad_norm": 19.875,
1625
+ "kd_loss": 0.00150456337723881,
1626
+ "learning_rate": 8.567828324914633e-06,
1627
+ "loss": 0.0539,
1628
+ "step": 810,
1629
+ "student_loss": 0.0004286724142730236,
1630
+ "teacher_loss": 0.0
1631
+ },
1632
+ {
1633
+ "epoch": 0.23500576701268744,
1634
+ "grad_norm": 7.0625,
1635
+ "kd_loss": 0.007362797390669584,
1636
+ "learning_rate": 8.548867819107803e-06,
1637
+ "loss": 0.0949,
1638
+ "step": 815,
1639
+ "student_loss": 0.001179418177343905,
1640
+ "teacher_loss": 0.0
1641
+ },
1642
+ {
1643
+ "epoch": 0.2364475201845444,
1644
+ "grad_norm": 6.6875,
1645
+ "kd_loss": 0.001827954431064427,
1646
+ "learning_rate": 8.52980390806131e-06,
1647
+ "loss": 0.0525,
1648
+ "step": 820,
1649
+ "student_loss": 0.004497961141169071,
1650
+ "teacher_loss": 0.0
1651
+ },
1652
+ {
1653
+ "epoch": 0.2378892733564014,
1654
+ "grad_norm": 10.1875,
1655
+ "kd_loss": 0.0011187309864908457,
1656
+ "learning_rate": 8.51063714725042e-06,
1657
+ "loss": 0.1075,
1658
+ "step": 825,
1659
+ "student_loss": 0.000924907042644918,
1660
+ "teacher_loss": 0.0
1661
+ },
1662
+ {
1663
+ "epoch": 0.23933102652825836,
1664
+ "grad_norm": 12.0625,
1665
+ "kd_loss": 0.12930721044540405,
1666
+ "learning_rate": 8.49136809514719e-06,
1667
+ "loss": 0.101,
1668
+ "step": 830,
1669
+ "student_loss": 0.012681624852120876,
1670
+ "teacher_loss": 0.0
1671
+ },
1672
+ {
1673
+ "epoch": 0.24077277970011535,
1674
+ "grad_norm": 22.125,
1675
+ "kd_loss": 0.0007281019352376461,
1676
+ "learning_rate": 8.471997313204183e-06,
1677
+ "loss": 0.0645,
1678
+ "step": 835,
1679
+ "student_loss": 0.01504411082714796,
1680
+ "teacher_loss": 0.0
1681
+ },
1682
+ {
1683
+ "epoch": 0.2422145328719723,
1684
+ "grad_norm": 12.0,
1685
+ "kd_loss": 0.0006879305001348257,
1686
+ "learning_rate": 8.452525365838123e-06,
1687
+ "loss": 0.0615,
1688
+ "step": 840,
1689
+ "student_loss": 0.0004384240601211786,
1690
+ "teacher_loss": 0.0
1691
+ },
1692
+ {
1693
+ "epoch": 0.2436562860438293,
1694
+ "grad_norm": 11.125,
1695
+ "kd_loss": 0.001608243677765131,
1696
+ "learning_rate": 8.432952820413447e-06,
1697
+ "loss": 0.0914,
1698
+ "step": 845,
1699
+ "student_loss": 0.0005417144275270402,
1700
+ "teacher_loss": 0.0
1701
+ },
1702
+ {
1703
+ "epoch": 0.24509803921568626,
1704
+ "grad_norm": 33.25,
1705
+ "kd_loss": 0.1373538076877594,
1706
+ "learning_rate": 8.41328024722577e-06,
1707
+ "loss": 0.1133,
1708
+ "step": 850,
1709
+ "student_loss": 0.01142928283661604,
1710
+ "teacher_loss": 0.0
1711
+ },
1712
+ {
1713
+ "epoch": 0.24653979238754326,
1714
+ "grad_norm": 10.3125,
1715
+ "kd_loss": 0.0016314623644575477,
1716
+ "learning_rate": 8.393508219485263e-06,
1717
+ "loss": 0.1024,
1718
+ "step": 855,
1719
+ "student_loss": 0.00022962426010053605,
1720
+ "teacher_loss": 0.0
1721
+ },
1722
+ {
1723
+ "epoch": 0.24798154555940022,
1724
+ "grad_norm": 5.90625,
1725
+ "kd_loss": 0.0019791526719927788,
1726
+ "learning_rate": 8.373637313299966e-06,
1727
+ "loss": 0.1224,
1728
+ "step": 860,
1729
+ "student_loss": 0.0005864988779649138,
1730
+ "teacher_loss": 0.0
1731
+ },
1732
+ {
1733
+ "epoch": 0.2494232987312572,
1734
+ "grad_norm": 9.875,
1735
+ "kd_loss": 0.000920691411010921,
1736
+ "learning_rate": 8.353668107658984e-06,
1737
+ "loss": 0.0633,
1738
+ "step": 865,
1739
+ "student_loss": 0.0002618279540911317,
1740
+ "teacher_loss": 0.0
1741
+ },
1742
+ {
1743
+ "epoch": 0.2508650519031142,
1744
+ "grad_norm": 75.0,
1745
+ "kd_loss": 0.41142162680625916,
1746
+ "learning_rate": 8.333601184415632e-06,
1747
+ "loss": 0.0753,
1748
+ "step": 870,
1749
+ "student_loss": 0.008613992482423782,
1750
+ "teacher_loss": 0.0
1751
+ },
1752
+ {
1753
+ "epoch": 0.25230680507497116,
1754
+ "grad_norm": 24.125,
1755
+ "kd_loss": 0.0006664465181529522,
1756
+ "learning_rate": 8.31343712827047e-06,
1757
+ "loss": 0.0742,
1758
+ "step": 875,
1759
+ "student_loss": 0.00016053368744906038,
1760
+ "teacher_loss": 0.0
1761
+ },
1762
+ {
1763
+ "epoch": 0.2537485582468281,
1764
+ "grad_norm": 5.84375,
1765
+ "kd_loss": 0.006322797387838364,
1766
+ "learning_rate": 8.293176526754274e-06,
1767
+ "loss": 0.1238,
1768
+ "step": 880,
1769
+ "student_loss": 0.000416541617596522,
1770
+ "teacher_loss": 0.0
1771
+ },
1772
+ {
1773
+ "epoch": 0.25519031141868515,
1774
+ "grad_norm": 12.75,
1775
+ "kd_loss": 0.19639839231967926,
1776
+ "learning_rate": 8.27281997021091e-06,
1777
+ "loss": 0.0568,
1778
+ "step": 885,
1779
+ "student_loss": 0.007913337089121342,
1780
+ "teacher_loss": 0.0
1781
+ },
1782
+ {
1783
+ "epoch": 0.2566320645905421,
1784
+ "grad_norm": 7.15625,
1785
+ "kd_loss": 0.001160801388323307,
1786
+ "learning_rate": 8.25236805178014e-06,
1787
+ "loss": 0.0727,
1788
+ "step": 890,
1789
+ "student_loss": 0.0003532449482008815,
1790
+ "teacher_loss": 0.0
1791
+ },
1792
+ {
1793
+ "epoch": 0.25807381776239907,
1794
+ "grad_norm": 12.6875,
1795
+ "kd_loss": 0.19797247648239136,
1796
+ "learning_rate": 8.231821367380335e-06,
1797
+ "loss": 0.128,
1798
+ "step": 895,
1799
+ "student_loss": 0.003061032621189952,
1800
+ "teacher_loss": 0.0
1801
+ },
1802
+ {
1803
+ "epoch": 0.25951557093425603,
1804
+ "grad_norm": 8.0,
1805
+ "kd_loss": 0.007570919580757618,
1806
+ "learning_rate": 8.211180515691108e-06,
1807
+ "loss": 0.0792,
1808
+ "step": 900,
1809
+ "student_loss": 0.033695392310619354,
1810
+ "teacher_loss": 0.0
1811
+ },
1812
+ {
1813
+ "epoch": 0.26095732410611305,
1814
+ "grad_norm": 11.3125,
1815
+ "kd_loss": 0.38567981123924255,
1816
+ "learning_rate": 8.190446098135883e-06,
1817
+ "loss": 0.0922,
1818
+ "step": 905,
1819
+ "student_loss": 0.028850680217146873,
1820
+ "teacher_loss": 0.0
1821
+ },
1822
+ {
1823
+ "epoch": 0.26239907727797,
1824
+ "grad_norm": 10.9375,
1825
+ "kd_loss": 0.0007043812656775117,
1826
+ "learning_rate": 8.16961871886435e-06,
1827
+ "loss": 0.0954,
1828
+ "step": 910,
1829
+ "student_loss": 0.0002878632803913206,
1830
+ "teacher_loss": 0.0
1831
+ },
1832
+ {
1833
+ "epoch": 0.263840830449827,
1834
+ "grad_norm": 8.1875,
1835
+ "kd_loss": 0.14511993527412415,
1836
+ "learning_rate": 8.148698984734887e-06,
1837
+ "loss": 0.0477,
1838
+ "step": 915,
1839
+ "student_loss": 0.0008710387046448886,
1840
+ "teacher_loss": 0.0
1841
+ },
1842
+ {
1843
+ "epoch": 0.26528258362168394,
1844
+ "grad_norm": 8.5,
1845
+ "kd_loss": 0.0384184792637825,
1846
+ "learning_rate": 8.127687505296852e-06,
1847
+ "loss": 0.0797,
1848
+ "step": 920,
1849
+ "student_loss": 0.016883913427591324,
1850
+ "teacher_loss": 0.0
1851
+ },
1852
+ {
1853
+ "epoch": 0.26672433679354096,
1854
+ "grad_norm": 4.4375,
1855
+ "kd_loss": 0.00027013575891032815,
1856
+ "learning_rate": 8.106584892772844e-06,
1857
+ "loss": 0.0756,
1858
+ "step": 925,
1859
+ "student_loss": 0.0002693244314286858,
1860
+ "teacher_loss": 0.0
1861
+ },
1862
+ {
1863
+ "epoch": 0.2681660899653979,
1864
+ "grad_norm": 10.0,
1865
+ "kd_loss": 0.0005406018462963402,
1866
+ "learning_rate": 8.08539176204085e-06,
1867
+ "loss": 0.0681,
1868
+ "step": 930,
1869
+ "student_loss": 0.025998031720519066,
1870
+ "teacher_loss": 0.0
1871
+ },
1872
+ {
1873
+ "epoch": 0.2696078431372549,
1874
+ "grad_norm": 7.84375,
1875
+ "kd_loss": 0.07199026644229889,
1876
+ "learning_rate": 8.064108730616336e-06,
1877
+ "loss": 0.0594,
1878
+ "step": 935,
1879
+ "student_loss": 0.0006435631075873971,
1880
+ "teacher_loss": 0.0
1881
+ },
1882
+ {
1883
+ "epoch": 0.2710495963091119,
1884
+ "grad_norm": 6.0625,
1885
+ "kd_loss": 0.006361419800668955,
1886
+ "learning_rate": 8.042736418634252e-06,
1887
+ "loss": 0.0734,
1888
+ "step": 940,
1889
+ "student_loss": 0.012877441942691803,
1890
+ "teacher_loss": 0.0
1891
+ },
1892
+ {
1893
+ "epoch": 0.27249134948096887,
1894
+ "grad_norm": 25.25,
1895
+ "kd_loss": 0.009065474383533001,
1896
+ "learning_rate": 8.02127544883096e-06,
1897
+ "loss": 0.0763,
1898
+ "step": 945,
1899
+ "student_loss": 0.007113016210496426,
1900
+ "teacher_loss": 0.0
1901
+ },
1902
+ {
1903
+ "epoch": 0.27393310265282583,
1904
+ "grad_norm": 1.890625,
1905
+ "kd_loss": 0.0012218133779242635,
1906
+ "learning_rate": 7.999726446526095e-06,
1907
+ "loss": 0.0542,
1908
+ "step": 950,
1909
+ "student_loss": 0.00028497542371042073,
1910
+ "teacher_loss": 0.0
1911
+ },
1912
+ {
1913
+ "epoch": 0.2753748558246828,
1914
+ "grad_norm": 8.8125,
1915
+ "kd_loss": 0.0014024431584402919,
1916
+ "learning_rate": 7.978090039604342e-06,
1917
+ "loss": 0.0548,
1918
+ "step": 955,
1919
+ "student_loss": 0.0005670087411999702,
1920
+ "teacher_loss": 0.0
1921
+ },
1922
+ {
1923
+ "epoch": 0.2768166089965398,
1924
+ "grad_norm": 26.5,
1925
+ "kd_loss": 0.006449506618082523,
1926
+ "learning_rate": 7.956366858497138e-06,
1927
+ "loss": 0.0734,
1928
+ "step": 960,
1929
+ "student_loss": 0.0003615773457568139,
1930
+ "teacher_loss": 0.0
1931
+ },
1932
+ {
1933
+ "epoch": 0.2782583621683968,
1934
+ "grad_norm": 16.625,
1935
+ "kd_loss": 0.001059827976860106,
1936
+ "learning_rate": 7.934557536164304e-06,
1937
+ "loss": 0.1002,
1938
+ "step": 965,
1939
+ "student_loss": 0.012110190466046333,
1940
+ "teacher_loss": 0.0
1941
+ },
1942
+ {
1943
+ "epoch": 0.27970011534025374,
1944
+ "grad_norm": 23.375,
1945
+ "kd_loss": 0.0003945316420868039,
1946
+ "learning_rate": 7.91266270807561e-06,
1947
+ "loss": 0.0942,
1948
+ "step": 970,
1949
+ "student_loss": 0.00027471804060041904,
1950
+ "teacher_loss": 0.0
1951
+ },
1952
+ {
1953
+ "epoch": 0.2811418685121107,
1954
+ "grad_norm": 12.0,
1955
+ "kd_loss": 0.006996847223490477,
1956
+ "learning_rate": 7.890683012192249e-06,
1957
+ "loss": 0.1016,
1958
+ "step": 975,
1959
+ "student_loss": 0.020073270425200462,
1960
+ "teacher_loss": 0.0
1961
+ },
1962
+ {
1963
+ "epoch": 0.2825836216839677,
1964
+ "grad_norm": 2.34375,
1965
+ "kd_loss": 0.0010718721896409988,
1966
+ "learning_rate": 7.868619088948246e-06,
1967
+ "loss": 0.0668,
1968
+ "step": 980,
1969
+ "student_loss": 0.000354983116267249,
1970
+ "teacher_loss": 0.0
1971
+ },
1972
+ {
1973
+ "epoch": 0.2840253748558247,
1974
+ "grad_norm": 7.9375,
1975
+ "kd_loss": 0.000552264042198658,
1976
+ "learning_rate": 7.846471581231814e-06,
1977
+ "loss": 0.1767,
1978
+ "step": 985,
1979
+ "student_loss": 0.000291457399725914,
1980
+ "teacher_loss": 0.0
1981
+ },
1982
+ {
1983
+ "epoch": 0.28546712802768165,
1984
+ "grad_norm": 21.375,
1985
+ "kd_loss": 0.0026221515145152807,
1986
+ "learning_rate": 7.824241134366604e-06,
1987
+ "loss": 0.0533,
1988
+ "step": 990,
1989
+ "student_loss": 0.00033598451409488916,
1990
+ "teacher_loss": 0.0
1991
+ },
1992
+ {
1993
+ "epoch": 0.2869088811995386,
1994
+ "grad_norm": 16.375,
1995
+ "kd_loss": 0.014461357146501541,
1996
+ "learning_rate": 7.801928396092908e-06,
1997
+ "loss": 0.0847,
1998
+ "step": 995,
1999
+ "student_loss": 0.006032541859894991,
2000
+ "teacher_loss": 0.0
2001
+ },
2002
+ {
2003
+ "epoch": 0.28835063437139563,
2004
+ "grad_norm": 3.90625,
2005
+ "kd_loss": 0.005895162001252174,
2006
+ "learning_rate": 7.779534016548791e-06,
2007
+ "loss": 0.0785,
2008
+ "step": 1000,
2009
+ "student_loss": 0.005668016150593758,
2010
+ "teacher_loss": 0.0
2011
+ }
2012
+ ],
2013
+ "logging_steps": 5,
2014
+ "max_steps": 3000,
2015
+ "num_input_tokens_seen": 0,
2016
+ "num_train_epochs": 1,
2017
+ "save_steps": 500,
2018
+ "stateful_callbacks": {
2019
+ "TrainerControl": {
2020
+ "args": {
2021
+ "should_epoch_stop": false,
2022
+ "should_evaluate": false,
2023
+ "should_log": false,
2024
+ "should_save": true,
2025
+ "should_training_stop": false
2026
+ },
2027
+ "attributes": {}
2028
+ }
2029
+ },
2030
+ "total_flos": 0.0,
2031
+ "train_batch_size": 1,
2032
+ "trial_name": null,
2033
+ "trial_params": null
2034
+ }
checkpoints/codi_frozen_1.5b_logit/checkpoint-1000/vocab.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoints/codi_frozen_1.5b_logit/checkpoint-500/added_tokens.json ADDED
@@ -0,0 +1,35 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "</tool_call>": 151658,
3
+ "<tool_call>": 151657,
4
+ "<|action_sep|>": 151670,
5
+ "<|arg_sep|>": 151671,
6
+ "<|box_end|>": 151649,
7
+ "<|box_start|>": 151648,
8
+ "<|call_sep|>": 151666,
9
+ "<|end_of_text|>": 151673,
10
+ "<|endoftext|>": 151643,
11
+ "<|exception_sep|>": 151669,
12
+ "<|file_sep|>": 151664,
13
+ "<|fim_middle|>": 151660,
14
+ "<|fim_pad|>": 151662,
15
+ "<|fim_prefix|>": 151659,
16
+ "<|fim_suffix|>": 151661,
17
+ "<|frame_sep|>": 151672,
18
+ "<|im_end|>": 151645,
19
+ "<|im_start|>": 151644,
20
+ "<|image_pad|>": 151655,
21
+ "<|latent_end|>": 151675,
22
+ "<|latent_start|>": 151674,
23
+ "<|line_sep|>": 151667,
24
+ "<|object_ref_end|>": 151647,
25
+ "<|object_ref_start|>": 151646,
26
+ "<|quad_end|>": 151651,
27
+ "<|quad_start|>": 151650,
28
+ "<|repo_name|>": 151663,
29
+ "<|return_sep|>": 151668,
30
+ "<|trace_context_start|>": 151665,
31
+ "<|video_pad|>": 151656,
32
+ "<|vision_end|>": 151653,
33
+ "<|vision_pad|>": 151654,
34
+ "<|vision_start|>": 151652
35
+ }
checkpoints/codi_frozen_1.5b_logit/checkpoint-500/chat_template.jinja ADDED
@@ -0,0 +1,54 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {%- if tools %}
2
+ {{- '<|im_start|>system\n' }}
3
+ {%- if messages[0]['role'] == 'system' %}
4
+ {{- messages[0]['content'] }}
5
+ {%- else %}
6
+ {{- 'You are Qwen, created by Alibaba Cloud. You are a helpful assistant.' }}
7
+ {%- endif %}
8
+ {{- "\n\n# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
9
+ {%- for tool in tools %}
10
+ {{- "\n" }}
11
+ {{- tool | tojson }}
12
+ {%- endfor %}
13
+ {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
14
+ {%- else %}
15
+ {%- if messages[0]['role'] == 'system' %}
16
+ {{- '<|im_start|>system\n' + messages[0]['content'] + '<|im_end|>\n' }}
17
+ {%- else %}
18
+ {{- '<|im_start|>system\nYou are Qwen, created by Alibaba Cloud. You are a helpful assistant.<|im_end|>\n' }}
19
+ {%- endif %}
20
+ {%- endif %}
21
+ {%- for message in messages %}
22
+ {%- if (message.role == "user") or (message.role == "system" and not loop.first) or (message.role == "assistant" and not message.tool_calls) %}
23
+ {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
24
+ {%- elif message.role == "assistant" %}
25
+ {{- '<|im_start|>' + message.role }}
26
+ {%- if message.content %}
27
+ {{- '\n' + message.content }}
28
+ {%- endif %}
29
+ {%- for tool_call in message.tool_calls %}
30
+ {%- if tool_call.function is defined %}
31
+ {%- set tool_call = tool_call.function %}
32
+ {%- endif %}
33
+ {{- '\n<tool_call>\n{"name": "' }}
34
+ {{- tool_call.name }}
35
+ {{- '", "arguments": ' }}
36
+ {{- tool_call.arguments | tojson }}
37
+ {{- '}\n</tool_call>' }}
38
+ {%- endfor %}
39
+ {{- '<|im_end|>\n' }}
40
+ {%- elif message.role == "tool" %}
41
+ {%- if (loop.index0 == 0) or (messages[loop.index0 - 1].role != "tool") %}
42
+ {{- '<|im_start|>user' }}
43
+ {%- endif %}
44
+ {{- '\n<tool_response>\n' }}
45
+ {{- message.content }}
46
+ {{- '\n</tool_response>' }}
47
+ {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
48
+ {{- '<|im_end|>\n' }}
49
+ {%- endif %}
50
+ {%- endif %}
51
+ {%- endfor %}
52
+ {%- if add_generation_prompt %}
53
+ {{- '<|im_start|>assistant\n' }}
54
+ {%- endif %}
checkpoints/codi_frozen_1.5b_logit/checkpoint-500/config.json ADDED
@@ -0,0 +1,58 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Qwen2ForCausalLM"
4
+ ],
5
+ "attention_dropout": 0.0,
6
+ "dtype": "bfloat16",
7
+ "eos_token_id": 151643,
8
+ "hidden_act": "silu",
9
+ "hidden_size": 1536,
10
+ "initializer_range": 0.02,
11
+ "intermediate_size": 8960,
12
+ "layer_types": [
13
+ "full_attention",
14
+ "full_attention",
15
+ "full_attention",
16
+ "full_attention",
17
+ "full_attention",
18
+ "full_attention",
19
+ "full_attention",
20
+ "full_attention",
21
+ "full_attention",
22
+ "full_attention",
23
+ "full_attention",
24
+ "full_attention",
25
+ "full_attention",
26
+ "full_attention",
27
+ "full_attention",
28
+ "full_attention",
29
+ "full_attention",
30
+ "full_attention",
31
+ "full_attention",
32
+ "full_attention",
33
+ "full_attention",
34
+ "full_attention",
35
+ "full_attention",
36
+ "full_attention",
37
+ "full_attention",
38
+ "full_attention",
39
+ "full_attention",
40
+ "full_attention"
41
+ ],
42
+ "max_position_embeddings": 32768,
43
+ "max_window_layers": 28,
44
+ "model_type": "qwen2",
45
+ "num_attention_heads": 12,
46
+ "num_hidden_layers": 28,
47
+ "num_key_value_heads": 2,
48
+ "pad_token_id": 151643,
49
+ "rms_norm_eps": 1e-06,
50
+ "rope_scaling": null,
51
+ "rope_theta": 1000000.0,
52
+ "sliding_window": null,
53
+ "tie_word_embeddings": true,
54
+ "transformers_version": "4.57.6",
55
+ "use_cache": true,
56
+ "use_sliding_window": false,
57
+ "vocab_size": 151676
58
+ }
checkpoints/codi_frozen_1.5b_logit/checkpoint-500/merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
checkpoints/codi_frozen_1.5b_logit/checkpoint-500/pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cc9ec41de66086cd428438196a2b8b65093fa7df242b033a7080a51605b7637f
3
+ size 3096275515
checkpoints/codi_frozen_1.5b_logit/checkpoint-500/special_tokens_map.json ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "additional_special_tokens": [
3
+ "<|im_start|>",
4
+ "<|im_end|>",
5
+ "<|object_ref_start|>",
6
+ "<|object_ref_end|>",
7
+ "<|box_start|>",
8
+ "<|box_end|>",
9
+ "<|quad_start|>",
10
+ "<|quad_end|>",
11
+ "<|vision_start|>",
12
+ "<|vision_end|>",
13
+ "<|vision_pad|>",
14
+ "<|image_pad|>",
15
+ "<|video_pad|>"
16
+ ],
17
+ "eos_token": {
18
+ "content": "<|endoftext|>",
19
+ "lstrip": false,
20
+ "normalized": false,
21
+ "rstrip": false,
22
+ "single_word": false
23
+ },
24
+ "pad_token": {
25
+ "content": "<|endoftext|>",
26
+ "lstrip": false,
27
+ "normalized": false,
28
+ "rstrip": false,
29
+ "single_word": false
30
+ }
31
+ }
checkpoints/codi_frozen_1.5b_logit/checkpoint-500/thought_projector.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a57da98eb4f7994bc0b45c2d2c5d4681bfe3d7940cb83bad0ef5cb8b5916d91e
3
+ size 9445953
checkpoints/codi_frozen_1.5b_logit/checkpoint-500/tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:83a790d654474f5dfe225f889afd0210313eb1083f942671f2c4b8e95a1c922b
3
+ size 11424004
checkpoints/codi_frozen_1.5b_logit/checkpoint-500/tokenizer_config.json ADDED
@@ -0,0 +1,295 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_bos_token": false,
3
+ "add_prefix_space": false,
4
+ "added_tokens_decoder": {
5
+ "151643": {
6
+ "content": "<|endoftext|>",
7
+ "lstrip": false,
8
+ "normalized": false,
9
+ "rstrip": false,
10
+ "single_word": false,
11
+ "special": true
12
+ },
13
+ "151644": {
14
+ "content": "<|im_start|>",
15
+ "lstrip": false,
16
+ "normalized": false,
17
+ "rstrip": false,
18
+ "single_word": false,
19
+ "special": true
20
+ },
21
+ "151645": {
22
+ "content": "<|im_end|>",
23
+ "lstrip": false,
24
+ "normalized": false,
25
+ "rstrip": false,
26
+ "single_word": false,
27
+ "special": true
28
+ },
29
+ "151646": {
30
+ "content": "<|object_ref_start|>",
31
+ "lstrip": false,
32
+ "normalized": false,
33
+ "rstrip": false,
34
+ "single_word": false,
35
+ "special": true
36
+ },
37
+ "151647": {
38
+ "content": "<|object_ref_end|>",
39
+ "lstrip": false,
40
+ "normalized": false,
41
+ "rstrip": false,
42
+ "single_word": false,
43
+ "special": true
44
+ },
45
+ "151648": {
46
+ "content": "<|box_start|>",
47
+ "lstrip": false,
48
+ "normalized": false,
49
+ "rstrip": false,
50
+ "single_word": false,
51
+ "special": true
52
+ },
53
+ "151649": {
54
+ "content": "<|box_end|>",
55
+ "lstrip": false,
56
+ "normalized": false,
57
+ "rstrip": false,
58
+ "single_word": false,
59
+ "special": true
60
+ },
61
+ "151650": {
62
+ "content": "<|quad_start|>",
63
+ "lstrip": false,
64
+ "normalized": false,
65
+ "rstrip": false,
66
+ "single_word": false,
67
+ "special": true
68
+ },
69
+ "151651": {
70
+ "content": "<|quad_end|>",
71
+ "lstrip": false,
72
+ "normalized": false,
73
+ "rstrip": false,
74
+ "single_word": false,
75
+ "special": true
76
+ },
77
+ "151652": {
78
+ "content": "<|vision_start|>",
79
+ "lstrip": false,
80
+ "normalized": false,
81
+ "rstrip": false,
82
+ "single_word": false,
83
+ "special": true
84
+ },
85
+ "151653": {
86
+ "content": "<|vision_end|>",
87
+ "lstrip": false,
88
+ "normalized": false,
89
+ "rstrip": false,
90
+ "single_word": false,
91
+ "special": true
92
+ },
93
+ "151654": {
94
+ "content": "<|vision_pad|>",
95
+ "lstrip": false,
96
+ "normalized": false,
97
+ "rstrip": false,
98
+ "single_word": false,
99
+ "special": true
100
+ },
101
+ "151655": {
102
+ "content": "<|image_pad|>",
103
+ "lstrip": false,
104
+ "normalized": false,
105
+ "rstrip": false,
106
+ "single_word": false,
107
+ "special": true
108
+ },
109
+ "151656": {
110
+ "content": "<|video_pad|>",
111
+ "lstrip": false,
112
+ "normalized": false,
113
+ "rstrip": false,
114
+ "single_word": false,
115
+ "special": true
116
+ },
117
+ "151657": {
118
+ "content": "<tool_call>",
119
+ "lstrip": false,
120
+ "normalized": false,
121
+ "rstrip": false,
122
+ "single_word": false,
123
+ "special": false
124
+ },
125
+ "151658": {
126
+ "content": "</tool_call>",
127
+ "lstrip": false,
128
+ "normalized": false,
129
+ "rstrip": false,
130
+ "single_word": false,
131
+ "special": false
132
+ },
133
+ "151659": {
134
+ "content": "<|fim_prefix|>",
135
+ "lstrip": false,
136
+ "normalized": false,
137
+ "rstrip": false,
138
+ "single_word": false,
139
+ "special": false
140
+ },
141
+ "151660": {
142
+ "content": "<|fim_middle|>",
143
+ "lstrip": false,
144
+ "normalized": false,
145
+ "rstrip": false,
146
+ "single_word": false,
147
+ "special": false
148
+ },
149
+ "151661": {
150
+ "content": "<|fim_suffix|>",
151
+ "lstrip": false,
152
+ "normalized": false,
153
+ "rstrip": false,
154
+ "single_word": false,
155
+ "special": false
156
+ },
157
+ "151662": {
158
+ "content": "<|fim_pad|>",
159
+ "lstrip": false,
160
+ "normalized": false,
161
+ "rstrip": false,
162
+ "single_word": false,
163
+ "special": false
164
+ },
165
+ "151663": {
166
+ "content": "<|repo_name|>",
167
+ "lstrip": false,
168
+ "normalized": false,
169
+ "rstrip": false,
170
+ "single_word": false,
171
+ "special": false
172
+ },
173
+ "151664": {
174
+ "content": "<|file_sep|>",
175
+ "lstrip": false,
176
+ "normalized": false,
177
+ "rstrip": false,
178
+ "single_word": false,
179
+ "special": false
180
+ },
181
+ "151665": {
182
+ "content": "<|trace_context_start|>",
183
+ "lstrip": false,
184
+ "normalized": false,
185
+ "rstrip": false,
186
+ "single_word": false,
187
+ "special": true
188
+ },
189
+ "151666": {
190
+ "content": "<|call_sep|>",
191
+ "lstrip": false,
192
+ "normalized": false,
193
+ "rstrip": false,
194
+ "single_word": false,
195
+ "special": true
196
+ },
197
+ "151667": {
198
+ "content": "<|line_sep|>",
199
+ "lstrip": false,
200
+ "normalized": false,
201
+ "rstrip": false,
202
+ "single_word": false,
203
+ "special": true
204
+ },
205
+ "151668": {
206
+ "content": "<|return_sep|>",
207
+ "lstrip": false,
208
+ "normalized": false,
209
+ "rstrip": false,
210
+ "single_word": false,
211
+ "special": true
212
+ },
213
+ "151669": {
214
+ "content": "<|exception_sep|>",
215
+ "lstrip": false,
216
+ "normalized": false,
217
+ "rstrip": false,
218
+ "single_word": false,
219
+ "special": true
220
+ },
221
+ "151670": {
222
+ "content": "<|action_sep|>",
223
+ "lstrip": false,
224
+ "normalized": false,
225
+ "rstrip": false,
226
+ "single_word": false,
227
+ "special": true
228
+ },
229
+ "151671": {
230
+ "content": "<|arg_sep|>",
231
+ "lstrip": false,
232
+ "normalized": false,
233
+ "rstrip": false,
234
+ "single_word": false,
235
+ "special": true
236
+ },
237
+ "151672": {
238
+ "content": "<|frame_sep|>",
239
+ "lstrip": false,
240
+ "normalized": false,
241
+ "rstrip": false,
242
+ "single_word": false,
243
+ "special": true
244
+ },
245
+ "151673": {
246
+ "content": "<|end_of_text|>",
247
+ "lstrip": false,
248
+ "normalized": false,
249
+ "rstrip": false,
250
+ "single_word": false,
251
+ "special": true
252
+ },
253
+ "151674": {
254
+ "content": "<|latent_start|>",
255
+ "lstrip": false,
256
+ "normalized": false,
257
+ "rstrip": false,
258
+ "single_word": false,
259
+ "special": true
260
+ },
261
+ "151675": {
262
+ "content": "<|latent_end|>",
263
+ "lstrip": false,
264
+ "normalized": false,
265
+ "rstrip": false,
266
+ "single_word": false,
267
+ "special": true
268
+ }
269
+ },
270
+ "additional_special_tokens": [
271
+ "<|im_start|>",
272
+ "<|im_end|>",
273
+ "<|object_ref_start|>",
274
+ "<|object_ref_end|>",
275
+ "<|box_start|>",
276
+ "<|box_end|>",
277
+ "<|quad_start|>",
278
+ "<|quad_end|>",
279
+ "<|vision_start|>",
280
+ "<|vision_end|>",
281
+ "<|vision_pad|>",
282
+ "<|image_pad|>",
283
+ "<|video_pad|>"
284
+ ],
285
+ "bos_token": null,
286
+ "clean_up_tokenization_spaces": false,
287
+ "eos_token": "<|endoftext|>",
288
+ "errors": "replace",
289
+ "extra_special_tokens": {},
290
+ "model_max_length": 32768,
291
+ "pad_token": "<|endoftext|>",
292
+ "split_special_tokens": false,
293
+ "tokenizer_class": "Qwen2Tokenizer",
294
+ "unk_token": null
295
+ }
checkpoints/codi_frozen_1.5b_logit/checkpoint-500/trainer_state.json ADDED
@@ -0,0 +1,1034 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 0.14417531718569782,
6
+ "eval_steps": 500,
7
+ "global_step": 500,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.0014417531718569781,
14
+ "grad_norm": 1952.0,
15
+ "kd_loss": 1.354009985923767,
16
+ "learning_rate": 4.444444444444445e-07,
17
+ "loss": 4.4194,
18
+ "step": 5,
19
+ "student_loss": 0.10154028236865997,
20
+ "teacher_loss": 0.0
21
+ },
22
+ {
23
+ "epoch": 0.0028835063437139563,
24
+ "grad_norm": 768.0,
25
+ "kd_loss": 2.1859679222106934,
26
+ "learning_rate": 1.0000000000000002e-06,
27
+ "loss": 4.072,
28
+ "step": 10,
29
+ "student_loss": 0.23161625862121582,
30
+ "teacher_loss": 0.0
31
+ },
32
+ {
33
+ "epoch": 0.004325259515570935,
34
+ "grad_norm": 676.0,
35
+ "kd_loss": 0.8974658250808716,
36
+ "learning_rate": 1.5555555555555558e-06,
37
+ "loss": 3.0092,
38
+ "step": 15,
39
+ "student_loss": 0.07354968786239624,
40
+ "teacher_loss": 0.0
41
+ },
42
+ {
43
+ "epoch": 0.0057670126874279125,
44
+ "grad_norm": 784.0,
45
+ "kd_loss": 5.459206581115723,
46
+ "learning_rate": 2.1111111111111114e-06,
47
+ "loss": 2.0182,
48
+ "step": 20,
49
+ "student_loss": 0.24417075514793396,
50
+ "teacher_loss": 0.0
51
+ },
52
+ {
53
+ "epoch": 0.00720876585928489,
54
+ "grad_norm": 132.0,
55
+ "kd_loss": 0.039198435842990875,
56
+ "learning_rate": 2.666666666666667e-06,
57
+ "loss": 0.9093,
58
+ "step": 25,
59
+ "student_loss": 0.3775336444377899,
60
+ "teacher_loss": 0.0
61
+ },
62
+ {
63
+ "epoch": 0.00865051903114187,
64
+ "grad_norm": 82.0,
65
+ "kd_loss": 0.01129975076764822,
66
+ "learning_rate": 3.2222222222222227e-06,
67
+ "loss": 0.6102,
68
+ "step": 30,
69
+ "student_loss": 0.48634102940559387,
70
+ "teacher_loss": 0.0
71
+ },
72
+ {
73
+ "epoch": 0.010092272202998846,
74
+ "grad_norm": 67.5,
75
+ "kd_loss": 0.01766686514019966,
76
+ "learning_rate": 3.777777777777778e-06,
77
+ "loss": 0.5753,
78
+ "step": 35,
79
+ "student_loss": 0.16448530554771423,
80
+ "teacher_loss": 0.0
81
+ },
82
+ {
83
+ "epoch": 0.011534025374855825,
84
+ "grad_norm": 66.0,
85
+ "kd_loss": 0.02908974140882492,
86
+ "learning_rate": 4.333333333333334e-06,
87
+ "loss": 0.441,
88
+ "step": 40,
89
+ "student_loss": 0.11244617402553558,
90
+ "teacher_loss": 0.0
91
+ },
92
+ {
93
+ "epoch": 0.012975778546712802,
94
+ "grad_norm": 68.0,
95
+ "kd_loss": 0.2916097640991211,
96
+ "learning_rate": 4.888888888888889e-06,
97
+ "loss": 0.2934,
98
+ "step": 45,
99
+ "student_loss": 0.07623665779829025,
100
+ "teacher_loss": 0.0
101
+ },
102
+ {
103
+ "epoch": 0.01441753171856978,
104
+ "grad_norm": 88.5,
105
+ "kd_loss": 0.19340172410011292,
106
+ "learning_rate": 5.444444444444445e-06,
107
+ "loss": 0.3841,
108
+ "step": 50,
109
+ "student_loss": 0.009455127641558647,
110
+ "teacher_loss": 0.0
111
+ },
112
+ {
113
+ "epoch": 0.015859284890426758,
114
+ "grad_norm": 115.0,
115
+ "kd_loss": 0.01024007610976696,
116
+ "learning_rate": 6e-06,
117
+ "loss": 0.2854,
118
+ "step": 55,
119
+ "student_loss": 0.0043172105215489864,
120
+ "teacher_loss": 0.0
121
+ },
122
+ {
123
+ "epoch": 0.01730103806228374,
124
+ "grad_norm": 57.75,
125
+ "kd_loss": 0.0018813807982951403,
126
+ "learning_rate": 6.555555555555556e-06,
127
+ "loss": 0.2133,
128
+ "step": 60,
129
+ "student_loss": 0.012564695440232754,
130
+ "teacher_loss": 0.0
131
+ },
132
+ {
133
+ "epoch": 0.018742791234140715,
134
+ "grad_norm": 56.0,
135
+ "kd_loss": 0.0018871173961088061,
136
+ "learning_rate": 7.111111111111112e-06,
137
+ "loss": 0.2158,
138
+ "step": 65,
139
+ "student_loss": 0.050387192517519,
140
+ "teacher_loss": 0.0
141
+ },
142
+ {
143
+ "epoch": 0.020184544405997693,
144
+ "grad_norm": 28.625,
145
+ "kd_loss": 0.4643922448158264,
146
+ "learning_rate": 7.666666666666667e-06,
147
+ "loss": 0.1818,
148
+ "step": 70,
149
+ "student_loss": 0.023501692339777946,
150
+ "teacher_loss": 0.0
151
+ },
152
+ {
153
+ "epoch": 0.02162629757785467,
154
+ "grad_norm": 43.0,
155
+ "kd_loss": 0.010926949791610241,
156
+ "learning_rate": 8.222222222222222e-06,
157
+ "loss": 0.3426,
158
+ "step": 75,
159
+ "student_loss": 0.0022108303382992744,
160
+ "teacher_loss": 0.0
161
+ },
162
+ {
163
+ "epoch": 0.02306805074971165,
164
+ "grad_norm": 13.4375,
165
+ "kd_loss": 0.004479927476495504,
166
+ "learning_rate": 8.777777777777778e-06,
167
+ "loss": 0.3733,
168
+ "step": 80,
169
+ "student_loss": 0.0041625467129051685,
170
+ "teacher_loss": 0.0
171
+ },
172
+ {
173
+ "epoch": 0.024509803921568627,
174
+ "grad_norm": 19.5,
175
+ "kd_loss": 0.003717327257618308,
176
+ "learning_rate": 9.333333333333334e-06,
177
+ "loss": 0.2466,
178
+ "step": 85,
179
+ "student_loss": 0.0008013551123440266,
180
+ "teacher_loss": 0.0
181
+ },
182
+ {
183
+ "epoch": 0.025951557093425604,
184
+ "grad_norm": 8.4375,
185
+ "kd_loss": 0.3134710490703583,
186
+ "learning_rate": 9.88888888888889e-06,
187
+ "loss": 0.2059,
188
+ "step": 90,
189
+ "student_loss": 0.016512297093868256,
190
+ "teacher_loss": 0.0
191
+ },
192
+ {
193
+ "epoch": 0.027393310265282585,
194
+ "grad_norm": 42.0,
195
+ "kd_loss": 0.0037511922419071198,
196
+ "learning_rate": 9.99995337991255e-06,
197
+ "loss": 0.1824,
198
+ "step": 95,
199
+ "student_loss": 0.014032241888344288,
200
+ "teacher_loss": 0.0
201
+ },
202
+ {
203
+ "epoch": 0.02883506343713956,
204
+ "grad_norm": 42.75,
205
+ "kd_loss": 0.0027243352960795164,
206
+ "learning_rate": 9.999763987297266e-06,
207
+ "loss": 0.1423,
208
+ "step": 100,
209
+ "student_loss": 0.0006586164236068726,
210
+ "teacher_loss": 0.0
211
+ },
212
+ {
213
+ "epoch": 0.03027681660899654,
214
+ "grad_norm": 22.75,
215
+ "kd_loss": 0.007741209119558334,
216
+ "learning_rate": 9.999428913912876e-06,
217
+ "loss": 0.1432,
218
+ "step": 105,
219
+ "student_loss": 0.006049849092960358,
220
+ "teacher_loss": 0.0
221
+ },
222
+ {
223
+ "epoch": 0.031718569780853516,
224
+ "grad_norm": 8.0,
225
+ "kd_loss": 0.002043877262622118,
226
+ "learning_rate": 9.998948169522587e-06,
227
+ "loss": 0.2012,
228
+ "step": 110,
229
+ "student_loss": 0.03380074352025986,
230
+ "teacher_loss": 0.0
231
+ },
232
+ {
233
+ "epoch": 0.03316032295271049,
234
+ "grad_norm": 12.9375,
235
+ "kd_loss": 0.13418245315551758,
236
+ "learning_rate": 9.998321768134101e-06,
237
+ "loss": 0.1955,
238
+ "step": 115,
239
+ "student_loss": 0.0011221985332667828,
240
+ "teacher_loss": 0.0
241
+ },
242
+ {
243
+ "epoch": 0.03460207612456748,
244
+ "grad_norm": 5.5,
245
+ "kd_loss": 0.00385228986851871,
246
+ "learning_rate": 9.997549727999212e-06,
247
+ "loss": 0.1011,
248
+ "step": 120,
249
+ "student_loss": 0.0008886058931238949,
250
+ "teacher_loss": 0.0
251
+ },
252
+ {
253
+ "epoch": 0.036043829296424454,
254
+ "grad_norm": 21.75,
255
+ "kd_loss": 0.8833059072494507,
256
+ "learning_rate": 9.996632071613257e-06,
257
+ "loss": 0.1431,
258
+ "step": 125,
259
+ "student_loss": 0.037908557802438736,
260
+ "teacher_loss": 0.0
261
+ },
262
+ {
263
+ "epoch": 0.03748558246828143,
264
+ "grad_norm": 22.75,
265
+ "kd_loss": 0.0020795725286006927,
266
+ "learning_rate": 9.995568825714479e-06,
267
+ "loss": 0.2122,
268
+ "step": 130,
269
+ "student_loss": 0.0007846044027246535,
270
+ "teacher_loss": 0.0
271
+ },
272
+ {
273
+ "epoch": 0.03892733564013841,
274
+ "grad_norm": 27.5,
275
+ "kd_loss": 0.6685954928398132,
276
+ "learning_rate": 9.994360021283235e-06,
277
+ "loss": 0.1744,
278
+ "step": 135,
279
+ "student_loss": 0.03987047076225281,
280
+ "teacher_loss": 0.0
281
+ },
282
+ {
283
+ "epoch": 0.040369088811995385,
284
+ "grad_norm": 6.34375,
285
+ "kd_loss": 0.0027170090470463037,
286
+ "learning_rate": 9.993005693541098e-06,
287
+ "loss": 0.1217,
288
+ "step": 140,
289
+ "student_loss": 0.0007884247461333871,
290
+ "teacher_loss": 0.0
291
+ },
292
+ {
293
+ "epoch": 0.04181084198385236,
294
+ "grad_norm": 15.625,
295
+ "kd_loss": 0.7305672764778137,
296
+ "learning_rate": 9.991505881949837e-06,
297
+ "loss": 0.1177,
298
+ "step": 145,
299
+ "student_loss": 0.032465774565935135,
300
+ "teacher_loss": 0.0
301
+ },
302
+ {
303
+ "epoch": 0.04325259515570934,
304
+ "grad_norm": 11.9375,
305
+ "kd_loss": 0.0009196309256367385,
306
+ "learning_rate": 9.989860630210251e-06,
307
+ "loss": 0.1145,
308
+ "step": 150,
309
+ "student_loss": 0.0006800334667786956,
310
+ "teacher_loss": 0.0
311
+ },
312
+ {
313
+ "epoch": 0.04469434832756632,
314
+ "grad_norm": 3.75,
315
+ "kd_loss": 0.0008867154247127473,
316
+ "learning_rate": 9.98806998626091e-06,
317
+ "loss": 0.1703,
318
+ "step": 155,
319
+ "student_loss": 0.000513731618411839,
320
+ "teacher_loss": 0.0
321
+ },
322
+ {
323
+ "epoch": 0.0461361014994233,
324
+ "grad_norm": 13.5625,
325
+ "kd_loss": 0.0031801851000636816,
326
+ "learning_rate": 9.98613400227676e-06,
327
+ "loss": 0.1881,
328
+ "step": 160,
329
+ "student_loss": 0.095796138048172,
330
+ "teacher_loss": 0.0
331
+ },
332
+ {
333
+ "epoch": 0.04757785467128028,
334
+ "grad_norm": 8.75,
335
+ "kd_loss": 0.009066148661077023,
336
+ "learning_rate": 9.984052734667586e-06,
337
+ "loss": 0.147,
338
+ "step": 165,
339
+ "student_loss": 0.00036973567330278456,
340
+ "teacher_loss": 0.0
341
+ },
342
+ {
343
+ "epoch": 0.049019607843137254,
344
+ "grad_norm": 19.875,
345
+ "kd_loss": 0.0007732495432719588,
346
+ "learning_rate": 9.981826244076387e-06,
347
+ "loss": 0.1159,
348
+ "step": 170,
349
+ "student_loss": 0.0006115734577178955,
350
+ "teacher_loss": 0.0
351
+ },
352
+ {
353
+ "epoch": 0.05046136101499423,
354
+ "grad_norm": 17.75,
355
+ "kd_loss": 0.001440449501387775,
356
+ "learning_rate": 9.979454595377594e-06,
357
+ "loss": 0.1439,
358
+ "step": 175,
359
+ "student_loss": 0.0018107936484739184,
360
+ "teacher_loss": 0.0
361
+ },
362
+ {
363
+ "epoch": 0.05190311418685121,
364
+ "grad_norm": 12.6875,
365
+ "kd_loss": 0.003575589507818222,
366
+ "learning_rate": 9.97693785767519e-06,
367
+ "loss": 0.0823,
368
+ "step": 180,
369
+ "student_loss": 0.0005937832174822688,
370
+ "teacher_loss": 0.0
371
+ },
372
+ {
373
+ "epoch": 0.05334486735870819,
374
+ "grad_norm": 13.8125,
375
+ "kd_loss": 0.002183184027671814,
376
+ "learning_rate": 9.9742761043007e-06,
377
+ "loss": 0.2085,
378
+ "step": 185,
379
+ "student_loss": 0.058680690824985504,
380
+ "teacher_loss": 0.0
381
+ },
382
+ {
383
+ "epoch": 0.05478662053056517,
384
+ "grad_norm": 7.9375,
385
+ "kd_loss": 0.13934114575386047,
386
+ "learning_rate": 9.971469412811032e-06,
387
+ "loss": 0.1039,
388
+ "step": 190,
389
+ "student_loss": 0.003180636093020439,
390
+ "teacher_loss": 0.0
391
+ },
392
+ {
393
+ "epoch": 0.056228373702422146,
394
+ "grad_norm": 14.5625,
395
+ "kd_loss": 0.2642575800418854,
396
+ "learning_rate": 9.968517864986247e-06,
397
+ "loss": 0.0963,
398
+ "step": 195,
399
+ "student_loss": 0.023328492417931557,
400
+ "teacher_loss": 0.0
401
+ },
402
+ {
403
+ "epoch": 0.05767012687427912,
404
+ "grad_norm": 6.1875,
405
+ "kd_loss": 0.0009822066640481353,
406
+ "learning_rate": 9.965421546827153e-06,
407
+ "loss": 0.1046,
408
+ "step": 200,
409
+ "student_loss": 0.000694891030434519,
410
+ "teacher_loss": 0.0
411
+ },
412
+ {
413
+ "epoch": 0.0591118800461361,
414
+ "grad_norm": 13.5,
415
+ "kd_loss": 0.0020439866930246353,
416
+ "learning_rate": 9.962180548552812e-06,
417
+ "loss": 0.0897,
418
+ "step": 205,
419
+ "student_loss": 0.042425841093063354,
420
+ "teacher_loss": 0.0
421
+ },
422
+ {
423
+ "epoch": 0.06055363321799308,
424
+ "grad_norm": 5.6875,
425
+ "kd_loss": 0.06946802884340286,
426
+ "learning_rate": 9.958794964597906e-06,
427
+ "loss": 0.0994,
428
+ "step": 210,
429
+ "student_loss": 0.0007058735354803503,
430
+ "teacher_loss": 0.0
431
+ },
432
+ {
433
+ "epoch": 0.061995386389850055,
434
+ "grad_norm": 8.75,
435
+ "kd_loss": 0.0022037022281438112,
436
+ "learning_rate": 9.955264893609986e-06,
437
+ "loss": 0.1323,
438
+ "step": 215,
439
+ "student_loss": 0.0006759354146197438,
440
+ "teacher_loss": 0.0
441
+ },
442
+ {
443
+ "epoch": 0.06343713956170703,
444
+ "grad_norm": 23.5,
445
+ "kd_loss": 0.11546128988265991,
446
+ "learning_rate": 9.951590438446597e-06,
447
+ "loss": 0.108,
448
+ "step": 220,
449
+ "student_loss": 0.04613906517624855,
450
+ "teacher_loss": 0.0
451
+ },
452
+ {
453
+ "epoch": 0.06487889273356401,
454
+ "grad_norm": 11.0,
455
+ "kd_loss": 0.006272319238632917,
456
+ "learning_rate": 9.947771706172285e-06,
457
+ "loss": 0.1205,
458
+ "step": 225,
459
+ "student_loss": 0.03350772336125374,
460
+ "teacher_loss": 0.0
461
+ },
462
+ {
463
+ "epoch": 0.06632064590542099,
464
+ "grad_norm": 24.25,
465
+ "kd_loss": 0.26141810417175293,
466
+ "learning_rate": 9.943808808055472e-06,
467
+ "loss": 0.0788,
468
+ "step": 230,
469
+ "student_loss": 0.003604925936087966,
470
+ "teacher_loss": 0.0
471
+ },
472
+ {
473
+ "epoch": 0.06776239907727798,
474
+ "grad_norm": 15.6875,
475
+ "kd_loss": 0.6798978447914124,
476
+ "learning_rate": 9.93970185956522e-06,
477
+ "loss": 0.0869,
478
+ "step": 235,
479
+ "student_loss": 0.015447350218892097,
480
+ "teacher_loss": 0.0
481
+ },
482
+ {
483
+ "epoch": 0.06920415224913495,
484
+ "grad_norm": 1.71875,
485
+ "kd_loss": 0.003386611584573984,
486
+ "learning_rate": 9.93545098036786e-06,
487
+ "loss": 0.1246,
488
+ "step": 240,
489
+ "student_loss": 0.0006252449820749462,
490
+ "teacher_loss": 0.0
491
+ },
492
+ {
493
+ "epoch": 0.07064590542099193,
494
+ "grad_norm": 21.25,
495
+ "kd_loss": 0.2356187403202057,
496
+ "learning_rate": 9.931056294323507e-06,
497
+ "loss": 0.146,
498
+ "step": 245,
499
+ "student_loss": 0.0012863262090831995,
500
+ "teacher_loss": 0.0
501
+ },
502
+ {
503
+ "epoch": 0.07208765859284891,
504
+ "grad_norm": 12.9375,
505
+ "kd_loss": 0.0037490571849048138,
506
+ "learning_rate": 9.926517929482454e-06,
507
+ "loss": 0.1485,
508
+ "step": 250,
509
+ "student_loss": 0.00040524674113839865,
510
+ "teacher_loss": 0.0
511
+ },
512
+ {
513
+ "epoch": 0.07352941176470588,
514
+ "grad_norm": 21.125,
515
+ "kd_loss": 0.0018789542373269796,
516
+ "learning_rate": 9.921836018081438e-06,
517
+ "loss": 0.056,
518
+ "step": 255,
519
+ "student_loss": 0.0003012118686456233,
520
+ "teacher_loss": 0.0
521
+ },
522
+ {
523
+ "epoch": 0.07497116493656286,
524
+ "grad_norm": 10.25,
525
+ "kd_loss": 0.008267119526863098,
526
+ "learning_rate": 9.917010696539793e-06,
527
+ "loss": 0.2125,
528
+ "step": 260,
529
+ "student_loss": 0.04526061564683914,
530
+ "teacher_loss": 0.0
531
+ },
532
+ {
533
+ "epoch": 0.07641291810841984,
534
+ "grad_norm": 4.8125,
535
+ "kd_loss": 0.0011836340418085456,
536
+ "learning_rate": 9.912042105455462e-06,
537
+ "loss": 0.1144,
538
+ "step": 265,
539
+ "student_loss": 0.00038986533763818443,
540
+ "teacher_loss": 0.0
541
+ },
542
+ {
543
+ "epoch": 0.07785467128027682,
544
+ "grad_norm": 12.875,
545
+ "kd_loss": 0.0008742965874262154,
546
+ "learning_rate": 9.906930389600917e-06,
547
+ "loss": 0.1514,
548
+ "step": 270,
549
+ "student_loss": 0.0005801369552500546,
550
+ "teacher_loss": 0.0
551
+ },
552
+ {
553
+ "epoch": 0.07929642445213379,
554
+ "grad_norm": 17.75,
555
+ "kd_loss": 0.0008876683423295617,
556
+ "learning_rate": 9.901675697918925e-06,
557
+ "loss": 0.1775,
558
+ "step": 275,
559
+ "student_loss": 0.053293921053409576,
560
+ "teacher_loss": 0.0
561
+ },
562
+ {
563
+ "epoch": 0.08073817762399077,
564
+ "grad_norm": 26.75,
565
+ "kd_loss": 0.002074585296213627,
566
+ "learning_rate": 9.896278183518216e-06,
567
+ "loss": 0.1494,
568
+ "step": 280,
569
+ "student_loss": 0.004262140020728111,
570
+ "teacher_loss": 0.0
571
+ },
572
+ {
573
+ "epoch": 0.08217993079584775,
574
+ "grad_norm": 7.65625,
575
+ "kd_loss": 0.0025187216233462095,
576
+ "learning_rate": 9.890738003669029e-06,
577
+ "loss": 0.1171,
578
+ "step": 285,
579
+ "student_loss": 0.0007169336313381791,
580
+ "teacher_loss": 0.0
581
+ },
582
+ {
583
+ "epoch": 0.08362168396770472,
584
+ "grad_norm": 21.375,
585
+ "kd_loss": 0.007369345054030418,
586
+ "learning_rate": 9.885055319798514e-06,
587
+ "loss": 0.1821,
588
+ "step": 290,
589
+ "student_loss": 0.0005684102652594447,
590
+ "teacher_loss": 0.0
591
+ },
592
+ {
593
+ "epoch": 0.0850634371395617,
594
+ "grad_norm": 16.0,
595
+ "kd_loss": 0.006104703992605209,
596
+ "learning_rate": 9.879230297486034e-06,
597
+ "loss": 0.1704,
598
+ "step": 295,
599
+ "student_loss": 0.0010396749712526798,
600
+ "teacher_loss": 0.0
601
+ },
602
+ {
603
+ "epoch": 0.08650519031141868,
604
+ "grad_norm": 23.375,
605
+ "kd_loss": 0.08903778344392776,
606
+ "learning_rate": 9.873263106458353e-06,
607
+ "loss": 0.1398,
608
+ "step": 300,
609
+ "student_loss": 0.014311299659311771,
610
+ "teacher_loss": 0.0
611
+ },
612
+ {
613
+ "epoch": 0.08794694348327567,
614
+ "grad_norm": 21.125,
615
+ "kd_loss": 0.003498164936900139,
616
+ "learning_rate": 9.86715392058467e-06,
617
+ "loss": 0.1304,
618
+ "step": 305,
619
+ "student_loss": 0.018731793388724327,
620
+ "teacher_loss": 0.0
621
+ },
622
+ {
623
+ "epoch": 0.08938869665513265,
624
+ "grad_norm": 22.25,
625
+ "kd_loss": 0.0002319709019502625,
626
+ "learning_rate": 9.860902917871566e-06,
627
+ "loss": 0.1595,
628
+ "step": 310,
629
+ "student_loss": 0.0005537475226446986,
630
+ "teacher_loss": 0.0
631
+ },
632
+ {
633
+ "epoch": 0.09083044982698962,
634
+ "grad_norm": 28.125,
635
+ "kd_loss": 0.012743989005684853,
636
+ "learning_rate": 9.854510280457819e-06,
637
+ "loss": 0.1062,
638
+ "step": 315,
639
+ "student_loss": 0.012689951807260513,
640
+ "teacher_loss": 0.0
641
+ },
642
+ {
643
+ "epoch": 0.0922722029988466,
644
+ "grad_norm": 18.625,
645
+ "kd_loss": 0.0014882351970300078,
646
+ "learning_rate": 9.847976194609083e-06,
647
+ "loss": 0.1479,
648
+ "step": 320,
649
+ "student_loss": 0.0005163275054655969,
650
+ "teacher_loss": 0.0
651
+ },
652
+ {
653
+ "epoch": 0.09371395617070358,
654
+ "grad_norm": 22.625,
655
+ "kd_loss": 0.3226740062236786,
656
+ "learning_rate": 9.841300850712479e-06,
657
+ "loss": 0.1206,
658
+ "step": 325,
659
+ "student_loss": 0.005041609983891249,
660
+ "teacher_loss": 0.0
661
+ },
662
+ {
663
+ "epoch": 0.09515570934256055,
664
+ "grad_norm": 10.125,
665
+ "kd_loss": 0.4770796298980713,
666
+ "learning_rate": 9.834484443271035e-06,
667
+ "loss": 0.0984,
668
+ "step": 330,
669
+ "student_loss": 0.015074173919856548,
670
+ "teacher_loss": 0.0
671
+ },
672
+ {
673
+ "epoch": 0.09659746251441753,
674
+ "grad_norm": 13.25,
675
+ "kd_loss": 0.000536711944732815,
676
+ "learning_rate": 9.827527170898019e-06,
677
+ "loss": 0.0722,
678
+ "step": 335,
679
+ "student_loss": 0.0006227240664884448,
680
+ "teacher_loss": 0.0
681
+ },
682
+ {
683
+ "epoch": 0.09803921568627451,
684
+ "grad_norm": 26.0,
685
+ "kd_loss": 0.12674546241760254,
686
+ "learning_rate": 9.820429236311158e-06,
687
+ "loss": 0.1458,
688
+ "step": 340,
689
+ "student_loss": 0.012428171001374722,
690
+ "teacher_loss": 0.0
691
+ },
692
+ {
693
+ "epoch": 0.09948096885813149,
694
+ "grad_norm": 8.375,
695
+ "kd_loss": 0.0027160593308508396,
696
+ "learning_rate": 9.81319084632673e-06,
697
+ "loss": 0.1197,
698
+ "step": 345,
699
+ "student_loss": 0.01698848232626915,
700
+ "teacher_loss": 0.0
701
+ },
702
+ {
703
+ "epoch": 0.10092272202998846,
704
+ "grad_norm": 25.5,
705
+ "kd_loss": 0.2688156068325043,
706
+ "learning_rate": 9.805812211853534e-06,
707
+ "loss": 0.0684,
708
+ "step": 350,
709
+ "student_loss": 0.006060065235942602,
710
+ "teacher_loss": 0.0
711
+ },
712
+ {
713
+ "epoch": 0.10236447520184544,
714
+ "grad_norm": 26.0,
715
+ "kd_loss": 0.0023566237650811672,
716
+ "learning_rate": 9.798293547886748e-06,
717
+ "loss": 0.1159,
718
+ "step": 355,
719
+ "student_loss": 0.0004638656973838806,
720
+ "teacher_loss": 0.0
721
+ },
722
+ {
723
+ "epoch": 0.10380622837370242,
724
+ "grad_norm": 9.125,
725
+ "kd_loss": 0.004494955763220787,
726
+ "learning_rate": 9.790635073501662e-06,
727
+ "loss": 0.082,
728
+ "step": 360,
729
+ "student_loss": 0.0004987778957001865,
730
+ "teacher_loss": 0.0
731
+ },
732
+ {
733
+ "epoch": 0.1052479815455594,
734
+ "grad_norm": 10.375,
735
+ "kd_loss": 0.0029149462934583426,
736
+ "learning_rate": 9.782837011847294e-06,
737
+ "loss": 0.1926,
738
+ "step": 365,
739
+ "student_loss": 0.043784309178590775,
740
+ "teacher_loss": 0.0
741
+ },
742
+ {
743
+ "epoch": 0.10668973471741638,
744
+ "grad_norm": 24.625,
745
+ "kd_loss": 0.07756392657756805,
746
+ "learning_rate": 9.774899590139897e-06,
747
+ "loss": 0.0849,
748
+ "step": 370,
749
+ "student_loss": 0.01941766031086445,
750
+ "teacher_loss": 0.0
751
+ },
752
+ {
753
+ "epoch": 0.10813148788927336,
754
+ "grad_norm": 13.125,
755
+ "kd_loss": 0.00261987978592515,
756
+ "learning_rate": 9.766823039656328e-06,
757
+ "loss": 0.1517,
758
+ "step": 375,
759
+ "student_loss": 0.03552848473191261,
760
+ "teacher_loss": 0.0
761
+ },
762
+ {
763
+ "epoch": 0.10957324106113034,
764
+ "grad_norm": 6.28125,
765
+ "kd_loss": 0.0021297987550497055,
766
+ "learning_rate": 9.758607595727309e-06,
767
+ "loss": 0.1632,
768
+ "step": 380,
769
+ "student_loss": 0.0005093910149298608,
770
+ "teacher_loss": 0.0
771
+ },
772
+ {
773
+ "epoch": 0.11101499423298732,
774
+ "grad_norm": 28.125,
775
+ "kd_loss": 0.012473474256694317,
776
+ "learning_rate": 9.75025349773058e-06,
777
+ "loss": 0.1271,
778
+ "step": 385,
779
+ "student_loss": 0.00041385076474398375,
780
+ "teacher_loss": 0.0
781
+ },
782
+ {
783
+ "epoch": 0.11245674740484429,
784
+ "grad_norm": 24.375,
785
+ "kd_loss": 0.0008915087673813105,
786
+ "learning_rate": 9.741760989083914e-06,
787
+ "loss": 0.0936,
788
+ "step": 390,
789
+ "student_loss": 0.00028736997046507895,
790
+ "teacher_loss": 0.0
791
+ },
792
+ {
793
+ "epoch": 0.11389850057670127,
794
+ "grad_norm": 12.875,
795
+ "kd_loss": 0.007614473812282085,
796
+ "learning_rate": 9.73313031723803e-06,
797
+ "loss": 0.0901,
798
+ "step": 395,
799
+ "student_loss": 0.0005106693133711815,
800
+ "teacher_loss": 0.0
801
+ },
802
+ {
803
+ "epoch": 0.11534025374855825,
804
+ "grad_norm": 11.25,
805
+ "kd_loss": 0.00347925559617579,
806
+ "learning_rate": 9.724361733669383e-06,
807
+ "loss": 0.0546,
808
+ "step": 400,
809
+ "student_loss": 0.0008912020130082965,
810
+ "teacher_loss": 0.0
811
+ },
812
+ {
813
+ "epoch": 0.11678200692041522,
814
+ "grad_norm": 8.0625,
815
+ "kd_loss": 0.0011954058427363634,
816
+ "learning_rate": 9.715455493872827e-06,
817
+ "loss": 0.1324,
818
+ "step": 405,
819
+ "student_loss": 0.004991908557713032,
820
+ "teacher_loss": 0.0
821
+ },
822
+ {
823
+ "epoch": 0.1182237600922722,
824
+ "grad_norm": 9.1875,
825
+ "kd_loss": 0.0025417145807296038,
826
+ "learning_rate": 9.70641185735419e-06,
827
+ "loss": 0.1483,
828
+ "step": 410,
829
+ "student_loss": 0.01691470481455326,
830
+ "teacher_loss": 0.0
831
+ },
832
+ {
833
+ "epoch": 0.11966551326412918,
834
+ "grad_norm": 13.3125,
835
+ "kd_loss": 0.003752017393708229,
836
+ "learning_rate": 9.697231087622691e-06,
837
+ "loss": 0.1511,
838
+ "step": 415,
839
+ "student_loss": 0.0010631532641127706,
840
+ "teacher_loss": 0.0
841
+ },
842
+ {
843
+ "epoch": 0.12110726643598616,
844
+ "grad_norm": 10.9375,
845
+ "kd_loss": 0.2554698586463928,
846
+ "learning_rate": 9.687913452183277e-06,
847
+ "loss": 0.1095,
848
+ "step": 420,
849
+ "student_loss": 0.06524818390607834,
850
+ "teacher_loss": 0.0
851
+ },
852
+ {
853
+ "epoch": 0.12254901960784313,
854
+ "grad_norm": 9.125,
855
+ "kd_loss": 0.0005501045961864293,
856
+ "learning_rate": 9.678459222528825e-06,
857
+ "loss": 0.0577,
858
+ "step": 425,
859
+ "student_loss": 0.0013150365557521582,
860
+ "teacher_loss": 0.0
861
+ },
862
+ {
863
+ "epoch": 0.12399077277970011,
864
+ "grad_norm": 10.0625,
865
+ "kd_loss": 0.00420792680233717,
866
+ "learning_rate": 9.668868674132224e-06,
867
+ "loss": 0.1381,
868
+ "step": 430,
869
+ "student_loss": 0.05358680710196495,
870
+ "teacher_loss": 0.0
871
+ },
872
+ {
873
+ "epoch": 0.1254325259515571,
874
+ "grad_norm": 14.625,
875
+ "kd_loss": 0.0016143789980560541,
876
+ "learning_rate": 9.65914208643836e-06,
877
+ "loss": 0.0654,
878
+ "step": 435,
879
+ "student_loss": 0.000772066181525588,
880
+ "teacher_loss": 0.0
881
+ },
882
+ {
883
+ "epoch": 0.12687427912341406,
884
+ "grad_norm": 7.15625,
885
+ "kd_loss": 0.0023973328061401844,
886
+ "learning_rate": 9.649279742855964e-06,
887
+ "loss": 0.1371,
888
+ "step": 440,
889
+ "student_loss": 0.025605041533708572,
890
+ "teacher_loss": 0.0
891
+ },
892
+ {
893
+ "epoch": 0.12831603229527105,
894
+ "grad_norm": 23.0,
895
+ "kd_loss": 0.0013417273294180632,
896
+ "learning_rate": 9.639281930749363e-06,
897
+ "loss": 0.0814,
898
+ "step": 445,
899
+ "student_loss": 0.03157199174165726,
900
+ "teacher_loss": 0.0
901
+ },
902
+ {
903
+ "epoch": 0.12975778546712802,
904
+ "grad_norm": 13.25,
905
+ "kd_loss": 0.0016620807582512498,
906
+ "learning_rate": 9.629148941430096e-06,
907
+ "loss": 0.1442,
908
+ "step": 450,
909
+ "student_loss": 0.0006721625686623156,
910
+ "teacher_loss": 0.0
911
+ },
912
+ {
913
+ "epoch": 0.131199538638985,
914
+ "grad_norm": 28.0,
915
+ "kd_loss": 0.0016946454998105764,
916
+ "learning_rate": 9.618881070148436e-06,
917
+ "loss": 0.1405,
918
+ "step": 455,
919
+ "student_loss": 0.0009172292775474489,
920
+ "teacher_loss": 0.0
921
+ },
922
+ {
923
+ "epoch": 0.13264129181084197,
924
+ "grad_norm": 10.125,
925
+ "kd_loss": 0.027760690078139305,
926
+ "learning_rate": 9.608478616084784e-06,
927
+ "loss": 0.0505,
928
+ "step": 460,
929
+ "student_loss": 0.0005534313968382776,
930
+ "teacher_loss": 0.0
931
+ },
932
+ {
933
+ "epoch": 0.13408304498269896,
934
+ "grad_norm": 14.25,
935
+ "kd_loss": 0.008209657855331898,
936
+ "learning_rate": 9.597941882340947e-06,
937
+ "loss": 0.1803,
938
+ "step": 465,
939
+ "student_loss": 0.000469625141704455,
940
+ "teacher_loss": 0.0
941
+ },
942
+ {
943
+ "epoch": 0.13552479815455595,
944
+ "grad_norm": 13.4375,
945
+ "kd_loss": 0.0006104095955379307,
946
+ "learning_rate": 9.587271175931313e-06,
947
+ "loss": 0.1148,
948
+ "step": 470,
949
+ "student_loss": 0.00037698415690101683,
950
+ "teacher_loss": 0.0
951
+ },
952
+ {
953
+ "epoch": 0.13696655132641292,
954
+ "grad_norm": 6.5625,
955
+ "kd_loss": 0.012027841061353683,
956
+ "learning_rate": 9.5764668077739e-06,
957
+ "loss": 0.0626,
958
+ "step": 475,
959
+ "student_loss": 0.0005685262149199843,
960
+ "teacher_loss": 0.0
961
+ },
962
+ {
963
+ "epoch": 0.1384083044982699,
964
+ "grad_norm": 9.0625,
965
+ "kd_loss": 0.0022413916885852814,
966
+ "learning_rate": 9.5655290926813e-06,
967
+ "loss": 0.099,
968
+ "step": 480,
969
+ "student_loss": 0.012239435687661171,
970
+ "teacher_loss": 0.0
971
+ },
972
+ {
973
+ "epoch": 0.13985005767012687,
974
+ "grad_norm": 25.75,
975
+ "kd_loss": 0.0016755703836679459,
976
+ "learning_rate": 9.554458349351507e-06,
977
+ "loss": 0.0891,
978
+ "step": 485,
979
+ "student_loss": 0.027401834726333618,
980
+ "teacher_loss": 0.0
981
+ },
982
+ {
983
+ "epoch": 0.14129181084198386,
984
+ "grad_norm": 29.875,
985
+ "kd_loss": 0.46647658944129944,
986
+ "learning_rate": 9.54325490035863e-06,
987
+ "loss": 0.1581,
988
+ "step": 490,
989
+ "student_loss": 0.0034644226543605328,
990
+ "teacher_loss": 0.0
991
+ },
992
+ {
993
+ "epoch": 0.14273356401384082,
994
+ "grad_norm": 12.0,
995
+ "kd_loss": 0.10091336816549301,
996
+ "learning_rate": 9.531919072143488e-06,
997
+ "loss": 0.1283,
998
+ "step": 495,
999
+ "student_loss": 0.0010429032845422626,
1000
+ "teacher_loss": 0.0
1001
+ },
1002
+ {
1003
+ "epoch": 0.14417531718569782,
1004
+ "grad_norm": 18.125,
1005
+ "kd_loss": 0.25293171405792236,
1006
+ "learning_rate": 9.520451195004107e-06,
1007
+ "loss": 0.0874,
1008
+ "step": 500,
1009
+ "student_loss": 0.0008772966684773564,
1010
+ "teacher_loss": 0.0
1011
+ }
1012
+ ],
1013
+ "logging_steps": 5,
1014
+ "max_steps": 3000,
1015
+ "num_input_tokens_seen": 0,
1016
+ "num_train_epochs": 1,
1017
+ "save_steps": 500,
1018
+ "stateful_callbacks": {
1019
+ "TrainerControl": {
1020
+ "args": {
1021
+ "should_epoch_stop": false,
1022
+ "should_evaluate": false,
1023
+ "should_log": false,
1024
+ "should_save": true,
1025
+ "should_training_stop": false
1026
+ },
1027
+ "attributes": {}
1028
+ }
1029
+ },
1030
+ "total_flos": 0.0,
1031
+ "train_batch_size": 1,
1032
+ "trial_name": null,
1033
+ "trial_params": null
1034
+ }
checkpoints/codi_frozen_1.5b_logit/checkpoint-500/vocab.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoints/codi_frozen_1.5b_logit/wandb/debug-internal.log ADDED
@@ -0,0 +1,11 @@
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"time":"2026-06-24T21:57:47.314707899+02:00","level":"INFO","msg":"wandb-core"}
2
+ {"time":"2026-06-24T21:57:47.314719289+02:00","level":"INFO","msg":"stream: starting","core version":"0.27.2"}
3
+ {"time":"2026-06-24T21:57:47.43578674+02:00","level":"WARN","msg":"featurechecker: GraphQL client is nil, skipping feature loading"}
4
+ {"time":"2026-06-24T21:57:47.435801873+02:00","level":"WARN","msg":"featurechecker: GraphQL client is nil, skipping feature loading"}
5
+ {"time":"2026-06-24T21:57:47.43581632+02:00","level":"INFO","msg":"stream: created new stream","id":"xw4jltek"}
6
+ {"time":"2026-06-24T21:57:47.435950364+02:00","level":"INFO","msg":"handler: started"}
7
+ {"time":"2026-06-24T21:57:47.43887291+02:00","level":"INFO","msg":"stream: started"}
8
+ {"time":"2026-06-24T21:57:47.438988244+02:00","level":"INFO","msg":"sender: started"}
9
+ {"time":"2026-06-24T21:57:47.438991475+02:00","level":"INFO","msg":"writer: started","stream_id":"xw4jltek"}
10
+ {"time":"2026-06-24T21:57:47.455584204+02:00","level":"WARN","msg":"featurechecker: GraphQL client is nil, skipping feature loading"}
11
+ {"time":"2026-06-24T21:57:47.45559365+02:00","level":"WARN","msg":"runupserter: server does not expand metric globs but the x_server_side_expand_glob_metrics setting is set; ignoring"}
checkpoints/codi_frozen_1.5b_logit/wandb/debug.log ADDED
@@ -0,0 +1,20 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ 2026-06-24 21:57:46,867 INFO MainThread:689585 [wandb_setup.py:_flush():81] Current SDK version is 0.27.2
2
+ 2026-06-24 21:57:46,867 INFO MainThread:689585 [wandb_setup.py:_flush():81] Configure stats pid to 689585
3
+ 2026-06-24 21:57:46,867 INFO MainThread:689585 [wandb_setup.py:_flush():81] Loading settings from environment variables
4
+ 2026-06-24 21:57:46,867 INFO MainThread:689585 [wandb_init.py:setup_run_log_directory():723] Logging user logs to model_weights/codi_frozen_1.5b_logit/wandb/offline-run-20260624_215746-xw4jltek/logs/debug.log
5
+ 2026-06-24 21:57:46,867 INFO MainThread:689585 [wandb_init.py:setup_run_log_directory():724] Logging internal logs to model_weights/codi_frozen_1.5b_logit/wandb/offline-run-20260624_215746-xw4jltek/logs/debug-internal.log
6
+ 2026-06-24 21:57:46,867 INFO MainThread:689585 [wandb_init.py:init():766] calling init triggers
7
+ 2026-06-24 21:57:46,867 INFO MainThread:689585 [wandb_init.py:init():771] wandb.init called with sweep_config: {}
8
+ config: {'model': 'model_weights/sft1.5b_lr1e5_bs32/checkpoint-6936', 'output_dir': 'model_weights/codi_frozen_1.5b_logit', 'sources': ['mbpp', 'humaneval', 'pyx'], 'cache_dir': 'data/cache/codi_train', 'n_samples': -1, 'max_seq_len': 3072, 'max_frames': -1, 'latent_steps': 1, 'epochs': 5.0, 'lr': 1e-05, 'batch_size': 1, 'grad_accum': 4, 'max_steps': 3000, 'save_steps': 500, 'alpha': 0.0, 'beta': 1.0, 'gamma': 1.0, 'kd_layers': [-1], 'frozen_teacher': 'model_weights/sft1.5b_lr1e5_bs32/checkpoint-6936', 'kd_target': 'logit', 'kd_temp': 2.0, 'single_anchor': False, 'ss_prob': 0.0, 'ss_ramp_frac': 0.5, '_wandb': {}}
9
+ 2026-06-24 21:57:46,867 INFO MainThread:689585 [wandb_init.py:init():814] starting backend
10
+ 2026-06-24 21:57:47,302 INFO MainThread:689585 [wandb_init.py:init():829] sending inform_init request
11
+ 2026-06-24 21:57:47,439 INFO MainThread:689585 [wandb_init.py:init():834] backend started and connected
12
+ 2026-06-24 21:57:47,441 INFO MainThread:689585 [wandb_init.py:init():904] updated telemetry
13
+ 2026-06-24 21:57:47,454 INFO MainThread:689585 [wandb_init.py:init():927] communicating run to backend with 90.0 second timeout
14
+ 2026-06-24 21:57:47,456 INFO MainThread:689585 [wandb_init.py:init():972] starting run threads in backend
15
+ 2026-06-24 21:57:47,902 INFO MainThread:689585 [wandb_run.py:_console_start():2523] atexit reg
16
+ 2026-06-24 21:57:47,902 INFO MainThread:689585 [wandb_run.py:_redirect():2373] redirect: wrap_raw
17
+ 2026-06-24 21:57:47,902 INFO MainThread:689585 [wandb_run.py:_redirect():2442] Wrapping output streams.
18
+ 2026-06-24 21:57:47,902 INFO MainThread:689585 [wandb_run.py:_redirect():2465] Redirects installed.
19
+ 2026-06-24 21:57:47,907 INFO MainThread:689585 [wandb_init.py:init():1010] run started, returning control to user process
20
+ 2026-06-24 21:57:52,397 INFO MainThread:689585 [wandb_run.py:_config_callback():1346] config_cb None None {'output_dir': 'model_weights/codi_frozen_1.5b_logit', 'overwrite_output_dir': False, 'do_train': False, 'do_eval': False, 'do_predict': False, 'eval_strategy': 'no', 'prediction_loss_only': False, 'per_device_train_batch_size': 1, 'per_device_eval_batch_size': 8, 'per_gpu_train_batch_size': None, 'per_gpu_eval_batch_size': None, 'gradient_accumulation_steps': 4, 'eval_accumulation_steps': None, 'eval_delay': 0, 'torch_empty_cache_steps': None, 'learning_rate': 1e-05, 'weight_decay': 0.1, 'adam_beta1': 0.9, 'adam_beta2': 0.999, 'adam_epsilon': 1e-08, 'max_grad_norm': 1.0, 'num_train_epochs': 5.0, 'max_steps': 3000, 'lr_scheduler_type': 'cosine', 'lr_scheduler_kwargs': None, 'warmup_ratio': 0.03, 'warmup_steps': 0, 'log_level': 'passive', 'log_level_replica': 'warning', 'log_on_each_node': True, 'logging_dir': 'model_weights/codi_frozen_1.5b_logit/runs/Jun24_21-57-47_alvis4-41', 'logging_strategy': 'steps', 'logging_first_step': False, 'logging_steps': 5, 'logging_nan_inf_filter': True, 'save_strategy': 'steps', 'save_steps': 500, 'save_total_limit': None, 'save_safetensors': True, 'save_on_each_node': False, 'save_only_model': False, 'restore_callback_states_from_checkpoint': False, 'no_cuda': False, 'use_cpu': False, 'use_mps_device': False, 'seed': 42, 'data_seed': None, 'jit_mode_eval': False, 'bf16': True, 'fp16': False, 'fp16_opt_level': 'O1', 'half_precision_backend': 'auto', 'bf16_full_eval': False, 'fp16_full_eval': False, 'tf32': None, 'local_rank': 0, 'ddp_backend': None, 'tpu_num_cores': None, 'tpu_metrics_debug': False, 'debug': [], 'dataloader_drop_last': False, 'eval_steps': None, 'dataloader_num_workers': 0, 'dataloader_prefetch_factor': None, 'past_index': -1, 'run_name': None, 'disable_tqdm': False, 'remove_unused_columns': False, 'label_names': [], 'load_best_model_at_end': False, 'metric_for_best_model': None, 'greater_is_better': None, 'ignore_data_skip': False, 'fsdp': [], 'fsdp_min_num_params': 0, 'fsdp_config': {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}, 'fsdp_transformer_layer_cls_to_wrap': None, 'accelerator_config': {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}, 'parallelism_config': None, 'deepspeed': None, 'label_smoothing_factor': 0.0, 'optim': 'paged_adamw_8bit', 'optim_args': None, 'adafactor': False, 'group_by_length': False, 'length_column_name': 'length', 'report_to': ['wandb'], 'project': 'huggingface', 'trackio_space_id': 'trackio', 'ddp_find_unused_parameters': False, 'ddp_bucket_cap_mb': None, 'ddp_broadcast_buffers': None, 'dataloader_pin_memory': True, 'dataloader_persistent_workers': False, 'skip_memory_metrics': True, 'use_legacy_prediction_loop': False, 'push_to_hub': False, 'resume_from_checkpoint': None, 'hub_model_id': None, 'hub_strategy': 'every_save', 'hub_token': '<HUB_TOKEN>', 'hub_private_repo': None, 'hub_always_push': False, 'hub_revision': None, 'gradient_checkpointing': False, 'gradient_checkpointing_kwargs': None, 'include_inputs_for_metrics': False, 'include_for_metrics': [], 'eval_do_concat_batches': True, 'fp16_backend': 'auto', 'push_to_hub_model_id': None, 'push_to_hub_organization': None, 'push_to_hub_token': '<PUSH_TO_HUB_TOKEN>', 'mp_parameters': '', 'auto_find_batch_size': False, 'full_determinism': False, 'torchdynamo': None, 'ray_scope': 'last', 'ddp_timeout': 1800, 'torch_compile': False, 'torch_compile_backend': None, 'torch_compile_mode': None, 'include_tokens_per_second': False, 'include_num_input_tokens_seen': 'no', 'neftune_noise_alpha': None, 'optim_target_modules': None, 'batch_eval_metrics': False, 'eval_on_start': False, 'use_liger_kernel': False, 'liger_kernel_config': None, 'eval_use_gather_object': False, 'average_tokens_across_devices': True}
checkpoints/codi_frozen_1.5b_logit/wandb/offline-run-20260624_215746-xw4jltek/files/requirements.txt ADDED
@@ -0,0 +1,229 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ wheel==0.47.0
2
+ triton==3.3.1
3
+ ptyprocess==0.7.0
4
+ kiwisolver==1.5.0
5
+ pyzmq==27.1.0
6
+ nvidia-cublas-cu12==12.6.4.1
7
+ python-dateutil==2.9.0.post0
8
+ cycler==0.12.1
9
+ rfc3339-validator==0.1.4
10
+ pluggy==1.6.0
11
+ grpclib==0.4.9
12
+ attrs==26.1.0
13
+ urllib3==2.7.0
14
+ cffi==2.0.0
15
+ openai==1.86.0
16
+ async-lru==2.3.0
17
+ cfgv==3.5.0
18
+ httpcore==1.0.9
19
+ bitsandbytes==0.49.2
20
+ iniconfig==2.3.0
21
+ nvidia-cuda-runtime-cu12==12.6.77
22
+ jupyterlab_server==2.28.0
23
+ latex2sympy2_extended==1.11.0
24
+ jupyter-events==0.12.1
25
+ docker==7.1.0
26
+ ipywidgets==8.1.8
27
+ blinker==1.9.0
28
+ cbor2==6.1.2
29
+ nvidia-cufft-cu12==11.3.0.4
30
+ nvidia-cusolver-cu12==11.7.1.2
31
+ multiprocess==0.70.16
32
+ anyio==4.14.0
33
+ itsdangerous==2.2.0
34
+ matplotlib==3.11.0
35
+ widgetsnbextension==4.0.15
36
+ watchfiles==1.2.0
37
+ swebench==4.1.0
38
+ blobfile==3.2.0
39
+ jupyter==1.1.1
40
+ certifi==2026.6.17
41
+ argon2-cffi==25.1.0
42
+ argon2-cffi-bindings==25.1.0
43
+ overrides==7.7.0
44
+ propcache==0.5.2
45
+ ipython==8.33.0
46
+ nbconvert==7.17.1
47
+ rich==15.0.0
48
+ asttokens==3.0.1
49
+ tokenizers==0.22.2
50
+ synchronicity==0.12.4
51
+ tqdm==4.68.3
52
+ nvidia-nvjitlink-cu12==12.6.85
53
+ nvidia-nvtx-cu12==12.6.77
54
+ jsonschema-specifications==2025.9.1
55
+ jupyterlab==4.5.8
56
+ pure_eval==0.2.3
57
+ uri-template==1.3.0
58
+ sentry-sdk==2.63.0
59
+ pycryptodomex==3.23.0
60
+ yarl==1.24.2
61
+ h2==4.3.0
62
+ h11==0.16.0
63
+ requests==2.34.2
64
+ Jinja2==3.1.6
65
+ xformers==0.0.31.post1
66
+ rfc3987-syntax==1.1.0
67
+ antlr4-python3-runtime==4.11.0
68
+ nvidia-cusparselt-cu12==0.6.3
69
+ safetensors==0.8.0
70
+ notebook_shim==0.2.4
71
+ Pygments==2.20.0
72
+ python-json-logger==4.1.0
73
+ isoduration==20.11.0
74
+ smmap==5.0.3
75
+ click==8.3.3
76
+ transformers==4.57.6
77
+ aiohappyeyeballs==2.6.2
78
+ regex==2026.5.9
79
+ jsonpointer==3.1.1
80
+ fqdn==1.5.1
81
+ jupyter-lsp==2.3.1
82
+ rfc3986-validator==0.1.1
83
+ prometheus_client==0.25.0
84
+ mistune==3.2.1
85
+ sentencepiece==0.2.1
86
+ types-toml==0.10.8.20260518
87
+ wcwidth==0.8.1
88
+ identify==2.6.19
89
+ protobuf==6.33.6
90
+ datasets==3.6.0
91
+ accelerate==1.14.0
92
+ networkx==3.6.1
93
+ pydantic==2.13.4
94
+ multidict==6.7.1
95
+ chardet==7.4.3
96
+ pytest==9.1.0
97
+ babel==2.18.0
98
+ pip==26.1.2
99
+ dill==0.3.8
100
+ referencing==0.37.0
101
+ pre_commit==4.6.0
102
+ numpy==2.4.6
103
+ math-verify==0.9.0
104
+ traitlets==5.15.1
105
+ nvidia-curand-cu12==10.3.7.77
106
+ websocket-client==1.9.0
107
+ hf-xet==1.5.1
108
+ unidiff==0.7.5
109
+ nest-asyncio2==1.7.2
110
+ typing_extensions==4.15.0
111
+ platformdirs==4.10.0
112
+ torch==2.7.1
113
+ tblib==3.2.2
114
+ sqlfluff==4.2.2
115
+ modal==1.5.0
116
+ ipykernel==7.3.0
117
+ msgspec==0.21.1
118
+ virtualenv==21.5.1
119
+ json5==0.14.0
120
+ pathlib_abc==0.5.2
121
+ aiosignal==1.4.0
122
+ moodist==0.5.1
123
+ rpds-py==2026.5.1
124
+ webencodings==0.5.1
125
+ sniffio==1.3.1
126
+ annotated-types==0.7.0
127
+ python-discovery==1.4.2
128
+ scipy==1.17.1
129
+ GitPython==3.1.50
130
+ defusedxml==0.7.1
131
+ nvidia-cufile-cu12==1.11.1.6
132
+ nvidia-nccl-cu12==2.26.2
133
+ sympy==1.14.0
134
+ tinycss2==1.5.1
135
+ soupsieve==2.8.4
136
+ ninja==1.13.0
137
+ python-dotenv==1.2.2
138
+ ghapi==1.0.13
139
+ jupyter_server_terminals==0.5.4
140
+ markdown-it-py==4.2.0
141
+ nbclient==0.11.0
142
+ mpmath==1.3.0
143
+ frozenlist==1.8.0
144
+ Werkzeug==3.1.8
145
+ pillow==12.2.0
146
+ executing==2.2.1
147
+ gitdb==4.0.12
148
+ pyparsing==3.3.2
149
+ setuptools==82.0.1
150
+ Flask==3.1.1
151
+ MarkupSafe==3.0.3
152
+ notebook==7.5.7
153
+ jupyter_core==5.9.1
154
+ jedi==0.20.0
155
+ mdurl==0.1.2
156
+ lark==1.3.1
157
+ bleach==6.4.0
158
+ hpack==4.1.0
159
+ google-re2==1.1.20240702
160
+ beautifulsoup4==4.15.0
161
+ jsonschema==4.26.0
162
+ psutil==7.2.2
163
+ nvidia-cuda-nvrtc-cu12==12.6.77
164
+ narwhals==2.22.1
165
+ fonttools==4.63.0
166
+ lxml==6.1.1
167
+ fastjsonschema==2.21.2
168
+ tornado==6.5.7
169
+ nodeenv==1.10.0
170
+ typing-inspection==0.4.2
171
+ httpx==0.28.1
172
+ distro==1.9.0
173
+ colorama==0.4.6
174
+ jupyter-console==6.6.3
175
+ jupyterlab_widgets==3.0.16
176
+ PyYAML==6.0.3
177
+ fsspec==2025.3.0
178
+ hyperframe==6.1.0
179
+ nvidia-cuda-cupti-cu12==12.6.80
180
+ charset-normalizer==3.4.7
181
+ wandb==0.27.2
182
+ nbformat==5.10.4
183
+ nvidia-cudnn-cu12==9.5.1.17
184
+ peft==0.19.1
185
+ pathspec==1.1.1
186
+ matplotlib-inline==0.2.2
187
+ huggingface_hub==0.36.2
188
+ pyarrow==24.0.0
189
+ nvidia-cusparse-cu12==12.5.4.2
190
+ cwm==0.0.1
191
+ prompt_toolkit==3.0.52
192
+ omegaconf==2.4.0.dev4
193
+ tiktoken==0.13.0
194
+ contourpy==1.3.3
195
+ comm==0.2.3
196
+ threadpoolctl==3.6.0
197
+ pydantic_core==2.46.4
198
+ tenacity==9.1.4
199
+ parso==0.8.7
200
+ pandas==3.0.3
201
+ webcolors==25.10.0
202
+ idna==3.18
203
+ terminado==0.18.1
204
+ pandocfilters==1.5.1
205
+ aiohttp==3.14.1
206
+ arrow==1.4.0
207
+ packaging==26.2
208
+ tzdata==2026.2
209
+ distlib==0.4.3
210
+ Send2Trash==2.1.0
211
+ jupyterlab_pygments==0.3.0
212
+ decorator==5.3.1
213
+ jupyter_server==2.20.0
214
+ scikit-learn==1.9.0
215
+ six==1.17.0
216
+ diff_cover==10.3.0
217
+ pycparser==3.0
218
+ types-certifi==2021.10.8.3
219
+ toml==0.10.2
220
+ stack-data==0.6.3
221
+ fastcore==1.13.5
222
+ filelock==3.29.4
223
+ xxhash==3.7.0
224
+ pexpect==4.9.0
225
+ jupyter_client==8.9.1
226
+ jiter==0.15.0
227
+ debugpy==1.8.21
228
+ universal_pathlib==0.3.10
229
+ joblib==1.5.3
checkpoints/codi_frozen_1.5b_logit/wandb/offline-run-20260624_215746-xw4jltek/logs/debug-core.log ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"time":"2026-06-24T21:57:47.122170036+02:00","level":"INFO","msg":"main: starting server","port-filename":"/local/tmp.6776481/tmpidl24fch/port-689585.txt","pid":689585,"detached":false,"idle-timeout":600000000000,"log-level":0,"disable-analytics":false,"shutdown-on-parent-exit":false,"enable-dcgm-profiling":false}
2
+ {"time":"2026-06-24T21:57:47.122763298+02:00","level":"INFO","msg":"server: will exit if parent process dies","ppid":689585}
3
+ {"time":"2026-06-24T21:57:47.122749828+02:00","level":"INFO","msg":"server: accepting connections","addr":{"Name":"/local/tmp.6776481/wandb-689585-689947-347648083/socket","Net":"unix"}}
4
+ {"time":"2026-06-24T21:57:47.301948548+02:00","level":"INFO","msg":"connection: ManageConnectionData: new connection created","id":"1(@)"}
5
+ {"time":"2026-06-24T21:57:47.313416025+02:00","level":"INFO","msg":"handleInformInit: received","streamId":"xw4jltek","id":"1(@)"}
6
+ {"time":"2026-06-24T21:57:47.438881346+02:00","level":"INFO","msg":"handleInformInit: stream started","streamId":"xw4jltek","id":"1(@)"}
7
+ {"time":"2026-06-25T08:28:06.446955702+02:00","level":"INFO","msg":"handleInformTeardown: server teardown initiated","id":"1(@)"}
8
+ {"time":"2026-06-25T08:28:06.447955343+02:00","level":"INFO","msg":"server is shutting down"}
9
+ {"time":"2026-06-25T08:28:06.448081437+02:00","level":"INFO","msg":"server: listener closed","addr":{"Name":"/local/tmp.6776481/wandb-689585-689947-347648083/socket","Net":"unix"}}
10
+ {"time":"2026-06-25T08:28:06.448204941+02:00","level":"INFO","msg":"processOutgoingData: finished","id":"1(@)"}
11
+ {"time":"2026-06-25T08:28:06.448216556+02:00","level":"INFO","msg":"connection: closing","id":"1(@)"}
12
+ {"time":"2026-06-25T08:28:06.448263144+02:00","level":"INFO","msg":"connection: closed successfully","id":"1(@)"}
13
+ {"time":"2026-06-25T08:28:06.463371252+02:00","level":"INFO","msg":"handleInformTeardown: server shutdown complete","id":"1(@)"}
14
+ {"time":"2026-06-25T08:28:06.463383211+02:00","level":"INFO","msg":"connection: ManageConnectionData: connection closed","id":"1(@)"}
15
+ {"time":"2026-06-25T08:28:06.463392135+02:00","level":"INFO","msg":"server is closed"}
checkpoints/codi_frozen_1.5b_logit/wandb/offline-run-20260624_215746-xw4jltek/logs/debug-internal.log ADDED
@@ -0,0 +1,11 @@
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"time":"2026-06-24T21:57:47.314707899+02:00","level":"INFO","msg":"wandb-core"}
2
+ {"time":"2026-06-24T21:57:47.314719289+02:00","level":"INFO","msg":"stream: starting","core version":"0.27.2"}
3
+ {"time":"2026-06-24T21:57:47.43578674+02:00","level":"WARN","msg":"featurechecker: GraphQL client is nil, skipping feature loading"}
4
+ {"time":"2026-06-24T21:57:47.435801873+02:00","level":"WARN","msg":"featurechecker: GraphQL client is nil, skipping feature loading"}
5
+ {"time":"2026-06-24T21:57:47.43581632+02:00","level":"INFO","msg":"stream: created new stream","id":"xw4jltek"}
6
+ {"time":"2026-06-24T21:57:47.435950364+02:00","level":"INFO","msg":"handler: started"}
7
+ {"time":"2026-06-24T21:57:47.43887291+02:00","level":"INFO","msg":"stream: started"}
8
+ {"time":"2026-06-24T21:57:47.438988244+02:00","level":"INFO","msg":"sender: started"}
9
+ {"time":"2026-06-24T21:57:47.438991475+02:00","level":"INFO","msg":"writer: started","stream_id":"xw4jltek"}
10
+ {"time":"2026-06-24T21:57:47.455584204+02:00","level":"WARN","msg":"featurechecker: GraphQL client is nil, skipping feature loading"}
11
+ {"time":"2026-06-24T21:57:47.45559365+02:00","level":"WARN","msg":"runupserter: server does not expand metric globs but the x_server_side_expand_glob_metrics setting is set; ignoring"}
checkpoints/codi_frozen_1.5b_logit/wandb/offline-run-20260624_215746-xw4jltek/logs/debug.log ADDED
@@ -0,0 +1,20 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ 2026-06-24 21:57:46,867 INFO MainThread:689585 [wandb_setup.py:_flush():81] Current SDK version is 0.27.2
2
+ 2026-06-24 21:57:46,867 INFO MainThread:689585 [wandb_setup.py:_flush():81] Configure stats pid to 689585
3
+ 2026-06-24 21:57:46,867 INFO MainThread:689585 [wandb_setup.py:_flush():81] Loading settings from environment variables
4
+ 2026-06-24 21:57:46,867 INFO MainThread:689585 [wandb_init.py:setup_run_log_directory():723] Logging user logs to model_weights/codi_frozen_1.5b_logit/wandb/offline-run-20260624_215746-xw4jltek/logs/debug.log
5
+ 2026-06-24 21:57:46,867 INFO MainThread:689585 [wandb_init.py:setup_run_log_directory():724] Logging internal logs to model_weights/codi_frozen_1.5b_logit/wandb/offline-run-20260624_215746-xw4jltek/logs/debug-internal.log
6
+ 2026-06-24 21:57:46,867 INFO MainThread:689585 [wandb_init.py:init():766] calling init triggers
7
+ 2026-06-24 21:57:46,867 INFO MainThread:689585 [wandb_init.py:init():771] wandb.init called with sweep_config: {}
8
+ config: {'model': 'model_weights/sft1.5b_lr1e5_bs32/checkpoint-6936', 'output_dir': 'model_weights/codi_frozen_1.5b_logit', 'sources': ['mbpp', 'humaneval', 'pyx'], 'cache_dir': 'data/cache/codi_train', 'n_samples': -1, 'max_seq_len': 3072, 'max_frames': -1, 'latent_steps': 1, 'epochs': 5.0, 'lr': 1e-05, 'batch_size': 1, 'grad_accum': 4, 'max_steps': 3000, 'save_steps': 500, 'alpha': 0.0, 'beta': 1.0, 'gamma': 1.0, 'kd_layers': [-1], 'frozen_teacher': 'model_weights/sft1.5b_lr1e5_bs32/checkpoint-6936', 'kd_target': 'logit', 'kd_temp': 2.0, 'single_anchor': False, 'ss_prob': 0.0, 'ss_ramp_frac': 0.5, '_wandb': {}}
9
+ 2026-06-24 21:57:46,867 INFO MainThread:689585 [wandb_init.py:init():814] starting backend
10
+ 2026-06-24 21:57:47,302 INFO MainThread:689585 [wandb_init.py:init():829] sending inform_init request
11
+ 2026-06-24 21:57:47,439 INFO MainThread:689585 [wandb_init.py:init():834] backend started and connected
12
+ 2026-06-24 21:57:47,441 INFO MainThread:689585 [wandb_init.py:init():904] updated telemetry
13
+ 2026-06-24 21:57:47,454 INFO MainThread:689585 [wandb_init.py:init():927] communicating run to backend with 90.0 second timeout
14
+ 2026-06-24 21:57:47,456 INFO MainThread:689585 [wandb_init.py:init():972] starting run threads in backend
15
+ 2026-06-24 21:57:47,902 INFO MainThread:689585 [wandb_run.py:_console_start():2523] atexit reg
16
+ 2026-06-24 21:57:47,902 INFO MainThread:689585 [wandb_run.py:_redirect():2373] redirect: wrap_raw
17
+ 2026-06-24 21:57:47,902 INFO MainThread:689585 [wandb_run.py:_redirect():2442] Wrapping output streams.
18
+ 2026-06-24 21:57:47,902 INFO MainThread:689585 [wandb_run.py:_redirect():2465] Redirects installed.
19
+ 2026-06-24 21:57:47,907 INFO MainThread:689585 [wandb_init.py:init():1010] run started, returning control to user process
20
+ 2026-06-24 21:57:52,397 INFO MainThread:689585 [wandb_run.py:_config_callback():1346] config_cb None None {'output_dir': 'model_weights/codi_frozen_1.5b_logit', 'overwrite_output_dir': False, 'do_train': False, 'do_eval': False, 'do_predict': False, 'eval_strategy': 'no', 'prediction_loss_only': False, 'per_device_train_batch_size': 1, 'per_device_eval_batch_size': 8, 'per_gpu_train_batch_size': None, 'per_gpu_eval_batch_size': None, 'gradient_accumulation_steps': 4, 'eval_accumulation_steps': None, 'eval_delay': 0, 'torch_empty_cache_steps': None, 'learning_rate': 1e-05, 'weight_decay': 0.1, 'adam_beta1': 0.9, 'adam_beta2': 0.999, 'adam_epsilon': 1e-08, 'max_grad_norm': 1.0, 'num_train_epochs': 5.0, 'max_steps': 3000, 'lr_scheduler_type': 'cosine', 'lr_scheduler_kwargs': None, 'warmup_ratio': 0.03, 'warmup_steps': 0, 'log_level': 'passive', 'log_level_replica': 'warning', 'log_on_each_node': True, 'logging_dir': 'model_weights/codi_frozen_1.5b_logit/runs/Jun24_21-57-47_alvis4-41', 'logging_strategy': 'steps', 'logging_first_step': False, 'logging_steps': 5, 'logging_nan_inf_filter': True, 'save_strategy': 'steps', 'save_steps': 500, 'save_total_limit': None, 'save_safetensors': True, 'save_on_each_node': False, 'save_only_model': False, 'restore_callback_states_from_checkpoint': False, 'no_cuda': False, 'use_cpu': False, 'use_mps_device': False, 'seed': 42, 'data_seed': None, 'jit_mode_eval': False, 'bf16': True, 'fp16': False, 'fp16_opt_level': 'O1', 'half_precision_backend': 'auto', 'bf16_full_eval': False, 'fp16_full_eval': False, 'tf32': None, 'local_rank': 0, 'ddp_backend': None, 'tpu_num_cores': None, 'tpu_metrics_debug': False, 'debug': [], 'dataloader_drop_last': False, 'eval_steps': None, 'dataloader_num_workers': 0, 'dataloader_prefetch_factor': None, 'past_index': -1, 'run_name': None, 'disable_tqdm': False, 'remove_unused_columns': False, 'label_names': [], 'load_best_model_at_end': False, 'metric_for_best_model': None, 'greater_is_better': None, 'ignore_data_skip': False, 'fsdp': [], 'fsdp_min_num_params': 0, 'fsdp_config': {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}, 'fsdp_transformer_layer_cls_to_wrap': None, 'accelerator_config': {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}, 'parallelism_config': None, 'deepspeed': None, 'label_smoothing_factor': 0.0, 'optim': 'paged_adamw_8bit', 'optim_args': None, 'adafactor': False, 'group_by_length': False, 'length_column_name': 'length', 'report_to': ['wandb'], 'project': 'huggingface', 'trackio_space_id': 'trackio', 'ddp_find_unused_parameters': False, 'ddp_bucket_cap_mb': None, 'ddp_broadcast_buffers': None, 'dataloader_pin_memory': True, 'dataloader_persistent_workers': False, 'skip_memory_metrics': True, 'use_legacy_prediction_loop': False, 'push_to_hub': False, 'resume_from_checkpoint': None, 'hub_model_id': None, 'hub_strategy': 'every_save', 'hub_token': '<HUB_TOKEN>', 'hub_private_repo': None, 'hub_always_push': False, 'hub_revision': None, 'gradient_checkpointing': False, 'gradient_checkpointing_kwargs': None, 'include_inputs_for_metrics': False, 'include_for_metrics': [], 'eval_do_concat_batches': True, 'fp16_backend': 'auto', 'push_to_hub_model_id': None, 'push_to_hub_organization': None, 'push_to_hub_token': '<PUSH_TO_HUB_TOKEN>', 'mp_parameters': '', 'auto_find_batch_size': False, 'full_determinism': False, 'torchdynamo': None, 'ray_scope': 'last', 'ddp_timeout': 1800, 'torch_compile': False, 'torch_compile_backend': None, 'torch_compile_mode': None, 'include_tokens_per_second': False, 'include_num_input_tokens_seen': 'no', 'neftune_noise_alpha': None, 'optim_target_modules': None, 'batch_eval_metrics': False, 'eval_on_start': False, 'use_liger_kernel': False, 'liger_kernel_config': None, 'eval_use_gather_object': False, 'average_tokens_across_devices': True}
checkpoints/codi_frozen_1.5b_logit/wandb/offline-run-20260624_215746-xw4jltek/run-xw4jltek.wandb ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:85a8c263c95b8437644ffbacd7699cb3fdfc80c5810c93699be39301f190191c
3
+ size 5439488