Hugging Face's logo

ASSERT-KTH
/
codi-trace

Safetensors
Model card Files
xet
 Community
sirui6011 commited on
Commit
d3bb630
·
verified ·
1 Parent(s): dc871d8

add checkpoints/codi3b_a1.0_b1.0_g1.0_ls1

Browse files
Files changed (41) hide show
  1. .gitattributes +4 -0
  2. checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-1000/added_tokens.json +35 -0
  3. checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-1000/chat_template.jinja +54 -0
  4. checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-1000/config.json +66 -0
  5. checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-1000/merges.txt +0 -0
  6. checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-1000/pytorch_model.bin +3 -0
  7. checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-1000/special_tokens_map.json +31 -0
  8. checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-1000/thought_projector.pt +3 -0
  9. checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-1000/tokenizer.json +3 -0
  10. checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-1000/tokenizer_config.json +295 -0
  11. checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-1000/trainer_state.json +2034 -0
  12. checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-1000/vocab.json +0 -0
  13. checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-1500/added_tokens.json +35 -0
  14. checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-1500/chat_template.jinja +54 -0
  15. checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-1500/config.json +66 -0
  16. checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-1500/merges.txt +0 -0
  17. checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-1500/pytorch_model.bin +3 -0
  18. checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-1500/special_tokens_map.json +31 -0
  19. checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-1500/thought_projector.pt +3 -0
  20. checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-1500/tokenizer.json +3 -0
  21. checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-1500/tokenizer_config.json +295 -0
  22. checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-1500/trainer_state.json +3034 -0
  23. checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-1500/vocab.json +0 -0
  24. checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-500/added_tokens.json +35 -0
  25. checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-500/chat_template.jinja +54 -0
  26. checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-500/config.json +66 -0
  27. checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-500/merges.txt +0 -0
  28. checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-500/pytorch_model.bin +3 -0
  29. checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-500/special_tokens_map.json +31 -0
  30. checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-500/thought_projector.pt +3 -0
  31. checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-500/tokenizer.json +3 -0
  32. checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-500/tokenizer_config.json +295 -0
  33. checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-500/trainer_state.json +1034 -0
  34. checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-500/vocab.json +0 -0
  35. checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/wandb/debug-internal.log +11 -0
  36. checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/wandb/debug.log +20 -0
  37. checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/wandb/offline-run-20260619_215302-txanrr5l/files/requirements.txt +222 -0
  38. checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/wandb/offline-run-20260619_215302-txanrr5l/logs/debug-core.log +12 -0
  39. checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/wandb/offline-run-20260619_215302-txanrr5l/logs/debug-internal.log +11 -0
  40. checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/wandb/offline-run-20260619_215302-txanrr5l/logs/debug.log +20 -0
  41. checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/wandb/offline-run-20260619_215302-txanrr5l/run-txanrr5l.wandb +3 -0
.gitattributes CHANGED
@@ -68,3 +68,7 @@ checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-400/tokenizer.json filter=lfs d
68
  checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-600/tokenizer.json filter=lfs diff=lfs merge=lfs -text
69
  checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-800/tokenizer.json filter=lfs diff=lfs merge=lfs -text
70
  checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/wandb/offline-run-20260620_205602-mgxj3qvh/run-mgxj3qvh.wandb filter=lfs diff=lfs merge=lfs -text
 
 
 
 
 
68
  checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-600/tokenizer.json filter=lfs diff=lfs merge=lfs -text
69
  checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-800/tokenizer.json filter=lfs diff=lfs merge=lfs -text
70
  checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/wandb/offline-run-20260620_205602-mgxj3qvh/run-mgxj3qvh.wandb filter=lfs diff=lfs merge=lfs -text
71
+ checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-1000/tokenizer.json filter=lfs diff=lfs merge=lfs -text
72
+ checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-1500/tokenizer.json filter=lfs diff=lfs merge=lfs -text
73
+ checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-500/tokenizer.json filter=lfs diff=lfs merge=lfs -text
74
+ checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/wandb/offline-run-20260619_215302-txanrr5l/run-txanrr5l.wandb filter=lfs diff=lfs merge=lfs -text
checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/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/codi3b_a1.0_b1.0_g1.0_ls1/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 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 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/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-1000/config.json ADDED
@@ -0,0 +1,66 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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": 2048,
10
+ "initializer_range": 0.02,
11
+ "intermediate_size": 11008,
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
+ "full_attention",
42
+ "full_attention",
43
+ "full_attention",
44
+ "full_attention",
45
+ "full_attention",
46
+ "full_attention",
47
+ "full_attention",
48
+ "full_attention"
49
+ ],
50
+ "max_position_embeddings": 32768,
51
+ "max_window_layers": 36,
52
+ "model_type": "qwen2",
53
+ "num_attention_heads": 16,
54
+ "num_hidden_layers": 36,
55
+ "num_key_value_heads": 2,
56
+ "pad_token_id": 151643,
57
+ "rms_norm_eps": 1e-06,
58
+ "rope_scaling": null,
59
+ "rope_theta": 1000000.0,
60
+ "sliding_window": null,
61
+ "tie_word_embeddings": true,
62
+ "transformers_version": "4.57.6",
63
+ "use_cache": true,
64
+ "use_sliding_window": false,
65
+ "vocab_size": 151676
66
+ }
checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-1000/merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-1000/pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e193492b97683551f3f7a26a0d3abee5364e78db12749005cb43325f25ffd05c
3
+ size 6187858991
checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/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/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-1000/thought_projector.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:564e5c949f450a26bf760d0980249bb82b0adbc2bd0142cc2772b36163fde6e9
3
+ size 16788033
checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/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/codi3b_a1.0_b1.0_g1.0_ls1/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/codi3b_a1.0_b1.0_g1.0_ls1/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": 42.0,
15
+ "kd_loss": 0.2894527316093445,
16
+ "learning_rate": 4.444444444444445e-07,
17
+ "loss": 0.5205,
18
+ "step": 5,
19
+ "student_loss": 0.12883709371089935,
20
+ "teacher_loss": 0.001858799601905048
21
+ },
22
+ {
23
+ "epoch": 0.0028835063437139563,
24
+ "grad_norm": 31.875,
25
+ "kd_loss": 0.2677682042121887,
26
+ "learning_rate": 1.0000000000000002e-06,
27
+ "loss": 0.4924,
28
+ "step": 10,
29
+ "student_loss": 0.17924517393112183,
30
+ "teacher_loss": 0.005514730233699083
31
+ },
32
+ {
33
+ "epoch": 0.004325259515570935,
34
+ "grad_norm": 33.25,
35
+ "kd_loss": 0.23119957745075226,
36
+ "learning_rate": 1.5555555555555558e-06,
37
+ "loss": 0.4551,
38
+ "step": 15,
39
+ "student_loss": 0.062010008841753006,
40
+ "teacher_loss": 0.0011852148454636335
41
+ },
42
+ {
43
+ "epoch": 0.0057670126874279125,
44
+ "grad_norm": 25.5,
45
+ "kd_loss": 0.20175831019878387,
46
+ "learning_rate": 2.1111111111111114e-06,
47
+ "loss": 0.4171,
48
+ "step": 20,
49
+ "student_loss": 0.08724690973758698,
50
+ "teacher_loss": 0.0005447491421364248
51
+ },
52
+ {
53
+ "epoch": 0.00720876585928489,
54
+ "grad_norm": 14.0,
55
+ "kd_loss": 0.20470531284809113,
56
+ "learning_rate": 2.666666666666667e-06,
57
+ "loss": 0.3355,
58
+ "step": 25,
59
+ "student_loss": 0.028787745162844658,
60
+ "teacher_loss": 0.0008465723367407918
61
+ },
62
+ {
63
+ "epoch": 0.00865051903114187,
64
+ "grad_norm": 7.59375,
65
+ "kd_loss": 0.17482763528823853,
66
+ "learning_rate": 3.2222222222222227e-06,
67
+ "loss": 0.2813,
68
+ "step": 30,
69
+ "student_loss": 0.04111301898956299,
70
+ "teacher_loss": 0.0003613036824390292
71
+ },
72
+ {
73
+ "epoch": 0.010092272202998846,
74
+ "grad_norm": 5.46875,
75
+ "kd_loss": 0.1520802527666092,
76
+ "learning_rate": 3.777777777777778e-06,
77
+ "loss": 0.2499,
78
+ "step": 35,
79
+ "student_loss": 0.029766706749796867,
80
+ "teacher_loss": 0.0002796741609927267
81
+ },
82
+ {
83
+ "epoch": 0.011534025374855825,
84
+ "grad_norm": 4.4375,
85
+ "kd_loss": 0.11269068717956543,
86
+ "learning_rate": 4.333333333333334e-06,
87
+ "loss": 0.2082,
88
+ "step": 40,
89
+ "student_loss": 0.05098738893866539,
90
+ "teacher_loss": 0.04165218397974968
91
+ },
92
+ {
93
+ "epoch": 0.012975778546712802,
94
+ "grad_norm": 3.078125,
95
+ "kd_loss": 0.20585855841636658,
96
+ "learning_rate": 4.888888888888889e-06,
97
+ "loss": 0.1605,
98
+ "step": 45,
99
+ "student_loss": 0.08690722286701202,
100
+ "teacher_loss": 0.0026482592802494764
101
+ },
102
+ {
103
+ "epoch": 0.01441753171856978,
104
+ "grad_norm": 2.734375,
105
+ "kd_loss": 0.074299156665802,
106
+ "learning_rate": 5.444444444444445e-06,
107
+ "loss": 0.1578,
108
+ "step": 50,
109
+ "student_loss": 0.017733801156282425,
110
+ "teacher_loss": 0.008056458085775375
111
+ },
112
+ {
113
+ "epoch": 0.015859284890426758,
114
+ "grad_norm": 2.78125,
115
+ "kd_loss": 0.062237780541181564,
116
+ "learning_rate": 6e-06,
117
+ "loss": 0.135,
118
+ "step": 55,
119
+ "student_loss": 0.0012035972904413939,
120
+ "teacher_loss": 0.000626895169261843
121
+ },
122
+ {
123
+ "epoch": 0.01730103806228374,
124
+ "grad_norm": 2.65625,
125
+ "kd_loss": 0.05602008104324341,
126
+ "learning_rate": 6.555555555555556e-06,
127
+ "loss": 0.1184,
128
+ "step": 60,
129
+ "student_loss": 0.002416773699223995,
130
+ "teacher_loss": 0.0013880858896300197
131
+ },
132
+ {
133
+ "epoch": 0.018742791234140715,
134
+ "grad_norm": 4.65625,
135
+ "kd_loss": 0.051335666328668594,
136
+ "learning_rate": 7.111111111111112e-06,
137
+ "loss": 0.1118,
138
+ "step": 65,
139
+ "student_loss": 0.027237623929977417,
140
+ "teacher_loss": 0.03812175989151001
141
+ },
142
+ {
143
+ "epoch": 0.020184544405997693,
144
+ "grad_norm": 1.7109375,
145
+ "kd_loss": 0.0913488045334816,
146
+ "learning_rate": 7.666666666666667e-06,
147
+ "loss": 0.1042,
148
+ "step": 70,
149
+ "student_loss": 0.015874937176704407,
150
+ "teacher_loss": 0.0013995245099067688
151
+ },
152
+ {
153
+ "epoch": 0.02162629757785467,
154
+ "grad_norm": 2.890625,
155
+ "kd_loss": 0.036495719105005264,
156
+ "learning_rate": 8.222222222222222e-06,
157
+ "loss": 0.0961,
158
+ "step": 75,
159
+ "student_loss": 0.0018275808542966843,
160
+ "teacher_loss": 0.000998618663288653
161
+ },
162
+ {
163
+ "epoch": 0.02306805074971165,
164
+ "grad_norm": 2.8125,
165
+ "kd_loss": 0.033367447555065155,
166
+ "learning_rate": 8.777777777777778e-06,
167
+ "loss": 0.0811,
168
+ "step": 80,
169
+ "student_loss": 0.001792682334780693,
170
+ "teacher_loss": 0.0010247478494420648
171
+ },
172
+ {
173
+ "epoch": 0.024509803921568627,
174
+ "grad_norm": 2.734375,
175
+ "kd_loss": 0.0299057699739933,
176
+ "learning_rate": 9.333333333333334e-06,
177
+ "loss": 0.0827,
178
+ "step": 85,
179
+ "student_loss": 0.0028650059830397367,
180
+ "teacher_loss": 0.004752026870846748
181
+ },
182
+ {
183
+ "epoch": 0.025951557093425604,
184
+ "grad_norm": 1.7890625,
185
+ "kd_loss": 0.08093900978565216,
186
+ "learning_rate": 9.88888888888889e-06,
187
+ "loss": 0.0755,
188
+ "step": 90,
189
+ "student_loss": 0.019452862441539764,
190
+ "teacher_loss": 0.0028147350531071424
191
+ },
192
+ {
193
+ "epoch": 0.027393310265282585,
194
+ "grad_norm": 1.4296875,
195
+ "kd_loss": 0.04993116110563278,
196
+ "learning_rate": 9.99995337991255e-06,
197
+ "loss": 0.0683,
198
+ "step": 95,
199
+ "student_loss": 0.009235645644366741,
200
+ "teacher_loss": 0.006103724706918001
201
+ },
202
+ {
203
+ "epoch": 0.02883506343713956,
204
+ "grad_norm": 1.546875,
205
+ "kd_loss": 0.02219005487859249,
206
+ "learning_rate": 9.999763987297266e-06,
207
+ "loss": 0.0591,
208
+ "step": 100,
209
+ "student_loss": 0.0009554760181345046,
210
+ "teacher_loss": 0.000685701088514179
211
+ },
212
+ {
213
+ "epoch": 0.03027681660899654,
214
+ "grad_norm": 2.53125,
215
+ "kd_loss": 0.03886059299111366,
216
+ "learning_rate": 9.999428913912876e-06,
217
+ "loss": 0.0632,
218
+ "step": 105,
219
+ "student_loss": 0.002928090514615178,
220
+ "teacher_loss": 0.00032716552959755063
221
+ },
222
+ {
223
+ "epoch": 0.031718569780853516,
224
+ "grad_norm": 2.171875,
225
+ "kd_loss": 0.04434794560074806,
226
+ "learning_rate": 9.998948169522587e-06,
227
+ "loss": 0.063,
228
+ "step": 110,
229
+ "student_loss": 0.033939410001039505,
230
+ "teacher_loss": 0.017983509227633476
231
+ },
232
+ {
233
+ "epoch": 0.03316032295271049,
234
+ "grad_norm": 1.3046875,
235
+ "kd_loss": 0.03771444410085678,
236
+ "learning_rate": 9.998321768134101e-06,
237
+ "loss": 0.0606,
238
+ "step": 115,
239
+ "student_loss": 0.0007961422670632601,
240
+ "teacher_loss": 0.00038347821100614965
241
+ },
242
+ {
243
+ "epoch": 0.03460207612456748,
244
+ "grad_norm": 1.6953125,
245
+ "kd_loss": 0.02649487555027008,
246
+ "learning_rate": 9.997549727999212e-06,
247
+ "loss": 0.0532,
248
+ "step": 120,
249
+ "student_loss": 0.0012731363531202078,
250
+ "teacher_loss": 0.0007500122301280499
251
+ },
252
+ {
253
+ "epoch": 0.036043829296424454,
254
+ "grad_norm": 2.09375,
255
+ "kd_loss": 0.057950329035520554,
256
+ "learning_rate": 9.996632071613257e-06,
257
+ "loss": 0.0601,
258
+ "step": 125,
259
+ "student_loss": 0.03109952248632908,
260
+ "teacher_loss": 0.000994477653875947
261
+ },
262
+ {
263
+ "epoch": 0.03748558246828143,
264
+ "grad_norm": 0.95703125,
265
+ "kd_loss": 0.026637699455022812,
266
+ "learning_rate": 9.995568825714479e-06,
267
+ "loss": 0.0574,
268
+ "step": 130,
269
+ "student_loss": 0.0008338625193573534,
270
+ "teacher_loss": 0.0006465786718763411
271
+ },
272
+ {
273
+ "epoch": 0.03892733564013841,
274
+ "grad_norm": 1.3515625,
275
+ "kd_loss": 0.04235714301466942,
276
+ "learning_rate": 9.994360021283235e-06,
277
+ "loss": 0.0611,
278
+ "step": 135,
279
+ "student_loss": 0.001498939236626029,
280
+ "teacher_loss": 0.0005972529761493206
281
+ },
282
+ {
283
+ "epoch": 0.040369088811995385,
284
+ "grad_norm": 1.46875,
285
+ "kd_loss": 0.024005092680454254,
286
+ "learning_rate": 9.993005693541098e-06,
287
+ "loss": 0.0562,
288
+ "step": 140,
289
+ "student_loss": 0.0008620141888968647,
290
+ "teacher_loss": 0.0004937360063195229
291
+ },
292
+ {
293
+ "epoch": 0.04181084198385236,
294
+ "grad_norm": 5.5,
295
+ "kd_loss": 0.07175353169441223,
296
+ "learning_rate": 9.991505881949837e-06,
297
+ "loss": 0.0539,
298
+ "step": 145,
299
+ "student_loss": 0.03820747137069702,
300
+ "teacher_loss": 0.006569729186594486
301
+ },
302
+ {
303
+ "epoch": 0.04325259515570934,
304
+ "grad_norm": 1.2109375,
305
+ "kd_loss": 0.016246870160102844,
306
+ "learning_rate": 9.989860630210251e-06,
307
+ "loss": 0.0538,
308
+ "step": 150,
309
+ "student_loss": 0.0014459972735494375,
310
+ "teacher_loss": 0.001147772534750402
311
+ },
312
+ {
313
+ "epoch": 0.04469434832756632,
314
+ "grad_norm": 1.4765625,
315
+ "kd_loss": 0.02643739990890026,
316
+ "learning_rate": 9.98806998626091e-06,
317
+ "loss": 0.0592,
318
+ "step": 155,
319
+ "student_loss": 0.0027810579631477594,
320
+ "teacher_loss": 0.000860289903357625
321
+ },
322
+ {
323
+ "epoch": 0.0461361014994233,
324
+ "grad_norm": 2.40625,
325
+ "kd_loss": 0.029045773670077324,
326
+ "learning_rate": 9.98613400227676e-06,
327
+ "loss": 0.0597,
328
+ "step": 160,
329
+ "student_loss": 0.08056973665952682,
330
+ "teacher_loss": 0.04562501981854439
331
+ },
332
+ {
333
+ "epoch": 0.04757785467128028,
334
+ "grad_norm": 1.4140625,
335
+ "kd_loss": 0.03424260392785072,
336
+ "learning_rate": 9.984052734667586e-06,
337
+ "loss": 0.0523,
338
+ "step": 165,
339
+ "student_loss": 0.0027138988953083754,
340
+ "teacher_loss": 0.005776015110313892
341
+ },
342
+ {
343
+ "epoch": 0.049019607843137254,
344
+ "grad_norm": 2.859375,
345
+ "kd_loss": 0.019065098837018013,
346
+ "learning_rate": 9.981826244076387e-06,
347
+ "loss": 0.0562,
348
+ "step": 170,
349
+ "student_loss": 0.0007321524317376316,
350
+ "teacher_loss": 0.000522188376635313
351
+ },
352
+ {
353
+ "epoch": 0.05046136101499423,
354
+ "grad_norm": 1.7578125,
355
+ "kd_loss": 0.027638504281640053,
356
+ "learning_rate": 9.979454595377594e-06,
357
+ "loss": 0.0529,
358
+ "step": 175,
359
+ "student_loss": 0.004792944528162479,
360
+ "teacher_loss": 0.000635432661511004
361
+ },
362
+ {
363
+ "epoch": 0.05190311418685121,
364
+ "grad_norm": 2.0,
365
+ "kd_loss": 0.03646935150027275,
366
+ "learning_rate": 9.97693785767519e-06,
367
+ "loss": 0.0514,
368
+ "step": 180,
369
+ "student_loss": 0.0014396726619452238,
370
+ "teacher_loss": 0.0003402946167625487
371
+ },
372
+ {
373
+ "epoch": 0.05334486735870819,
374
+ "grad_norm": 1.453125,
375
+ "kd_loss": 0.03992177173495293,
376
+ "learning_rate": 9.9742761043007e-06,
377
+ "loss": 0.0582,
378
+ "step": 185,
379
+ "student_loss": 0.03759889304637909,
380
+ "teacher_loss": 0.00029035221086815
381
+ },
382
+ {
383
+ "epoch": 0.05478662053056517,
384
+ "grad_norm": 3.890625,
385
+ "kd_loss": 0.05405907332897186,
386
+ "learning_rate": 9.971469412811032e-06,
387
+ "loss": 0.0527,
388
+ "step": 190,
389
+ "student_loss": 0.004218568559736013,
390
+ "teacher_loss": 0.0009858696721494198
391
+ },
392
+ {
393
+ "epoch": 0.056228373702422146,
394
+ "grad_norm": 2.375,
395
+ "kd_loss": 0.05194515734910965,
396
+ "learning_rate": 9.968517864986247e-06,
397
+ "loss": 0.0524,
398
+ "step": 195,
399
+ "student_loss": 0.005763455759733915,
400
+ "teacher_loss": 0.0004817036970052868
401
+ },
402
+ {
403
+ "epoch": 0.05767012687427912,
404
+ "grad_norm": 1.2109375,
405
+ "kd_loss": 0.036061037331819534,
406
+ "learning_rate": 9.965421546827153e-06,
407
+ "loss": 0.064,
408
+ "step": 200,
409
+ "student_loss": 0.0005494994111359119,
410
+ "teacher_loss": 0.00020005257101729512
411
+ },
412
+ {
413
+ "epoch": 0.0591118800461361,
414
+ "grad_norm": 2.40625,
415
+ "kd_loss": 0.02808363176882267,
416
+ "learning_rate": 9.962180548552812e-06,
417
+ "loss": 0.0577,
418
+ "step": 205,
419
+ "student_loss": 0.03726515918970108,
420
+ "teacher_loss": 0.028193596750497818
421
+ },
422
+ {
423
+ "epoch": 0.06055363321799308,
424
+ "grad_norm": 4.15625,
425
+ "kd_loss": 0.03680839017033577,
426
+ "learning_rate": 9.958794964597906e-06,
427
+ "loss": 0.0515,
428
+ "step": 210,
429
+ "student_loss": 0.0011522703571245074,
430
+ "teacher_loss": 0.0004914712626487017
431
+ },
432
+ {
433
+ "epoch": 0.061995386389850055,
434
+ "grad_norm": 5.65625,
435
+ "kd_loss": 0.037980884313583374,
436
+ "learning_rate": 9.955264893609986e-06,
437
+ "loss": 0.0559,
438
+ "step": 215,
439
+ "student_loss": 0.0005635322886519134,
440
+ "teacher_loss": 0.00027164584025740623
441
+ },
442
+ {
443
+ "epoch": 0.06343713956170703,
444
+ "grad_norm": 4.0,
445
+ "kd_loss": 0.04362543299794197,
446
+ "learning_rate": 9.951590438446597e-06,
447
+ "loss": 0.0507,
448
+ "step": 220,
449
+ "student_loss": 0.032083068042993546,
450
+ "teacher_loss": 0.010480552911758423
451
+ },
452
+ {
453
+ "epoch": 0.06487889273356401,
454
+ "grad_norm": 1.3125,
455
+ "kd_loss": 0.02105102129280567,
456
+ "learning_rate": 9.947771706172285e-06,
457
+ "loss": 0.0605,
458
+ "step": 225,
459
+ "student_loss": 0.038467925041913986,
460
+ "teacher_loss": 0.0018907957710325718
461
+ },
462
+ {
463
+ "epoch": 0.06632064590542099,
464
+ "grad_norm": 2.25,
465
+ "kd_loss": 0.04121282696723938,
466
+ "learning_rate": 9.943808808055472e-06,
467
+ "loss": 0.0495,
468
+ "step": 230,
469
+ "student_loss": 0.01008139830082655,
470
+ "teacher_loss": 0.003380328882485628
471
+ },
472
+ {
473
+ "epoch": 0.06776239907727798,
474
+ "grad_norm": 9.5,
475
+ "kd_loss": 0.07618479430675507,
476
+ "learning_rate": 9.93970185956522e-06,
477
+ "loss": 0.0516,
478
+ "step": 235,
479
+ "student_loss": 0.010011663660407066,
480
+ "teacher_loss": 0.0004287679912522435
481
+ },
482
+ {
483
+ "epoch": 0.06920415224913495,
484
+ "grad_norm": 2.796875,
485
+ "kd_loss": 0.031842563301324844,
486
+ "learning_rate": 9.93545098036786e-06,
487
+ "loss": 0.0567,
488
+ "step": 240,
489
+ "student_loss": 0.0031024671625345945,
490
+ "teacher_loss": 0.0003865428443532437
491
+ },
492
+ {
493
+ "epoch": 0.07064590542099193,
494
+ "grad_norm": 1.921875,
495
+ "kd_loss": 0.01385761983692646,
496
+ "learning_rate": 9.931056294323507e-06,
497
+ "loss": 0.048,
498
+ "step": 245,
499
+ "student_loss": 0.011989925056695938,
500
+ "teacher_loss": 0.015163863077759743
501
+ },
502
+ {
503
+ "epoch": 0.07208765859284891,
504
+ "grad_norm": 1.8125,
505
+ "kd_loss": 0.020028578117489815,
506
+ "learning_rate": 9.926517929482454e-06,
507
+ "loss": 0.0514,
508
+ "step": 250,
509
+ "student_loss": 0.0005340924835763872,
510
+ "teacher_loss": 0.000291565025690943
511
+ },
512
+ {
513
+ "epoch": 0.07352941176470588,
514
+ "grad_norm": 1.8359375,
515
+ "kd_loss": 0.01568070985376835,
516
+ "learning_rate": 9.921836018081438e-06,
517
+ "loss": 0.0512,
518
+ "step": 255,
519
+ "student_loss": 0.0003255972405895591,
520
+ "teacher_loss": 0.0003198981285095215
521
+ },
522
+ {
523
+ "epoch": 0.07497116493656286,
524
+ "grad_norm": 1.65625,
525
+ "kd_loss": 0.03392383083701134,
526
+ "learning_rate": 9.917010696539793e-06,
527
+ "loss": 0.0577,
528
+ "step": 260,
529
+ "student_loss": 0.0405547134578228,
530
+ "teacher_loss": 0.03275086730718613
531
+ },
532
+ {
533
+ "epoch": 0.07641291810841984,
534
+ "grad_norm": 4.96875,
535
+ "kd_loss": 0.018212661147117615,
536
+ "learning_rate": 9.912042105455462e-06,
537
+ "loss": 0.0565,
538
+ "step": 265,
539
+ "student_loss": 0.0005924144061282277,
540
+ "teacher_loss": 0.0004515529435593635
541
+ },
542
+ {
543
+ "epoch": 0.07785467128027682,
544
+ "grad_norm": 1.484375,
545
+ "kd_loss": 0.020866869017481804,
546
+ "learning_rate": 9.906930389600917e-06,
547
+ "loss": 0.0593,
548
+ "step": 270,
549
+ "student_loss": 0.0013341928133741021,
550
+ "teacher_loss": 0.0011696964502334595
551
+ },
552
+ {
553
+ "epoch": 0.07929642445213379,
554
+ "grad_norm": 2.1875,
555
+ "kd_loss": 0.02725479006767273,
556
+ "learning_rate": 9.901675697918925e-06,
557
+ "loss": 0.0511,
558
+ "step": 275,
559
+ "student_loss": 0.048831257969141006,
560
+ "teacher_loss": 0.035170599818229675
561
+ },
562
+ {
563
+ "epoch": 0.08073817762399077,
564
+ "grad_norm": 4.75,
565
+ "kd_loss": 0.03362221270799637,
566
+ "learning_rate": 9.896278183518216e-06,
567
+ "loss": 0.0509,
568
+ "step": 280,
569
+ "student_loss": 0.0004928835551254451,
570
+ "teacher_loss": 0.00017192006635013968
571
+ },
572
+ {
573
+ "epoch": 0.08217993079584775,
574
+ "grad_norm": 2.921875,
575
+ "kd_loss": 0.021101269870996475,
576
+ "learning_rate": 9.890738003669029e-06,
577
+ "loss": 0.0572,
578
+ "step": 285,
579
+ "student_loss": 0.000973534828517586,
580
+ "teacher_loss": 0.0008148897904902697
581
+ },
582
+ {
583
+ "epoch": 0.08362168396770472,
584
+ "grad_norm": 2.203125,
585
+ "kd_loss": 0.017731280997395515,
586
+ "learning_rate": 9.885055319798514e-06,
587
+ "loss": 0.0582,
588
+ "step": 290,
589
+ "student_loss": 0.014440718106925488,
590
+ "teacher_loss": 0.012271302752196789
591
+ },
592
+ {
593
+ "epoch": 0.0850634371395617,
594
+ "grad_norm": 2.46875,
595
+ "kd_loss": 0.01968582533299923,
596
+ "learning_rate": 9.879230297486034e-06,
597
+ "loss": 0.0561,
598
+ "step": 295,
599
+ "student_loss": 0.0010608865413814783,
600
+ "teacher_loss": 0.0008463782141916454
601
+ },
602
+ {
603
+ "epoch": 0.08650519031141868,
604
+ "grad_norm": 3.34375,
605
+ "kd_loss": 0.05237875133752823,
606
+ "learning_rate": 9.873263106458353e-06,
607
+ "loss": 0.0526,
608
+ "step": 300,
609
+ "student_loss": 0.017360921949148178,
610
+ "teacher_loss": 0.001965837087482214
611
+ },
612
+ {
613
+ "epoch": 0.08794694348327567,
614
+ "grad_norm": 2.296875,
615
+ "kd_loss": 0.023132195696234703,
616
+ "learning_rate": 9.86715392058467e-06,
617
+ "loss": 0.0485,
618
+ "step": 305,
619
+ "student_loss": 0.007244116626679897,
620
+ "teacher_loss": 0.003649009158834815
621
+ },
622
+ {
623
+ "epoch": 0.08938869665513265,
624
+ "grad_norm": 2.671875,
625
+ "kd_loss": 0.020483316853642464,
626
+ "learning_rate": 9.860902917871566e-06,
627
+ "loss": 0.0535,
628
+ "step": 310,
629
+ "student_loss": 0.0007183508132584393,
630
+ "teacher_loss": 0.0004938851925544441
631
+ },
632
+ {
633
+ "epoch": 0.09083044982698962,
634
+ "grad_norm": 2.828125,
635
+ "kd_loss": 0.04086384177207947,
636
+ "learning_rate": 9.854510280457819e-06,
637
+ "loss": 0.0521,
638
+ "step": 315,
639
+ "student_loss": 0.010588191449642181,
640
+ "teacher_loss": 0.00029907570569776
641
+ },
642
+ {
643
+ "epoch": 0.0922722029988466,
644
+ "grad_norm": 2.75,
645
+ "kd_loss": 0.026799874380230904,
646
+ "learning_rate": 9.847976194609083e-06,
647
+ "loss": 0.055,
648
+ "step": 320,
649
+ "student_loss": 0.0006185687379911542,
650
+ "teacher_loss": 0.0003583572688512504
651
+ },
652
+ {
653
+ "epoch": 0.09371395617070358,
654
+ "grad_norm": 2.265625,
655
+ "kd_loss": 0.07637664675712585,
656
+ "learning_rate": 9.841300850712479e-06,
657
+ "loss": 0.0516,
658
+ "step": 325,
659
+ "student_loss": 0.008596767671406269,
660
+ "teacher_loss": 0.0225156731903553
661
+ },
662
+ {
663
+ "epoch": 0.09515570934256055,
664
+ "grad_norm": 1.75,
665
+ "kd_loss": 0.0531776025891304,
666
+ "learning_rate": 9.834484443271035e-06,
667
+ "loss": 0.0557,
668
+ "step": 330,
669
+ "student_loss": 0.019897613674402237,
670
+ "teacher_loss": 0.004191068932414055
671
+ },
672
+ {
673
+ "epoch": 0.09659746251441753,
674
+ "grad_norm": 1.8125,
675
+ "kd_loss": 0.013695458881556988,
676
+ "learning_rate": 9.827527170898019e-06,
677
+ "loss": 0.0488,
678
+ "step": 335,
679
+ "student_loss": 0.0016651605255901814,
680
+ "teacher_loss": 0.0014800777425989509
681
+ },
682
+ {
683
+ "epoch": 0.09803921568627451,
684
+ "grad_norm": 2.3125,
685
+ "kd_loss": 0.027426283806562424,
686
+ "learning_rate": 9.820429236311158e-06,
687
+ "loss": 0.0567,
688
+ "step": 340,
689
+ "student_loss": 0.010532001033425331,
690
+ "teacher_loss": 0.006886747200042009
691
+ },
692
+ {
693
+ "epoch": 0.09948096885813149,
694
+ "grad_norm": 1.8359375,
695
+ "kd_loss": 0.03429238498210907,
696
+ "learning_rate": 9.81319084632673e-06,
697
+ "loss": 0.0564,
698
+ "step": 345,
699
+ "student_loss": 0.007421151269227266,
700
+ "teacher_loss": 0.0005970117053948343
701
+ },
702
+ {
703
+ "epoch": 0.10092272202998846,
704
+ "grad_norm": 2.546875,
705
+ "kd_loss": 0.058330416679382324,
706
+ "learning_rate": 9.805812211853534e-06,
707
+ "loss": 0.0473,
708
+ "step": 350,
709
+ "student_loss": 0.002645631320774555,
710
+ "teacher_loss": 0.001440234947949648
711
+ },
712
+ {
713
+ "epoch": 0.10236447520184544,
714
+ "grad_norm": 2.046875,
715
+ "kd_loss": 0.016124732792377472,
716
+ "learning_rate": 9.798293547886748e-06,
717
+ "loss": 0.0571,
718
+ "step": 355,
719
+ "student_loss": 0.001094816834665835,
720
+ "teacher_loss": 0.0007638070383109152
721
+ },
722
+ {
723
+ "epoch": 0.10380622837370242,
724
+ "grad_norm": 1.765625,
725
+ "kd_loss": 0.015412062406539917,
726
+ "learning_rate": 9.790635073501662e-06,
727
+ "loss": 0.0541,
728
+ "step": 360,
729
+ "student_loss": 0.0010126332053914666,
730
+ "teacher_loss": 0.0007303679012693465
731
+ },
732
+ {
733
+ "epoch": 0.1052479815455594,
734
+ "grad_norm": 1.328125,
735
+ "kd_loss": 0.0318637378513813,
736
+ "learning_rate": 9.782837011847294e-06,
737
+ "loss": 0.0555,
738
+ "step": 365,
739
+ "student_loss": 0.043804578483104706,
740
+ "teacher_loss": 0.0287578534334898
741
+ },
742
+ {
743
+ "epoch": 0.10668973471741638,
744
+ "grad_norm": 3.734375,
745
+ "kd_loss": 0.054416049271821976,
746
+ "learning_rate": 9.774899590139897e-06,
747
+ "loss": 0.064,
748
+ "step": 370,
749
+ "student_loss": 0.02294199727475643,
750
+ "teacher_loss": 0.0003071785031352192
751
+ },
752
+ {
753
+ "epoch": 0.10813148788927336,
754
+ "grad_norm": 3.09375,
755
+ "kd_loss": 0.03518553078174591,
756
+ "learning_rate": 9.766823039656328e-06,
757
+ "loss": 0.0564,
758
+ "step": 375,
759
+ "student_loss": 0.03153789043426514,
760
+ "teacher_loss": 0.020613783970475197
761
+ },
762
+ {
763
+ "epoch": 0.10957324106113034,
764
+ "grad_norm": 4.40625,
765
+ "kd_loss": 0.021105539053678513,
766
+ "learning_rate": 9.758607595727309e-06,
767
+ "loss": 0.0537,
768
+ "step": 380,
769
+ "student_loss": 0.000557993131224066,
770
+ "teacher_loss": 0.0003853706002701074
771
+ },
772
+ {
773
+ "epoch": 0.11101499423298732,
774
+ "grad_norm": 2.03125,
775
+ "kd_loss": 0.03399880975484848,
776
+ "learning_rate": 9.75025349773058e-06,
777
+ "loss": 0.0504,
778
+ "step": 385,
779
+ "student_loss": 0.0006929709343239665,
780
+ "teacher_loss": 0.0005223862244747579
781
+ },
782
+ {
783
+ "epoch": 0.11245674740484429,
784
+ "grad_norm": 1.2421875,
785
+ "kd_loss": 0.022061804309487343,
786
+ "learning_rate": 9.741760989083914e-06,
787
+ "loss": 0.0473,
788
+ "step": 390,
789
+ "student_loss": 0.0003254064649809152,
790
+ "teacher_loss": 0.0003073395637329668
791
+ },
792
+ {
793
+ "epoch": 0.11389850057670127,
794
+ "grad_norm": 1.9453125,
795
+ "kd_loss": 0.021170753985643387,
796
+ "learning_rate": 9.73313031723803e-06,
797
+ "loss": 0.0519,
798
+ "step": 395,
799
+ "student_loss": 0.0008620450389571488,
800
+ "teacher_loss": 0.0008799576899036765
801
+ },
802
+ {
803
+ "epoch": 0.11534025374855825,
804
+ "grad_norm": 1.6875,
805
+ "kd_loss": 0.02962590754032135,
806
+ "learning_rate": 9.724361733669383e-06,
807
+ "loss": 0.0462,
808
+ "step": 400,
809
+ "student_loss": 0.0009699641377665102,
810
+ "teacher_loss": 0.0002524392039049417
811
+ },
812
+ {
813
+ "epoch": 0.11678200692041522,
814
+ "grad_norm": 1.4140625,
815
+ "kd_loss": 0.02815704233944416,
816
+ "learning_rate": 9.715455493872827e-06,
817
+ "loss": 0.0563,
818
+ "step": 405,
819
+ "student_loss": 0.0135729368776083,
820
+ "teacher_loss": 0.00044868470286019146
821
+ },
822
+ {
823
+ "epoch": 0.1182237600922722,
824
+ "grad_norm": 1.8125,
825
+ "kd_loss": 0.03630191460251808,
826
+ "learning_rate": 9.70641185735419e-06,
827
+ "loss": 0.0593,
828
+ "step": 410,
829
+ "student_loss": 0.014492504298686981,
830
+ "teacher_loss": 0.007929121144115925
831
+ },
832
+ {
833
+ "epoch": 0.11966551326412918,
834
+ "grad_norm": 1.6171875,
835
+ "kd_loss": 0.026567015796899796,
836
+ "learning_rate": 9.697231087622691e-06,
837
+ "loss": 0.0585,
838
+ "step": 415,
839
+ "student_loss": 0.004891898948699236,
840
+ "teacher_loss": 0.003953997977077961
841
+ },
842
+ {
843
+ "epoch": 0.12110726643598616,
844
+ "grad_norm": 2.890625,
845
+ "kd_loss": 0.04854276403784752,
846
+ "learning_rate": 9.687913452183277e-06,
847
+ "loss": 0.0584,
848
+ "step": 420,
849
+ "student_loss": 0.04526115953922272,
850
+ "teacher_loss": 0.023706616833806038
851
+ },
852
+ {
853
+ "epoch": 0.12254901960784313,
854
+ "grad_norm": 2.71875,
855
+ "kd_loss": 0.014143871143460274,
856
+ "learning_rate": 9.678459222528825e-06,
857
+ "loss": 0.0539,
858
+ "step": 425,
859
+ "student_loss": 0.0028342436999082565,
860
+ "teacher_loss": 0.003036739770323038
861
+ },
862
+ {
863
+ "epoch": 0.12399077277970011,
864
+ "grad_norm": 2.546875,
865
+ "kd_loss": 0.05789174139499664,
866
+ "learning_rate": 9.668868674132224e-06,
867
+ "loss": 0.0637,
868
+ "step": 430,
869
+ "student_loss": 0.058899883180856705,
870
+ "teacher_loss": 0.064610056579113
871
+ },
872
+ {
873
+ "epoch": 0.1254325259515571,
874
+ "grad_norm": 2.25,
875
+ "kd_loss": 0.021487534046173096,
876
+ "learning_rate": 9.65914208643836e-06,
877
+ "loss": 0.0537,
878
+ "step": 435,
879
+ "student_loss": 0.0009387442842125893,
880
+ "teacher_loss": 0.0007441287161782384
881
+ },
882
+ {
883
+ "epoch": 0.12687427912341406,
884
+ "grad_norm": 6.78125,
885
+ "kd_loss": 0.03997809439897537,
886
+ "learning_rate": 9.649279742855964e-06,
887
+ "loss": 0.0599,
888
+ "step": 440,
889
+ "student_loss": 0.012818059884011745,
890
+ "teacher_loss": 0.007239806931465864
891
+ },
892
+ {
893
+ "epoch": 0.12831603229527105,
894
+ "grad_norm": 2.375,
895
+ "kd_loss": 0.037731364369392395,
896
+ "learning_rate": 9.639281930749363e-06,
897
+ "loss": 0.0586,
898
+ "step": 445,
899
+ "student_loss": 0.03573564440011978,
900
+ "teacher_loss": 0.01907653547823429
901
+ },
902
+ {
903
+ "epoch": 0.12975778546712802,
904
+ "grad_norm": 2.546875,
905
+ "kd_loss": 0.05183478444814682,
906
+ "learning_rate": 9.629148941430096e-06,
907
+ "loss": 0.0617,
908
+ "step": 450,
909
+ "student_loss": 0.002695092000067234,
910
+ "teacher_loss": 0.0027623483911156654
911
+ },
912
+ {
913
+ "epoch": 0.131199538638985,
914
+ "grad_norm": 1.6015625,
915
+ "kd_loss": 0.029434863477945328,
916
+ "learning_rate": 9.618881070148436e-06,
917
+ "loss": 0.0615,
918
+ "step": 455,
919
+ "student_loss": 0.00108905672095716,
920
+ "teacher_loss": 0.0005926509038545191
921
+ },
922
+ {
923
+ "epoch": 0.13264129181084197,
924
+ "grad_norm": 1.5703125,
925
+ "kd_loss": 0.05266507714986801,
926
+ "learning_rate": 9.608478616084784e-06,
927
+ "loss": 0.0548,
928
+ "step": 460,
929
+ "student_loss": 0.016866859048604965,
930
+ "teacher_loss": 0.0005994263920001686
931
+ },
932
+ {
933
+ "epoch": 0.13408304498269896,
934
+ "grad_norm": 4.90625,
935
+ "kd_loss": 0.04152460768818855,
936
+ "learning_rate": 9.597941882340947e-06,
937
+ "loss": 0.0648,
938
+ "step": 465,
939
+ "student_loss": 0.0035157979000359774,
940
+ "teacher_loss": 0.0008752731955610216
941
+ },
942
+ {
943
+ "epoch": 0.13552479815455595,
944
+ "grad_norm": 1.5546875,
945
+ "kd_loss": 0.03477834537625313,
946
+ "learning_rate": 9.587271175931313e-06,
947
+ "loss": 0.0637,
948
+ "step": 470,
949
+ "student_loss": 0.0005260187899693847,
950
+ "teacher_loss": 0.00038832571590319276
951
+ },
952
+ {
953
+ "epoch": 0.13696655132641292,
954
+ "grad_norm": 2.90625,
955
+ "kd_loss": 0.040360964834690094,
956
+ "learning_rate": 9.5764668077739e-06,
957
+ "loss": 0.0567,
958
+ "step": 475,
959
+ "student_loss": 0.0007983366376720369,
960
+ "teacher_loss": 0.0005927964230068028
961
+ },
962
+ {
963
+ "epoch": 0.1384083044982699,
964
+ "grad_norm": 3.3125,
965
+ "kd_loss": 0.04144829884171486,
966
+ "learning_rate": 9.5655290926813e-06,
967
+ "loss": 0.0762,
968
+ "step": 480,
969
+ "student_loss": 0.03820875287055969,
970
+ "teacher_loss": 0.02910608798265457
971
+ },
972
+ {
973
+ "epoch": 0.13985005767012687,
974
+ "grad_norm": 3.25,
975
+ "kd_loss": 0.0166642889380455,
976
+ "learning_rate": 9.554458349351507e-06,
977
+ "loss": 0.0627,
978
+ "step": 485,
979
+ "student_loss": 0.001466175657697022,
980
+ "teacher_loss": 0.0013084234669804573
981
+ },
982
+ {
983
+ "epoch": 0.14129181084198386,
984
+ "grad_norm": 3.5,
985
+ "kd_loss": 0.025684630498290062,
986
+ "learning_rate": 9.54325490035863e-06,
987
+ "loss": 0.0737,
988
+ "step": 490,
989
+ "student_loss": 0.0014865982811897993,
990
+ "teacher_loss": 0.0012143527856096625
991
+ },
992
+ {
993
+ "epoch": 0.14273356401384082,
994
+ "grad_norm": 2.359375,
995
+ "kd_loss": 0.0510389544069767,
996
+ "learning_rate": 9.531919072143488e-06,
997
+ "loss": 0.0776,
998
+ "step": 495,
999
+ "student_loss": 0.004116421099752188,
1000
+ "teacher_loss": 0.002551821991801262
1001
+ },
1002
+ {
1003
+ "epoch": 0.14417531718569782,
1004
+ "grad_norm": 2.109375,
1005
+ "kd_loss": 0.03247147053480148,
1006
+ "learning_rate": 9.520451195004107e-06,
1007
+ "loss": 0.0637,
1008
+ "step": 500,
1009
+ "student_loss": 0.0007873536669649184,
1010
+ "teacher_loss": 0.0004125205450691283
1011
+ },
1012
+ {
1013
+ "epoch": 0.14561707035755478,
1014
+ "grad_norm": 5.40625,
1015
+ "kd_loss": 0.02945433370769024,
1016
+ "learning_rate": 9.508851603086094e-06,
1017
+ "loss": 0.0726,
1018
+ "step": 505,
1019
+ "student_loss": 0.000638218130916357,
1020
+ "teacher_loss": 0.0005516612436622381
1021
+ },
1022
+ {
1023
+ "epoch": 0.14705882352941177,
1024
+ "grad_norm": 3.15625,
1025
+ "kd_loss": 0.024534931406378746,
1026
+ "learning_rate": 9.497120634372894e-06,
1027
+ "loss": 0.0615,
1028
+ "step": 510,
1029
+ "student_loss": 0.00046345568262040615,
1030
+ "teacher_loss": 0.0005228539230301976
1031
+ },
1032
+ {
1033
+ "epoch": 0.14850057670126873,
1034
+ "grad_norm": 2.171875,
1035
+ "kd_loss": 0.030015984550118446,
1036
+ "learning_rate": 9.48525863067595e-06,
1037
+ "loss": 0.0552,
1038
+ "step": 515,
1039
+ "student_loss": 0.01863456331193447,
1040
+ "teacher_loss": 0.00035437109181657434
1041
+ },
1042
+ {
1043
+ "epoch": 0.14994232987312572,
1044
+ "grad_norm": 1.96875,
1045
+ "kd_loss": 0.09350574016571045,
1046
+ "learning_rate": 9.473265937624748e-06,
1047
+ "loss": 0.0617,
1048
+ "step": 520,
1049
+ "student_loss": 0.041941773146390915,
1050
+ "teacher_loss": 0.01209901925176382
1051
+ },
1052
+ {
1053
+ "epoch": 0.1513840830449827,
1054
+ "grad_norm": 2.296875,
1055
+ "kd_loss": 0.03124225325882435,
1056
+ "learning_rate": 9.461142904656722e-06,
1057
+ "loss": 0.0637,
1058
+ "step": 525,
1059
+ "student_loss": 0.0007001934573054314,
1060
+ "teacher_loss": 0.0003092395782005042
1061
+ },
1062
+ {
1063
+ "epoch": 0.15282583621683968,
1064
+ "grad_norm": 2.953125,
1065
+ "kd_loss": 0.027622882276773453,
1066
+ "learning_rate": 9.448889885007105e-06,
1067
+ "loss": 0.0592,
1068
+ "step": 530,
1069
+ "student_loss": 0.001552732428535819,
1070
+ "teacher_loss": 0.0020485566928982735
1071
+ },
1072
+ {
1073
+ "epoch": 0.15426758938869667,
1074
+ "grad_norm": 3.765625,
1075
+ "kd_loss": 0.03841177374124527,
1076
+ "learning_rate": 9.436507235698613e-06,
1077
+ "loss": 0.0555,
1078
+ "step": 535,
1079
+ "student_loss": 0.001078014844097197,
1080
+ "teacher_loss": 0.0004558424116112292
1081
+ },
1082
+ {
1083
+ "epoch": 0.15570934256055363,
1084
+ "grad_norm": 3.15625,
1085
+ "kd_loss": 0.06665755063295364,
1086
+ "learning_rate": 9.423995317531049e-06,
1087
+ "loss": 0.0545,
1088
+ "step": 540,
1089
+ "student_loss": 0.01969953626394272,
1090
+ "teacher_loss": 0.00044258436537347734
1091
+ },
1092
+ {
1093
+ "epoch": 0.15715109573241062,
1094
+ "grad_norm": 4.75,
1095
+ "kd_loss": 0.025856027379631996,
1096
+ "learning_rate": 9.4113544950708e-06,
1097
+ "loss": 0.0538,
1098
+ "step": 545,
1099
+ "student_loss": 0.0025607882998883724,
1100
+ "teacher_loss": 0.0022856788709759712
1101
+ },
1102
+ {
1103
+ "epoch": 0.15859284890426759,
1104
+ "grad_norm": 2.015625,
1105
+ "kd_loss": 0.07671085000038147,
1106
+ "learning_rate": 9.398585136640195e-06,
1107
+ "loss": 0.0534,
1108
+ "step": 550,
1109
+ "student_loss": 0.0003819255216512829,
1110
+ "teacher_loss": 0.004131191875785589
1111
+ },
1112
+ {
1113
+ "epoch": 0.16003460207612458,
1114
+ "grad_norm": 2.421875,
1115
+ "kd_loss": 0.08971834182739258,
1116
+ "learning_rate": 9.38568761430679e-06,
1117
+ "loss": 0.0616,
1118
+ "step": 555,
1119
+ "student_loss": 0.03680429980158806,
1120
+ "teacher_loss": 0.001880571013316512
1121
+ },
1122
+ {
1123
+ "epoch": 0.16147635524798154,
1124
+ "grad_norm": 2.703125,
1125
+ "kd_loss": 0.049756698310375214,
1126
+ "learning_rate": 9.372662303872514e-06,
1127
+ "loss": 0.0684,
1128
+ "step": 560,
1129
+ "student_loss": 0.002375077921897173,
1130
+ "teacher_loss": 0.0005416898638941348
1131
+ },
1132
+ {
1133
+ "epoch": 0.16291810841983853,
1134
+ "grad_norm": 2.875,
1135
+ "kd_loss": 0.058007437735795975,
1136
+ "learning_rate": 9.359509584862735e-06,
1137
+ "loss": 0.0676,
1138
+ "step": 565,
1139
+ "student_loss": 0.02881671115756035,
1140
+ "teacher_loss": 0.01986512541770935
1141
+ },
1142
+ {
1143
+ "epoch": 0.1643598615916955,
1144
+ "grad_norm": 3.890625,
1145
+ "kd_loss": 0.0371030792593956,
1146
+ "learning_rate": 9.346229840515186e-06,
1147
+ "loss": 0.0695,
1148
+ "step": 570,
1149
+ "student_loss": 0.011267436668276787,
1150
+ "teacher_loss": 0.007594536989927292
1151
+ },
1152
+ {
1153
+ "epoch": 0.16580161476355249,
1154
+ "grad_norm": 2.109375,
1155
+ "kd_loss": 0.030466603115200996,
1156
+ "learning_rate": 9.332823457768805e-06,
1157
+ "loss": 0.064,
1158
+ "step": 575,
1159
+ "student_loss": 0.001199263846501708,
1160
+ "teacher_loss": 0.0006413141381926835
1161
+ },
1162
+ {
1163
+ "epoch": 0.16724336793540945,
1164
+ "grad_norm": 2.796875,
1165
+ "kd_loss": 0.029692523181438446,
1166
+ "learning_rate": 9.31929082725246e-06,
1167
+ "loss": 0.0567,
1168
+ "step": 580,
1169
+ "student_loss": 0.00196670601144433,
1170
+ "teacher_loss": 0.0029784676153212786
1171
+ },
1172
+ {
1173
+ "epoch": 0.16868512110726644,
1174
+ "grad_norm": 1.703125,
1175
+ "kd_loss": 0.03761596605181694,
1176
+ "learning_rate": 9.305632343273567e-06,
1177
+ "loss": 0.0541,
1178
+ "step": 585,
1179
+ "student_loss": 0.0009754559141583741,
1180
+ "teacher_loss": 0.0006102125626057386
1181
+ },
1182
+ {
1183
+ "epoch": 0.1701268742791234,
1184
+ "grad_norm": 4.03125,
1185
+ "kd_loss": 0.03190089017152786,
1186
+ "learning_rate": 9.291848403806604e-06,
1187
+ "loss": 0.057,
1188
+ "step": 590,
1189
+ "student_loss": 0.0020897286012768745,
1190
+ "teacher_loss": 0.0024630518164485693
1191
+ },
1192
+ {
1193
+ "epoch": 0.1715686274509804,
1194
+ "grad_norm": 4.59375,
1195
+ "kd_loss": 0.023459656164050102,
1196
+ "learning_rate": 9.277939410481507e-06,
1197
+ "loss": 0.0593,
1198
+ "step": 595,
1199
+ "student_loss": 0.0007652693311683834,
1200
+ "teacher_loss": 0.0006856726831756532
1201
+ },
1202
+ {
1203
+ "epoch": 0.17301038062283736,
1204
+ "grad_norm": 3.09375,
1205
+ "kd_loss": 0.04589663818478584,
1206
+ "learning_rate": 9.263905768571977e-06,
1207
+ "loss": 0.0633,
1208
+ "step": 600,
1209
+ "student_loss": 0.04291452467441559,
1210
+ "teacher_loss": 0.005533033981919289
1211
+ },
1212
+ {
1213
+ "epoch": 0.17445213379469435,
1214
+ "grad_norm": 1.84375,
1215
+ "kd_loss": 0.05738772079348564,
1216
+ "learning_rate": 9.249747886983664e-06,
1217
+ "loss": 0.0601,
1218
+ "step": 605,
1219
+ "student_loss": 0.08045658469200134,
1220
+ "teacher_loss": 0.033129699528217316
1221
+ },
1222
+ {
1223
+ "epoch": 0.17589388696655134,
1224
+ "grad_norm": 1.921875,
1225
+ "kd_loss": 0.04765104874968529,
1226
+ "learning_rate": 9.235466178242255e-06,
1227
+ "loss": 0.0583,
1228
+ "step": 610,
1229
+ "student_loss": 0.013285127468407154,
1230
+ "teacher_loss": 0.0006613412988372147
1231
+ },
1232
+ {
1233
+ "epoch": 0.1773356401384083,
1234
+ "grad_norm": 2.9375,
1235
+ "kd_loss": 0.026093661785125732,
1236
+ "learning_rate": 9.221061058481456e-06,
1237
+ "loss": 0.0556,
1238
+ "step": 615,
1239
+ "student_loss": 0.0002871236647479236,
1240
+ "teacher_loss": 0.0002521740389056504
1241
+ },
1242
+ {
1243
+ "epoch": 0.1787773933102653,
1244
+ "grad_norm": 2.84375,
1245
+ "kd_loss": 0.03619556128978729,
1246
+ "learning_rate": 9.206532947430865e-06,
1247
+ "loss": 0.0678,
1248
+ "step": 620,
1249
+ "student_loss": 0.011322817765176296,
1250
+ "teacher_loss": 0.0073880418203771114
1251
+ },
1252
+ {
1253
+ "epoch": 0.18021914648212226,
1254
+ "grad_norm": 2.609375,
1255
+ "kd_loss": 0.04490155354142189,
1256
+ "learning_rate": 9.191882268403743e-06,
1257
+ "loss": 0.0533,
1258
+ "step": 625,
1259
+ "student_loss": 0.011955109424889088,
1260
+ "teacher_loss": 0.005125494673848152
1261
+ },
1262
+ {
1263
+ "epoch": 0.18166089965397925,
1264
+ "grad_norm": 1.8359375,
1265
+ "kd_loss": 0.04392609745264053,
1266
+ "learning_rate": 9.177109448284673e-06,
1267
+ "loss": 0.056,
1268
+ "step": 630,
1269
+ "student_loss": 0.0005792961455881596,
1270
+ "teacher_loss": 0.0003546806110534817
1271
+ },
1272
+ {
1273
+ "epoch": 0.1831026528258362,
1274
+ "grad_norm": 1.5078125,
1275
+ "kd_loss": 0.03447414189577103,
1276
+ "learning_rate": 9.162214917517139e-06,
1277
+ "loss": 0.083,
1278
+ "step": 635,
1279
+ "student_loss": 0.001471965340897441,
1280
+ "teacher_loss": 0.00033435679506510496
1281
+ },
1282
+ {
1283
+ "epoch": 0.1845444059976932,
1284
+ "grad_norm": 3.1875,
1285
+ "kd_loss": 0.03496161848306656,
1286
+ "learning_rate": 9.14719911009096e-06,
1287
+ "loss": 0.0529,
1288
+ "step": 640,
1289
+ "student_loss": 0.0007727449410595,
1290
+ "teacher_loss": 0.0010425432119518518
1291
+ },
1292
+ {
1293
+ "epoch": 0.18598615916955016,
1294
+ "grad_norm": 2.015625,
1295
+ "kd_loss": 0.04637427628040314,
1296
+ "learning_rate": 9.132062463529666e-06,
1297
+ "loss": 0.0523,
1298
+ "step": 645,
1299
+ "student_loss": 0.0006955766584724188,
1300
+ "teacher_loss": 0.0006944125052541494
1301
+ },
1302
+ {
1303
+ "epoch": 0.18742791234140715,
1304
+ "grad_norm": 2.625,
1305
+ "kd_loss": 0.024316728115081787,
1306
+ "learning_rate": 9.11680541887774e-06,
1307
+ "loss": 0.0522,
1308
+ "step": 650,
1309
+ "student_loss": 0.001512147020548582,
1310
+ "teacher_loss": 0.0013077100738883018
1311
+ },
1312
+ {
1313
+ "epoch": 0.18886966551326412,
1314
+ "grad_norm": 1.8671875,
1315
+ "kd_loss": 0.04993291571736336,
1316
+ "learning_rate": 9.101428420687759e-06,
1317
+ "loss": 0.0543,
1318
+ "step": 655,
1319
+ "student_loss": 0.0008968978654593229,
1320
+ "teacher_loss": 0.00030134135158732533
1321
+ },
1322
+ {
1323
+ "epoch": 0.1903114186851211,
1324
+ "grad_norm": 1.8828125,
1325
+ "kd_loss": 0.025667058303952217,
1326
+ "learning_rate": 9.085931917007464e-06,
1327
+ "loss": 0.0602,
1328
+ "step": 660,
1329
+ "student_loss": 0.0007511115982197225,
1330
+ "teacher_loss": 0.0007162842084653676
1331
+ },
1332
+ {
1333
+ "epoch": 0.19175317185697807,
1334
+ "grad_norm": 2.734375,
1335
+ "kd_loss": 0.08388909697532654,
1336
+ "learning_rate": 9.070316359366681e-06,
1337
+ "loss": 0.0631,
1338
+ "step": 665,
1339
+ "student_loss": 0.023106243461370468,
1340
+ "teacher_loss": 0.0046136160381138325
1341
+ },
1342
+ {
1343
+ "epoch": 0.19319492502883506,
1344
+ "grad_norm": 2.390625,
1345
+ "kd_loss": 0.03343624249100685,
1346
+ "learning_rate": 9.054582202764175e-06,
1347
+ "loss": 0.0589,
1348
+ "step": 670,
1349
+ "student_loss": 0.02935577742755413,
1350
+ "teacher_loss": 0.014301146380603313
1351
+ },
1352
+ {
1353
+ "epoch": 0.19463667820069205,
1354
+ "grad_norm": 1.546875,
1355
+ "kd_loss": 0.018453417345881462,
1356
+ "learning_rate": 9.038729905654393e-06,
1357
+ "loss": 0.0585,
1358
+ "step": 675,
1359
+ "student_loss": 0.001494016731157899,
1360
+ "teacher_loss": 0.0013811474200338125
1361
+ },
1362
+ {
1363
+ "epoch": 0.19607843137254902,
1364
+ "grad_norm": 4.78125,
1365
+ "kd_loss": 0.06428319215774536,
1366
+ "learning_rate": 9.022759929934109e-06,
1367
+ "loss": 0.0534,
1368
+ "step": 680,
1369
+ "student_loss": 0.10977048426866531,
1370
+ "teacher_loss": 0.03891151025891304
1371
+ },
1372
+ {
1373
+ "epoch": 0.197520184544406,
1374
+ "grad_norm": 1.890625,
1375
+ "kd_loss": 0.03628097102046013,
1376
+ "learning_rate": 9.006672740928952e-06,
1377
+ "loss": 0.058,
1378
+ "step": 685,
1379
+ "student_loss": 0.04102587699890137,
1380
+ "teacher_loss": 0.03290959447622299
1381
+ },
1382
+ {
1383
+ "epoch": 0.19896193771626297,
1384
+ "grad_norm": 3.40625,
1385
+ "kd_loss": 0.05668497458100319,
1386
+ "learning_rate": 8.990468807379861e-06,
1387
+ "loss": 0.0598,
1388
+ "step": 690,
1389
+ "student_loss": 0.01099018007516861,
1390
+ "teacher_loss": 0.011626017279922962
1391
+ },
1392
+ {
1393
+ "epoch": 0.20040369088811996,
1394
+ "grad_norm": 7.59375,
1395
+ "kd_loss": 0.04181699454784393,
1396
+ "learning_rate": 8.974148601429426e-06,
1397
+ "loss": 0.0613,
1398
+ "step": 695,
1399
+ "student_loss": 0.002078200224786997,
1400
+ "teacher_loss": 0.005999539513140917
1401
+ },
1402
+ {
1403
+ "epoch": 0.20184544405997693,
1404
+ "grad_norm": 2.03125,
1405
+ "kd_loss": 0.025250906124711037,
1406
+ "learning_rate": 8.957712598608123e-06,
1407
+ "loss": 0.0559,
1408
+ "step": 700,
1409
+ "student_loss": 0.0004931570729240775,
1410
+ "teacher_loss": 0.00033929728670045733
1411
+ },
1412
+ {
1413
+ "epoch": 0.20328719723183392,
1414
+ "grad_norm": 2.921875,
1415
+ "kd_loss": 0.024194616824388504,
1416
+ "learning_rate": 8.941161277820462e-06,
1417
+ "loss": 0.057,
1418
+ "step": 705,
1419
+ "student_loss": 0.006048702169209719,
1420
+ "teacher_loss": 0.002249963115900755
1421
+ },
1422
+ {
1423
+ "epoch": 0.20472895040369088,
1424
+ "grad_norm": 1.625,
1425
+ "kd_loss": 0.0612892284989357,
1426
+ "learning_rate": 8.924495121331038e-06,
1427
+ "loss": 0.0608,
1428
+ "step": 710,
1429
+ "student_loss": 0.003674024948850274,
1430
+ "teacher_loss": 0.005100577138364315
1431
+ },
1432
+ {
1433
+ "epoch": 0.20617070357554787,
1434
+ "grad_norm": 1.7578125,
1435
+ "kd_loss": 0.02638155035674572,
1436
+ "learning_rate": 8.907714614750473e-06,
1437
+ "loss": 0.0647,
1438
+ "step": 715,
1439
+ "student_loss": 0.002913597971200943,
1440
+ "teacher_loss": 0.004206113051623106
1441
+ },
1442
+ {
1443
+ "epoch": 0.20761245674740483,
1444
+ "grad_norm": 3.171875,
1445
+ "kd_loss": 0.028725113719701767,
1446
+ "learning_rate": 8.890820247021263e-06,
1447
+ "loss": 0.0585,
1448
+ "step": 720,
1449
+ "student_loss": 0.0016089101554825902,
1450
+ "teacher_loss": 0.00185672368388623
1451
+ },
1452
+ {
1453
+ "epoch": 0.20905420991926182,
1454
+ "grad_norm": 12.375,
1455
+ "kd_loss": 0.023859383538365364,
1456
+ "learning_rate": 8.873812510403545e-06,
1457
+ "loss": 0.0489,
1458
+ "step": 725,
1459
+ "student_loss": 0.0008256085566245019,
1460
+ "teacher_loss": 0.0009837058605626225
1461
+ },
1462
+ {
1463
+ "epoch": 0.2104959630911188,
1464
+ "grad_norm": 2.703125,
1465
+ "kd_loss": 0.038513269275426865,
1466
+ "learning_rate": 8.85669190046074e-06,
1467
+ "loss": 0.0554,
1468
+ "step": 730,
1469
+ "student_loss": 0.014817601069808006,
1470
+ "teacher_loss": 0.002030330477282405
1471
+ },
1472
+ {
1473
+ "epoch": 0.21193771626297578,
1474
+ "grad_norm": 5.40625,
1475
+ "kd_loss": 0.037664614617824554,
1476
+ "learning_rate": 8.839458916045117e-06,
1477
+ "loss": 0.0648,
1478
+ "step": 735,
1479
+ "student_loss": 0.0006673415191471577,
1480
+ "teacher_loss": 0.00018354821077082306
1481
+ },
1482
+ {
1483
+ "epoch": 0.21337946943483277,
1484
+ "grad_norm": 3.515625,
1485
+ "kd_loss": 0.04300172999501228,
1486
+ "learning_rate": 8.822114059283267e-06,
1487
+ "loss": 0.0534,
1488
+ "step": 740,
1489
+ "student_loss": 0.0023371612187474966,
1490
+ "teacher_loss": 0.0002555273531470448
1491
+ },
1492
+ {
1493
+ "epoch": 0.21482122260668973,
1494
+ "grad_norm": 2.125,
1495
+ "kd_loss": 0.03269065171480179,
1496
+ "learning_rate": 8.804657835561456e-06,
1497
+ "loss": 0.0573,
1498
+ "step": 745,
1499
+ "student_loss": 0.04052459076046944,
1500
+ "teacher_loss": 0.019421465694904327
1501
+ },
1502
+ {
1503
+ "epoch": 0.21626297577854672,
1504
+ "grad_norm": 2.75,
1505
+ "kd_loss": 0.06529466807842255,
1506
+ "learning_rate": 8.787090753510917e-06,
1507
+ "loss": 0.055,
1508
+ "step": 750,
1509
+ "student_loss": 0.01737082377076149,
1510
+ "teacher_loss": 0.004820111207664013
1511
+ },
1512
+ {
1513
+ "epoch": 0.2177047289504037,
1514
+ "grad_norm": 3.296875,
1515
+ "kd_loss": 0.029180845245718956,
1516
+ "learning_rate": 8.76941332499301e-06,
1517
+ "loss": 0.0646,
1518
+ "step": 755,
1519
+ "student_loss": 0.036368243396282196,
1520
+ "teacher_loss": 0.00810588151216507
1521
+ },
1522
+ {
1523
+ "epoch": 0.21914648212226068,
1524
+ "grad_norm": 4.09375,
1525
+ "kd_loss": 0.07944276928901672,
1526
+ "learning_rate": 8.751626065084328e-06,
1527
+ "loss": 0.0552,
1528
+ "step": 760,
1529
+ "student_loss": 0.04477182775735855,
1530
+ "teacher_loss": 0.0025220662355422974
1531
+ },
1532
+ {
1533
+ "epoch": 0.22058823529411764,
1534
+ "grad_norm": 3.484375,
1535
+ "kd_loss": 0.07944491505622864,
1536
+ "learning_rate": 8.733729492061675e-06,
1537
+ "loss": 0.0633,
1538
+ "step": 765,
1539
+ "student_loss": 0.003388631623238325,
1540
+ "teacher_loss": 0.0013585498090833426
1541
+ },
1542
+ {
1543
+ "epoch": 0.22202998846597463,
1544
+ "grad_norm": 4.375,
1545
+ "kd_loss": 0.06357012689113617,
1546
+ "learning_rate": 8.715724127386971e-06,
1547
+ "loss": 0.0583,
1548
+ "step": 770,
1549
+ "student_loss": 0.0046012806706130505,
1550
+ "teacher_loss": 0.0021095508709549904
1551
+ },
1552
+ {
1553
+ "epoch": 0.2234717416378316,
1554
+ "grad_norm": 1.7890625,
1555
+ "kd_loss": 0.04382927715778351,
1556
+ "learning_rate": 8.697610495692055e-06,
1557
+ "loss": 0.0529,
1558
+ "step": 775,
1559
+ "student_loss": 0.003970274236053228,
1560
+ "teacher_loss": 0.0015163600910454988
1561
+ },
1562
+ {
1563
+ "epoch": 0.22491349480968859,
1564
+ "grad_norm": 2.734375,
1565
+ "kd_loss": 0.04260809347033501,
1566
+ "learning_rate": 8.679389124763396e-06,
1567
+ "loss": 0.0618,
1568
+ "step": 780,
1569
+ "student_loss": 0.00038051677984185517,
1570
+ "teacher_loss": 0.00029593356885015965
1571
+ },
1572
+ {
1573
+ "epoch": 0.22635524798154555,
1574
+ "grad_norm": 4.5,
1575
+ "kd_loss": 0.01874377392232418,
1576
+ "learning_rate": 8.661060545526724e-06,
1577
+ "loss": 0.0542,
1578
+ "step": 785,
1579
+ "student_loss": 0.0004492719890549779,
1580
+ "teacher_loss": 0.0009444112074561417
1581
+ },
1582
+ {
1583
+ "epoch": 0.22779700115340254,
1584
+ "grad_norm": 2.71875,
1585
+ "kd_loss": 0.01880568452179432,
1586
+ "learning_rate": 8.64262529203155e-06,
1587
+ "loss": 0.0507,
1588
+ "step": 790,
1589
+ "student_loss": 0.0022801393643021584,
1590
+ "teacher_loss": 0.002699216129258275
1591
+ },
1592
+ {
1593
+ "epoch": 0.2292387543252595,
1594
+ "grad_norm": 1.8984375,
1595
+ "kd_loss": 0.03094569221138954,
1596
+ "learning_rate": 8.624083901435613e-06,
1597
+ "loss": 0.0562,
1598
+ "step": 795,
1599
+ "student_loss": 0.00043586210813373327,
1600
+ "teacher_loss": 0.0003660838701762259
1601
+ },
1602
+ {
1603
+ "epoch": 0.2306805074971165,
1604
+ "grad_norm": 2.40625,
1605
+ "kd_loss": 0.06601320207118988,
1606
+ "learning_rate": 8.60543691398922e-06,
1607
+ "loss": 0.0653,
1608
+ "step": 800,
1609
+ "student_loss": 0.08038675785064697,
1610
+ "teacher_loss": 0.03730867803096771
1611
+ },
1612
+ {
1613
+ "epoch": 0.23212226066897348,
1614
+ "grad_norm": 2.34375,
1615
+ "kd_loss": 0.020229965448379517,
1616
+ "learning_rate": 8.586684873019513e-06,
1617
+ "loss": 0.0617,
1618
+ "step": 805,
1619
+ "student_loss": 0.000572382181417197,
1620
+ "teacher_loss": 0.0004885804955847561
1621
+ },
1622
+ {
1623
+ "epoch": 0.23356401384083045,
1624
+ "grad_norm": 42.25,
1625
+ "kd_loss": 0.040298011153936386,
1626
+ "learning_rate": 8.567828324914633e-06,
1627
+ "loss": 0.0805,
1628
+ "step": 810,
1629
+ "student_loss": 0.004066553432494402,
1630
+ "teacher_loss": 0.001117456704378128
1631
+ },
1632
+ {
1633
+ "epoch": 0.23500576701268744,
1634
+ "grad_norm": 4.125,
1635
+ "kd_loss": 0.022938063368201256,
1636
+ "learning_rate": 8.548867819107803e-06,
1637
+ "loss": 0.0552,
1638
+ "step": 815,
1639
+ "student_loss": 0.001487394911237061,
1640
+ "teacher_loss": 0.010924072004854679
1641
+ },
1642
+ {
1643
+ "epoch": 0.2364475201845444,
1644
+ "grad_norm": 5.75,
1645
+ "kd_loss": 0.04013598710298538,
1646
+ "learning_rate": 8.52980390806131e-06,
1647
+ "loss": 0.0571,
1648
+ "step": 820,
1649
+ "student_loss": 0.0090996865183115,
1650
+ "teacher_loss": 0.0024097838904708624
1651
+ },
1652
+ {
1653
+ "epoch": 0.2378892733564014,
1654
+ "grad_norm": 1.84375,
1655
+ "kd_loss": 0.044994160532951355,
1656
+ "learning_rate": 8.51063714725042e-06,
1657
+ "loss": 0.0547,
1658
+ "step": 825,
1659
+ "student_loss": 0.0013581844978034496,
1660
+ "teacher_loss": 0.016404999420046806
1661
+ },
1662
+ {
1663
+ "epoch": 0.23933102652825836,
1664
+ "grad_norm": 1.4375,
1665
+ "kd_loss": 0.03672877326607704,
1666
+ "learning_rate": 8.49136809514719e-06,
1667
+ "loss": 0.056,
1668
+ "step": 830,
1669
+ "student_loss": 0.015440435148775578,
1670
+ "teacher_loss": 0.0067474255338311195
1671
+ },
1672
+ {
1673
+ "epoch": 0.24077277970011535,
1674
+ "grad_norm": 3.03125,
1675
+ "kd_loss": 0.03259000554680824,
1676
+ "learning_rate": 8.471997313204183e-06,
1677
+ "loss": 0.0503,
1678
+ "step": 835,
1679
+ "student_loss": 0.013825266622006893,
1680
+ "teacher_loss": 0.004111866001039743
1681
+ },
1682
+ {
1683
+ "epoch": 0.2422145328719723,
1684
+ "grad_norm": 2.640625,
1685
+ "kd_loss": 0.043639760464429855,
1686
+ "learning_rate": 8.452525365838123e-06,
1687
+ "loss": 0.0549,
1688
+ "step": 840,
1689
+ "student_loss": 0.00048281383351422846,
1690
+ "teacher_loss": 0.0002513812214601785
1691
+ },
1692
+ {
1693
+ "epoch": 0.2436562860438293,
1694
+ "grad_norm": 5.21875,
1695
+ "kd_loss": 0.02270587347447872,
1696
+ "learning_rate": 8.432952820413447e-06,
1697
+ "loss": 0.0739,
1698
+ "step": 845,
1699
+ "student_loss": 0.0007846739026717842,
1700
+ "teacher_loss": 0.0005552724469453096
1701
+ },
1702
+ {
1703
+ "epoch": 0.24509803921568626,
1704
+ "grad_norm": 5.53125,
1705
+ "kd_loss": 0.04634132981300354,
1706
+ "learning_rate": 8.41328024722577e-06,
1707
+ "loss": 0.0569,
1708
+ "step": 850,
1709
+ "student_loss": 0.008840148337185383,
1710
+ "teacher_loss": 0.0004619469400495291
1711
+ },
1712
+ {
1713
+ "epoch": 0.24653979238754326,
1714
+ "grad_norm": 1.8046875,
1715
+ "kd_loss": 0.029315784573554993,
1716
+ "learning_rate": 8.393508219485263e-06,
1717
+ "loss": 0.0552,
1718
+ "step": 855,
1719
+ "student_loss": 0.0003546437656041235,
1720
+ "teacher_loss": 0.0002995182294398546
1721
+ },
1722
+ {
1723
+ "epoch": 0.24798154555940022,
1724
+ "grad_norm": 3.046875,
1725
+ "kd_loss": 0.03230588883161545,
1726
+ "learning_rate": 8.373637313299966e-06,
1727
+ "loss": 0.0642,
1728
+ "step": 860,
1729
+ "student_loss": 0.0017227281350642443,
1730
+ "teacher_loss": 0.001343526761047542
1731
+ },
1732
+ {
1733
+ "epoch": 0.2494232987312572,
1734
+ "grad_norm": 5.0625,
1735
+ "kd_loss": 0.03644983097910881,
1736
+ "learning_rate": 8.353668107658984e-06,
1737
+ "loss": 0.0589,
1738
+ "step": 865,
1739
+ "student_loss": 0.0006424355087801814,
1740
+ "teacher_loss": 0.0005446228315122426
1741
+ },
1742
+ {
1743
+ "epoch": 0.2508650519031142,
1744
+ "grad_norm": 2.484375,
1745
+ "kd_loss": 0.03575681149959564,
1746
+ "learning_rate": 8.333601184415632e-06,
1747
+ "loss": 0.0575,
1748
+ "step": 870,
1749
+ "student_loss": 0.007697219029068947,
1750
+ "teacher_loss": 0.003998692613095045
1751
+ },
1752
+ {
1753
+ "epoch": 0.25230680507497116,
1754
+ "grad_norm": 2.5625,
1755
+ "kd_loss": 0.023254452273249626,
1756
+ "learning_rate": 8.31343712827047e-06,
1757
+ "loss": 0.0538,
1758
+ "step": 875,
1759
+ "student_loss": 0.0005534820375032723,
1760
+ "teacher_loss": 0.000275202066404745
1761
+ },
1762
+ {
1763
+ "epoch": 0.2537485582468281,
1764
+ "grad_norm": 2.0,
1765
+ "kd_loss": 0.031032033264636993,
1766
+ "learning_rate": 8.293176526754274e-06,
1767
+ "loss": 0.0555,
1768
+ "step": 880,
1769
+ "student_loss": 0.004417644813656807,
1770
+ "teacher_loss": 0.0007619113312102854
1771
+ },
1772
+ {
1773
+ "epoch": 0.25519031141868515,
1774
+ "grad_norm": 2.15625,
1775
+ "kd_loss": 0.0855787992477417,
1776
+ "learning_rate": 8.27281997021091e-06,
1777
+ "loss": 0.0489,
1778
+ "step": 885,
1779
+ "student_loss": 0.02222822979092598,
1780
+ "teacher_loss": 0.04898800328373909
1781
+ },
1782
+ {
1783
+ "epoch": 0.2566320645905421,
1784
+ "grad_norm": 3.28125,
1785
+ "kd_loss": 0.019980890676379204,
1786
+ "learning_rate": 8.25236805178014e-06,
1787
+ "loss": 0.0566,
1788
+ "step": 890,
1789
+ "student_loss": 0.0005770166753791273,
1790
+ "teacher_loss": 0.000528016360476613
1791
+ },
1792
+ {
1793
+ "epoch": 0.25807381776239907,
1794
+ "grad_norm": 2.09375,
1795
+ "kd_loss": 0.04792998731136322,
1796
+ "learning_rate": 8.231821367380335e-06,
1797
+ "loss": 0.056,
1798
+ "step": 895,
1799
+ "student_loss": 0.021013708785176277,
1800
+ "teacher_loss": 0.009786177426576614
1801
+ },
1802
+ {
1803
+ "epoch": 0.25951557093425603,
1804
+ "grad_norm": 2.0,
1805
+ "kd_loss": 0.033532388508319855,
1806
+ "learning_rate": 8.211180515691108e-06,
1807
+ "loss": 0.0589,
1808
+ "step": 900,
1809
+ "student_loss": 0.030665317550301552,
1810
+ "teacher_loss": 0.022832820191979408
1811
+ },
1812
+ {
1813
+ "epoch": 0.26095732410611305,
1814
+ "grad_norm": 4.5625,
1815
+ "kd_loss": 0.02638239599764347,
1816
+ "learning_rate": 8.190446098135883e-06,
1817
+ "loss": 0.0525,
1818
+ "step": 905,
1819
+ "student_loss": 0.0071715242229402065,
1820
+ "teacher_loss": 0.00487996730953455
1821
+ },
1822
+ {
1823
+ "epoch": 0.26239907727797,
1824
+ "grad_norm": 4.0,
1825
+ "kd_loss": 0.02014465443789959,
1826
+ "learning_rate": 8.16961871886435e-06,
1827
+ "loss": 0.0524,
1828
+ "step": 910,
1829
+ "student_loss": 0.0006020800792612135,
1830
+ "teacher_loss": 0.0006278998334892094
1831
+ },
1832
+ {
1833
+ "epoch": 0.263840830449827,
1834
+ "grad_norm": 5.15625,
1835
+ "kd_loss": 0.06602304428815842,
1836
+ "learning_rate": 8.148698984734887e-06,
1837
+ "loss": 0.0533,
1838
+ "step": 915,
1839
+ "student_loss": 0.009310321882367134,
1840
+ "teacher_loss": 0.0008082063868641853
1841
+ },
1842
+ {
1843
+ "epoch": 0.26528258362168394,
1844
+ "grad_norm": 3.390625,
1845
+ "kd_loss": 0.02289159595966339,
1846
+ "learning_rate": 8.127687505296852e-06,
1847
+ "loss": 0.0543,
1848
+ "step": 920,
1849
+ "student_loss": 0.013857186771929264,
1850
+ "teacher_loss": 0.01195952482521534
1851
+ },
1852
+ {
1853
+ "epoch": 0.26672433679354096,
1854
+ "grad_norm": 1.5390625,
1855
+ "kd_loss": 0.022193575277924538,
1856
+ "learning_rate": 8.106584892772844e-06,
1857
+ "loss": 0.0526,
1858
+ "step": 925,
1859
+ "student_loss": 0.0005213135736994445,
1860
+ "teacher_loss": 0.0004576026403810829
1861
+ },
1862
+ {
1863
+ "epoch": 0.2681660899653979,
1864
+ "grad_norm": 4.21875,
1865
+ "kd_loss": 0.03260066732764244,
1866
+ "learning_rate": 8.08539176204085e-06,
1867
+ "loss": 0.0538,
1868
+ "step": 930,
1869
+ "student_loss": 0.0016611138125881553,
1870
+ "teacher_loss": 0.0023257529828697443
1871
+ },
1872
+ {
1873
+ "epoch": 0.2696078431372549,
1874
+ "grad_norm": 1.84375,
1875
+ "kd_loss": 0.027795003727078438,
1876
+ "learning_rate": 8.064108730616336e-06,
1877
+ "loss": 0.0528,
1878
+ "step": 935,
1879
+ "student_loss": 0.005895544774830341,
1880
+ "teacher_loss": 0.005227226763963699
1881
+ },
1882
+ {
1883
+ "epoch": 0.2710495963091119,
1884
+ "grad_norm": 2.828125,
1885
+ "kd_loss": 0.04935586079955101,
1886
+ "learning_rate": 8.042736418634252e-06,
1887
+ "loss": 0.0585,
1888
+ "step": 940,
1889
+ "student_loss": 0.005737821571528912,
1890
+ "teacher_loss": 0.0018308532889932394
1891
+ },
1892
+ {
1893
+ "epoch": 0.27249134948096887,
1894
+ "grad_norm": 2.46875,
1895
+ "kd_loss": 0.03260369971394539,
1896
+ "learning_rate": 8.02127544883096e-06,
1897
+ "loss": 0.0599,
1898
+ "step": 945,
1899
+ "student_loss": 0.004811516497284174,
1900
+ "teacher_loss": 0.0023290570825338364
1901
+ },
1902
+ {
1903
+ "epoch": 0.27393310265282583,
1904
+ "grad_norm": 2.546875,
1905
+ "kd_loss": 0.032012615352869034,
1906
+ "learning_rate": 7.999726446526095e-06,
1907
+ "loss": 0.0523,
1908
+ "step": 950,
1909
+ "student_loss": 0.00035348188248462975,
1910
+ "teacher_loss": 0.0002836557396221906
1911
+ },
1912
+ {
1913
+ "epoch": 0.2753748558246828,
1914
+ "grad_norm": 5.0,
1915
+ "kd_loss": 0.034856293350458145,
1916
+ "learning_rate": 7.978090039604342e-06,
1917
+ "loss": 0.0543,
1918
+ "step": 955,
1919
+ "student_loss": 0.001993444049730897,
1920
+ "teacher_loss": 0.0004989230656065047
1921
+ },
1922
+ {
1923
+ "epoch": 0.2768166089965398,
1924
+ "grad_norm": 3.5625,
1925
+ "kd_loss": 0.03832929953932762,
1926
+ "learning_rate": 7.956366858497138e-06,
1927
+ "loss": 0.0557,
1928
+ "step": 960,
1929
+ "student_loss": 0.0037265366408973932,
1930
+ "teacher_loss": 0.0020040564704686403
1931
+ },
1932
+ {
1933
+ "epoch": 0.2782583621683968,
1934
+ "grad_norm": 3.9375,
1935
+ "kd_loss": 0.04956424608826637,
1936
+ "learning_rate": 7.934557536164304e-06,
1937
+ "loss": 0.0506,
1938
+ "step": 965,
1939
+ "student_loss": 0.010483872145414352,
1940
+ "teacher_loss": 0.00675132917240262
1941
+ },
1942
+ {
1943
+ "epoch": 0.27970011534025374,
1944
+ "grad_norm": 2.53125,
1945
+ "kd_loss": 0.0348704569041729,
1946
+ "learning_rate": 7.91266270807561e-06,
1947
+ "loss": 0.0546,
1948
+ "step": 970,
1949
+ "student_loss": 0.00043184924288652837,
1950
+ "teacher_loss": 0.000216901971725747
1951
+ },
1952
+ {
1953
+ "epoch": 0.2811418685121107,
1954
+ "grad_norm": 3.84375,
1955
+ "kd_loss": 0.023248419165611267,
1956
+ "learning_rate": 7.890683012192249e-06,
1957
+ "loss": 0.0501,
1958
+ "step": 975,
1959
+ "student_loss": 0.012200110591948032,
1960
+ "teacher_loss": 0.0006010584183968604
1961
+ },
1962
+ {
1963
+ "epoch": 0.2825836216839677,
1964
+ "grad_norm": 3.078125,
1965
+ "kd_loss": 0.037069566547870636,
1966
+ "learning_rate": 7.868619088948246e-06,
1967
+ "loss": 0.0547,
1968
+ "step": 980,
1969
+ "student_loss": 0.00222848542034626,
1970
+ "teacher_loss": 0.000363580125849694
1971
+ },
1972
+ {
1973
+ "epoch": 0.2840253748558247,
1974
+ "grad_norm": 6.34375,
1975
+ "kd_loss": 0.031835682690143585,
1976
+ "learning_rate": 7.846471581231814e-06,
1977
+ "loss": 0.0565,
1978
+ "step": 985,
1979
+ "student_loss": 0.0003109886310994625,
1980
+ "teacher_loss": 0.00021889382333029062
1981
+ },
1982
+ {
1983
+ "epoch": 0.28546712802768165,
1984
+ "grad_norm": 5.65625,
1985
+ "kd_loss": 0.03366367891430855,
1986
+ "learning_rate": 7.824241134366604e-06,
1987
+ "loss": 0.0487,
1988
+ "step": 990,
1989
+ "student_loss": 0.00030151859391480684,
1990
+ "teacher_loss": 0.00022411555983126163
1991
+ },
1992
+ {
1993
+ "epoch": 0.2869088811995386,
1994
+ "grad_norm": 5.71875,
1995
+ "kd_loss": 0.04830598831176758,
1996
+ "learning_rate": 7.801928396092908e-06,
1997
+ "loss": 0.0574,
1998
+ "step": 995,
1999
+ "student_loss": 0.0025664782151579857,
2000
+ "teacher_loss": 0.002282222732901573
2001
+ },
2002
+ {
2003
+ "epoch": 0.28835063437139563,
2004
+ "grad_norm": 2.9375,
2005
+ "kd_loss": 0.03720923885703087,
2006
+ "learning_rate": 7.779534016548791e-06,
2007
+ "loss": 0.0541,
2008
+ "step": 1000,
2009
+ "student_loss": 0.005910244304686785,
2010
+ "teacher_loss": 0.00032394094159826636
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/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-1000/vocab.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-1500/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/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-1500/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 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 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/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-1500/config.json ADDED
@@ -0,0 +1,66 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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": 2048,
10
+ "initializer_range": 0.02,
11
+ "intermediate_size": 11008,
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
+ "full_attention",
42
+ "full_attention",
43
+ "full_attention",
44
+ "full_attention",
45
+ "full_attention",
46
+ "full_attention",
47
+ "full_attention",
48
+ "full_attention"
49
+ ],
50
+ "max_position_embeddings": 32768,
51
+ "max_window_layers": 36,
52
+ "model_type": "qwen2",
53
+ "num_attention_heads": 16,
54
+ "num_hidden_layers": 36,
55
+ "num_key_value_heads": 2,
56
+ "pad_token_id": 151643,
57
+ "rms_norm_eps": 1e-06,
58
+ "rope_scaling": null,
59
+ "rope_theta": 1000000.0,
60
+ "sliding_window": null,
61
+ "tie_word_embeddings": true,
62
+ "transformers_version": "4.57.6",
63
+ "use_cache": true,
64
+ "use_sliding_window": false,
65
+ "vocab_size": 151676
66
+ }
checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-1500/merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-1500/pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:57fc214f69c227235f740a6012c2838ed1423437c7345ebc6ecb7aaca41619df
3
+ size 6187858991
checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-1500/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/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-1500/thought_projector.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:eae88c4fef07a4780b10d5a7f348af079985700976c96797d804b5ebf58f6f3a
3
+ size 16788033
checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-1500/tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:83a790d654474f5dfe225f889afd0210313eb1083f942671f2c4b8e95a1c922b
3
+ size 11424004
checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-1500/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/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-1500/trainer_state.json ADDED
@@ -0,0 +1,3034 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 0.43252595155709345,
6
+ "eval_steps": 500,
7
+ "global_step": 1500,
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": 42.0,
15
+ "kd_loss": 0.2894527316093445,
16
+ "learning_rate": 4.444444444444445e-07,
17
+ "loss": 0.5205,
18
+ "step": 5,
19
+ "student_loss": 0.12883709371089935,
20
+ "teacher_loss": 0.001858799601905048
21
+ },
22
+ {
23
+ "epoch": 0.0028835063437139563,
24
+ "grad_norm": 31.875,
25
+ "kd_loss": 0.2677682042121887,
26
+ "learning_rate": 1.0000000000000002e-06,
27
+ "loss": 0.4924,
28
+ "step": 10,
29
+ "student_loss": 0.17924517393112183,
30
+ "teacher_loss": 0.005514730233699083
31
+ },
32
+ {
33
+ "epoch": 0.004325259515570935,
34
+ "grad_norm": 33.25,
35
+ "kd_loss": 0.23119957745075226,
36
+ "learning_rate": 1.5555555555555558e-06,
37
+ "loss": 0.4551,
38
+ "step": 15,
39
+ "student_loss": 0.062010008841753006,
40
+ "teacher_loss": 0.0011852148454636335
41
+ },
42
+ {
43
+ "epoch": 0.0057670126874279125,
44
+ "grad_norm": 25.5,
45
+ "kd_loss": 0.20175831019878387,
46
+ "learning_rate": 2.1111111111111114e-06,
47
+ "loss": 0.4171,
48
+ "step": 20,
49
+ "student_loss": 0.08724690973758698,
50
+ "teacher_loss": 0.0005447491421364248
51
+ },
52
+ {
53
+ "epoch": 0.00720876585928489,
54
+ "grad_norm": 14.0,
55
+ "kd_loss": 0.20470531284809113,
56
+ "learning_rate": 2.666666666666667e-06,
57
+ "loss": 0.3355,
58
+ "step": 25,
59
+ "student_loss": 0.028787745162844658,
60
+ "teacher_loss": 0.0008465723367407918
61
+ },
62
+ {
63
+ "epoch": 0.00865051903114187,
64
+ "grad_norm": 7.59375,
65
+ "kd_loss": 0.17482763528823853,
66
+ "learning_rate": 3.2222222222222227e-06,
67
+ "loss": 0.2813,
68
+ "step": 30,
69
+ "student_loss": 0.04111301898956299,
70
+ "teacher_loss": 0.0003613036824390292
71
+ },
72
+ {
73
+ "epoch": 0.010092272202998846,
74
+ "grad_norm": 5.46875,
75
+ "kd_loss": 0.1520802527666092,
76
+ "learning_rate": 3.777777777777778e-06,
77
+ "loss": 0.2499,
78
+ "step": 35,
79
+ "student_loss": 0.029766706749796867,
80
+ "teacher_loss": 0.0002796741609927267
81
+ },
82
+ {
83
+ "epoch": 0.011534025374855825,
84
+ "grad_norm": 4.4375,
85
+ "kd_loss": 0.11269068717956543,
86
+ "learning_rate": 4.333333333333334e-06,
87
+ "loss": 0.2082,
88
+ "step": 40,
89
+ "student_loss": 0.05098738893866539,
90
+ "teacher_loss": 0.04165218397974968
91
+ },
92
+ {
93
+ "epoch": 0.012975778546712802,
94
+ "grad_norm": 3.078125,
95
+ "kd_loss": 0.20585855841636658,
96
+ "learning_rate": 4.888888888888889e-06,
97
+ "loss": 0.1605,
98
+ "step": 45,
99
+ "student_loss": 0.08690722286701202,
100
+ "teacher_loss": 0.0026482592802494764
101
+ },
102
+ {
103
+ "epoch": 0.01441753171856978,
104
+ "grad_norm": 2.734375,
105
+ "kd_loss": 0.074299156665802,
106
+ "learning_rate": 5.444444444444445e-06,
107
+ "loss": 0.1578,
108
+ "step": 50,
109
+ "student_loss": 0.017733801156282425,
110
+ "teacher_loss": 0.008056458085775375
111
+ },
112
+ {
113
+ "epoch": 0.015859284890426758,
114
+ "grad_norm": 2.78125,
115
+ "kd_loss": 0.062237780541181564,
116
+ "learning_rate": 6e-06,
117
+ "loss": 0.135,
118
+ "step": 55,
119
+ "student_loss": 0.0012035972904413939,
120
+ "teacher_loss": 0.000626895169261843
121
+ },
122
+ {
123
+ "epoch": 0.01730103806228374,
124
+ "grad_norm": 2.65625,
125
+ "kd_loss": 0.05602008104324341,
126
+ "learning_rate": 6.555555555555556e-06,
127
+ "loss": 0.1184,
128
+ "step": 60,
129
+ "student_loss": 0.002416773699223995,
130
+ "teacher_loss": 0.0013880858896300197
131
+ },
132
+ {
133
+ "epoch": 0.018742791234140715,
134
+ "grad_norm": 4.65625,
135
+ "kd_loss": 0.051335666328668594,
136
+ "learning_rate": 7.111111111111112e-06,
137
+ "loss": 0.1118,
138
+ "step": 65,
139
+ "student_loss": 0.027237623929977417,
140
+ "teacher_loss": 0.03812175989151001
141
+ },
142
+ {
143
+ "epoch": 0.020184544405997693,
144
+ "grad_norm": 1.7109375,
145
+ "kd_loss": 0.0913488045334816,
146
+ "learning_rate": 7.666666666666667e-06,
147
+ "loss": 0.1042,
148
+ "step": 70,
149
+ "student_loss": 0.015874937176704407,
150
+ "teacher_loss": 0.0013995245099067688
151
+ },
152
+ {
153
+ "epoch": 0.02162629757785467,
154
+ "grad_norm": 2.890625,
155
+ "kd_loss": 0.036495719105005264,
156
+ "learning_rate": 8.222222222222222e-06,
157
+ "loss": 0.0961,
158
+ "step": 75,
159
+ "student_loss": 0.0018275808542966843,
160
+ "teacher_loss": 0.000998618663288653
161
+ },
162
+ {
163
+ "epoch": 0.02306805074971165,
164
+ "grad_norm": 2.8125,
165
+ "kd_loss": 0.033367447555065155,
166
+ "learning_rate": 8.777777777777778e-06,
167
+ "loss": 0.0811,
168
+ "step": 80,
169
+ "student_loss": 0.001792682334780693,
170
+ "teacher_loss": 0.0010247478494420648
171
+ },
172
+ {
173
+ "epoch": 0.024509803921568627,
174
+ "grad_norm": 2.734375,
175
+ "kd_loss": 0.0299057699739933,
176
+ "learning_rate": 9.333333333333334e-06,
177
+ "loss": 0.0827,
178
+ "step": 85,
179
+ "student_loss": 0.0028650059830397367,
180
+ "teacher_loss": 0.004752026870846748
181
+ },
182
+ {
183
+ "epoch": 0.025951557093425604,
184
+ "grad_norm": 1.7890625,
185
+ "kd_loss": 0.08093900978565216,
186
+ "learning_rate": 9.88888888888889e-06,
187
+ "loss": 0.0755,
188
+ "step": 90,
189
+ "student_loss": 0.019452862441539764,
190
+ "teacher_loss": 0.0028147350531071424
191
+ },
192
+ {
193
+ "epoch": 0.027393310265282585,
194
+ "grad_norm": 1.4296875,
195
+ "kd_loss": 0.04993116110563278,
196
+ "learning_rate": 9.99995337991255e-06,
197
+ "loss": 0.0683,
198
+ "step": 95,
199
+ "student_loss": 0.009235645644366741,
200
+ "teacher_loss": 0.006103724706918001
201
+ },
202
+ {
203
+ "epoch": 0.02883506343713956,
204
+ "grad_norm": 1.546875,
205
+ "kd_loss": 0.02219005487859249,
206
+ "learning_rate": 9.999763987297266e-06,
207
+ "loss": 0.0591,
208
+ "step": 100,
209
+ "student_loss": 0.0009554760181345046,
210
+ "teacher_loss": 0.000685701088514179
211
+ },
212
+ {
213
+ "epoch": 0.03027681660899654,
214
+ "grad_norm": 2.53125,
215
+ "kd_loss": 0.03886059299111366,
216
+ "learning_rate": 9.999428913912876e-06,
217
+ "loss": 0.0632,
218
+ "step": 105,
219
+ "student_loss": 0.002928090514615178,
220
+ "teacher_loss": 0.00032716552959755063
221
+ },
222
+ {
223
+ "epoch": 0.031718569780853516,
224
+ "grad_norm": 2.171875,
225
+ "kd_loss": 0.04434794560074806,
226
+ "learning_rate": 9.998948169522587e-06,
227
+ "loss": 0.063,
228
+ "step": 110,
229
+ "student_loss": 0.033939410001039505,
230
+ "teacher_loss": 0.017983509227633476
231
+ },
232
+ {
233
+ "epoch": 0.03316032295271049,
234
+ "grad_norm": 1.3046875,
235
+ "kd_loss": 0.03771444410085678,
236
+ "learning_rate": 9.998321768134101e-06,
237
+ "loss": 0.0606,
238
+ "step": 115,
239
+ "student_loss": 0.0007961422670632601,
240
+ "teacher_loss": 0.00038347821100614965
241
+ },
242
+ {
243
+ "epoch": 0.03460207612456748,
244
+ "grad_norm": 1.6953125,
245
+ "kd_loss": 0.02649487555027008,
246
+ "learning_rate": 9.997549727999212e-06,
247
+ "loss": 0.0532,
248
+ "step": 120,
249
+ "student_loss": 0.0012731363531202078,
250
+ "teacher_loss": 0.0007500122301280499
251
+ },
252
+ {
253
+ "epoch": 0.036043829296424454,
254
+ "grad_norm": 2.09375,
255
+ "kd_loss": 0.057950329035520554,
256
+ "learning_rate": 9.996632071613257e-06,
257
+ "loss": 0.0601,
258
+ "step": 125,
259
+ "student_loss": 0.03109952248632908,
260
+ "teacher_loss": 0.000994477653875947
261
+ },
262
+ {
263
+ "epoch": 0.03748558246828143,
264
+ "grad_norm": 0.95703125,
265
+ "kd_loss": 0.026637699455022812,
266
+ "learning_rate": 9.995568825714479e-06,
267
+ "loss": 0.0574,
268
+ "step": 130,
269
+ "student_loss": 0.0008338625193573534,
270
+ "teacher_loss": 0.0006465786718763411
271
+ },
272
+ {
273
+ "epoch": 0.03892733564013841,
274
+ "grad_norm": 1.3515625,
275
+ "kd_loss": 0.04235714301466942,
276
+ "learning_rate": 9.994360021283235e-06,
277
+ "loss": 0.0611,
278
+ "step": 135,
279
+ "student_loss": 0.001498939236626029,
280
+ "teacher_loss": 0.0005972529761493206
281
+ },
282
+ {
283
+ "epoch": 0.040369088811995385,
284
+ "grad_norm": 1.46875,
285
+ "kd_loss": 0.024005092680454254,
286
+ "learning_rate": 9.993005693541098e-06,
287
+ "loss": 0.0562,
288
+ "step": 140,
289
+ "student_loss": 0.0008620141888968647,
290
+ "teacher_loss": 0.0004937360063195229
291
+ },
292
+ {
293
+ "epoch": 0.04181084198385236,
294
+ "grad_norm": 5.5,
295
+ "kd_loss": 0.07175353169441223,
296
+ "learning_rate": 9.991505881949837e-06,
297
+ "loss": 0.0539,
298
+ "step": 145,
299
+ "student_loss": 0.03820747137069702,
300
+ "teacher_loss": 0.006569729186594486
301
+ },
302
+ {
303
+ "epoch": 0.04325259515570934,
304
+ "grad_norm": 1.2109375,
305
+ "kd_loss": 0.016246870160102844,
306
+ "learning_rate": 9.989860630210251e-06,
307
+ "loss": 0.0538,
308
+ "step": 150,
309
+ "student_loss": 0.0014459972735494375,
310
+ "teacher_loss": 0.001147772534750402
311
+ },
312
+ {
313
+ "epoch": 0.04469434832756632,
314
+ "grad_norm": 1.4765625,
315
+ "kd_loss": 0.02643739990890026,
316
+ "learning_rate": 9.98806998626091e-06,
317
+ "loss": 0.0592,
318
+ "step": 155,
319
+ "student_loss": 0.0027810579631477594,
320
+ "teacher_loss": 0.000860289903357625
321
+ },
322
+ {
323
+ "epoch": 0.0461361014994233,
324
+ "grad_norm": 2.40625,
325
+ "kd_loss": 0.029045773670077324,
326
+ "learning_rate": 9.98613400227676e-06,
327
+ "loss": 0.0597,
328
+ "step": 160,
329
+ "student_loss": 0.08056973665952682,
330
+ "teacher_loss": 0.04562501981854439
331
+ },
332
+ {
333
+ "epoch": 0.04757785467128028,
334
+ "grad_norm": 1.4140625,
335
+ "kd_loss": 0.03424260392785072,
336
+ "learning_rate": 9.984052734667586e-06,
337
+ "loss": 0.0523,
338
+ "step": 165,
339
+ "student_loss": 0.0027138988953083754,
340
+ "teacher_loss": 0.005776015110313892
341
+ },
342
+ {
343
+ "epoch": 0.049019607843137254,
344
+ "grad_norm": 2.859375,
345
+ "kd_loss": 0.019065098837018013,
346
+ "learning_rate": 9.981826244076387e-06,
347
+ "loss": 0.0562,
348
+ "step": 170,
349
+ "student_loss": 0.0007321524317376316,
350
+ "teacher_loss": 0.000522188376635313
351
+ },
352
+ {
353
+ "epoch": 0.05046136101499423,
354
+ "grad_norm": 1.7578125,
355
+ "kd_loss": 0.027638504281640053,
356
+ "learning_rate": 9.979454595377594e-06,
357
+ "loss": 0.0529,
358
+ "step": 175,
359
+ "student_loss": 0.004792944528162479,
360
+ "teacher_loss": 0.000635432661511004
361
+ },
362
+ {
363
+ "epoch": 0.05190311418685121,
364
+ "grad_norm": 2.0,
365
+ "kd_loss": 0.03646935150027275,
366
+ "learning_rate": 9.97693785767519e-06,
367
+ "loss": 0.0514,
368
+ "step": 180,
369
+ "student_loss": 0.0014396726619452238,
370
+ "teacher_loss": 0.0003402946167625487
371
+ },
372
+ {
373
+ "epoch": 0.05334486735870819,
374
+ "grad_norm": 1.453125,
375
+ "kd_loss": 0.03992177173495293,
376
+ "learning_rate": 9.9742761043007e-06,
377
+ "loss": 0.0582,
378
+ "step": 185,
379
+ "student_loss": 0.03759889304637909,
380
+ "teacher_loss": 0.00029035221086815
381
+ },
382
+ {
383
+ "epoch": 0.05478662053056517,
384
+ "grad_norm": 3.890625,
385
+ "kd_loss": 0.05405907332897186,
386
+ "learning_rate": 9.971469412811032e-06,
387
+ "loss": 0.0527,
388
+ "step": 190,
389
+ "student_loss": 0.004218568559736013,
390
+ "teacher_loss": 0.0009858696721494198
391
+ },
392
+ {
393
+ "epoch": 0.056228373702422146,
394
+ "grad_norm": 2.375,
395
+ "kd_loss": 0.05194515734910965,
396
+ "learning_rate": 9.968517864986247e-06,
397
+ "loss": 0.0524,
398
+ "step": 195,
399
+ "student_loss": 0.005763455759733915,
400
+ "teacher_loss": 0.0004817036970052868
401
+ },
402
+ {
403
+ "epoch": 0.05767012687427912,
404
+ "grad_norm": 1.2109375,
405
+ "kd_loss": 0.036061037331819534,
406
+ "learning_rate": 9.965421546827153e-06,
407
+ "loss": 0.064,
408
+ "step": 200,
409
+ "student_loss": 0.0005494994111359119,
410
+ "teacher_loss": 0.00020005257101729512
411
+ },
412
+ {
413
+ "epoch": 0.0591118800461361,
414
+ "grad_norm": 2.40625,
415
+ "kd_loss": 0.02808363176882267,
416
+ "learning_rate": 9.962180548552812e-06,
417
+ "loss": 0.0577,
418
+ "step": 205,
419
+ "student_loss": 0.03726515918970108,
420
+ "teacher_loss": 0.028193596750497818
421
+ },
422
+ {
423
+ "epoch": 0.06055363321799308,
424
+ "grad_norm": 4.15625,
425
+ "kd_loss": 0.03680839017033577,
426
+ "learning_rate": 9.958794964597906e-06,
427
+ "loss": 0.0515,
428
+ "step": 210,
429
+ "student_loss": 0.0011522703571245074,
430
+ "teacher_loss": 0.0004914712626487017
431
+ },
432
+ {
433
+ "epoch": 0.061995386389850055,
434
+ "grad_norm": 5.65625,
435
+ "kd_loss": 0.037980884313583374,
436
+ "learning_rate": 9.955264893609986e-06,
437
+ "loss": 0.0559,
438
+ "step": 215,
439
+ "student_loss": 0.0005635322886519134,
440
+ "teacher_loss": 0.00027164584025740623
441
+ },
442
+ {
443
+ "epoch": 0.06343713956170703,
444
+ "grad_norm": 4.0,
445
+ "kd_loss": 0.04362543299794197,
446
+ "learning_rate": 9.951590438446597e-06,
447
+ "loss": 0.0507,
448
+ "step": 220,
449
+ "student_loss": 0.032083068042993546,
450
+ "teacher_loss": 0.010480552911758423
451
+ },
452
+ {
453
+ "epoch": 0.06487889273356401,
454
+ "grad_norm": 1.3125,
455
+ "kd_loss": 0.02105102129280567,
456
+ "learning_rate": 9.947771706172285e-06,
457
+ "loss": 0.0605,
458
+ "step": 225,
459
+ "student_loss": 0.038467925041913986,
460
+ "teacher_loss": 0.0018907957710325718
461
+ },
462
+ {
463
+ "epoch": 0.06632064590542099,
464
+ "grad_norm": 2.25,
465
+ "kd_loss": 0.04121282696723938,
466
+ "learning_rate": 9.943808808055472e-06,
467
+ "loss": 0.0495,
468
+ "step": 230,
469
+ "student_loss": 0.01008139830082655,
470
+ "teacher_loss": 0.003380328882485628
471
+ },
472
+ {
473
+ "epoch": 0.06776239907727798,
474
+ "grad_norm": 9.5,
475
+ "kd_loss": 0.07618479430675507,
476
+ "learning_rate": 9.93970185956522e-06,
477
+ "loss": 0.0516,
478
+ "step": 235,
479
+ "student_loss": 0.010011663660407066,
480
+ "teacher_loss": 0.0004287679912522435
481
+ },
482
+ {
483
+ "epoch": 0.06920415224913495,
484
+ "grad_norm": 2.796875,
485
+ "kd_loss": 0.031842563301324844,
486
+ "learning_rate": 9.93545098036786e-06,
487
+ "loss": 0.0567,
488
+ "step": 240,
489
+ "student_loss": 0.0031024671625345945,
490
+ "teacher_loss": 0.0003865428443532437
491
+ },
492
+ {
493
+ "epoch": 0.07064590542099193,
494
+ "grad_norm": 1.921875,
495
+ "kd_loss": 0.01385761983692646,
496
+ "learning_rate": 9.931056294323507e-06,
497
+ "loss": 0.048,
498
+ "step": 245,
499
+ "student_loss": 0.011989925056695938,
500
+ "teacher_loss": 0.015163863077759743
501
+ },
502
+ {
503
+ "epoch": 0.07208765859284891,
504
+ "grad_norm": 1.8125,
505
+ "kd_loss": 0.020028578117489815,
506
+ "learning_rate": 9.926517929482454e-06,
507
+ "loss": 0.0514,
508
+ "step": 250,
509
+ "student_loss": 0.0005340924835763872,
510
+ "teacher_loss": 0.000291565025690943
511
+ },
512
+ {
513
+ "epoch": 0.07352941176470588,
514
+ "grad_norm": 1.8359375,
515
+ "kd_loss": 0.01568070985376835,
516
+ "learning_rate": 9.921836018081438e-06,
517
+ "loss": 0.0512,
518
+ "step": 255,
519
+ "student_loss": 0.0003255972405895591,
520
+ "teacher_loss": 0.0003198981285095215
521
+ },
522
+ {
523
+ "epoch": 0.07497116493656286,
524
+ "grad_norm": 1.65625,
525
+ "kd_loss": 0.03392383083701134,
526
+ "learning_rate": 9.917010696539793e-06,
527
+ "loss": 0.0577,
528
+ "step": 260,
529
+ "student_loss": 0.0405547134578228,
530
+ "teacher_loss": 0.03275086730718613
531
+ },
532
+ {
533
+ "epoch": 0.07641291810841984,
534
+ "grad_norm": 4.96875,
535
+ "kd_loss": 0.018212661147117615,
536
+ "learning_rate": 9.912042105455462e-06,
537
+ "loss": 0.0565,
538
+ "step": 265,
539
+ "student_loss": 0.0005924144061282277,
540
+ "teacher_loss": 0.0004515529435593635
541
+ },
542
+ {
543
+ "epoch": 0.07785467128027682,
544
+ "grad_norm": 1.484375,
545
+ "kd_loss": 0.020866869017481804,
546
+ "learning_rate": 9.906930389600917e-06,
547
+ "loss": 0.0593,
548
+ "step": 270,
549
+ "student_loss": 0.0013341928133741021,
550
+ "teacher_loss": 0.0011696964502334595
551
+ },
552
+ {
553
+ "epoch": 0.07929642445213379,
554
+ "grad_norm": 2.1875,
555
+ "kd_loss": 0.02725479006767273,
556
+ "learning_rate": 9.901675697918925e-06,
557
+ "loss": 0.0511,
558
+ "step": 275,
559
+ "student_loss": 0.048831257969141006,
560
+ "teacher_loss": 0.035170599818229675
561
+ },
562
+ {
563
+ "epoch": 0.08073817762399077,
564
+ "grad_norm": 4.75,
565
+ "kd_loss": 0.03362221270799637,
566
+ "learning_rate": 9.896278183518216e-06,
567
+ "loss": 0.0509,
568
+ "step": 280,
569
+ "student_loss": 0.0004928835551254451,
570
+ "teacher_loss": 0.00017192006635013968
571
+ },
572
+ {
573
+ "epoch": 0.08217993079584775,
574
+ "grad_norm": 2.921875,
575
+ "kd_loss": 0.021101269870996475,
576
+ "learning_rate": 9.890738003669029e-06,
577
+ "loss": 0.0572,
578
+ "step": 285,
579
+ "student_loss": 0.000973534828517586,
580
+ "teacher_loss": 0.0008148897904902697
581
+ },
582
+ {
583
+ "epoch": 0.08362168396770472,
584
+ "grad_norm": 2.203125,
585
+ "kd_loss": 0.017731280997395515,
586
+ "learning_rate": 9.885055319798514e-06,
587
+ "loss": 0.0582,
588
+ "step": 290,
589
+ "student_loss": 0.014440718106925488,
590
+ "teacher_loss": 0.012271302752196789
591
+ },
592
+ {
593
+ "epoch": 0.0850634371395617,
594
+ "grad_norm": 2.46875,
595
+ "kd_loss": 0.01968582533299923,
596
+ "learning_rate": 9.879230297486034e-06,
597
+ "loss": 0.0561,
598
+ "step": 295,
599
+ "student_loss": 0.0010608865413814783,
600
+ "teacher_loss": 0.0008463782141916454
601
+ },
602
+ {
603
+ "epoch": 0.08650519031141868,
604
+ "grad_norm": 3.34375,
605
+ "kd_loss": 0.05237875133752823,
606
+ "learning_rate": 9.873263106458353e-06,
607
+ "loss": 0.0526,
608
+ "step": 300,
609
+ "student_loss": 0.017360921949148178,
610
+ "teacher_loss": 0.001965837087482214
611
+ },
612
+ {
613
+ "epoch": 0.08794694348327567,
614
+ "grad_norm": 2.296875,
615
+ "kd_loss": 0.023132195696234703,
616
+ "learning_rate": 9.86715392058467e-06,
617
+ "loss": 0.0485,
618
+ "step": 305,
619
+ "student_loss": 0.007244116626679897,
620
+ "teacher_loss": 0.003649009158834815
621
+ },
622
+ {
623
+ "epoch": 0.08938869665513265,
624
+ "grad_norm": 2.671875,
625
+ "kd_loss": 0.020483316853642464,
626
+ "learning_rate": 9.860902917871566e-06,
627
+ "loss": 0.0535,
628
+ "step": 310,
629
+ "student_loss": 0.0007183508132584393,
630
+ "teacher_loss": 0.0004938851925544441
631
+ },
632
+ {
633
+ "epoch": 0.09083044982698962,
634
+ "grad_norm": 2.828125,
635
+ "kd_loss": 0.04086384177207947,
636
+ "learning_rate": 9.854510280457819e-06,
637
+ "loss": 0.0521,
638
+ "step": 315,
639
+ "student_loss": 0.010588191449642181,
640
+ "teacher_loss": 0.00029907570569776
641
+ },
642
+ {
643
+ "epoch": 0.0922722029988466,
644
+ "grad_norm": 2.75,
645
+ "kd_loss": 0.026799874380230904,
646
+ "learning_rate": 9.847976194609083e-06,
647
+ "loss": 0.055,
648
+ "step": 320,
649
+ "student_loss": 0.0006185687379911542,
650
+ "teacher_loss": 0.0003583572688512504
651
+ },
652
+ {
653
+ "epoch": 0.09371395617070358,
654
+ "grad_norm": 2.265625,
655
+ "kd_loss": 0.07637664675712585,
656
+ "learning_rate": 9.841300850712479e-06,
657
+ "loss": 0.0516,
658
+ "step": 325,
659
+ "student_loss": 0.008596767671406269,
660
+ "teacher_loss": 0.0225156731903553
661
+ },
662
+ {
663
+ "epoch": 0.09515570934256055,
664
+ "grad_norm": 1.75,
665
+ "kd_loss": 0.0531776025891304,
666
+ "learning_rate": 9.834484443271035e-06,
667
+ "loss": 0.0557,
668
+ "step": 330,
669
+ "student_loss": 0.019897613674402237,
670
+ "teacher_loss": 0.004191068932414055
671
+ },
672
+ {
673
+ "epoch": 0.09659746251441753,
674
+ "grad_norm": 1.8125,
675
+ "kd_loss": 0.013695458881556988,
676
+ "learning_rate": 9.827527170898019e-06,
677
+ "loss": 0.0488,
678
+ "step": 335,
679
+ "student_loss": 0.0016651605255901814,
680
+ "teacher_loss": 0.0014800777425989509
681
+ },
682
+ {
683
+ "epoch": 0.09803921568627451,
684
+ "grad_norm": 2.3125,
685
+ "kd_loss": 0.027426283806562424,
686
+ "learning_rate": 9.820429236311158e-06,
687
+ "loss": 0.0567,
688
+ "step": 340,
689
+ "student_loss": 0.010532001033425331,
690
+ "teacher_loss": 0.006886747200042009
691
+ },
692
+ {
693
+ "epoch": 0.09948096885813149,
694
+ "grad_norm": 1.8359375,
695
+ "kd_loss": 0.03429238498210907,
696
+ "learning_rate": 9.81319084632673e-06,
697
+ "loss": 0.0564,
698
+ "step": 345,
699
+ "student_loss": 0.007421151269227266,
700
+ "teacher_loss": 0.0005970117053948343
701
+ },
702
+ {
703
+ "epoch": 0.10092272202998846,
704
+ "grad_norm": 2.546875,
705
+ "kd_loss": 0.058330416679382324,
706
+ "learning_rate": 9.805812211853534e-06,
707
+ "loss": 0.0473,
708
+ "step": 350,
709
+ "student_loss": 0.002645631320774555,
710
+ "teacher_loss": 0.001440234947949648
711
+ },
712
+ {
713
+ "epoch": 0.10236447520184544,
714
+ "grad_norm": 2.046875,
715
+ "kd_loss": 0.016124732792377472,
716
+ "learning_rate": 9.798293547886748e-06,
717
+ "loss": 0.0571,
718
+ "step": 355,
719
+ "student_loss": 0.001094816834665835,
720
+ "teacher_loss": 0.0007638070383109152
721
+ },
722
+ {
723
+ "epoch": 0.10380622837370242,
724
+ "grad_norm": 1.765625,
725
+ "kd_loss": 0.015412062406539917,
726
+ "learning_rate": 9.790635073501662e-06,
727
+ "loss": 0.0541,
728
+ "step": 360,
729
+ "student_loss": 0.0010126332053914666,
730
+ "teacher_loss": 0.0007303679012693465
731
+ },
732
+ {
733
+ "epoch": 0.1052479815455594,
734
+ "grad_norm": 1.328125,
735
+ "kd_loss": 0.0318637378513813,
736
+ "learning_rate": 9.782837011847294e-06,
737
+ "loss": 0.0555,
738
+ "step": 365,
739
+ "student_loss": 0.043804578483104706,
740
+ "teacher_loss": 0.0287578534334898
741
+ },
742
+ {
743
+ "epoch": 0.10668973471741638,
744
+ "grad_norm": 3.734375,
745
+ "kd_loss": 0.054416049271821976,
746
+ "learning_rate": 9.774899590139897e-06,
747
+ "loss": 0.064,
748
+ "step": 370,
749
+ "student_loss": 0.02294199727475643,
750
+ "teacher_loss": 0.0003071785031352192
751
+ },
752
+ {
753
+ "epoch": 0.10813148788927336,
754
+ "grad_norm": 3.09375,
755
+ "kd_loss": 0.03518553078174591,
756
+ "learning_rate": 9.766823039656328e-06,
757
+ "loss": 0.0564,
758
+ "step": 375,
759
+ "student_loss": 0.03153789043426514,
760
+ "teacher_loss": 0.020613783970475197
761
+ },
762
+ {
763
+ "epoch": 0.10957324106113034,
764
+ "grad_norm": 4.40625,
765
+ "kd_loss": 0.021105539053678513,
766
+ "learning_rate": 9.758607595727309e-06,
767
+ "loss": 0.0537,
768
+ "step": 380,
769
+ "student_loss": 0.000557993131224066,
770
+ "teacher_loss": 0.0003853706002701074
771
+ },
772
+ {
773
+ "epoch": 0.11101499423298732,
774
+ "grad_norm": 2.03125,
775
+ "kd_loss": 0.03399880975484848,
776
+ "learning_rate": 9.75025349773058e-06,
777
+ "loss": 0.0504,
778
+ "step": 385,
779
+ "student_loss": 0.0006929709343239665,
780
+ "teacher_loss": 0.0005223862244747579
781
+ },
782
+ {
783
+ "epoch": 0.11245674740484429,
784
+ "grad_norm": 1.2421875,
785
+ "kd_loss": 0.022061804309487343,
786
+ "learning_rate": 9.741760989083914e-06,
787
+ "loss": 0.0473,
788
+ "step": 390,
789
+ "student_loss": 0.0003254064649809152,
790
+ "teacher_loss": 0.0003073395637329668
791
+ },
792
+ {
793
+ "epoch": 0.11389850057670127,
794
+ "grad_norm": 1.9453125,
795
+ "kd_loss": 0.021170753985643387,
796
+ "learning_rate": 9.73313031723803e-06,
797
+ "loss": 0.0519,
798
+ "step": 395,
799
+ "student_loss": 0.0008620450389571488,
800
+ "teacher_loss": 0.0008799576899036765
801
+ },
802
+ {
803
+ "epoch": 0.11534025374855825,
804
+ "grad_norm": 1.6875,
805
+ "kd_loss": 0.02962590754032135,
806
+ "learning_rate": 9.724361733669383e-06,
807
+ "loss": 0.0462,
808
+ "step": 400,
809
+ "student_loss": 0.0009699641377665102,
810
+ "teacher_loss": 0.0002524392039049417
811
+ },
812
+ {
813
+ "epoch": 0.11678200692041522,
814
+ "grad_norm": 1.4140625,
815
+ "kd_loss": 0.02815704233944416,
816
+ "learning_rate": 9.715455493872827e-06,
817
+ "loss": 0.0563,
818
+ "step": 405,
819
+ "student_loss": 0.0135729368776083,
820
+ "teacher_loss": 0.00044868470286019146
821
+ },
822
+ {
823
+ "epoch": 0.1182237600922722,
824
+ "grad_norm": 1.8125,
825
+ "kd_loss": 0.03630191460251808,
826
+ "learning_rate": 9.70641185735419e-06,
827
+ "loss": 0.0593,
828
+ "step": 410,
829
+ "student_loss": 0.014492504298686981,
830
+ "teacher_loss": 0.007929121144115925
831
+ },
832
+ {
833
+ "epoch": 0.11966551326412918,
834
+ "grad_norm": 1.6171875,
835
+ "kd_loss": 0.026567015796899796,
836
+ "learning_rate": 9.697231087622691e-06,
837
+ "loss": 0.0585,
838
+ "step": 415,
839
+ "student_loss": 0.004891898948699236,
840
+ "teacher_loss": 0.003953997977077961
841
+ },
842
+ {
843
+ "epoch": 0.12110726643598616,
844
+ "grad_norm": 2.890625,
845
+ "kd_loss": 0.04854276403784752,
846
+ "learning_rate": 9.687913452183277e-06,
847
+ "loss": 0.0584,
848
+ "step": 420,
849
+ "student_loss": 0.04526115953922272,
850
+ "teacher_loss": 0.023706616833806038
851
+ },
852
+ {
853
+ "epoch": 0.12254901960784313,
854
+ "grad_norm": 2.71875,
855
+ "kd_loss": 0.014143871143460274,
856
+ "learning_rate": 9.678459222528825e-06,
857
+ "loss": 0.0539,
858
+ "step": 425,
859
+ "student_loss": 0.0028342436999082565,
860
+ "teacher_loss": 0.003036739770323038
861
+ },
862
+ {
863
+ "epoch": 0.12399077277970011,
864
+ "grad_norm": 2.546875,
865
+ "kd_loss": 0.05789174139499664,
866
+ "learning_rate": 9.668868674132224e-06,
867
+ "loss": 0.0637,
868
+ "step": 430,
869
+ "student_loss": 0.058899883180856705,
870
+ "teacher_loss": 0.064610056579113
871
+ },
872
+ {
873
+ "epoch": 0.1254325259515571,
874
+ "grad_norm": 2.25,
875
+ "kd_loss": 0.021487534046173096,
876
+ "learning_rate": 9.65914208643836e-06,
877
+ "loss": 0.0537,
878
+ "step": 435,
879
+ "student_loss": 0.0009387442842125893,
880
+ "teacher_loss": 0.0007441287161782384
881
+ },
882
+ {
883
+ "epoch": 0.12687427912341406,
884
+ "grad_norm": 6.78125,
885
+ "kd_loss": 0.03997809439897537,
886
+ "learning_rate": 9.649279742855964e-06,
887
+ "loss": 0.0599,
888
+ "step": 440,
889
+ "student_loss": 0.012818059884011745,
890
+ "teacher_loss": 0.007239806931465864
891
+ },
892
+ {
893
+ "epoch": 0.12831603229527105,
894
+ "grad_norm": 2.375,
895
+ "kd_loss": 0.037731364369392395,
896
+ "learning_rate": 9.639281930749363e-06,
897
+ "loss": 0.0586,
898
+ "step": 445,
899
+ "student_loss": 0.03573564440011978,
900
+ "teacher_loss": 0.01907653547823429
901
+ },
902
+ {
903
+ "epoch": 0.12975778546712802,
904
+ "grad_norm": 2.546875,
905
+ "kd_loss": 0.05183478444814682,
906
+ "learning_rate": 9.629148941430096e-06,
907
+ "loss": 0.0617,
908
+ "step": 450,
909
+ "student_loss": 0.002695092000067234,
910
+ "teacher_loss": 0.0027623483911156654
911
+ },
912
+ {
913
+ "epoch": 0.131199538638985,
914
+ "grad_norm": 1.6015625,
915
+ "kd_loss": 0.029434863477945328,
916
+ "learning_rate": 9.618881070148436e-06,
917
+ "loss": 0.0615,
918
+ "step": 455,
919
+ "student_loss": 0.00108905672095716,
920
+ "teacher_loss": 0.0005926509038545191
921
+ },
922
+ {
923
+ "epoch": 0.13264129181084197,
924
+ "grad_norm": 1.5703125,
925
+ "kd_loss": 0.05266507714986801,
926
+ "learning_rate": 9.608478616084784e-06,
927
+ "loss": 0.0548,
928
+ "step": 460,
929
+ "student_loss": 0.016866859048604965,
930
+ "teacher_loss": 0.0005994263920001686
931
+ },
932
+ {
933
+ "epoch": 0.13408304498269896,
934
+ "grad_norm": 4.90625,
935
+ "kd_loss": 0.04152460768818855,
936
+ "learning_rate": 9.597941882340947e-06,
937
+ "loss": 0.0648,
938
+ "step": 465,
939
+ "student_loss": 0.0035157979000359774,
940
+ "teacher_loss": 0.0008752731955610216
941
+ },
942
+ {
943
+ "epoch": 0.13552479815455595,
944
+ "grad_norm": 1.5546875,
945
+ "kd_loss": 0.03477834537625313,
946
+ "learning_rate": 9.587271175931313e-06,
947
+ "loss": 0.0637,
948
+ "step": 470,
949
+ "student_loss": 0.0005260187899693847,
950
+ "teacher_loss": 0.00038832571590319276
951
+ },
952
+ {
953
+ "epoch": 0.13696655132641292,
954
+ "grad_norm": 2.90625,
955
+ "kd_loss": 0.040360964834690094,
956
+ "learning_rate": 9.5764668077739e-06,
957
+ "loss": 0.0567,
958
+ "step": 475,
959
+ "student_loss": 0.0007983366376720369,
960
+ "teacher_loss": 0.0005927964230068028
961
+ },
962
+ {
963
+ "epoch": 0.1384083044982699,
964
+ "grad_norm": 3.3125,
965
+ "kd_loss": 0.04144829884171486,
966
+ "learning_rate": 9.5655290926813e-06,
967
+ "loss": 0.0762,
968
+ "step": 480,
969
+ "student_loss": 0.03820875287055969,
970
+ "teacher_loss": 0.02910608798265457
971
+ },
972
+ {
973
+ "epoch": 0.13985005767012687,
974
+ "grad_norm": 3.25,
975
+ "kd_loss": 0.0166642889380455,
976
+ "learning_rate": 9.554458349351507e-06,
977
+ "loss": 0.0627,
978
+ "step": 485,
979
+ "student_loss": 0.001466175657697022,
980
+ "teacher_loss": 0.0013084234669804573
981
+ },
982
+ {
983
+ "epoch": 0.14129181084198386,
984
+ "grad_norm": 3.5,
985
+ "kd_loss": 0.025684630498290062,
986
+ "learning_rate": 9.54325490035863e-06,
987
+ "loss": 0.0737,
988
+ "step": 490,
989
+ "student_loss": 0.0014865982811897993,
990
+ "teacher_loss": 0.0012143527856096625
991
+ },
992
+ {
993
+ "epoch": 0.14273356401384082,
994
+ "grad_norm": 2.359375,
995
+ "kd_loss": 0.0510389544069767,
996
+ "learning_rate": 9.531919072143488e-06,
997
+ "loss": 0.0776,
998
+ "step": 495,
999
+ "student_loss": 0.004116421099752188,
1000
+ "teacher_loss": 0.002551821991801262
1001
+ },
1002
+ {
1003
+ "epoch": 0.14417531718569782,
1004
+ "grad_norm": 2.109375,
1005
+ "kd_loss": 0.03247147053480148,
1006
+ "learning_rate": 9.520451195004107e-06,
1007
+ "loss": 0.0637,
1008
+ "step": 500,
1009
+ "student_loss": 0.0007873536669649184,
1010
+ "teacher_loss": 0.0004125205450691283
1011
+ },
1012
+ {
1013
+ "epoch": 0.14561707035755478,
1014
+ "grad_norm": 5.40625,
1015
+ "kd_loss": 0.02945433370769024,
1016
+ "learning_rate": 9.508851603086094e-06,
1017
+ "loss": 0.0726,
1018
+ "step": 505,
1019
+ "student_loss": 0.000638218130916357,
1020
+ "teacher_loss": 0.0005516612436622381
1021
+ },
1022
+ {
1023
+ "epoch": 0.14705882352941177,
1024
+ "grad_norm": 3.15625,
1025
+ "kd_loss": 0.024534931406378746,
1026
+ "learning_rate": 9.497120634372894e-06,
1027
+ "loss": 0.0615,
1028
+ "step": 510,
1029
+ "student_loss": 0.00046345568262040615,
1030
+ "teacher_loss": 0.0005228539230301976
1031
+ },
1032
+ {
1033
+ "epoch": 0.14850057670126873,
1034
+ "grad_norm": 2.171875,
1035
+ "kd_loss": 0.030015984550118446,
1036
+ "learning_rate": 9.48525863067595e-06,
1037
+ "loss": 0.0552,
1038
+ "step": 515,
1039
+ "student_loss": 0.01863456331193447,
1040
+ "teacher_loss": 0.00035437109181657434
1041
+ },
1042
+ {
1043
+ "epoch": 0.14994232987312572,
1044
+ "grad_norm": 1.96875,
1045
+ "kd_loss": 0.09350574016571045,
1046
+ "learning_rate": 9.473265937624748e-06,
1047
+ "loss": 0.0617,
1048
+ "step": 520,
1049
+ "student_loss": 0.041941773146390915,
1050
+ "teacher_loss": 0.01209901925176382
1051
+ },
1052
+ {
1053
+ "epoch": 0.1513840830449827,
1054
+ "grad_norm": 2.296875,
1055
+ "kd_loss": 0.03124225325882435,
1056
+ "learning_rate": 9.461142904656722e-06,
1057
+ "loss": 0.0637,
1058
+ "step": 525,
1059
+ "student_loss": 0.0007001934573054314,
1060
+ "teacher_loss": 0.0003092395782005042
1061
+ },
1062
+ {
1063
+ "epoch": 0.15282583621683968,
1064
+ "grad_norm": 2.953125,
1065
+ "kd_loss": 0.027622882276773453,
1066
+ "learning_rate": 9.448889885007105e-06,
1067
+ "loss": 0.0592,
1068
+ "step": 530,
1069
+ "student_loss": 0.001552732428535819,
1070
+ "teacher_loss": 0.0020485566928982735
1071
+ },
1072
+ {
1073
+ "epoch": 0.15426758938869667,
1074
+ "grad_norm": 3.765625,
1075
+ "kd_loss": 0.03841177374124527,
1076
+ "learning_rate": 9.436507235698613e-06,
1077
+ "loss": 0.0555,
1078
+ "step": 535,
1079
+ "student_loss": 0.001078014844097197,
1080
+ "teacher_loss": 0.0004558424116112292
1081
+ },
1082
+ {
1083
+ "epoch": 0.15570934256055363,
1084
+ "grad_norm": 3.15625,
1085
+ "kd_loss": 0.06665755063295364,
1086
+ "learning_rate": 9.423995317531049e-06,
1087
+ "loss": 0.0545,
1088
+ "step": 540,
1089
+ "student_loss": 0.01969953626394272,
1090
+ "teacher_loss": 0.00044258436537347734
1091
+ },
1092
+ {
1093
+ "epoch": 0.15715109573241062,
1094
+ "grad_norm": 4.75,
1095
+ "kd_loss": 0.025856027379631996,
1096
+ "learning_rate": 9.4113544950708e-06,
1097
+ "loss": 0.0538,
1098
+ "step": 545,
1099
+ "student_loss": 0.0025607882998883724,
1100
+ "teacher_loss": 0.0022856788709759712
1101
+ },
1102
+ {
1103
+ "epoch": 0.15859284890426759,
1104
+ "grad_norm": 2.015625,
1105
+ "kd_loss": 0.07671085000038147,
1106
+ "learning_rate": 9.398585136640195e-06,
1107
+ "loss": 0.0534,
1108
+ "step": 550,
1109
+ "student_loss": 0.0003819255216512829,
1110
+ "teacher_loss": 0.004131191875785589
1111
+ },
1112
+ {
1113
+ "epoch": 0.16003460207612458,
1114
+ "grad_norm": 2.421875,
1115
+ "kd_loss": 0.08971834182739258,
1116
+ "learning_rate": 9.38568761430679e-06,
1117
+ "loss": 0.0616,
1118
+ "step": 555,
1119
+ "student_loss": 0.03680429980158806,
1120
+ "teacher_loss": 0.001880571013316512
1121
+ },
1122
+ {
1123
+ "epoch": 0.16147635524798154,
1124
+ "grad_norm": 2.703125,
1125
+ "kd_loss": 0.049756698310375214,
1126
+ "learning_rate": 9.372662303872514e-06,
1127
+ "loss": 0.0684,
1128
+ "step": 560,
1129
+ "student_loss": 0.002375077921897173,
1130
+ "teacher_loss": 0.0005416898638941348
1131
+ },
1132
+ {
1133
+ "epoch": 0.16291810841983853,
1134
+ "grad_norm": 2.875,
1135
+ "kd_loss": 0.058007437735795975,
1136
+ "learning_rate": 9.359509584862735e-06,
1137
+ "loss": 0.0676,
1138
+ "step": 565,
1139
+ "student_loss": 0.02881671115756035,
1140
+ "teacher_loss": 0.01986512541770935
1141
+ },
1142
+ {
1143
+ "epoch": 0.1643598615916955,
1144
+ "grad_norm": 3.890625,
1145
+ "kd_loss": 0.0371030792593956,
1146
+ "learning_rate": 9.346229840515186e-06,
1147
+ "loss": 0.0695,
1148
+ "step": 570,
1149
+ "student_loss": 0.011267436668276787,
1150
+ "teacher_loss": 0.007594536989927292
1151
+ },
1152
+ {
1153
+ "epoch": 0.16580161476355249,
1154
+ "grad_norm": 2.109375,
1155
+ "kd_loss": 0.030466603115200996,
1156
+ "learning_rate": 9.332823457768805e-06,
1157
+ "loss": 0.064,
1158
+ "step": 575,
1159
+ "student_loss": 0.001199263846501708,
1160
+ "teacher_loss": 0.0006413141381926835
1161
+ },
1162
+ {
1163
+ "epoch": 0.16724336793540945,
1164
+ "grad_norm": 2.796875,
1165
+ "kd_loss": 0.029692523181438446,
1166
+ "learning_rate": 9.31929082725246e-06,
1167
+ "loss": 0.0567,
1168
+ "step": 580,
1169
+ "student_loss": 0.00196670601144433,
1170
+ "teacher_loss": 0.0029784676153212786
1171
+ },
1172
+ {
1173
+ "epoch": 0.16868512110726644,
1174
+ "grad_norm": 1.703125,
1175
+ "kd_loss": 0.03761596605181694,
1176
+ "learning_rate": 9.305632343273567e-06,
1177
+ "loss": 0.0541,
1178
+ "step": 585,
1179
+ "student_loss": 0.0009754559141583741,
1180
+ "teacher_loss": 0.0006102125626057386
1181
+ },
1182
+ {
1183
+ "epoch": 0.1701268742791234,
1184
+ "grad_norm": 4.03125,
1185
+ "kd_loss": 0.03190089017152786,
1186
+ "learning_rate": 9.291848403806604e-06,
1187
+ "loss": 0.057,
1188
+ "step": 590,
1189
+ "student_loss": 0.0020897286012768745,
1190
+ "teacher_loss": 0.0024630518164485693
1191
+ },
1192
+ {
1193
+ "epoch": 0.1715686274509804,
1194
+ "grad_norm": 4.59375,
1195
+ "kd_loss": 0.023459656164050102,
1196
+ "learning_rate": 9.277939410481507e-06,
1197
+ "loss": 0.0593,
1198
+ "step": 595,
1199
+ "student_loss": 0.0007652693311683834,
1200
+ "teacher_loss": 0.0006856726831756532
1201
+ },
1202
+ {
1203
+ "epoch": 0.17301038062283736,
1204
+ "grad_norm": 3.09375,
1205
+ "kd_loss": 0.04589663818478584,
1206
+ "learning_rate": 9.263905768571977e-06,
1207
+ "loss": 0.0633,
1208
+ "step": 600,
1209
+ "student_loss": 0.04291452467441559,
1210
+ "teacher_loss": 0.005533033981919289
1211
+ },
1212
+ {
1213
+ "epoch": 0.17445213379469435,
1214
+ "grad_norm": 1.84375,
1215
+ "kd_loss": 0.05738772079348564,
1216
+ "learning_rate": 9.249747886983664e-06,
1217
+ "loss": 0.0601,
1218
+ "step": 605,
1219
+ "student_loss": 0.08045658469200134,
1220
+ "teacher_loss": 0.033129699528217316
1221
+ },
1222
+ {
1223
+ "epoch": 0.17589388696655134,
1224
+ "grad_norm": 1.921875,
1225
+ "kd_loss": 0.04765104874968529,
1226
+ "learning_rate": 9.235466178242255e-06,
1227
+ "loss": 0.0583,
1228
+ "step": 610,
1229
+ "student_loss": 0.013285127468407154,
1230
+ "teacher_loss": 0.0006613412988372147
1231
+ },
1232
+ {
1233
+ "epoch": 0.1773356401384083,
1234
+ "grad_norm": 2.9375,
1235
+ "kd_loss": 0.026093661785125732,
1236
+ "learning_rate": 9.221061058481456e-06,
1237
+ "loss": 0.0556,
1238
+ "step": 615,
1239
+ "student_loss": 0.0002871236647479236,
1240
+ "teacher_loss": 0.0002521740389056504
1241
+ },
1242
+ {
1243
+ "epoch": 0.1787773933102653,
1244
+ "grad_norm": 2.84375,
1245
+ "kd_loss": 0.03619556128978729,
1246
+ "learning_rate": 9.206532947430865e-06,
1247
+ "loss": 0.0678,
1248
+ "step": 620,
1249
+ "student_loss": 0.011322817765176296,
1250
+ "teacher_loss": 0.0073880418203771114
1251
+ },
1252
+ {
1253
+ "epoch": 0.18021914648212226,
1254
+ "grad_norm": 2.609375,
1255
+ "kd_loss": 0.04490155354142189,
1256
+ "learning_rate": 9.191882268403743e-06,
1257
+ "loss": 0.0533,
1258
+ "step": 625,
1259
+ "student_loss": 0.011955109424889088,
1260
+ "teacher_loss": 0.005125494673848152
1261
+ },
1262
+ {
1263
+ "epoch": 0.18166089965397925,
1264
+ "grad_norm": 1.8359375,
1265
+ "kd_loss": 0.04392609745264053,
1266
+ "learning_rate": 9.177109448284673e-06,
1267
+ "loss": 0.056,
1268
+ "step": 630,
1269
+ "student_loss": 0.0005792961455881596,
1270
+ "teacher_loss": 0.0003546806110534817
1271
+ },
1272
+ {
1273
+ "epoch": 0.1831026528258362,
1274
+ "grad_norm": 1.5078125,
1275
+ "kd_loss": 0.03447414189577103,
1276
+ "learning_rate": 9.162214917517139e-06,
1277
+ "loss": 0.083,
1278
+ "step": 635,
1279
+ "student_loss": 0.001471965340897441,
1280
+ "teacher_loss": 0.00033435679506510496
1281
+ },
1282
+ {
1283
+ "epoch": 0.1845444059976932,
1284
+ "grad_norm": 3.1875,
1285
+ "kd_loss": 0.03496161848306656,
1286
+ "learning_rate": 9.14719911009096e-06,
1287
+ "loss": 0.0529,
1288
+ "step": 640,
1289
+ "student_loss": 0.0007727449410595,
1290
+ "teacher_loss": 0.0010425432119518518
1291
+ },
1292
+ {
1293
+ "epoch": 0.18598615916955016,
1294
+ "grad_norm": 2.015625,
1295
+ "kd_loss": 0.04637427628040314,
1296
+ "learning_rate": 9.132062463529666e-06,
1297
+ "loss": 0.0523,
1298
+ "step": 645,
1299
+ "student_loss": 0.0006955766584724188,
1300
+ "teacher_loss": 0.0006944125052541494
1301
+ },
1302
+ {
1303
+ "epoch": 0.18742791234140715,
1304
+ "grad_norm": 2.625,
1305
+ "kd_loss": 0.024316728115081787,
1306
+ "learning_rate": 9.11680541887774e-06,
1307
+ "loss": 0.0522,
1308
+ "step": 650,
1309
+ "student_loss": 0.001512147020548582,
1310
+ "teacher_loss": 0.0013077100738883018
1311
+ },
1312
+ {
1313
+ "epoch": 0.18886966551326412,
1314
+ "grad_norm": 1.8671875,
1315
+ "kd_loss": 0.04993291571736336,
1316
+ "learning_rate": 9.101428420687759e-06,
1317
+ "loss": 0.0543,
1318
+ "step": 655,
1319
+ "student_loss": 0.0008968978654593229,
1320
+ "teacher_loss": 0.00030134135158732533
1321
+ },
1322
+ {
1323
+ "epoch": 0.1903114186851211,
1324
+ "grad_norm": 1.8828125,
1325
+ "kd_loss": 0.025667058303952217,
1326
+ "learning_rate": 9.085931917007464e-06,
1327
+ "loss": 0.0602,
1328
+ "step": 660,
1329
+ "student_loss": 0.0007511115982197225,
1330
+ "teacher_loss": 0.0007162842084653676
1331
+ },
1332
+ {
1333
+ "epoch": 0.19175317185697807,
1334
+ "grad_norm": 2.734375,
1335
+ "kd_loss": 0.08388909697532654,
1336
+ "learning_rate": 9.070316359366681e-06,
1337
+ "loss": 0.0631,
1338
+ "step": 665,
1339
+ "student_loss": 0.023106243461370468,
1340
+ "teacher_loss": 0.0046136160381138325
1341
+ },
1342
+ {
1343
+ "epoch": 0.19319492502883506,
1344
+ "grad_norm": 2.390625,
1345
+ "kd_loss": 0.03343624249100685,
1346
+ "learning_rate": 9.054582202764175e-06,
1347
+ "loss": 0.0589,
1348
+ "step": 670,
1349
+ "student_loss": 0.02935577742755413,
1350
+ "teacher_loss": 0.014301146380603313
1351
+ },
1352
+ {
1353
+ "epoch": 0.19463667820069205,
1354
+ "grad_norm": 1.546875,
1355
+ "kd_loss": 0.018453417345881462,
1356
+ "learning_rate": 9.038729905654393e-06,
1357
+ "loss": 0.0585,
1358
+ "step": 675,
1359
+ "student_loss": 0.001494016731157899,
1360
+ "teacher_loss": 0.0013811474200338125
1361
+ },
1362
+ {
1363
+ "epoch": 0.19607843137254902,
1364
+ "grad_norm": 4.78125,
1365
+ "kd_loss": 0.06428319215774536,
1366
+ "learning_rate": 9.022759929934109e-06,
1367
+ "loss": 0.0534,
1368
+ "step": 680,
1369
+ "student_loss": 0.10977048426866531,
1370
+ "teacher_loss": 0.03891151025891304
1371
+ },
1372
+ {
1373
+ "epoch": 0.197520184544406,
1374
+ "grad_norm": 1.890625,
1375
+ "kd_loss": 0.03628097102046013,
1376
+ "learning_rate": 9.006672740928952e-06,
1377
+ "loss": 0.058,
1378
+ "step": 685,
1379
+ "student_loss": 0.04102587699890137,
1380
+ "teacher_loss": 0.03290959447622299
1381
+ },
1382
+ {
1383
+ "epoch": 0.19896193771626297,
1384
+ "grad_norm": 3.40625,
1385
+ "kd_loss": 0.05668497458100319,
1386
+ "learning_rate": 8.990468807379861e-06,
1387
+ "loss": 0.0598,
1388
+ "step": 690,
1389
+ "student_loss": 0.01099018007516861,
1390
+ "teacher_loss": 0.011626017279922962
1391
+ },
1392
+ {
1393
+ "epoch": 0.20040369088811996,
1394
+ "grad_norm": 7.59375,
1395
+ "kd_loss": 0.04181699454784393,
1396
+ "learning_rate": 8.974148601429426e-06,
1397
+ "loss": 0.0613,
1398
+ "step": 695,
1399
+ "student_loss": 0.002078200224786997,
1400
+ "teacher_loss": 0.005999539513140917
1401
+ },
1402
+ {
1403
+ "epoch": 0.20184544405997693,
1404
+ "grad_norm": 2.03125,
1405
+ "kd_loss": 0.025250906124711037,
1406
+ "learning_rate": 8.957712598608123e-06,
1407
+ "loss": 0.0559,
1408
+ "step": 700,
1409
+ "student_loss": 0.0004931570729240775,
1410
+ "teacher_loss": 0.00033929728670045733
1411
+ },
1412
+ {
1413
+ "epoch": 0.20328719723183392,
1414
+ "grad_norm": 2.921875,
1415
+ "kd_loss": 0.024194616824388504,
1416
+ "learning_rate": 8.941161277820462e-06,
1417
+ "loss": 0.057,
1418
+ "step": 705,
1419
+ "student_loss": 0.006048702169209719,
1420
+ "teacher_loss": 0.002249963115900755
1421
+ },
1422
+ {
1423
+ "epoch": 0.20472895040369088,
1424
+ "grad_norm": 1.625,
1425
+ "kd_loss": 0.0612892284989357,
1426
+ "learning_rate": 8.924495121331038e-06,
1427
+ "loss": 0.0608,
1428
+ "step": 710,
1429
+ "student_loss": 0.003674024948850274,
1430
+ "teacher_loss": 0.005100577138364315
1431
+ },
1432
+ {
1433
+ "epoch": 0.20617070357554787,
1434
+ "grad_norm": 1.7578125,
1435
+ "kd_loss": 0.02638155035674572,
1436
+ "learning_rate": 8.907714614750473e-06,
1437
+ "loss": 0.0647,
1438
+ "step": 715,
1439
+ "student_loss": 0.002913597971200943,
1440
+ "teacher_loss": 0.004206113051623106
1441
+ },
1442
+ {
1443
+ "epoch": 0.20761245674740483,
1444
+ "grad_norm": 3.171875,
1445
+ "kd_loss": 0.028725113719701767,
1446
+ "learning_rate": 8.890820247021263e-06,
1447
+ "loss": 0.0585,
1448
+ "step": 720,
1449
+ "student_loss": 0.0016089101554825902,
1450
+ "teacher_loss": 0.00185672368388623
1451
+ },
1452
+ {
1453
+ "epoch": 0.20905420991926182,
1454
+ "grad_norm": 12.375,
1455
+ "kd_loss": 0.023859383538365364,
1456
+ "learning_rate": 8.873812510403545e-06,
1457
+ "loss": 0.0489,
1458
+ "step": 725,
1459
+ "student_loss": 0.0008256085566245019,
1460
+ "teacher_loss": 0.0009837058605626225
1461
+ },
1462
+ {
1463
+ "epoch": 0.2104959630911188,
1464
+ "grad_norm": 2.703125,
1465
+ "kd_loss": 0.038513269275426865,
1466
+ "learning_rate": 8.85669190046074e-06,
1467
+ "loss": 0.0554,
1468
+ "step": 730,
1469
+ "student_loss": 0.014817601069808006,
1470
+ "teacher_loss": 0.002030330477282405
1471
+ },
1472
+ {
1473
+ "epoch": 0.21193771626297578,
1474
+ "grad_norm": 5.40625,
1475
+ "kd_loss": 0.037664614617824554,
1476
+ "learning_rate": 8.839458916045117e-06,
1477
+ "loss": 0.0648,
1478
+ "step": 735,
1479
+ "student_loss": 0.0006673415191471577,
1480
+ "teacher_loss": 0.00018354821077082306
1481
+ },
1482
+ {
1483
+ "epoch": 0.21337946943483277,
1484
+ "grad_norm": 3.515625,
1485
+ "kd_loss": 0.04300172999501228,
1486
+ "learning_rate": 8.822114059283267e-06,
1487
+ "loss": 0.0534,
1488
+ "step": 740,
1489
+ "student_loss": 0.0023371612187474966,
1490
+ "teacher_loss": 0.0002555273531470448
1491
+ },
1492
+ {
1493
+ "epoch": 0.21482122260668973,
1494
+ "grad_norm": 2.125,
1495
+ "kd_loss": 0.03269065171480179,
1496
+ "learning_rate": 8.804657835561456e-06,
1497
+ "loss": 0.0573,
1498
+ "step": 745,
1499
+ "student_loss": 0.04052459076046944,
1500
+ "teacher_loss": 0.019421465694904327
1501
+ },
1502
+ {
1503
+ "epoch": 0.21626297577854672,
1504
+ "grad_norm": 2.75,
1505
+ "kd_loss": 0.06529466807842255,
1506
+ "learning_rate": 8.787090753510917e-06,
1507
+ "loss": 0.055,
1508
+ "step": 750,
1509
+ "student_loss": 0.01737082377076149,
1510
+ "teacher_loss": 0.004820111207664013
1511
+ },
1512
+ {
1513
+ "epoch": 0.2177047289504037,
1514
+ "grad_norm": 3.296875,
1515
+ "kd_loss": 0.029180845245718956,
1516
+ "learning_rate": 8.76941332499301e-06,
1517
+ "loss": 0.0646,
1518
+ "step": 755,
1519
+ "student_loss": 0.036368243396282196,
1520
+ "teacher_loss": 0.00810588151216507
1521
+ },
1522
+ {
1523
+ "epoch": 0.21914648212226068,
1524
+ "grad_norm": 4.09375,
1525
+ "kd_loss": 0.07944276928901672,
1526
+ "learning_rate": 8.751626065084328e-06,
1527
+ "loss": 0.0552,
1528
+ "step": 760,
1529
+ "student_loss": 0.04477182775735855,
1530
+ "teacher_loss": 0.0025220662355422974
1531
+ },
1532
+ {
1533
+ "epoch": 0.22058823529411764,
1534
+ "grad_norm": 3.484375,
1535
+ "kd_loss": 0.07944491505622864,
1536
+ "learning_rate": 8.733729492061675e-06,
1537
+ "loss": 0.0633,
1538
+ "step": 765,
1539
+ "student_loss": 0.003388631623238325,
1540
+ "teacher_loss": 0.0013585498090833426
1541
+ },
1542
+ {
1543
+ "epoch": 0.22202998846597463,
1544
+ "grad_norm": 4.375,
1545
+ "kd_loss": 0.06357012689113617,
1546
+ "learning_rate": 8.715724127386971e-06,
1547
+ "loss": 0.0583,
1548
+ "step": 770,
1549
+ "student_loss": 0.0046012806706130505,
1550
+ "teacher_loss": 0.0021095508709549904
1551
+ },
1552
+ {
1553
+ "epoch": 0.2234717416378316,
1554
+ "grad_norm": 1.7890625,
1555
+ "kd_loss": 0.04382927715778351,
1556
+ "learning_rate": 8.697610495692055e-06,
1557
+ "loss": 0.0529,
1558
+ "step": 775,
1559
+ "student_loss": 0.003970274236053228,
1560
+ "teacher_loss": 0.0015163600910454988
1561
+ },
1562
+ {
1563
+ "epoch": 0.22491349480968859,
1564
+ "grad_norm": 2.734375,
1565
+ "kd_loss": 0.04260809347033501,
1566
+ "learning_rate": 8.679389124763396e-06,
1567
+ "loss": 0.0618,
1568
+ "step": 780,
1569
+ "student_loss": 0.00038051677984185517,
1570
+ "teacher_loss": 0.00029593356885015965
1571
+ },
1572
+ {
1573
+ "epoch": 0.22635524798154555,
1574
+ "grad_norm": 4.5,
1575
+ "kd_loss": 0.01874377392232418,
1576
+ "learning_rate": 8.661060545526724e-06,
1577
+ "loss": 0.0542,
1578
+ "step": 785,
1579
+ "student_loss": 0.0004492719890549779,
1580
+ "teacher_loss": 0.0009444112074561417
1581
+ },
1582
+ {
1583
+ "epoch": 0.22779700115340254,
1584
+ "grad_norm": 2.71875,
1585
+ "kd_loss": 0.01880568452179432,
1586
+ "learning_rate": 8.64262529203155e-06,
1587
+ "loss": 0.0507,
1588
+ "step": 790,
1589
+ "student_loss": 0.0022801393643021584,
1590
+ "teacher_loss": 0.002699216129258275
1591
+ },
1592
+ {
1593
+ "epoch": 0.2292387543252595,
1594
+ "grad_norm": 1.8984375,
1595
+ "kd_loss": 0.03094569221138954,
1596
+ "learning_rate": 8.624083901435613e-06,
1597
+ "loss": 0.0562,
1598
+ "step": 795,
1599
+ "student_loss": 0.00043586210813373327,
1600
+ "teacher_loss": 0.0003660838701762259
1601
+ },
1602
+ {
1603
+ "epoch": 0.2306805074971165,
1604
+ "grad_norm": 2.40625,
1605
+ "kd_loss": 0.06601320207118988,
1606
+ "learning_rate": 8.60543691398922e-06,
1607
+ "loss": 0.0653,
1608
+ "step": 800,
1609
+ "student_loss": 0.08038675785064697,
1610
+ "teacher_loss": 0.03730867803096771
1611
+ },
1612
+ {
1613
+ "epoch": 0.23212226066897348,
1614
+ "grad_norm": 2.34375,
1615
+ "kd_loss": 0.020229965448379517,
1616
+ "learning_rate": 8.586684873019513e-06,
1617
+ "loss": 0.0617,
1618
+ "step": 805,
1619
+ "student_loss": 0.000572382181417197,
1620
+ "teacher_loss": 0.0004885804955847561
1621
+ },
1622
+ {
1623
+ "epoch": 0.23356401384083045,
1624
+ "grad_norm": 42.25,
1625
+ "kd_loss": 0.040298011153936386,
1626
+ "learning_rate": 8.567828324914633e-06,
1627
+ "loss": 0.0805,
1628
+ "step": 810,
1629
+ "student_loss": 0.004066553432494402,
1630
+ "teacher_loss": 0.001117456704378128
1631
+ },
1632
+ {
1633
+ "epoch": 0.23500576701268744,
1634
+ "grad_norm": 4.125,
1635
+ "kd_loss": 0.022938063368201256,
1636
+ "learning_rate": 8.548867819107803e-06,
1637
+ "loss": 0.0552,
1638
+ "step": 815,
1639
+ "student_loss": 0.001487394911237061,
1640
+ "teacher_loss": 0.010924072004854679
1641
+ },
1642
+ {
1643
+ "epoch": 0.2364475201845444,
1644
+ "grad_norm": 5.75,
1645
+ "kd_loss": 0.04013598710298538,
1646
+ "learning_rate": 8.52980390806131e-06,
1647
+ "loss": 0.0571,
1648
+ "step": 820,
1649
+ "student_loss": 0.0090996865183115,
1650
+ "teacher_loss": 0.0024097838904708624
1651
+ },
1652
+ {
1653
+ "epoch": 0.2378892733564014,
1654
+ "grad_norm": 1.84375,
1655
+ "kd_loss": 0.044994160532951355,
1656
+ "learning_rate": 8.51063714725042e-06,
1657
+ "loss": 0.0547,
1658
+ "step": 825,
1659
+ "student_loss": 0.0013581844978034496,
1660
+ "teacher_loss": 0.016404999420046806
1661
+ },
1662
+ {
1663
+ "epoch": 0.23933102652825836,
1664
+ "grad_norm": 1.4375,
1665
+ "kd_loss": 0.03672877326607704,
1666
+ "learning_rate": 8.49136809514719e-06,
1667
+ "loss": 0.056,
1668
+ "step": 830,
1669
+ "student_loss": 0.015440435148775578,
1670
+ "teacher_loss": 0.0067474255338311195
1671
+ },
1672
+ {
1673
+ "epoch": 0.24077277970011535,
1674
+ "grad_norm": 3.03125,
1675
+ "kd_loss": 0.03259000554680824,
1676
+ "learning_rate": 8.471997313204183e-06,
1677
+ "loss": 0.0503,
1678
+ "step": 835,
1679
+ "student_loss": 0.013825266622006893,
1680
+ "teacher_loss": 0.004111866001039743
1681
+ },
1682
+ {
1683
+ "epoch": 0.2422145328719723,
1684
+ "grad_norm": 2.640625,
1685
+ "kd_loss": 0.043639760464429855,
1686
+ "learning_rate": 8.452525365838123e-06,
1687
+ "loss": 0.0549,
1688
+ "step": 840,
1689
+ "student_loss": 0.00048281383351422846,
1690
+ "teacher_loss": 0.0002513812214601785
1691
+ },
1692
+ {
1693
+ "epoch": 0.2436562860438293,
1694
+ "grad_norm": 5.21875,
1695
+ "kd_loss": 0.02270587347447872,
1696
+ "learning_rate": 8.432952820413447e-06,
1697
+ "loss": 0.0739,
1698
+ "step": 845,
1699
+ "student_loss": 0.0007846739026717842,
1700
+ "teacher_loss": 0.0005552724469453096
1701
+ },
1702
+ {
1703
+ "epoch": 0.24509803921568626,
1704
+ "grad_norm": 5.53125,
1705
+ "kd_loss": 0.04634132981300354,
1706
+ "learning_rate": 8.41328024722577e-06,
1707
+ "loss": 0.0569,
1708
+ "step": 850,
1709
+ "student_loss": 0.008840148337185383,
1710
+ "teacher_loss": 0.0004619469400495291
1711
+ },
1712
+ {
1713
+ "epoch": 0.24653979238754326,
1714
+ "grad_norm": 1.8046875,
1715
+ "kd_loss": 0.029315784573554993,
1716
+ "learning_rate": 8.393508219485263e-06,
1717
+ "loss": 0.0552,
1718
+ "step": 855,
1719
+ "student_loss": 0.0003546437656041235,
1720
+ "teacher_loss": 0.0002995182294398546
1721
+ },
1722
+ {
1723
+ "epoch": 0.24798154555940022,
1724
+ "grad_norm": 3.046875,
1725
+ "kd_loss": 0.03230588883161545,
1726
+ "learning_rate": 8.373637313299966e-06,
1727
+ "loss": 0.0642,
1728
+ "step": 860,
1729
+ "student_loss": 0.0017227281350642443,
1730
+ "teacher_loss": 0.001343526761047542
1731
+ },
1732
+ {
1733
+ "epoch": 0.2494232987312572,
1734
+ "grad_norm": 5.0625,
1735
+ "kd_loss": 0.03644983097910881,
1736
+ "learning_rate": 8.353668107658984e-06,
1737
+ "loss": 0.0589,
1738
+ "step": 865,
1739
+ "student_loss": 0.0006424355087801814,
1740
+ "teacher_loss": 0.0005446228315122426
1741
+ },
1742
+ {
1743
+ "epoch": 0.2508650519031142,
1744
+ "grad_norm": 2.484375,
1745
+ "kd_loss": 0.03575681149959564,
1746
+ "learning_rate": 8.333601184415632e-06,
1747
+ "loss": 0.0575,
1748
+ "step": 870,
1749
+ "student_loss": 0.007697219029068947,
1750
+ "teacher_loss": 0.003998692613095045
1751
+ },
1752
+ {
1753
+ "epoch": 0.25230680507497116,
1754
+ "grad_norm": 2.5625,
1755
+ "kd_loss": 0.023254452273249626,
1756
+ "learning_rate": 8.31343712827047e-06,
1757
+ "loss": 0.0538,
1758
+ "step": 875,
1759
+ "student_loss": 0.0005534820375032723,
1760
+ "teacher_loss": 0.000275202066404745
1761
+ },
1762
+ {
1763
+ "epoch": 0.2537485582468281,
1764
+ "grad_norm": 2.0,
1765
+ "kd_loss": 0.031032033264636993,
1766
+ "learning_rate": 8.293176526754274e-06,
1767
+ "loss": 0.0555,
1768
+ "step": 880,
1769
+ "student_loss": 0.004417644813656807,
1770
+ "teacher_loss": 0.0007619113312102854
1771
+ },
1772
+ {
1773
+ "epoch": 0.25519031141868515,
1774
+ "grad_norm": 2.15625,
1775
+ "kd_loss": 0.0855787992477417,
1776
+ "learning_rate": 8.27281997021091e-06,
1777
+ "loss": 0.0489,
1778
+ "step": 885,
1779
+ "student_loss": 0.02222822979092598,
1780
+ "teacher_loss": 0.04898800328373909
1781
+ },
1782
+ {
1783
+ "epoch": 0.2566320645905421,
1784
+ "grad_norm": 3.28125,
1785
+ "kd_loss": 0.019980890676379204,
1786
+ "learning_rate": 8.25236805178014e-06,
1787
+ "loss": 0.0566,
1788
+ "step": 890,
1789
+ "student_loss": 0.0005770166753791273,
1790
+ "teacher_loss": 0.000528016360476613
1791
+ },
1792
+ {
1793
+ "epoch": 0.25807381776239907,
1794
+ "grad_norm": 2.09375,
1795
+ "kd_loss": 0.04792998731136322,
1796
+ "learning_rate": 8.231821367380335e-06,
1797
+ "loss": 0.056,
1798
+ "step": 895,
1799
+ "student_loss": 0.021013708785176277,
1800
+ "teacher_loss": 0.009786177426576614
1801
+ },
1802
+ {
1803
+ "epoch": 0.25951557093425603,
1804
+ "grad_norm": 2.0,
1805
+ "kd_loss": 0.033532388508319855,
1806
+ "learning_rate": 8.211180515691108e-06,
1807
+ "loss": 0.0589,
1808
+ "step": 900,
1809
+ "student_loss": 0.030665317550301552,
1810
+ "teacher_loss": 0.022832820191979408
1811
+ },
1812
+ {
1813
+ "epoch": 0.26095732410611305,
1814
+ "grad_norm": 4.5625,
1815
+ "kd_loss": 0.02638239599764347,
1816
+ "learning_rate": 8.190446098135883e-06,
1817
+ "loss": 0.0525,
1818
+ "step": 905,
1819
+ "student_loss": 0.0071715242229402065,
1820
+ "teacher_loss": 0.00487996730953455
1821
+ },
1822
+ {
1823
+ "epoch": 0.26239907727797,
1824
+ "grad_norm": 4.0,
1825
+ "kd_loss": 0.02014465443789959,
1826
+ "learning_rate": 8.16961871886435e-06,
1827
+ "loss": 0.0524,
1828
+ "step": 910,
1829
+ "student_loss": 0.0006020800792612135,
1830
+ "teacher_loss": 0.0006278998334892094
1831
+ },
1832
+ {
1833
+ "epoch": 0.263840830449827,
1834
+ "grad_norm": 5.15625,
1835
+ "kd_loss": 0.06602304428815842,
1836
+ "learning_rate": 8.148698984734887e-06,
1837
+ "loss": 0.0533,
1838
+ "step": 915,
1839
+ "student_loss": 0.009310321882367134,
1840
+ "teacher_loss": 0.0008082063868641853
1841
+ },
1842
+ {
1843
+ "epoch": 0.26528258362168394,
1844
+ "grad_norm": 3.390625,
1845
+ "kd_loss": 0.02289159595966339,
1846
+ "learning_rate": 8.127687505296852e-06,
1847
+ "loss": 0.0543,
1848
+ "step": 920,
1849
+ "student_loss": 0.013857186771929264,
1850
+ "teacher_loss": 0.01195952482521534
1851
+ },
1852
+ {
1853
+ "epoch": 0.26672433679354096,
1854
+ "grad_norm": 1.5390625,
1855
+ "kd_loss": 0.022193575277924538,
1856
+ "learning_rate": 8.106584892772844e-06,
1857
+ "loss": 0.0526,
1858
+ "step": 925,
1859
+ "student_loss": 0.0005213135736994445,
1860
+ "teacher_loss": 0.0004576026403810829
1861
+ },
1862
+ {
1863
+ "epoch": 0.2681660899653979,
1864
+ "grad_norm": 4.21875,
1865
+ "kd_loss": 0.03260066732764244,
1866
+ "learning_rate": 8.08539176204085e-06,
1867
+ "loss": 0.0538,
1868
+ "step": 930,
1869
+ "student_loss": 0.0016611138125881553,
1870
+ "teacher_loss": 0.0023257529828697443
1871
+ },
1872
+ {
1873
+ "epoch": 0.2696078431372549,
1874
+ "grad_norm": 1.84375,
1875
+ "kd_loss": 0.027795003727078438,
1876
+ "learning_rate": 8.064108730616336e-06,
1877
+ "loss": 0.0528,
1878
+ "step": 935,
1879
+ "student_loss": 0.005895544774830341,
1880
+ "teacher_loss": 0.005227226763963699
1881
+ },
1882
+ {
1883
+ "epoch": 0.2710495963091119,
1884
+ "grad_norm": 2.828125,
1885
+ "kd_loss": 0.04935586079955101,
1886
+ "learning_rate": 8.042736418634252e-06,
1887
+ "loss": 0.0585,
1888
+ "step": 940,
1889
+ "student_loss": 0.005737821571528912,
1890
+ "teacher_loss": 0.0018308532889932394
1891
+ },
1892
+ {
1893
+ "epoch": 0.27249134948096887,
1894
+ "grad_norm": 2.46875,
1895
+ "kd_loss": 0.03260369971394539,
1896
+ "learning_rate": 8.02127544883096e-06,
1897
+ "loss": 0.0599,
1898
+ "step": 945,
1899
+ "student_loss": 0.004811516497284174,
1900
+ "teacher_loss": 0.0023290570825338364
1901
+ },
1902
+ {
1903
+ "epoch": 0.27393310265282583,
1904
+ "grad_norm": 2.546875,
1905
+ "kd_loss": 0.032012615352869034,
1906
+ "learning_rate": 7.999726446526095e-06,
1907
+ "loss": 0.0523,
1908
+ "step": 950,
1909
+ "student_loss": 0.00035348188248462975,
1910
+ "teacher_loss": 0.0002836557396221906
1911
+ },
1912
+ {
1913
+ "epoch": 0.2753748558246828,
1914
+ "grad_norm": 5.0,
1915
+ "kd_loss": 0.034856293350458145,
1916
+ "learning_rate": 7.978090039604342e-06,
1917
+ "loss": 0.0543,
1918
+ "step": 955,
1919
+ "student_loss": 0.001993444049730897,
1920
+ "teacher_loss": 0.0004989230656065047
1921
+ },
1922
+ {
1923
+ "epoch": 0.2768166089965398,
1924
+ "grad_norm": 3.5625,
1925
+ "kd_loss": 0.03832929953932762,
1926
+ "learning_rate": 7.956366858497138e-06,
1927
+ "loss": 0.0557,
1928
+ "step": 960,
1929
+ "student_loss": 0.0037265366408973932,
1930
+ "teacher_loss": 0.0020040564704686403
1931
+ },
1932
+ {
1933
+ "epoch": 0.2782583621683968,
1934
+ "grad_norm": 3.9375,
1935
+ "kd_loss": 0.04956424608826637,
1936
+ "learning_rate": 7.934557536164304e-06,
1937
+ "loss": 0.0506,
1938
+ "step": 965,
1939
+ "student_loss": 0.010483872145414352,
1940
+ "teacher_loss": 0.00675132917240262
1941
+ },
1942
+ {
1943
+ "epoch": 0.27970011534025374,
1944
+ "grad_norm": 2.53125,
1945
+ "kd_loss": 0.0348704569041729,
1946
+ "learning_rate": 7.91266270807561e-06,
1947
+ "loss": 0.0546,
1948
+ "step": 970,
1949
+ "student_loss": 0.00043184924288652837,
1950
+ "teacher_loss": 0.000216901971725747
1951
+ },
1952
+ {
1953
+ "epoch": 0.2811418685121107,
1954
+ "grad_norm": 3.84375,
1955
+ "kd_loss": 0.023248419165611267,
1956
+ "learning_rate": 7.890683012192249e-06,
1957
+ "loss": 0.0501,
1958
+ "step": 975,
1959
+ "student_loss": 0.012200110591948032,
1960
+ "teacher_loss": 0.0006010584183968604
1961
+ },
1962
+ {
1963
+ "epoch": 0.2825836216839677,
1964
+ "grad_norm": 3.078125,
1965
+ "kd_loss": 0.037069566547870636,
1966
+ "learning_rate": 7.868619088948246e-06,
1967
+ "loss": 0.0547,
1968
+ "step": 980,
1969
+ "student_loss": 0.00222848542034626,
1970
+ "teacher_loss": 0.000363580125849694
1971
+ },
1972
+ {
1973
+ "epoch": 0.2840253748558247,
1974
+ "grad_norm": 6.34375,
1975
+ "kd_loss": 0.031835682690143585,
1976
+ "learning_rate": 7.846471581231814e-06,
1977
+ "loss": 0.0565,
1978
+ "step": 985,
1979
+ "student_loss": 0.0003109886310994625,
1980
+ "teacher_loss": 0.00021889382333029062
1981
+ },
1982
+ {
1983
+ "epoch": 0.28546712802768165,
1984
+ "grad_norm": 5.65625,
1985
+ "kd_loss": 0.03366367891430855,
1986
+ "learning_rate": 7.824241134366604e-06,
1987
+ "loss": 0.0487,
1988
+ "step": 990,
1989
+ "student_loss": 0.00030151859391480684,
1990
+ "teacher_loss": 0.00022411555983126163
1991
+ },
1992
+ {
1993
+ "epoch": 0.2869088811995386,
1994
+ "grad_norm": 5.71875,
1995
+ "kd_loss": 0.04830598831176758,
1996
+ "learning_rate": 7.801928396092908e-06,
1997
+ "loss": 0.0574,
1998
+ "step": 995,
1999
+ "student_loss": 0.0025664782151579857,
2000
+ "teacher_loss": 0.002282222732901573
2001
+ },
2002
+ {
2003
+ "epoch": 0.28835063437139563,
2004
+ "grad_norm": 2.9375,
2005
+ "kd_loss": 0.03720923885703087,
2006
+ "learning_rate": 7.779534016548791e-06,
2007
+ "loss": 0.0541,
2008
+ "step": 1000,
2009
+ "student_loss": 0.005910244304686785,
2010
+ "teacher_loss": 0.00032394094159826636
2011
+ },
2012
+ {
2013
+ "epoch": 0.2897923875432526,
2014
+ "grad_norm": 4.28125,
2015
+ "kd_loss": 0.10840308666229248,
2016
+ "learning_rate": 7.75705864825114e-06,
2017
+ "loss": 0.063,
2018
+ "step": 1005,
2019
+ "student_loss": 0.04282848909497261,
2020
+ "teacher_loss": 0.06654597073793411
2021
+ },
2022
+ {
2023
+ "epoch": 0.29123414071510956,
2024
+ "grad_norm": 2.203125,
2025
+ "kd_loss": 0.01996036432683468,
2026
+ "learning_rate": 7.734502946076656e-06,
2027
+ "loss": 0.0535,
2028
+ "step": 1010,
2029
+ "student_loss": 0.0006448886124417186,
2030
+ "teacher_loss": 0.0006763582350686193
2031
+ },
2032
+ {
2033
+ "epoch": 0.2926758938869666,
2034
+ "grad_norm": 1.78125,
2035
+ "kd_loss": 0.07583313435316086,
2036
+ "learning_rate": 7.711867567242769e-06,
2037
+ "loss": 0.0611,
2038
+ "step": 1015,
2039
+ "student_loss": 0.019403517246246338,
2040
+ "teacher_loss": 0.007980776019394398
2041
+ },
2042
+ {
2043
+ "epoch": 0.29411764705882354,
2044
+ "grad_norm": 2.234375,
2045
+ "kd_loss": 0.03309423103928566,
2046
+ "learning_rate": 7.689153171288487e-06,
2047
+ "loss": 0.0518,
2048
+ "step": 1020,
2049
+ "student_loss": 0.04360064119100571,
2050
+ "teacher_loss": 0.028901230543851852
2051
+ },
2052
+ {
2053
+ "epoch": 0.2955594002306805,
2054
+ "grad_norm": 7.4375,
2055
+ "kd_loss": 0.03392940014600754,
2056
+ "learning_rate": 7.666360420055188e-06,
2057
+ "loss": 0.055,
2058
+ "step": 1025,
2059
+ "student_loss": 0.0003429439675528556,
2060
+ "teacher_loss": 0.00022535354946739972
2061
+ },
2062
+ {
2063
+ "epoch": 0.29700115340253747,
2064
+ "grad_norm": 4.65625,
2065
+ "kd_loss": 0.04872604459524155,
2066
+ "learning_rate": 7.643489977667327e-06,
2067
+ "loss": 0.0636,
2068
+ "step": 1030,
2069
+ "student_loss": 0.010639332234859467,
2070
+ "teacher_loss": 0.0026095418725162745
2071
+ },
2072
+ {
2073
+ "epoch": 0.2984429065743945,
2074
+ "grad_norm": 5.5,
2075
+ "kd_loss": 0.04310382157564163,
2076
+ "learning_rate": 7.6205425105130855e-06,
2077
+ "loss": 0.0569,
2078
+ "step": 1035,
2079
+ "student_loss": 0.001403859001584351,
2080
+ "teacher_loss": 0.0019977327901870012
2081
+ },
2082
+ {
2083
+ "epoch": 0.29988465974625145,
2084
+ "grad_norm": 6.375,
2085
+ "kd_loss": 0.028113925829529762,
2086
+ "learning_rate": 7.597518687224959e-06,
2087
+ "loss": 0.0585,
2088
+ "step": 1040,
2089
+ "student_loss": 0.01497615035623312,
2090
+ "teacher_loss": 0.0008363696397282183
2091
+ },
2092
+ {
2093
+ "epoch": 0.3013264129181084,
2094
+ "grad_norm": 4.5,
2095
+ "kd_loss": 0.024579087272286415,
2096
+ "learning_rate": 7.574419178660269e-06,
2097
+ "loss": 0.06,
2098
+ "step": 1045,
2099
+ "student_loss": 0.0012051565572619438,
2100
+ "teacher_loss": 0.0010785430204123259
2101
+ },
2102
+ {
2103
+ "epoch": 0.3027681660899654,
2104
+ "grad_norm": 2.109375,
2105
+ "kd_loss": 0.03319098800420761,
2106
+ "learning_rate": 7.551244657881618e-06,
2107
+ "loss": 0.0587,
2108
+ "step": 1050,
2109
+ "student_loss": 0.014071808196604252,
2110
+ "teacher_loss": 0.0008901043329387903
2111
+ },
2112
+ {
2113
+ "epoch": 0.3042099192618224,
2114
+ "grad_norm": 3.515625,
2115
+ "kd_loss": 0.03324364870786667,
2116
+ "learning_rate": 7.527995800137287e-06,
2117
+ "loss": 0.0525,
2118
+ "step": 1055,
2119
+ "student_loss": 0.0006993751740083098,
2120
+ "teacher_loss": 0.0006332803168334067
2121
+ },
2122
+ {
2123
+ "epoch": 0.30565167243367936,
2124
+ "grad_norm": 2.15625,
2125
+ "kd_loss": 0.03496374934911728,
2126
+ "learning_rate": 7.504673282841544e-06,
2127
+ "loss": 0.0548,
2128
+ "step": 1060,
2129
+ "student_loss": 0.0087258480489254,
2130
+ "teacher_loss": 0.00019599376537371427
2131
+ },
2132
+ {
2133
+ "epoch": 0.3070934256055363,
2134
+ "grad_norm": 1.953125,
2135
+ "kd_loss": 0.10101212561130524,
2136
+ "learning_rate": 7.481277785554918e-06,
2137
+ "loss": 0.055,
2138
+ "step": 1065,
2139
+ "student_loss": 0.05201160907745361,
2140
+ "teacher_loss": 0.004846156574785709
2141
+ },
2142
+ {
2143
+ "epoch": 0.30853517877739334,
2144
+ "grad_norm": 2.484375,
2145
+ "kd_loss": 0.046249356120824814,
2146
+ "learning_rate": 7.457809989964393e-06,
2147
+ "loss": 0.0613,
2148
+ "step": 1070,
2149
+ "student_loss": 0.0042078858241438866,
2150
+ "teacher_loss": 0.00018545822240412235
2151
+ },
2152
+ {
2153
+ "epoch": 0.3099769319492503,
2154
+ "grad_norm": 5.4375,
2155
+ "kd_loss": 0.02874414436519146,
2156
+ "learning_rate": 7.434270579863549e-06,
2157
+ "loss": 0.0583,
2158
+ "step": 1075,
2159
+ "student_loss": 0.0013546549016609788,
2160
+ "teacher_loss": 0.0010448966640979052
2161
+ },
2162
+ {
2163
+ "epoch": 0.31141868512110726,
2164
+ "grad_norm": 3.234375,
2165
+ "kd_loss": 0.024736085906624794,
2166
+ "learning_rate": 7.4106602411326345e-06,
2167
+ "loss": 0.0552,
2168
+ "step": 1080,
2169
+ "student_loss": 0.0006483947508968413,
2170
+ "teacher_loss": 0.008884440176188946
2171
+ },
2172
+ {
2173
+ "epoch": 0.3128604382929642,
2174
+ "grad_norm": 1.421875,
2175
+ "kd_loss": 0.022130126133561134,
2176
+ "learning_rate": 7.386979661718585e-06,
2177
+ "loss": 0.0555,
2178
+ "step": 1085,
2179
+ "student_loss": 0.00040961572085507214,
2180
+ "teacher_loss": 0.0004247698816470802
2181
+ },
2182
+ {
2183
+ "epoch": 0.31430219146482125,
2184
+ "grad_norm": 7.4375,
2185
+ "kd_loss": 0.05692844092845917,
2186
+ "learning_rate": 7.363229531614973e-06,
2187
+ "loss": 0.0572,
2188
+ "step": 1090,
2189
+ "student_loss": 0.018242081627249718,
2190
+ "teacher_loss": 0.012068298645317554
2191
+ },
2192
+ {
2193
+ "epoch": 0.3157439446366782,
2194
+ "grad_norm": 2.796875,
2195
+ "kd_loss": 0.03296171501278877,
2196
+ "learning_rate": 7.339410542841906e-06,
2197
+ "loss": 0.0585,
2198
+ "step": 1095,
2199
+ "student_loss": 0.03956712409853935,
2200
+ "teacher_loss": 0.007221496198326349
2201
+ },
2202
+ {
2203
+ "epoch": 0.31718569780853517,
2204
+ "grad_norm": 2.515625,
2205
+ "kd_loss": 0.029358811676502228,
2206
+ "learning_rate": 7.315523389425867e-06,
2207
+ "loss": 0.0577,
2208
+ "step": 1100,
2209
+ "student_loss": 0.0012080130400136113,
2210
+ "teacher_loss": 0.0003019682480953634
2211
+ },
2212
+ {
2213
+ "epoch": 0.31862745098039214,
2214
+ "grad_norm": 1.703125,
2215
+ "kd_loss": 0.03217841684818268,
2216
+ "learning_rate": 7.291568767379484e-06,
2217
+ "loss": 0.0508,
2218
+ "step": 1105,
2219
+ "student_loss": 0.0016615891363471746,
2220
+ "teacher_loss": 0.00029494339833036065
2221
+ },
2222
+ {
2223
+ "epoch": 0.32006920415224915,
2224
+ "grad_norm": 3.640625,
2225
+ "kd_loss": 0.05884777754545212,
2226
+ "learning_rate": 7.267547374681259e-06,
2227
+ "loss": 0.0515,
2228
+ "step": 1110,
2229
+ "student_loss": 0.0011524326400831342,
2230
+ "teacher_loss": 0.0009470232180319726
2231
+ },
2232
+ {
2233
+ "epoch": 0.3215109573241061,
2234
+ "grad_norm": 1.921875,
2235
+ "kd_loss": 0.03904154151678085,
2236
+ "learning_rate": 7.24345991125522e-06,
2237
+ "loss": 0.0554,
2238
+ "step": 1115,
2239
+ "student_loss": 0.0006270820158533752,
2240
+ "teacher_loss": 0.000263675901805982
2241
+ },
2242
+ {
2243
+ "epoch": 0.3229527104959631,
2244
+ "grad_norm": 6.71875,
2245
+ "kd_loss": 0.025044631212949753,
2246
+ "learning_rate": 7.219307078950536e-06,
2247
+ "loss": 0.0535,
2248
+ "step": 1120,
2249
+ "student_loss": 0.0013725879834964871,
2250
+ "teacher_loss": 0.0007414240972138941
2251
+ },
2252
+ {
2253
+ "epoch": 0.32439446366782004,
2254
+ "grad_norm": 3.609375,
2255
+ "kd_loss": 0.04100983217358589,
2256
+ "learning_rate": 7.195089581521064e-06,
2257
+ "loss": 0.059,
2258
+ "step": 1125,
2259
+ "student_loss": 0.04167897254228592,
2260
+ "teacher_loss": 0.02281663566827774
2261
+ },
2262
+ {
2263
+ "epoch": 0.32583621683967706,
2264
+ "grad_norm": 3.5,
2265
+ "kd_loss": 0.06311588734388351,
2266
+ "learning_rate": 7.170808124604842e-06,
2267
+ "loss": 0.0532,
2268
+ "step": 1130,
2269
+ "student_loss": 0.025650445371866226,
2270
+ "teacher_loss": 0.001013300847262144
2271
+ },
2272
+ {
2273
+ "epoch": 0.327277970011534,
2274
+ "grad_norm": 4.28125,
2275
+ "kd_loss": 0.06815233081579208,
2276
+ "learning_rate": 7.14646341570353e-06,
2277
+ "loss": 0.0581,
2278
+ "step": 1135,
2279
+ "student_loss": 0.022493448108434677,
2280
+ "teacher_loss": 0.005548724438995123
2281
+ },
2282
+ {
2283
+ "epoch": 0.328719723183391,
2284
+ "grad_norm": 2.015625,
2285
+ "kd_loss": 0.015753168612718582,
2286
+ "learning_rate": 7.122056164161795e-06,
2287
+ "loss": 0.0521,
2288
+ "step": 1140,
2289
+ "student_loss": 0.008625597693026066,
2290
+ "teacher_loss": 0.007858009077608585
2291
+ },
2292
+ {
2293
+ "epoch": 0.330161476355248,
2294
+ "grad_norm": 1.625,
2295
+ "kd_loss": 0.02173934131860733,
2296
+ "learning_rate": 7.097587081146636e-06,
2297
+ "loss": 0.0469,
2298
+ "step": 1145,
2299
+ "student_loss": 0.0009621863136999309,
2300
+ "teacher_loss": 0.0009391510975547135
2301
+ },
2302
+ {
2303
+ "epoch": 0.33160322952710497,
2304
+ "grad_norm": 3.1875,
2305
+ "kd_loss": 0.029764069244265556,
2306
+ "learning_rate": 7.073056879626681e-06,
2307
+ "loss": 0.0622,
2308
+ "step": 1150,
2309
+ "student_loss": 0.0005636148853227496,
2310
+ "teacher_loss": 0.0004029673582408577
2311
+ },
2312
+ {
2313
+ "epoch": 0.33304498269896193,
2314
+ "grad_norm": 5.4375,
2315
+ "kd_loss": 0.03274824470281601,
2316
+ "learning_rate": 7.048466274351389e-06,
2317
+ "loss": 0.0496,
2318
+ "step": 1155,
2319
+ "student_loss": 0.006183917168527842,
2320
+ "teacher_loss": 0.0001937657070811838
2321
+ },
2322
+ {
2323
+ "epoch": 0.3344867358708189,
2324
+ "grad_norm": 5.28125,
2325
+ "kd_loss": 0.03574247658252716,
2326
+ "learning_rate": 7.023815981830236e-06,
2327
+ "loss": 0.0634,
2328
+ "step": 1160,
2329
+ "student_loss": 0.030602917075157166,
2330
+ "teacher_loss": 0.02470449171960354
2331
+ },
2332
+ {
2333
+ "epoch": 0.3359284890426759,
2334
+ "grad_norm": 2.484375,
2335
+ "kd_loss": 0.032244376838207245,
2336
+ "learning_rate": 6.999106720311846e-06,
2337
+ "loss": 0.0495,
2338
+ "step": 1165,
2339
+ "student_loss": 0.05181184783577919,
2340
+ "teacher_loss": 0.00039351946907117963
2341
+ },
2342
+ {
2343
+ "epoch": 0.3373702422145329,
2344
+ "grad_norm": 2.734375,
2345
+ "kd_loss": 0.029991688206791878,
2346
+ "learning_rate": 6.974339209763043e-06,
2347
+ "loss": 0.0531,
2348
+ "step": 1170,
2349
+ "student_loss": 0.019939549267292023,
2350
+ "teacher_loss": 0.017931494861841202
2351
+ },
2352
+ {
2353
+ "epoch": 0.33881199538638984,
2354
+ "grad_norm": 5.78125,
2355
+ "kd_loss": 0.042389221489429474,
2356
+ "learning_rate": 6.949514171847891e-06,
2357
+ "loss": 0.0519,
2358
+ "step": 1175,
2359
+ "student_loss": 0.0012790773762390018,
2360
+ "teacher_loss": 0.0032734181731939316
2361
+ },
2362
+ {
2363
+ "epoch": 0.3402537485582468,
2364
+ "grad_norm": 3.15625,
2365
+ "kd_loss": 0.029558181762695312,
2366
+ "learning_rate": 6.924632329906657e-06,
2367
+ "loss": 0.0563,
2368
+ "step": 1180,
2369
+ "student_loss": 0.06338590383529663,
2370
+ "teacher_loss": 0.05036116763949394
2371
+ },
2372
+ {
2373
+ "epoch": 0.3416955017301038,
2374
+ "grad_norm": 2.15625,
2375
+ "kd_loss": 0.03158871456980705,
2376
+ "learning_rate": 6.899694408934734e-06,
2377
+ "loss": 0.0547,
2378
+ "step": 1185,
2379
+ "student_loss": 0.006782465614378452,
2380
+ "teacher_loss": 0.0041842032223939896
2381
+ },
2382
+ {
2383
+ "epoch": 0.3431372549019608,
2384
+ "grad_norm": 2.140625,
2385
+ "kd_loss": 0.03334498405456543,
2386
+ "learning_rate": 6.874701135561524e-06,
2387
+ "loss": 0.0544,
2388
+ "step": 1190,
2389
+ "student_loss": 0.0004166806465946138,
2390
+ "teacher_loss": 0.00023417123884428293
2391
+ },
2392
+ {
2393
+ "epoch": 0.34457900807381775,
2394
+ "grad_norm": 2.21875,
2395
+ "kd_loss": 0.029722491279244423,
2396
+ "learning_rate": 6.849653238029261e-06,
2397
+ "loss": 0.0539,
2398
+ "step": 1195,
2399
+ "student_loss": 0.0012941990280523896,
2400
+ "teacher_loss": 0.0004922052612528205
2401
+ },
2402
+ {
2403
+ "epoch": 0.3460207612456747,
2404
+ "grad_norm": 2.703125,
2405
+ "kd_loss": 0.029284365475177765,
2406
+ "learning_rate": 6.824551446171788e-06,
2407
+ "loss": 0.0623,
2408
+ "step": 1200,
2409
+ "student_loss": 0.000534356979187578,
2410
+ "teacher_loss": 0.00057708885287866
2411
+ },
2412
+ {
2413
+ "epoch": 0.34746251441753173,
2414
+ "grad_norm": 4.4375,
2415
+ "kd_loss": 0.03979990631341934,
2416
+ "learning_rate": 6.7993964913932975e-06,
2417
+ "loss": 0.0541,
2418
+ "step": 1205,
2419
+ "student_loss": 0.006738252472132444,
2420
+ "teacher_loss": 0.000260715140029788
2421
+ },
2422
+ {
2423
+ "epoch": 0.3489042675893887,
2424
+ "grad_norm": 2.78125,
2425
+ "kd_loss": 0.031953390687704086,
2426
+ "learning_rate": 6.774189106647021e-06,
2427
+ "loss": 0.0558,
2428
+ "step": 1210,
2429
+ "student_loss": 0.0004896536120213568,
2430
+ "teacher_loss": 0.00031862995820119977
2431
+ },
2432
+ {
2433
+ "epoch": 0.35034602076124566,
2434
+ "grad_norm": 4.59375,
2435
+ "kd_loss": 0.042156774550676346,
2436
+ "learning_rate": 6.748930026413865e-06,
2437
+ "loss": 0.0556,
2438
+ "step": 1215,
2439
+ "student_loss": 0.008981169201433659,
2440
+ "teacher_loss": 0.001311102183535695
2441
+ },
2442
+ {
2443
+ "epoch": 0.3517877739331027,
2444
+ "grad_norm": 8.5,
2445
+ "kd_loss": 0.04532252624630928,
2446
+ "learning_rate": 6.7236199866810185e-06,
2447
+ "loss": 0.0574,
2448
+ "step": 1220,
2449
+ "student_loss": 0.03215564414858818,
2450
+ "teacher_loss": 0.010875841602683067
2451
+ },
2452
+ {
2453
+ "epoch": 0.35322952710495964,
2454
+ "grad_norm": 5.0,
2455
+ "kd_loss": 0.02971484698355198,
2456
+ "learning_rate": 6.698259724920503e-06,
2457
+ "loss": 0.0567,
2458
+ "step": 1225,
2459
+ "student_loss": 0.0010156335774809122,
2460
+ "teacher_loss": 0.0021725716069340706
2461
+ },
2462
+ {
2463
+ "epoch": 0.3546712802768166,
2464
+ "grad_norm": 6.09375,
2465
+ "kd_loss": 0.017630236223340034,
2466
+ "learning_rate": 6.672849980067685e-06,
2467
+ "loss": 0.058,
2468
+ "step": 1230,
2469
+ "student_loss": 0.0004538622742984444,
2470
+ "teacher_loss": 0.0004401269252412021
2471
+ },
2472
+ {
2473
+ "epoch": 0.35611303344867357,
2474
+ "grad_norm": 3.6875,
2475
+ "kd_loss": 0.018272139132022858,
2476
+ "learning_rate": 6.647391492499746e-06,
2477
+ "loss": 0.0547,
2478
+ "step": 1235,
2479
+ "student_loss": 0.010347398929297924,
2480
+ "teacher_loss": 0.009122287854552269
2481
+ },
2482
+ {
2483
+ "epoch": 0.3575547866205306,
2484
+ "grad_norm": 2.75,
2485
+ "kd_loss": 0.0222186092287302,
2486
+ "learning_rate": 6.621885004014113e-06,
2487
+ "loss": 0.0663,
2488
+ "step": 1240,
2489
+ "student_loss": 0.0006056490237824619,
2490
+ "teacher_loss": 0.0005078864633105695
2491
+ },
2492
+ {
2493
+ "epoch": 0.35899653979238755,
2494
+ "grad_norm": 2.703125,
2495
+ "kd_loss": 0.03117276541888714,
2496
+ "learning_rate": 6.596331257806837e-06,
2497
+ "loss": 0.0565,
2498
+ "step": 1245,
2499
+ "student_loss": 0.0005393949686549604,
2500
+ "teacher_loss": 0.0004884015652351081
2501
+ },
2502
+ {
2503
+ "epoch": 0.3604382929642445,
2504
+ "grad_norm": 2.875,
2505
+ "kd_loss": 0.04197624325752258,
2506
+ "learning_rate": 6.570730998450945e-06,
2507
+ "loss": 0.0555,
2508
+ "step": 1250,
2509
+ "student_loss": 0.022176889702677727,
2510
+ "teacher_loss": 0.0035029917489737272
2511
+ },
2512
+ {
2513
+ "epoch": 0.3618800461361015,
2514
+ "grad_norm": 2.21875,
2515
+ "kd_loss": 0.029811320826411247,
2516
+ "learning_rate": 6.545084971874738e-06,
2517
+ "loss": 0.0585,
2518
+ "step": 1255,
2519
+ "student_loss": 0.016485337167978287,
2520
+ "teacher_loss": 0.024284513667225838
2521
+ },
2522
+ {
2523
+ "epoch": 0.3633217993079585,
2524
+ "grad_norm": 2.390625,
2525
+ "kd_loss": 0.026033582165837288,
2526
+ "learning_rate": 6.519393925340067e-06,
2527
+ "loss": 0.0609,
2528
+ "step": 1260,
2529
+ "student_loss": 0.0003027021884918213,
2530
+ "teacher_loss": 0.00038418604526668787
2531
+ },
2532
+ {
2533
+ "epoch": 0.36476355247981546,
2534
+ "grad_norm": 2.046875,
2535
+ "kd_loss": 0.04303126409649849,
2536
+ "learning_rate": 6.49365860742055e-06,
2537
+ "loss": 0.0593,
2538
+ "step": 1265,
2539
+ "student_loss": 0.009110042825341225,
2540
+ "teacher_loss": 0.020518532022833824
2541
+ },
2542
+ {
2543
+ "epoch": 0.3662053056516724,
2544
+ "grad_norm": 3.1875,
2545
+ "kd_loss": 0.023909209296107292,
2546
+ "learning_rate": 6.467879767979764e-06,
2547
+ "loss": 0.0525,
2548
+ "step": 1270,
2549
+ "student_loss": 0.0005614465335384011,
2550
+ "teacher_loss": 0.0004355542187113315
2551
+ },
2552
+ {
2553
+ "epoch": 0.36764705882352944,
2554
+ "grad_norm": 2.609375,
2555
+ "kd_loss": 0.026815881952643394,
2556
+ "learning_rate": 6.442058158149396e-06,
2557
+ "loss": 0.048,
2558
+ "step": 1275,
2559
+ "student_loss": 0.02352437563240528,
2560
+ "teacher_loss": 0.01322677731513977
2561
+ },
2562
+ {
2563
+ "epoch": 0.3690888119953864,
2564
+ "grad_norm": 2.15625,
2565
+ "kd_loss": 0.03451206162571907,
2566
+ "learning_rate": 6.4161945303073535e-06,
2567
+ "loss": 0.0516,
2568
+ "step": 1280,
2569
+ "student_loss": 0.0010189125314354897,
2570
+ "teacher_loss": 0.0003286921710241586
2571
+ },
2572
+ {
2573
+ "epoch": 0.37053056516724336,
2574
+ "grad_norm": 4.96875,
2575
+ "kd_loss": 0.025017155334353447,
2576
+ "learning_rate": 6.390289638055851e-06,
2577
+ "loss": 0.0548,
2578
+ "step": 1285,
2579
+ "student_loss": 0.0006344349822029471,
2580
+ "teacher_loss": 0.00043440310400910676
2581
+ },
2582
+ {
2583
+ "epoch": 0.3719723183391003,
2584
+ "grad_norm": 4.375,
2585
+ "kd_loss": 0.030148381367325783,
2586
+ "learning_rate": 6.364344236199441e-06,
2587
+ "loss": 0.0575,
2588
+ "step": 1290,
2589
+ "student_loss": 0.04187524691224098,
2590
+ "teacher_loss": 0.027156909927725792
2591
+ },
2592
+ {
2593
+ "epoch": 0.37341407151095735,
2594
+ "grad_norm": 5.65625,
2595
+ "kd_loss": 0.0274452343583107,
2596
+ "learning_rate": 6.3383590807230264e-06,
2597
+ "loss": 0.0624,
2598
+ "step": 1295,
2599
+ "student_loss": 0.000639778096228838,
2600
+ "teacher_loss": 0.008077265694737434
2601
+ },
2602
+ {
2603
+ "epoch": 0.3748558246828143,
2604
+ "grad_norm": 9.0,
2605
+ "kd_loss": 0.027542022988200188,
2606
+ "learning_rate": 6.3123349287698345e-06,
2607
+ "loss": 0.0503,
2608
+ "step": 1300,
2609
+ "student_loss": 0.0009962976910173893,
2610
+ "teacher_loss": 0.0005908433231525123
2611
+ },
2612
+ {
2613
+ "epoch": 0.3762975778546713,
2614
+ "grad_norm": 4.625,
2615
+ "kd_loss": 0.08821658045053482,
2616
+ "learning_rate": 6.286272538619351e-06,
2617
+ "loss": 0.0601,
2618
+ "step": 1305,
2619
+ "student_loss": 0.0014918975066393614,
2620
+ "teacher_loss": 0.28248000144958496
2621
+ },
2622
+ {
2623
+ "epoch": 0.37773933102652824,
2624
+ "grad_norm": 4.34375,
2625
+ "kd_loss": 0.022815359756350517,
2626
+ "learning_rate": 6.260172669665233e-06,
2627
+ "loss": 0.0522,
2628
+ "step": 1310,
2629
+ "student_loss": 0.0007336288690567017,
2630
+ "teacher_loss": 0.0006467080675065517
2631
+ },
2632
+ {
2633
+ "epoch": 0.37918108419838525,
2634
+ "grad_norm": 2.6875,
2635
+ "kd_loss": 0.026514001190662384,
2636
+ "learning_rate": 6.234036082393171e-06,
2637
+ "loss": 0.058,
2638
+ "step": 1315,
2639
+ "student_loss": 0.0030276973266154528,
2640
+ "teacher_loss": 0.0014054755447432399
2641
+ },
2642
+ {
2643
+ "epoch": 0.3806228373702422,
2644
+ "grad_norm": 3.390625,
2645
+ "kd_loss": 0.0344613678753376,
2646
+ "learning_rate": 6.207863538358741e-06,
2647
+ "loss": 0.0525,
2648
+ "step": 1320,
2649
+ "student_loss": 0.0002873508492484689,
2650
+ "teacher_loss": 0.00020539635443128645
2651
+ },
2652
+ {
2653
+ "epoch": 0.3820645905420992,
2654
+ "grad_norm": 3.921875,
2655
+ "kd_loss": 0.06418075412511826,
2656
+ "learning_rate": 6.181655800165207e-06,
2657
+ "loss": 0.0559,
2658
+ "step": 1325,
2659
+ "student_loss": 0.0010924733942374587,
2660
+ "teacher_loss": 0.0010531607549637556
2661
+ },
2662
+ {
2663
+ "epoch": 0.38350634371395614,
2664
+ "grad_norm": 8.25,
2665
+ "kd_loss": 0.04847671091556549,
2666
+ "learning_rate": 6.155413631441307e-06,
2667
+ "loss": 0.0542,
2668
+ "step": 1330,
2669
+ "student_loss": 0.006549970246851444,
2670
+ "teacher_loss": 0.00026613776572048664
2671
+ },
2672
+ {
2673
+ "epoch": 0.38494809688581316,
2674
+ "grad_norm": 3.78125,
2675
+ "kd_loss": 0.03629009798169136,
2676
+ "learning_rate": 6.129137796818997e-06,
2677
+ "loss": 0.0542,
2678
+ "step": 1335,
2679
+ "student_loss": 0.0005170853110030293,
2680
+ "teacher_loss": 0.0003051331441383809
2681
+ },
2682
+ {
2683
+ "epoch": 0.3863898500576701,
2684
+ "grad_norm": 2.90625,
2685
+ "kd_loss": 0.056832052767276764,
2686
+ "learning_rate": 6.102829061911176e-06,
2687
+ "loss": 0.059,
2688
+ "step": 1340,
2689
+ "student_loss": 0.01520230621099472,
2690
+ "teacher_loss": 0.006385981570929289
2691
+ },
2692
+ {
2693
+ "epoch": 0.3878316032295271,
2694
+ "grad_norm": 2.765625,
2695
+ "kd_loss": 0.0216154083609581,
2696
+ "learning_rate": 6.076488193289375e-06,
2697
+ "loss": 0.0571,
2698
+ "step": 1345,
2699
+ "student_loss": 0.0007815634598955512,
2700
+ "teacher_loss": 0.000781836686655879
2701
+ },
2702
+ {
2703
+ "epoch": 0.3892733564013841,
2704
+ "grad_norm": 2.1875,
2705
+ "kd_loss": 0.03685383126139641,
2706
+ "learning_rate": 6.050115958461423e-06,
2707
+ "loss": 0.0531,
2708
+ "step": 1350,
2709
+ "student_loss": 0.0021377273369580507,
2710
+ "teacher_loss": 0.0013112404849380255
2711
+ },
2712
+ {
2713
+ "epoch": 0.39071510957324107,
2714
+ "grad_norm": 2.890625,
2715
+ "kd_loss": 0.051180627197027206,
2716
+ "learning_rate": 6.02371312584908e-06,
2717
+ "loss": 0.0538,
2718
+ "step": 1355,
2719
+ "student_loss": 0.0007495570462197065,
2720
+ "teacher_loss": 0.00022196845384314656
2721
+ },
2722
+ {
2723
+ "epoch": 0.39215686274509803,
2724
+ "grad_norm": 4.03125,
2725
+ "kd_loss": 0.0423092395067215,
2726
+ "learning_rate": 5.997280464765655e-06,
2727
+ "loss": 0.057,
2728
+ "step": 1360,
2729
+ "student_loss": 0.0018419226398691535,
2730
+ "teacher_loss": 0.0012569584650918841
2731
+ },
2732
+ {
2733
+ "epoch": 0.393598615916955,
2734
+ "grad_norm": 7.53125,
2735
+ "kd_loss": 0.04048175364732742,
2736
+ "learning_rate": 5.970818745393579e-06,
2737
+ "loss": 0.0552,
2738
+ "step": 1365,
2739
+ "student_loss": 0.041127726435661316,
2740
+ "teacher_loss": 0.021389057859778404
2741
+ },
2742
+ {
2743
+ "epoch": 0.395040369088812,
2744
+ "grad_norm": 4.84375,
2745
+ "kd_loss": 0.03260928764939308,
2746
+ "learning_rate": 5.9443287387619754e-06,
2747
+ "loss": 0.0604,
2748
+ "step": 1370,
2749
+ "student_loss": 0.03800129145383835,
2750
+ "teacher_loss": 0.02555379644036293
2751
+ },
2752
+ {
2753
+ "epoch": 0.396482122260669,
2754
+ "grad_norm": 4.9375,
2755
+ "kd_loss": 0.07199908047914505,
2756
+ "learning_rate": 5.9178112167241805e-06,
2757
+ "loss": 0.0574,
2758
+ "step": 1375,
2759
+ "student_loss": 0.027047595009207726,
2760
+ "teacher_loss": 0.0034769964404404163
2761
+ },
2762
+ {
2763
+ "epoch": 0.39792387543252594,
2764
+ "grad_norm": 3.578125,
2765
+ "kd_loss": 0.02310655638575554,
2766
+ "learning_rate": 5.8912669519352725e-06,
2767
+ "loss": 0.0556,
2768
+ "step": 1380,
2769
+ "student_loss": 0.004811471328139305,
2770
+ "teacher_loss": 0.0035251553636044264
2771
+ },
2772
+ {
2773
+ "epoch": 0.3993656286043829,
2774
+ "grad_norm": 3.9375,
2775
+ "kd_loss": 0.03349766880273819,
2776
+ "learning_rate": 5.864696717829539e-06,
2777
+ "loss": 0.0575,
2778
+ "step": 1385,
2779
+ "student_loss": 0.016697963699698448,
2780
+ "teacher_loss": 0.011566993780434132
2781
+ },
2782
+ {
2783
+ "epoch": 0.4008073817762399,
2784
+ "grad_norm": 6.15625,
2785
+ "kd_loss": 0.039346568286418915,
2786
+ "learning_rate": 5.838101288597951e-06,
2787
+ "loss": 0.0538,
2788
+ "step": 1390,
2789
+ "student_loss": 0.0009079024312086403,
2790
+ "teacher_loss": 0.0006242908420972526
2791
+ },
2792
+ {
2793
+ "epoch": 0.4022491349480969,
2794
+ "grad_norm": 2.59375,
2795
+ "kd_loss": 0.042282599955797195,
2796
+ "learning_rate": 5.8114814391656046e-06,
2797
+ "loss": 0.0605,
2798
+ "step": 1395,
2799
+ "student_loss": 0.0006360080442391336,
2800
+ "teacher_loss": 0.00035487968125380576
2801
+ },
2802
+ {
2803
+ "epoch": 0.40369088811995385,
2804
+ "grad_norm": 5.21875,
2805
+ "kd_loss": 0.029627664014697075,
2806
+ "learning_rate": 5.78483794516914e-06,
2807
+ "loss": 0.0541,
2808
+ "step": 1400,
2809
+ "student_loss": 0.0007953746826387942,
2810
+ "teacher_loss": 0.0003530785324983299
2811
+ },
2812
+ {
2813
+ "epoch": 0.40513264129181087,
2814
+ "grad_norm": 2.484375,
2815
+ "kd_loss": 0.03120785765349865,
2816
+ "learning_rate": 5.75817158293414e-06,
2817
+ "loss": 0.0607,
2818
+ "step": 1405,
2819
+ "student_loss": 0.00045174715342000127,
2820
+ "teacher_loss": 0.000379163830075413
2821
+ },
2822
+ {
2823
+ "epoch": 0.40657439446366783,
2824
+ "grad_norm": 5.21875,
2825
+ "kd_loss": 0.030944446101784706,
2826
+ "learning_rate": 5.731483129452514e-06,
2827
+ "loss": 0.0521,
2828
+ "step": 1410,
2829
+ "student_loss": 0.00312256277538836,
2830
+ "teacher_loss": 0.0016517802141606808
2831
+ },
2832
+ {
2833
+ "epoch": 0.4080161476355248,
2834
+ "grad_norm": 3.140625,
2835
+ "kd_loss": 0.02221444807946682,
2836
+ "learning_rate": 5.704773362359854e-06,
2837
+ "loss": 0.0522,
2838
+ "step": 1415,
2839
+ "student_loss": 0.004858232103288174,
2840
+ "teacher_loss": 0.01279450673609972
2841
+ },
2842
+ {
2843
+ "epoch": 0.40945790080738176,
2844
+ "grad_norm": 4.875,
2845
+ "kd_loss": 0.042620472609996796,
2846
+ "learning_rate": 5.678043059912776e-06,
2847
+ "loss": 0.0594,
2848
+ "step": 1420,
2849
+ "student_loss": 0.0003699050284922123,
2850
+ "teacher_loss": 0.0008315423037856817
2851
+ },
2852
+ {
2853
+ "epoch": 0.4108996539792388,
2854
+ "grad_norm": 2.34375,
2855
+ "kd_loss": 0.0454087108373642,
2856
+ "learning_rate": 5.6512930009662524e-06,
2857
+ "loss": 0.0574,
2858
+ "step": 1425,
2859
+ "student_loss": 0.01310674473643303,
2860
+ "teacher_loss": 0.00045387831050902605
2861
+ },
2862
+ {
2863
+ "epoch": 0.41234140715109574,
2864
+ "grad_norm": 2.40625,
2865
+ "kd_loss": 0.028434637933969498,
2866
+ "learning_rate": 5.624523964950903e-06,
2867
+ "loss": 0.0494,
2868
+ "step": 1430,
2869
+ "student_loss": 0.0004379544116090983,
2870
+ "teacher_loss": 0.0003040166338905692
2871
+ },
2872
+ {
2873
+ "epoch": 0.4137831603229527,
2874
+ "grad_norm": 3.765625,
2875
+ "kd_loss": 0.02056478150188923,
2876
+ "learning_rate": 5.597736731850295e-06,
2877
+ "loss": 0.0527,
2878
+ "step": 1435,
2879
+ "student_loss": 0.0010475446470081806,
2880
+ "teacher_loss": 0.0007806891808286309
2881
+ },
2882
+ {
2883
+ "epoch": 0.41522491349480967,
2884
+ "grad_norm": 4.5,
2885
+ "kd_loss": 0.032079726457595825,
2886
+ "learning_rate": 5.570932082178219e-06,
2887
+ "loss": 0.058,
2888
+ "step": 1440,
2889
+ "student_loss": 0.005438623018562794,
2890
+ "teacher_loss": 0.00016811551176942885
2891
+ },
2892
+ {
2893
+ "epoch": 0.4166666666666667,
2894
+ "grad_norm": 2.09375,
2895
+ "kd_loss": 0.06181689724326134,
2896
+ "learning_rate": 5.5441107969559315e-06,
2897
+ "loss": 0.0517,
2898
+ "step": 1445,
2899
+ "student_loss": 0.011423390358686447,
2900
+ "teacher_loss": 0.004737222567200661
2901
+ },
2902
+ {
2903
+ "epoch": 0.41810841983852365,
2904
+ "grad_norm": 2.0625,
2905
+ "kd_loss": 0.034419264644384384,
2906
+ "learning_rate": 5.517273657689419e-06,
2907
+ "loss": 0.0523,
2908
+ "step": 1450,
2909
+ "student_loss": 0.004626914393156767,
2910
+ "teacher_loss": 0.002724819350987673
2911
+ },
2912
+ {
2913
+ "epoch": 0.4195501730103806,
2914
+ "grad_norm": 3.078125,
2915
+ "kd_loss": 0.03091241978108883,
2916
+ "learning_rate": 5.490421446346608e-06,
2917
+ "loss": 0.0531,
2918
+ "step": 1455,
2919
+ "student_loss": 0.00043606918188743293,
2920
+ "teacher_loss": 0.00032708875369280577
2921
+ },
2922
+ {
2923
+ "epoch": 0.4209919261822376,
2924
+ "grad_norm": 2.78125,
2925
+ "kd_loss": 0.0346670001745224,
2926
+ "learning_rate": 5.463554945334589e-06,
2927
+ "loss": 0.0566,
2928
+ "step": 1460,
2929
+ "student_loss": 0.000474665459478274,
2930
+ "teacher_loss": 0.00014825857942923903
2931
+ },
2932
+ {
2933
+ "epoch": 0.4224336793540946,
2934
+ "grad_norm": 3.40625,
2935
+ "kd_loss": 0.02378292940557003,
2936
+ "learning_rate": 5.43667493747682e-06,
2937
+ "loss": 0.0725,
2938
+ "step": 1465,
2939
+ "student_loss": 0.014978962950408459,
2940
+ "teacher_loss": 0.01323742512613535
2941
+ },
2942
+ {
2943
+ "epoch": 0.42387543252595156,
2944
+ "grad_norm": 4.0625,
2945
+ "kd_loss": 0.0335330069065094,
2946
+ "learning_rate": 5.409782205990317e-06,
2947
+ "loss": 0.0699,
2948
+ "step": 1470,
2949
+ "student_loss": 0.0011153829982504249,
2950
+ "teacher_loss": 0.0039979517459869385
2951
+ },
2952
+ {
2953
+ "epoch": 0.4253171856978085,
2954
+ "grad_norm": 2.65625,
2955
+ "kd_loss": 0.01991354115307331,
2956
+ "learning_rate": 5.3828775344628245e-06,
2957
+ "loss": 0.0494,
2958
+ "step": 1475,
2959
+ "student_loss": 0.0005757836042903364,
2960
+ "teacher_loss": 0.0005198651342652738
2961
+ },
2962
+ {
2963
+ "epoch": 0.42675893886966554,
2964
+ "grad_norm": 2.875,
2965
+ "kd_loss": 0.017849793657660484,
2966
+ "learning_rate": 5.355961706829997e-06,
2967
+ "loss": 0.0682,
2968
+ "step": 1480,
2969
+ "student_loss": 0.000734495057258755,
2970
+ "teacher_loss": 0.0007692194194532931
2971
+ },
2972
+ {
2973
+ "epoch": 0.4282006920415225,
2974
+ "grad_norm": 13.0,
2975
+ "kd_loss": 0.015589379705488682,
2976
+ "learning_rate": 5.329035507352548e-06,
2977
+ "loss": 0.0562,
2978
+ "step": 1485,
2979
+ "student_loss": 0.000377491902327165,
2980
+ "teacher_loss": 0.0003673851606436074
2981
+ },
2982
+ {
2983
+ "epoch": 0.42964244521337946,
2984
+ "grad_norm": 3.515625,
2985
+ "kd_loss": 0.015487972646951675,
2986
+ "learning_rate": 5.3020997205933985e-06,
2987
+ "loss": 0.0532,
2988
+ "step": 1490,
2989
+ "student_loss": 0.009750030003488064,
2990
+ "teacher_loss": 0.008933374658226967
2991
+ },
2992
+ {
2993
+ "epoch": 0.43108419838523643,
2994
+ "grad_norm": 2.671875,
2995
+ "kd_loss": 0.0228768028318882,
2996
+ "learning_rate": 5.275155131394825e-06,
2997
+ "loss": 0.0571,
2998
+ "step": 1495,
2999
+ "student_loss": 0.0008404748514294624,
3000
+ "teacher_loss": 0.0005730246775783598
3001
+ },
3002
+ {
3003
+ "epoch": 0.43252595155709345,
3004
+ "grad_norm": 3.71875,
3005
+ "kd_loss": 0.07327655702829361,
3006
+ "learning_rate": 5.248202524855578e-06,
3007
+ "loss": 0.0543,
3008
+ "step": 1500,
3009
+ "student_loss": 0.014728349633514881,
3010
+ "teacher_loss": 0.004950221627950668
3011
+ }
3012
+ ],
3013
+ "logging_steps": 5,
3014
+ "max_steps": 3000,
3015
+ "num_input_tokens_seen": 0,
3016
+ "num_train_epochs": 1,
3017
+ "save_steps": 500,
3018
+ "stateful_callbacks": {
3019
+ "TrainerControl": {
3020
+ "args": {
3021
+ "should_epoch_stop": false,
3022
+ "should_evaluate": false,
3023
+ "should_log": false,
3024
+ "should_save": true,
3025
+ "should_training_stop": false
3026
+ },
3027
+ "attributes": {}
3028
+ }
3029
+ },
3030
+ "total_flos": 0.0,
3031
+ "train_batch_size": 1,
3032
+ "trial_name": null,
3033
+ "trial_params": null
3034
+ }
checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-1500/vocab.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/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/codi3b_a1.0_b1.0_g1.0_ls1/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 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 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/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-500/config.json ADDED
@@ -0,0 +1,66 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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": 2048,
10
+ "initializer_range": 0.02,
11
+ "intermediate_size": 11008,
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
+ "full_attention",
42
+ "full_attention",
43
+ "full_attention",
44
+ "full_attention",
45
+ "full_attention",
46
+ "full_attention",
47
+ "full_attention",
48
+ "full_attention"
49
+ ],
50
+ "max_position_embeddings": 32768,
51
+ "max_window_layers": 36,
52
+ "model_type": "qwen2",
53
+ "num_attention_heads": 16,
54
+ "num_hidden_layers": 36,
55
+ "num_key_value_heads": 2,
56
+ "pad_token_id": 151643,
57
+ "rms_norm_eps": 1e-06,
58
+ "rope_scaling": null,
59
+ "rope_theta": 1000000.0,
60
+ "sliding_window": null,
61
+ "tie_word_embeddings": true,
62
+ "transformers_version": "4.57.6",
63
+ "use_cache": true,
64
+ "use_sliding_window": false,
65
+ "vocab_size": 151676
66
+ }
checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-500/merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-500/pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0810eff1d4453fc9064c2982101f18a59c4299176995a4e0e38536fc9353f0ce
3
+ size 6187858991
checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/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/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-500/thought_projector.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f34568355545596b04c2f50c68774e815d252de3392056e392f6386820cc6ba6
3
+ size 16788033
checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/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/codi3b_a1.0_b1.0_g1.0_ls1/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/codi3b_a1.0_b1.0_g1.0_ls1/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": 42.0,
15
+ "kd_loss": 0.2894527316093445,
16
+ "learning_rate": 4.444444444444445e-07,
17
+ "loss": 0.5205,
18
+ "step": 5,
19
+ "student_loss": 0.12883709371089935,
20
+ "teacher_loss": 0.001858799601905048
21
+ },
22
+ {
23
+ "epoch": 0.0028835063437139563,
24
+ "grad_norm": 31.875,
25
+ "kd_loss": 0.2677682042121887,
26
+ "learning_rate": 1.0000000000000002e-06,
27
+ "loss": 0.4924,
28
+ "step": 10,
29
+ "student_loss": 0.17924517393112183,
30
+ "teacher_loss": 0.005514730233699083
31
+ },
32
+ {
33
+ "epoch": 0.004325259515570935,
34
+ "grad_norm": 33.25,
35
+ "kd_loss": 0.23119957745075226,
36
+ "learning_rate": 1.5555555555555558e-06,
37
+ "loss": 0.4551,
38
+ "step": 15,
39
+ "student_loss": 0.062010008841753006,
40
+ "teacher_loss": 0.0011852148454636335
41
+ },
42
+ {
43
+ "epoch": 0.0057670126874279125,
44
+ "grad_norm": 25.5,
45
+ "kd_loss": 0.20175831019878387,
46
+ "learning_rate": 2.1111111111111114e-06,
47
+ "loss": 0.4171,
48
+ "step": 20,
49
+ "student_loss": 0.08724690973758698,
50
+ "teacher_loss": 0.0005447491421364248
51
+ },
52
+ {
53
+ "epoch": 0.00720876585928489,
54
+ "grad_norm": 14.0,
55
+ "kd_loss": 0.20470531284809113,
56
+ "learning_rate": 2.666666666666667e-06,
57
+ "loss": 0.3355,
58
+ "step": 25,
59
+ "student_loss": 0.028787745162844658,
60
+ "teacher_loss": 0.0008465723367407918
61
+ },
62
+ {
63
+ "epoch": 0.00865051903114187,
64
+ "grad_norm": 7.59375,
65
+ "kd_loss": 0.17482763528823853,
66
+ "learning_rate": 3.2222222222222227e-06,
67
+ "loss": 0.2813,
68
+ "step": 30,
69
+ "student_loss": 0.04111301898956299,
70
+ "teacher_loss": 0.0003613036824390292
71
+ },
72
+ {
73
+ "epoch": 0.010092272202998846,
74
+ "grad_norm": 5.46875,
75
+ "kd_loss": 0.1520802527666092,
76
+ "learning_rate": 3.777777777777778e-06,
77
+ "loss": 0.2499,
78
+ "step": 35,
79
+ "student_loss": 0.029766706749796867,
80
+ "teacher_loss": 0.0002796741609927267
81
+ },
82
+ {
83
+ "epoch": 0.011534025374855825,
84
+ "grad_norm": 4.4375,
85
+ "kd_loss": 0.11269068717956543,
86
+ "learning_rate": 4.333333333333334e-06,
87
+ "loss": 0.2082,
88
+ "step": 40,
89
+ "student_loss": 0.05098738893866539,
90
+ "teacher_loss": 0.04165218397974968
91
+ },
92
+ {
93
+ "epoch": 0.012975778546712802,
94
+ "grad_norm": 3.078125,
95
+ "kd_loss": 0.20585855841636658,
96
+ "learning_rate": 4.888888888888889e-06,
97
+ "loss": 0.1605,
98
+ "step": 45,
99
+ "student_loss": 0.08690722286701202,
100
+ "teacher_loss": 0.0026482592802494764
101
+ },
102
+ {
103
+ "epoch": 0.01441753171856978,
104
+ "grad_norm": 2.734375,
105
+ "kd_loss": 0.074299156665802,
106
+ "learning_rate": 5.444444444444445e-06,
107
+ "loss": 0.1578,
108
+ "step": 50,
109
+ "student_loss": 0.017733801156282425,
110
+ "teacher_loss": 0.008056458085775375
111
+ },
112
+ {
113
+ "epoch": 0.015859284890426758,
114
+ "grad_norm": 2.78125,
115
+ "kd_loss": 0.062237780541181564,
116
+ "learning_rate": 6e-06,
117
+ "loss": 0.135,
118
+ "step": 55,
119
+ "student_loss": 0.0012035972904413939,
120
+ "teacher_loss": 0.000626895169261843
121
+ },
122
+ {
123
+ "epoch": 0.01730103806228374,
124
+ "grad_norm": 2.65625,
125
+ "kd_loss": 0.05602008104324341,
126
+ "learning_rate": 6.555555555555556e-06,
127
+ "loss": 0.1184,
128
+ "step": 60,
129
+ "student_loss": 0.002416773699223995,
130
+ "teacher_loss": 0.0013880858896300197
131
+ },
132
+ {
133
+ "epoch": 0.018742791234140715,
134
+ "grad_norm": 4.65625,
135
+ "kd_loss": 0.051335666328668594,
136
+ "learning_rate": 7.111111111111112e-06,
137
+ "loss": 0.1118,
138
+ "step": 65,
139
+ "student_loss": 0.027237623929977417,
140
+ "teacher_loss": 0.03812175989151001
141
+ },
142
+ {
143
+ "epoch": 0.020184544405997693,
144
+ "grad_norm": 1.7109375,
145
+ "kd_loss": 0.0913488045334816,
146
+ "learning_rate": 7.666666666666667e-06,
147
+ "loss": 0.1042,
148
+ "step": 70,
149
+ "student_loss": 0.015874937176704407,
150
+ "teacher_loss": 0.0013995245099067688
151
+ },
152
+ {
153
+ "epoch": 0.02162629757785467,
154
+ "grad_norm": 2.890625,
155
+ "kd_loss": 0.036495719105005264,
156
+ "learning_rate": 8.222222222222222e-06,
157
+ "loss": 0.0961,
158
+ "step": 75,
159
+ "student_loss": 0.0018275808542966843,
160
+ "teacher_loss": 0.000998618663288653
161
+ },
162
+ {
163
+ "epoch": 0.02306805074971165,
164
+ "grad_norm": 2.8125,
165
+ "kd_loss": 0.033367447555065155,
166
+ "learning_rate": 8.777777777777778e-06,
167
+ "loss": 0.0811,
168
+ "step": 80,
169
+ "student_loss": 0.001792682334780693,
170
+ "teacher_loss": 0.0010247478494420648
171
+ },
172
+ {
173
+ "epoch": 0.024509803921568627,
174
+ "grad_norm": 2.734375,
175
+ "kd_loss": 0.0299057699739933,
176
+ "learning_rate": 9.333333333333334e-06,
177
+ "loss": 0.0827,
178
+ "step": 85,
179
+ "student_loss": 0.0028650059830397367,
180
+ "teacher_loss": 0.004752026870846748
181
+ },
182
+ {
183
+ "epoch": 0.025951557093425604,
184
+ "grad_norm": 1.7890625,
185
+ "kd_loss": 0.08093900978565216,
186
+ "learning_rate": 9.88888888888889e-06,
187
+ "loss": 0.0755,
188
+ "step": 90,
189
+ "student_loss": 0.019452862441539764,
190
+ "teacher_loss": 0.0028147350531071424
191
+ },
192
+ {
193
+ "epoch": 0.027393310265282585,
194
+ "grad_norm": 1.4296875,
195
+ "kd_loss": 0.04993116110563278,
196
+ "learning_rate": 9.99995337991255e-06,
197
+ "loss": 0.0683,
198
+ "step": 95,
199
+ "student_loss": 0.009235645644366741,
200
+ "teacher_loss": 0.006103724706918001
201
+ },
202
+ {
203
+ "epoch": 0.02883506343713956,
204
+ "grad_norm": 1.546875,
205
+ "kd_loss": 0.02219005487859249,
206
+ "learning_rate": 9.999763987297266e-06,
207
+ "loss": 0.0591,
208
+ "step": 100,
209
+ "student_loss": 0.0009554760181345046,
210
+ "teacher_loss": 0.000685701088514179
211
+ },
212
+ {
213
+ "epoch": 0.03027681660899654,
214
+ "grad_norm": 2.53125,
215
+ "kd_loss": 0.03886059299111366,
216
+ "learning_rate": 9.999428913912876e-06,
217
+ "loss": 0.0632,
218
+ "step": 105,
219
+ "student_loss": 0.002928090514615178,
220
+ "teacher_loss": 0.00032716552959755063
221
+ },
222
+ {
223
+ "epoch": 0.031718569780853516,
224
+ "grad_norm": 2.171875,
225
+ "kd_loss": 0.04434794560074806,
226
+ "learning_rate": 9.998948169522587e-06,
227
+ "loss": 0.063,
228
+ "step": 110,
229
+ "student_loss": 0.033939410001039505,
230
+ "teacher_loss": 0.017983509227633476
231
+ },
232
+ {
233
+ "epoch": 0.03316032295271049,
234
+ "grad_norm": 1.3046875,
235
+ "kd_loss": 0.03771444410085678,
236
+ "learning_rate": 9.998321768134101e-06,
237
+ "loss": 0.0606,
238
+ "step": 115,
239
+ "student_loss": 0.0007961422670632601,
240
+ "teacher_loss": 0.00038347821100614965
241
+ },
242
+ {
243
+ "epoch": 0.03460207612456748,
244
+ "grad_norm": 1.6953125,
245
+ "kd_loss": 0.02649487555027008,
246
+ "learning_rate": 9.997549727999212e-06,
247
+ "loss": 0.0532,
248
+ "step": 120,
249
+ "student_loss": 0.0012731363531202078,
250
+ "teacher_loss": 0.0007500122301280499
251
+ },
252
+ {
253
+ "epoch": 0.036043829296424454,
254
+ "grad_norm": 2.09375,
255
+ "kd_loss": 0.057950329035520554,
256
+ "learning_rate": 9.996632071613257e-06,
257
+ "loss": 0.0601,
258
+ "step": 125,
259
+ "student_loss": 0.03109952248632908,
260
+ "teacher_loss": 0.000994477653875947
261
+ },
262
+ {
263
+ "epoch": 0.03748558246828143,
264
+ "grad_norm": 0.95703125,
265
+ "kd_loss": 0.026637699455022812,
266
+ "learning_rate": 9.995568825714479e-06,
267
+ "loss": 0.0574,
268
+ "step": 130,
269
+ "student_loss": 0.0008338625193573534,
270
+ "teacher_loss": 0.0006465786718763411
271
+ },
272
+ {
273
+ "epoch": 0.03892733564013841,
274
+ "grad_norm": 1.3515625,
275
+ "kd_loss": 0.04235714301466942,
276
+ "learning_rate": 9.994360021283235e-06,
277
+ "loss": 0.0611,
278
+ "step": 135,
279
+ "student_loss": 0.001498939236626029,
280
+ "teacher_loss": 0.0005972529761493206
281
+ },
282
+ {
283
+ "epoch": 0.040369088811995385,
284
+ "grad_norm": 1.46875,
285
+ "kd_loss": 0.024005092680454254,
286
+ "learning_rate": 9.993005693541098e-06,
287
+ "loss": 0.0562,
288
+ "step": 140,
289
+ "student_loss": 0.0008620141888968647,
290
+ "teacher_loss": 0.0004937360063195229
291
+ },
292
+ {
293
+ "epoch": 0.04181084198385236,
294
+ "grad_norm": 5.5,
295
+ "kd_loss": 0.07175353169441223,
296
+ "learning_rate": 9.991505881949837e-06,
297
+ "loss": 0.0539,
298
+ "step": 145,
299
+ "student_loss": 0.03820747137069702,
300
+ "teacher_loss": 0.006569729186594486
301
+ },
302
+ {
303
+ "epoch": 0.04325259515570934,
304
+ "grad_norm": 1.2109375,
305
+ "kd_loss": 0.016246870160102844,
306
+ "learning_rate": 9.989860630210251e-06,
307
+ "loss": 0.0538,
308
+ "step": 150,
309
+ "student_loss": 0.0014459972735494375,
310
+ "teacher_loss": 0.001147772534750402
311
+ },
312
+ {
313
+ "epoch": 0.04469434832756632,
314
+ "grad_norm": 1.4765625,
315
+ "kd_loss": 0.02643739990890026,
316
+ "learning_rate": 9.98806998626091e-06,
317
+ "loss": 0.0592,
318
+ "step": 155,
319
+ "student_loss": 0.0027810579631477594,
320
+ "teacher_loss": 0.000860289903357625
321
+ },
322
+ {
323
+ "epoch": 0.0461361014994233,
324
+ "grad_norm": 2.40625,
325
+ "kd_loss": 0.029045773670077324,
326
+ "learning_rate": 9.98613400227676e-06,
327
+ "loss": 0.0597,
328
+ "step": 160,
329
+ "student_loss": 0.08056973665952682,
330
+ "teacher_loss": 0.04562501981854439
331
+ },
332
+ {
333
+ "epoch": 0.04757785467128028,
334
+ "grad_norm": 1.4140625,
335
+ "kd_loss": 0.03424260392785072,
336
+ "learning_rate": 9.984052734667586e-06,
337
+ "loss": 0.0523,
338
+ "step": 165,
339
+ "student_loss": 0.0027138988953083754,
340
+ "teacher_loss": 0.005776015110313892
341
+ },
342
+ {
343
+ "epoch": 0.049019607843137254,
344
+ "grad_norm": 2.859375,
345
+ "kd_loss": 0.019065098837018013,
346
+ "learning_rate": 9.981826244076387e-06,
347
+ "loss": 0.0562,
348
+ "step": 170,
349
+ "student_loss": 0.0007321524317376316,
350
+ "teacher_loss": 0.000522188376635313
351
+ },
352
+ {
353
+ "epoch": 0.05046136101499423,
354
+ "grad_norm": 1.7578125,
355
+ "kd_loss": 0.027638504281640053,
356
+ "learning_rate": 9.979454595377594e-06,
357
+ "loss": 0.0529,
358
+ "step": 175,
359
+ "student_loss": 0.004792944528162479,
360
+ "teacher_loss": 0.000635432661511004
361
+ },
362
+ {
363
+ "epoch": 0.05190311418685121,
364
+ "grad_norm": 2.0,
365
+ "kd_loss": 0.03646935150027275,
366
+ "learning_rate": 9.97693785767519e-06,
367
+ "loss": 0.0514,
368
+ "step": 180,
369
+ "student_loss": 0.0014396726619452238,
370
+ "teacher_loss": 0.0003402946167625487
371
+ },
372
+ {
373
+ "epoch": 0.05334486735870819,
374
+ "grad_norm": 1.453125,
375
+ "kd_loss": 0.03992177173495293,
376
+ "learning_rate": 9.9742761043007e-06,
377
+ "loss": 0.0582,
378
+ "step": 185,
379
+ "student_loss": 0.03759889304637909,
380
+ "teacher_loss": 0.00029035221086815
381
+ },
382
+ {
383
+ "epoch": 0.05478662053056517,
384
+ "grad_norm": 3.890625,
385
+ "kd_loss": 0.05405907332897186,
386
+ "learning_rate": 9.971469412811032e-06,
387
+ "loss": 0.0527,
388
+ "step": 190,
389
+ "student_loss": 0.004218568559736013,
390
+ "teacher_loss": 0.0009858696721494198
391
+ },
392
+ {
393
+ "epoch": 0.056228373702422146,
394
+ "grad_norm": 2.375,
395
+ "kd_loss": 0.05194515734910965,
396
+ "learning_rate": 9.968517864986247e-06,
397
+ "loss": 0.0524,
398
+ "step": 195,
399
+ "student_loss": 0.005763455759733915,
400
+ "teacher_loss": 0.0004817036970052868
401
+ },
402
+ {
403
+ "epoch": 0.05767012687427912,
404
+ "grad_norm": 1.2109375,
405
+ "kd_loss": 0.036061037331819534,
406
+ "learning_rate": 9.965421546827153e-06,
407
+ "loss": 0.064,
408
+ "step": 200,
409
+ "student_loss": 0.0005494994111359119,
410
+ "teacher_loss": 0.00020005257101729512
411
+ },
412
+ {
413
+ "epoch": 0.0591118800461361,
414
+ "grad_norm": 2.40625,
415
+ "kd_loss": 0.02808363176882267,
416
+ "learning_rate": 9.962180548552812e-06,
417
+ "loss": 0.0577,
418
+ "step": 205,
419
+ "student_loss": 0.03726515918970108,
420
+ "teacher_loss": 0.028193596750497818
421
+ },
422
+ {
423
+ "epoch": 0.06055363321799308,
424
+ "grad_norm": 4.15625,
425
+ "kd_loss": 0.03680839017033577,
426
+ "learning_rate": 9.958794964597906e-06,
427
+ "loss": 0.0515,
428
+ "step": 210,
429
+ "student_loss": 0.0011522703571245074,
430
+ "teacher_loss": 0.0004914712626487017
431
+ },
432
+ {
433
+ "epoch": 0.061995386389850055,
434
+ "grad_norm": 5.65625,
435
+ "kd_loss": 0.037980884313583374,
436
+ "learning_rate": 9.955264893609986e-06,
437
+ "loss": 0.0559,
438
+ "step": 215,
439
+ "student_loss": 0.0005635322886519134,
440
+ "teacher_loss": 0.00027164584025740623
441
+ },
442
+ {
443
+ "epoch": 0.06343713956170703,
444
+ "grad_norm": 4.0,
445
+ "kd_loss": 0.04362543299794197,
446
+ "learning_rate": 9.951590438446597e-06,
447
+ "loss": 0.0507,
448
+ "step": 220,
449
+ "student_loss": 0.032083068042993546,
450
+ "teacher_loss": 0.010480552911758423
451
+ },
452
+ {
453
+ "epoch": 0.06487889273356401,
454
+ "grad_norm": 1.3125,
455
+ "kd_loss": 0.02105102129280567,
456
+ "learning_rate": 9.947771706172285e-06,
457
+ "loss": 0.0605,
458
+ "step": 225,
459
+ "student_loss": 0.038467925041913986,
460
+ "teacher_loss": 0.0018907957710325718
461
+ },
462
+ {
463
+ "epoch": 0.06632064590542099,
464
+ "grad_norm": 2.25,
465
+ "kd_loss": 0.04121282696723938,
466
+ "learning_rate": 9.943808808055472e-06,
467
+ "loss": 0.0495,
468
+ "step": 230,
469
+ "student_loss": 0.01008139830082655,
470
+ "teacher_loss": 0.003380328882485628
471
+ },
472
+ {
473
+ "epoch": 0.06776239907727798,
474
+ "grad_norm": 9.5,
475
+ "kd_loss": 0.07618479430675507,
476
+ "learning_rate": 9.93970185956522e-06,
477
+ "loss": 0.0516,
478
+ "step": 235,
479
+ "student_loss": 0.010011663660407066,
480
+ "teacher_loss": 0.0004287679912522435
481
+ },
482
+ {
483
+ "epoch": 0.06920415224913495,
484
+ "grad_norm": 2.796875,
485
+ "kd_loss": 0.031842563301324844,
486
+ "learning_rate": 9.93545098036786e-06,
487
+ "loss": 0.0567,
488
+ "step": 240,
489
+ "student_loss": 0.0031024671625345945,
490
+ "teacher_loss": 0.0003865428443532437
491
+ },
492
+ {
493
+ "epoch": 0.07064590542099193,
494
+ "grad_norm": 1.921875,
495
+ "kd_loss": 0.01385761983692646,
496
+ "learning_rate": 9.931056294323507e-06,
497
+ "loss": 0.048,
498
+ "step": 245,
499
+ "student_loss": 0.011989925056695938,
500
+ "teacher_loss": 0.015163863077759743
501
+ },
502
+ {
503
+ "epoch": 0.07208765859284891,
504
+ "grad_norm": 1.8125,
505
+ "kd_loss": 0.020028578117489815,
506
+ "learning_rate": 9.926517929482454e-06,
507
+ "loss": 0.0514,
508
+ "step": 250,
509
+ "student_loss": 0.0005340924835763872,
510
+ "teacher_loss": 0.000291565025690943
511
+ },
512
+ {
513
+ "epoch": 0.07352941176470588,
514
+ "grad_norm": 1.8359375,
515
+ "kd_loss": 0.01568070985376835,
516
+ "learning_rate": 9.921836018081438e-06,
517
+ "loss": 0.0512,
518
+ "step": 255,
519
+ "student_loss": 0.0003255972405895591,
520
+ "teacher_loss": 0.0003198981285095215
521
+ },
522
+ {
523
+ "epoch": 0.07497116493656286,
524
+ "grad_norm": 1.65625,
525
+ "kd_loss": 0.03392383083701134,
526
+ "learning_rate": 9.917010696539793e-06,
527
+ "loss": 0.0577,
528
+ "step": 260,
529
+ "student_loss": 0.0405547134578228,
530
+ "teacher_loss": 0.03275086730718613
531
+ },
532
+ {
533
+ "epoch": 0.07641291810841984,
534
+ "grad_norm": 4.96875,
535
+ "kd_loss": 0.018212661147117615,
536
+ "learning_rate": 9.912042105455462e-06,
537
+ "loss": 0.0565,
538
+ "step": 265,
539
+ "student_loss": 0.0005924144061282277,
540
+ "teacher_loss": 0.0004515529435593635
541
+ },
542
+ {
543
+ "epoch": 0.07785467128027682,
544
+ "grad_norm": 1.484375,
545
+ "kd_loss": 0.020866869017481804,
546
+ "learning_rate": 9.906930389600917e-06,
547
+ "loss": 0.0593,
548
+ "step": 270,
549
+ "student_loss": 0.0013341928133741021,
550
+ "teacher_loss": 0.0011696964502334595
551
+ },
552
+ {
553
+ "epoch": 0.07929642445213379,
554
+ "grad_norm": 2.1875,
555
+ "kd_loss": 0.02725479006767273,
556
+ "learning_rate": 9.901675697918925e-06,
557
+ "loss": 0.0511,
558
+ "step": 275,
559
+ "student_loss": 0.048831257969141006,
560
+ "teacher_loss": 0.035170599818229675
561
+ },
562
+ {
563
+ "epoch": 0.08073817762399077,
564
+ "grad_norm": 4.75,
565
+ "kd_loss": 0.03362221270799637,
566
+ "learning_rate": 9.896278183518216e-06,
567
+ "loss": 0.0509,
568
+ "step": 280,
569
+ "student_loss": 0.0004928835551254451,
570
+ "teacher_loss": 0.00017192006635013968
571
+ },
572
+ {
573
+ "epoch": 0.08217993079584775,
574
+ "grad_norm": 2.921875,
575
+ "kd_loss": 0.021101269870996475,
576
+ "learning_rate": 9.890738003669029e-06,
577
+ "loss": 0.0572,
578
+ "step": 285,
579
+ "student_loss": 0.000973534828517586,
580
+ "teacher_loss": 0.0008148897904902697
581
+ },
582
+ {
583
+ "epoch": 0.08362168396770472,
584
+ "grad_norm": 2.203125,
585
+ "kd_loss": 0.017731280997395515,
586
+ "learning_rate": 9.885055319798514e-06,
587
+ "loss": 0.0582,
588
+ "step": 290,
589
+ "student_loss": 0.014440718106925488,
590
+ "teacher_loss": 0.012271302752196789
591
+ },
592
+ {
593
+ "epoch": 0.0850634371395617,
594
+ "grad_norm": 2.46875,
595
+ "kd_loss": 0.01968582533299923,
596
+ "learning_rate": 9.879230297486034e-06,
597
+ "loss": 0.0561,
598
+ "step": 295,
599
+ "student_loss": 0.0010608865413814783,
600
+ "teacher_loss": 0.0008463782141916454
601
+ },
602
+ {
603
+ "epoch": 0.08650519031141868,
604
+ "grad_norm": 3.34375,
605
+ "kd_loss": 0.05237875133752823,
606
+ "learning_rate": 9.873263106458353e-06,
607
+ "loss": 0.0526,
608
+ "step": 300,
609
+ "student_loss": 0.017360921949148178,
610
+ "teacher_loss": 0.001965837087482214
611
+ },
612
+ {
613
+ "epoch": 0.08794694348327567,
614
+ "grad_norm": 2.296875,
615
+ "kd_loss": 0.023132195696234703,
616
+ "learning_rate": 9.86715392058467e-06,
617
+ "loss": 0.0485,
618
+ "step": 305,
619
+ "student_loss": 0.007244116626679897,
620
+ "teacher_loss": 0.003649009158834815
621
+ },
622
+ {
623
+ "epoch": 0.08938869665513265,
624
+ "grad_norm": 2.671875,
625
+ "kd_loss": 0.020483316853642464,
626
+ "learning_rate": 9.860902917871566e-06,
627
+ "loss": 0.0535,
628
+ "step": 310,
629
+ "student_loss": 0.0007183508132584393,
630
+ "teacher_loss": 0.0004938851925544441
631
+ },
632
+ {
633
+ "epoch": 0.09083044982698962,
634
+ "grad_norm": 2.828125,
635
+ "kd_loss": 0.04086384177207947,
636
+ "learning_rate": 9.854510280457819e-06,
637
+ "loss": 0.0521,
638
+ "step": 315,
639
+ "student_loss": 0.010588191449642181,
640
+ "teacher_loss": 0.00029907570569776
641
+ },
642
+ {
643
+ "epoch": 0.0922722029988466,
644
+ "grad_norm": 2.75,
645
+ "kd_loss": 0.026799874380230904,
646
+ "learning_rate": 9.847976194609083e-06,
647
+ "loss": 0.055,
648
+ "step": 320,
649
+ "student_loss": 0.0006185687379911542,
650
+ "teacher_loss": 0.0003583572688512504
651
+ },
652
+ {
653
+ "epoch": 0.09371395617070358,
654
+ "grad_norm": 2.265625,
655
+ "kd_loss": 0.07637664675712585,
656
+ "learning_rate": 9.841300850712479e-06,
657
+ "loss": 0.0516,
658
+ "step": 325,
659
+ "student_loss": 0.008596767671406269,
660
+ "teacher_loss": 0.0225156731903553
661
+ },
662
+ {
663
+ "epoch": 0.09515570934256055,
664
+ "grad_norm": 1.75,
665
+ "kd_loss": 0.0531776025891304,
666
+ "learning_rate": 9.834484443271035e-06,
667
+ "loss": 0.0557,
668
+ "step": 330,
669
+ "student_loss": 0.019897613674402237,
670
+ "teacher_loss": 0.004191068932414055
671
+ },
672
+ {
673
+ "epoch": 0.09659746251441753,
674
+ "grad_norm": 1.8125,
675
+ "kd_loss": 0.013695458881556988,
676
+ "learning_rate": 9.827527170898019e-06,
677
+ "loss": 0.0488,
678
+ "step": 335,
679
+ "student_loss": 0.0016651605255901814,
680
+ "teacher_loss": 0.0014800777425989509
681
+ },
682
+ {
683
+ "epoch": 0.09803921568627451,
684
+ "grad_norm": 2.3125,
685
+ "kd_loss": 0.027426283806562424,
686
+ "learning_rate": 9.820429236311158e-06,
687
+ "loss": 0.0567,
688
+ "step": 340,
689
+ "student_loss": 0.010532001033425331,
690
+ "teacher_loss": 0.006886747200042009
691
+ },
692
+ {
693
+ "epoch": 0.09948096885813149,
694
+ "grad_norm": 1.8359375,
695
+ "kd_loss": 0.03429238498210907,
696
+ "learning_rate": 9.81319084632673e-06,
697
+ "loss": 0.0564,
698
+ "step": 345,
699
+ "student_loss": 0.007421151269227266,
700
+ "teacher_loss": 0.0005970117053948343
701
+ },
702
+ {
703
+ "epoch": 0.10092272202998846,
704
+ "grad_norm": 2.546875,
705
+ "kd_loss": 0.058330416679382324,
706
+ "learning_rate": 9.805812211853534e-06,
707
+ "loss": 0.0473,
708
+ "step": 350,
709
+ "student_loss": 0.002645631320774555,
710
+ "teacher_loss": 0.001440234947949648
711
+ },
712
+ {
713
+ "epoch": 0.10236447520184544,
714
+ "grad_norm": 2.046875,
715
+ "kd_loss": 0.016124732792377472,
716
+ "learning_rate": 9.798293547886748e-06,
717
+ "loss": 0.0571,
718
+ "step": 355,
719
+ "student_loss": 0.001094816834665835,
720
+ "teacher_loss": 0.0007638070383109152
721
+ },
722
+ {
723
+ "epoch": 0.10380622837370242,
724
+ "grad_norm": 1.765625,
725
+ "kd_loss": 0.015412062406539917,
726
+ "learning_rate": 9.790635073501662e-06,
727
+ "loss": 0.0541,
728
+ "step": 360,
729
+ "student_loss": 0.0010126332053914666,
730
+ "teacher_loss": 0.0007303679012693465
731
+ },
732
+ {
733
+ "epoch": 0.1052479815455594,
734
+ "grad_norm": 1.328125,
735
+ "kd_loss": 0.0318637378513813,
736
+ "learning_rate": 9.782837011847294e-06,
737
+ "loss": 0.0555,
738
+ "step": 365,
739
+ "student_loss": 0.043804578483104706,
740
+ "teacher_loss": 0.0287578534334898
741
+ },
742
+ {
743
+ "epoch": 0.10668973471741638,
744
+ "grad_norm": 3.734375,
745
+ "kd_loss": 0.054416049271821976,
746
+ "learning_rate": 9.774899590139897e-06,
747
+ "loss": 0.064,
748
+ "step": 370,
749
+ "student_loss": 0.02294199727475643,
750
+ "teacher_loss": 0.0003071785031352192
751
+ },
752
+ {
753
+ "epoch": 0.10813148788927336,
754
+ "grad_norm": 3.09375,
755
+ "kd_loss": 0.03518553078174591,
756
+ "learning_rate": 9.766823039656328e-06,
757
+ "loss": 0.0564,
758
+ "step": 375,
759
+ "student_loss": 0.03153789043426514,
760
+ "teacher_loss": 0.020613783970475197
761
+ },
762
+ {
763
+ "epoch": 0.10957324106113034,
764
+ "grad_norm": 4.40625,
765
+ "kd_loss": 0.021105539053678513,
766
+ "learning_rate": 9.758607595727309e-06,
767
+ "loss": 0.0537,
768
+ "step": 380,
769
+ "student_loss": 0.000557993131224066,
770
+ "teacher_loss": 0.0003853706002701074
771
+ },
772
+ {
773
+ "epoch": 0.11101499423298732,
774
+ "grad_norm": 2.03125,
775
+ "kd_loss": 0.03399880975484848,
776
+ "learning_rate": 9.75025349773058e-06,
777
+ "loss": 0.0504,
778
+ "step": 385,
779
+ "student_loss": 0.0006929709343239665,
780
+ "teacher_loss": 0.0005223862244747579
781
+ },
782
+ {
783
+ "epoch": 0.11245674740484429,
784
+ "grad_norm": 1.2421875,
785
+ "kd_loss": 0.022061804309487343,
786
+ "learning_rate": 9.741760989083914e-06,
787
+ "loss": 0.0473,
788
+ "step": 390,
789
+ "student_loss": 0.0003254064649809152,
790
+ "teacher_loss": 0.0003073395637329668
791
+ },
792
+ {
793
+ "epoch": 0.11389850057670127,
794
+ "grad_norm": 1.9453125,
795
+ "kd_loss": 0.021170753985643387,
796
+ "learning_rate": 9.73313031723803e-06,
797
+ "loss": 0.0519,
798
+ "step": 395,
799
+ "student_loss": 0.0008620450389571488,
800
+ "teacher_loss": 0.0008799576899036765
801
+ },
802
+ {
803
+ "epoch": 0.11534025374855825,
804
+ "grad_norm": 1.6875,
805
+ "kd_loss": 0.02962590754032135,
806
+ "learning_rate": 9.724361733669383e-06,
807
+ "loss": 0.0462,
808
+ "step": 400,
809
+ "student_loss": 0.0009699641377665102,
810
+ "teacher_loss": 0.0002524392039049417
811
+ },
812
+ {
813
+ "epoch": 0.11678200692041522,
814
+ "grad_norm": 1.4140625,
815
+ "kd_loss": 0.02815704233944416,
816
+ "learning_rate": 9.715455493872827e-06,
817
+ "loss": 0.0563,
818
+ "step": 405,
819
+ "student_loss": 0.0135729368776083,
820
+ "teacher_loss": 0.00044868470286019146
821
+ },
822
+ {
823
+ "epoch": 0.1182237600922722,
824
+ "grad_norm": 1.8125,
825
+ "kd_loss": 0.03630191460251808,
826
+ "learning_rate": 9.70641185735419e-06,
827
+ "loss": 0.0593,
828
+ "step": 410,
829
+ "student_loss": 0.014492504298686981,
830
+ "teacher_loss": 0.007929121144115925
831
+ },
832
+ {
833
+ "epoch": 0.11966551326412918,
834
+ "grad_norm": 1.6171875,
835
+ "kd_loss": 0.026567015796899796,
836
+ "learning_rate": 9.697231087622691e-06,
837
+ "loss": 0.0585,
838
+ "step": 415,
839
+ "student_loss": 0.004891898948699236,
840
+ "teacher_loss": 0.003953997977077961
841
+ },
842
+ {
843
+ "epoch": 0.12110726643598616,
844
+ "grad_norm": 2.890625,
845
+ "kd_loss": 0.04854276403784752,
846
+ "learning_rate": 9.687913452183277e-06,
847
+ "loss": 0.0584,
848
+ "step": 420,
849
+ "student_loss": 0.04526115953922272,
850
+ "teacher_loss": 0.023706616833806038
851
+ },
852
+ {
853
+ "epoch": 0.12254901960784313,
854
+ "grad_norm": 2.71875,
855
+ "kd_loss": 0.014143871143460274,
856
+ "learning_rate": 9.678459222528825e-06,
857
+ "loss": 0.0539,
858
+ "step": 425,
859
+ "student_loss": 0.0028342436999082565,
860
+ "teacher_loss": 0.003036739770323038
861
+ },
862
+ {
863
+ "epoch": 0.12399077277970011,
864
+ "grad_norm": 2.546875,
865
+ "kd_loss": 0.05789174139499664,
866
+ "learning_rate": 9.668868674132224e-06,
867
+ "loss": 0.0637,
868
+ "step": 430,
869
+ "student_loss": 0.058899883180856705,
870
+ "teacher_loss": 0.064610056579113
871
+ },
872
+ {
873
+ "epoch": 0.1254325259515571,
874
+ "grad_norm": 2.25,
875
+ "kd_loss": 0.021487534046173096,
876
+ "learning_rate": 9.65914208643836e-06,
877
+ "loss": 0.0537,
878
+ "step": 435,
879
+ "student_loss": 0.0009387442842125893,
880
+ "teacher_loss": 0.0007441287161782384
881
+ },
882
+ {
883
+ "epoch": 0.12687427912341406,
884
+ "grad_norm": 6.78125,
885
+ "kd_loss": 0.03997809439897537,
886
+ "learning_rate": 9.649279742855964e-06,
887
+ "loss": 0.0599,
888
+ "step": 440,
889
+ "student_loss": 0.012818059884011745,
890
+ "teacher_loss": 0.007239806931465864
891
+ },
892
+ {
893
+ "epoch": 0.12831603229527105,
894
+ "grad_norm": 2.375,
895
+ "kd_loss": 0.037731364369392395,
896
+ "learning_rate": 9.639281930749363e-06,
897
+ "loss": 0.0586,
898
+ "step": 445,
899
+ "student_loss": 0.03573564440011978,
900
+ "teacher_loss": 0.01907653547823429
901
+ },
902
+ {
903
+ "epoch": 0.12975778546712802,
904
+ "grad_norm": 2.546875,
905
+ "kd_loss": 0.05183478444814682,
906
+ "learning_rate": 9.629148941430096e-06,
907
+ "loss": 0.0617,
908
+ "step": 450,
909
+ "student_loss": 0.002695092000067234,
910
+ "teacher_loss": 0.0027623483911156654
911
+ },
912
+ {
913
+ "epoch": 0.131199538638985,
914
+ "grad_norm": 1.6015625,
915
+ "kd_loss": 0.029434863477945328,
916
+ "learning_rate": 9.618881070148436e-06,
917
+ "loss": 0.0615,
918
+ "step": 455,
919
+ "student_loss": 0.00108905672095716,
920
+ "teacher_loss": 0.0005926509038545191
921
+ },
922
+ {
923
+ "epoch": 0.13264129181084197,
924
+ "grad_norm": 1.5703125,
925
+ "kd_loss": 0.05266507714986801,
926
+ "learning_rate": 9.608478616084784e-06,
927
+ "loss": 0.0548,
928
+ "step": 460,
929
+ "student_loss": 0.016866859048604965,
930
+ "teacher_loss": 0.0005994263920001686
931
+ },
932
+ {
933
+ "epoch": 0.13408304498269896,
934
+ "grad_norm": 4.90625,
935
+ "kd_loss": 0.04152460768818855,
936
+ "learning_rate": 9.597941882340947e-06,
937
+ "loss": 0.0648,
938
+ "step": 465,
939
+ "student_loss": 0.0035157979000359774,
940
+ "teacher_loss": 0.0008752731955610216
941
+ },
942
+ {
943
+ "epoch": 0.13552479815455595,
944
+ "grad_norm": 1.5546875,
945
+ "kd_loss": 0.03477834537625313,
946
+ "learning_rate": 9.587271175931313e-06,
947
+ "loss": 0.0637,
948
+ "step": 470,
949
+ "student_loss": 0.0005260187899693847,
950
+ "teacher_loss": 0.00038832571590319276
951
+ },
952
+ {
953
+ "epoch": 0.13696655132641292,
954
+ "grad_norm": 2.90625,
955
+ "kd_loss": 0.040360964834690094,
956
+ "learning_rate": 9.5764668077739e-06,
957
+ "loss": 0.0567,
958
+ "step": 475,
959
+ "student_loss": 0.0007983366376720369,
960
+ "teacher_loss": 0.0005927964230068028
961
+ },
962
+ {
963
+ "epoch": 0.1384083044982699,
964
+ "grad_norm": 3.3125,
965
+ "kd_loss": 0.04144829884171486,
966
+ "learning_rate": 9.5655290926813e-06,
967
+ "loss": 0.0762,
968
+ "step": 480,
969
+ "student_loss": 0.03820875287055969,
970
+ "teacher_loss": 0.02910608798265457
971
+ },
972
+ {
973
+ "epoch": 0.13985005767012687,
974
+ "grad_norm": 3.25,
975
+ "kd_loss": 0.0166642889380455,
976
+ "learning_rate": 9.554458349351507e-06,
977
+ "loss": 0.0627,
978
+ "step": 485,
979
+ "student_loss": 0.001466175657697022,
980
+ "teacher_loss": 0.0013084234669804573
981
+ },
982
+ {
983
+ "epoch": 0.14129181084198386,
984
+ "grad_norm": 3.5,
985
+ "kd_loss": 0.025684630498290062,
986
+ "learning_rate": 9.54325490035863e-06,
987
+ "loss": 0.0737,
988
+ "step": 490,
989
+ "student_loss": 0.0014865982811897993,
990
+ "teacher_loss": 0.0012143527856096625
991
+ },
992
+ {
993
+ "epoch": 0.14273356401384082,
994
+ "grad_norm": 2.359375,
995
+ "kd_loss": 0.0510389544069767,
996
+ "learning_rate": 9.531919072143488e-06,
997
+ "loss": 0.0776,
998
+ "step": 495,
999
+ "student_loss": 0.004116421099752188,
1000
+ "teacher_loss": 0.002551821991801262
1001
+ },
1002
+ {
1003
+ "epoch": 0.14417531718569782,
1004
+ "grad_norm": 2.109375,
1005
+ "kd_loss": 0.03247147053480148,
1006
+ "learning_rate": 9.520451195004107e-06,
1007
+ "loss": 0.0637,
1008
+ "step": 500,
1009
+ "student_loss": 0.0007873536669649184,
1010
+ "teacher_loss": 0.0004125205450691283
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/codi3b_a1.0_b1.0_g1.0_ls1/checkpoint-500/vocab.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/wandb/debug-internal.log ADDED
@@ -0,0 +1,11 @@
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"time":"2026-06-19T21:53:03.291282981+02:00","level":"INFO","msg":"wandb-core"}
2
+ {"time":"2026-06-19T21:53:03.291300227+02:00","level":"INFO","msg":"stream: starting","core version":"0.27.2"}
3
+ {"time":"2026-06-19T21:53:03.416887341+02:00","level":"WARN","msg":"featurechecker: GraphQL client is nil, skipping feature loading"}
4
+ {"time":"2026-06-19T21:53:03.416909079+02:00","level":"WARN","msg":"featurechecker: GraphQL client is nil, skipping feature loading"}
5
+ {"time":"2026-06-19T21:53:03.416925111+02:00","level":"INFO","msg":"stream: created new stream","id":"txanrr5l"}
6
+ {"time":"2026-06-19T21:53:03.417036212+02:00","level":"INFO","msg":"handler: started"}
7
+ {"time":"2026-06-19T21:53:03.418690716+02:00","level":"INFO","msg":"stream: started"}
8
+ {"time":"2026-06-19T21:53:03.418814769+02:00","level":"INFO","msg":"writer: started","stream_id":"txanrr5l"}
9
+ {"time":"2026-06-19T21:53:03.418845474+02:00","level":"INFO","msg":"sender: started"}
10
+ {"time":"2026-06-19T21:53:03.439760344+02:00","level":"WARN","msg":"featurechecker: GraphQL client is nil, skipping feature loading"}
11
+ {"time":"2026-06-19T21:53:03.439773104+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/codi3b_a1.0_b1.0_g1.0_ls1/wandb/debug.log ADDED
@@ -0,0 +1,20 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ 2026-06-19 21:53:02,842 INFO MainThread:2910938 [wandb_setup.py:_flush():81] Current SDK version is 0.27.2
2
+ 2026-06-19 21:53:02,842 INFO MainThread:2910938 [wandb_setup.py:_flush():81] Configure stats pid to 2910938
3
+ 2026-06-19 21:53:02,842 INFO MainThread:2910938 [wandb_setup.py:_flush():81] Loading settings from environment variables
4
+ 2026-06-19 21:53:02,842 INFO MainThread:2910938 [wandb_init.py:setup_run_log_directory():723] Logging user logs to model_weights/codi-multi-3b/wandb/offline-run-20260619_215302-txanrr5l/logs/debug.log
5
+ 2026-06-19 21:53:02,843 INFO MainThread:2910938 [wandb_init.py:setup_run_log_directory():724] Logging internal logs to model_weights/codi-multi-3b/wandb/offline-run-20260619_215302-txanrr5l/logs/debug-internal.log
6
+ 2026-06-19 21:53:02,843 INFO MainThread:2910938 [wandb_init.py:init():766] calling init triggers
7
+ 2026-06-19 21:53:02,843 INFO MainThread:2910938 [wandb_init.py:init():771] wandb.init called with sweep_config: {}
8
+ config: {'model': 'model_weights/sft-coder-3b/checkpoint-3000', 'output_dir': 'model_weights/codi-multi-3b', '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': 1.0, 'beta': 1.0, 'gamma': 1.0, 'kd_layers': None, 'single_anchor': False, '_wandb': {}}
9
+ 2026-06-19 21:53:02,843 INFO MainThread:2910938 [wandb_init.py:init():814] starting backend
10
+ 2026-06-19 21:53:03,277 INFO MainThread:2910938 [wandb_init.py:init():829] sending inform_init request
11
+ 2026-06-19 21:53:03,423 INFO MainThread:2910938 [wandb_init.py:init():834] backend started and connected
12
+ 2026-06-19 21:53:03,425 INFO MainThread:2910938 [wandb_init.py:init():904] updated telemetry
13
+ 2026-06-19 21:53:03,438 INFO MainThread:2910938 [wandb_init.py:init():927] communicating run to backend with 90.0 second timeout
14
+ 2026-06-19 21:53:03,440 INFO MainThread:2910938 [wandb_init.py:init():972] starting run threads in backend
15
+ 2026-06-19 21:53:03,864 INFO MainThread:2910938 [wandb_run.py:_console_start():2523] atexit reg
16
+ 2026-06-19 21:53:03,864 INFO MainThread:2910938 [wandb_run.py:_redirect():2373] redirect: wrap_raw
17
+ 2026-06-19 21:53:03,864 INFO MainThread:2910938 [wandb_run.py:_redirect():2442] Wrapping output streams.
18
+ 2026-06-19 21:53:03,864 INFO MainThread:2910938 [wandb_run.py:_redirect():2465] Redirects installed.
19
+ 2026-06-19 21:53:03,871 INFO MainThread:2910938 [wandb_init.py:init():1010] run started, returning control to user process
20
+ 2026-06-19 21:53:09,123 INFO MainThread:2910938 [wandb_run.py:_config_callback():1346] config_cb None None {'output_dir': 'model_weights/codi-multi-3b', '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-multi-3b/runs/Jun19_21-53-03_alvis4-40', '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/codi3b_a1.0_b1.0_g1.0_ls1/wandb/offline-run-20260619_215302-txanrr5l/files/requirements.txt ADDED
@@ -0,0 +1,222 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ wheel==0.47.0
2
+ triton==3.3.1
3
+ ptyprocess==0.7.0
4
+ pyzmq==27.1.0
5
+ nvidia-cublas-cu12==12.6.4.1
6
+ python-dateutil==2.9.0.post0
7
+ rfc3339-validator==0.1.4
8
+ pluggy==1.6.0
9
+ grpclib==0.4.9
10
+ attrs==26.1.0
11
+ urllib3==2.7.0
12
+ cffi==2.0.0
13
+ openai==1.86.0
14
+ async-lru==2.3.0
15
+ cfgv==3.5.0
16
+ httpcore==1.0.9
17
+ bitsandbytes==0.49.2
18
+ iniconfig==2.3.0
19
+ nvidia-cuda-runtime-cu12==12.6.77
20
+ jupyterlab_server==2.28.0
21
+ latex2sympy2_extended==1.11.0
22
+ jupyter-events==0.12.1
23
+ docker==7.1.0
24
+ ipywidgets==8.1.8
25
+ blinker==1.9.0
26
+ cbor2==6.1.2
27
+ nvidia-cufft-cu12==11.3.0.4
28
+ nvidia-cusolver-cu12==11.7.1.2
29
+ multiprocess==0.70.16
30
+ anyio==4.14.0
31
+ itsdangerous==2.2.0
32
+ widgetsnbextension==4.0.15
33
+ watchfiles==1.2.0
34
+ swebench==4.1.0
35
+ blobfile==3.2.0
36
+ jupyter==1.1.1
37
+ certifi==2026.6.17
38
+ argon2-cffi==25.1.0
39
+ argon2-cffi-bindings==25.1.0
40
+ overrides==7.7.0
41
+ propcache==0.5.2
42
+ ipython==8.33.0
43
+ nbconvert==7.17.1
44
+ rich==15.0.0
45
+ asttokens==3.0.1
46
+ tokenizers==0.22.2
47
+ synchronicity==0.12.4
48
+ tqdm==4.68.3
49
+ nvidia-nvjitlink-cu12==12.6.85
50
+ nvidia-nvtx-cu12==12.6.77
51
+ jsonschema-specifications==2025.9.1
52
+ jupyterlab==4.5.8
53
+ pure_eval==0.2.3
54
+ uri-template==1.3.0
55
+ sentry-sdk==2.63.0
56
+ pycryptodomex==3.23.0
57
+ yarl==1.24.2
58
+ h2==4.3.0
59
+ h11==0.16.0
60
+ requests==2.34.2
61
+ Jinja2==3.1.6
62
+ xformers==0.0.31.post1
63
+ rfc3987-syntax==1.1.0
64
+ antlr4-python3-runtime==4.11.0
65
+ nvidia-cusparselt-cu12==0.6.3
66
+ safetensors==0.8.0
67
+ notebook_shim==0.2.4
68
+ Pygments==2.20.0
69
+ python-json-logger==4.1.0
70
+ isoduration==20.11.0
71
+ smmap==5.0.3
72
+ click==8.3.3
73
+ transformers==4.57.6
74
+ aiohappyeyeballs==2.6.2
75
+ regex==2026.5.9
76
+ jsonpointer==3.1.1
77
+ fqdn==1.5.1
78
+ jupyter-lsp==2.3.1
79
+ rfc3986-validator==0.1.1
80
+ prometheus_client==0.25.0
81
+ mistune==3.2.1
82
+ sentencepiece==0.2.1
83
+ types-toml==0.10.8.20260518
84
+ wcwidth==0.8.1
85
+ identify==2.6.19
86
+ protobuf==6.33.6
87
+ datasets==3.6.0
88
+ accelerate==1.14.0
89
+ networkx==3.6.1
90
+ pydantic==2.13.4
91
+ multidict==6.7.1
92
+ chardet==7.4.3
93
+ pytest==9.1.0
94
+ babel==2.18.0
95
+ pip==26.1.2
96
+ dill==0.3.8
97
+ referencing==0.37.0
98
+ pre_commit==4.6.0
99
+ numpy==2.4.6
100
+ math-verify==0.9.0
101
+ traitlets==5.15.1
102
+ nvidia-curand-cu12==10.3.7.77
103
+ websocket-client==1.9.0
104
+ hf-xet==1.5.1
105
+ unidiff==0.7.5
106
+ nest-asyncio2==1.7.2
107
+ typing_extensions==4.15.0
108
+ platformdirs==4.10.0
109
+ torch==2.7.1
110
+ tblib==3.2.2
111
+ sqlfluff==4.2.2
112
+ modal==1.5.0
113
+ ipykernel==7.3.0
114
+ msgspec==0.21.1
115
+ virtualenv==21.5.1
116
+ json5==0.14.0
117
+ pathlib_abc==0.5.2
118
+ aiosignal==1.4.0
119
+ moodist==0.5.1
120
+ rpds-py==2026.5.1
121
+ webencodings==0.5.1
122
+ sniffio==1.3.1
123
+ annotated-types==0.7.0
124
+ python-discovery==1.4.2
125
+ scipy==1.17.1
126
+ GitPython==3.1.50
127
+ defusedxml==0.7.1
128
+ nvidia-cufile-cu12==1.11.1.6
129
+ nvidia-nccl-cu12==2.26.2
130
+ sympy==1.14.0
131
+ tinycss2==1.5.1
132
+ soupsieve==2.8.4
133
+ ninja==1.13.0
134
+ python-dotenv==1.2.2
135
+ ghapi==1.0.13
136
+ jupyter_server_terminals==0.5.4
137
+ markdown-it-py==4.2.0
138
+ nbclient==0.11.0
139
+ mpmath==1.3.0
140
+ frozenlist==1.8.0
141
+ Werkzeug==3.1.8
142
+ executing==2.2.1
143
+ gitdb==4.0.12
144
+ setuptools==82.0.1
145
+ Flask==3.1.1
146
+ MarkupSafe==3.0.3
147
+ notebook==7.5.7
148
+ jupyter_core==5.9.1
149
+ jedi==0.20.0
150
+ mdurl==0.1.2
151
+ lark==1.3.1
152
+ bleach==6.4.0
153
+ hpack==4.1.0
154
+ google-re2==1.1.20240702
155
+ beautifulsoup4==4.15.0
156
+ jsonschema==4.26.0
157
+ psutil==7.2.2
158
+ nvidia-cuda-nvrtc-cu12==12.6.77
159
+ narwhals==2.22.1
160
+ lxml==6.1.1
161
+ fastjsonschema==2.21.2
162
+ tornado==6.5.7
163
+ nodeenv==1.10.0
164
+ typing-inspection==0.4.2
165
+ httpx==0.28.1
166
+ distro==1.9.0
167
+ colorama==0.4.6
168
+ jupyter-console==6.6.3
169
+ jupyterlab_widgets==3.0.16
170
+ PyYAML==6.0.3
171
+ fsspec==2025.3.0
172
+ hyperframe==6.1.0
173
+ nvidia-cuda-cupti-cu12==12.6.80
174
+ charset-normalizer==3.4.7
175
+ wandb==0.27.2
176
+ nbformat==5.10.4
177
+ nvidia-cudnn-cu12==9.5.1.17
178
+ peft==0.19.1
179
+ pathspec==1.1.1
180
+ matplotlib-inline==0.2.2
181
+ huggingface_hub==0.36.2
182
+ pyarrow==24.0.0
183
+ nvidia-cusparse-cu12==12.5.4.2
184
+ cwm==0.0.1
185
+ prompt_toolkit==3.0.52
186
+ omegaconf==2.4.0.dev4
187
+ tiktoken==0.13.0
188
+ comm==0.2.3
189
+ threadpoolctl==3.6.0
190
+ pydantic_core==2.46.4
191
+ tenacity==9.1.4
192
+ parso==0.8.7
193
+ pandas==3.0.3
194
+ webcolors==25.10.0
195
+ idna==3.18
196
+ terminado==0.18.1
197
+ pandocfilters==1.5.1
198
+ aiohttp==3.14.1
199
+ arrow==1.4.0
200
+ packaging==26.2
201
+ tzdata==2026.2
202
+ distlib==0.4.3
203
+ Send2Trash==2.1.0
204
+ jupyterlab_pygments==0.3.0
205
+ decorator==5.3.1
206
+ jupyter_server==2.20.0
207
+ scikit-learn==1.9.0
208
+ six==1.17.0
209
+ diff_cover==10.3.0
210
+ pycparser==3.0
211
+ types-certifi==2021.10.8.3
212
+ toml==0.10.2
213
+ stack-data==0.6.3
214
+ fastcore==1.13.5
215
+ filelock==3.29.4
216
+ xxhash==3.7.0
217
+ pexpect==4.9.0
218
+ jupyter_client==8.9.1
219
+ jiter==0.15.0
220
+ debugpy==1.8.21
221
+ universal_pathlib==0.3.10
222
+ joblib==1.5.3
checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/wandb/offline-run-20260619_215302-txanrr5l/logs/debug-core.log ADDED
@@ -0,0 +1,12 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"time":"2026-06-19T21:53:03.094929463+02:00","level":"INFO","msg":"main: starting server","port-filename":"/local/tmp.6767841/tmptc5jh63i/port-2910938.txt","pid":2910938,"detached":false,"idle-timeout":600000000000,"log-level":0,"disable-analytics":false,"shutdown-on-parent-exit":false,"enable-dcgm-profiling":false}
2
+ {"time":"2026-06-19T21:53:03.095783268+02:00","level":"INFO","msg":"server: will exit if parent process dies","ppid":2910938}
3
+ {"time":"2026-06-19T21:53:03.095779775+02:00","level":"INFO","msg":"server: accepting connections","addr":{"Name":"/local/tmp.6767841/wandb-2910938-2911302-2012556736/socket","Net":"unix"}}
4
+ {"time":"2026-06-19T21:53:03.10475933+02:00","level":"INFO","msg":"main: starting server","port-filename":"/local/tmp.6767840/tmpzkap0g5c/port-906312.txt","pid":906312,"detached":false,"idle-timeout":600000000000,"log-level":0,"disable-analytics":false,"shutdown-on-parent-exit":false,"enable-dcgm-profiling":false}
5
+ {"time":"2026-06-19T21:53:03.108209915+02:00","level":"INFO","msg":"server: accepting connections","addr":{"Name":"/local/tmp.6767840/wandb-906312-906672-88537129/socket","Net":"unix"}}
6
+ {"time":"2026-06-19T21:53:03.108404089+02:00","level":"INFO","msg":"server: will exit if parent process dies","ppid":906312}
7
+ {"time":"2026-06-19T21:53:03.276797796+02:00","level":"INFO","msg":"connection: ManageConnectionData: new connection created","id":"1(@)"}
8
+ {"time":"2026-06-19T21:53:03.2833645+02:00","level":"INFO","msg":"connection: ManageConnectionData: new connection created","id":"1(@)"}
9
+ {"time":"2026-06-19T21:53:03.287130346+02:00","level":"INFO","msg":"handleInformInit: received","streamId":"txanrr5l","id":"1(@)"}
10
+ {"time":"2026-06-19T21:53:03.293822712+02:00","level":"INFO","msg":"handleInformInit: received","streamId":"q1wgok6t","id":"1(@)"}
11
+ {"time":"2026-06-19T21:53:03.418697302+02:00","level":"INFO","msg":"handleInformInit: stream started","streamId":"txanrr5l","id":"1(@)"}
12
+ {"time":"2026-06-19T21:53:03.421786591+02:00","level":"INFO","msg":"handleInformInit: stream started","streamId":"q1wgok6t","id":"1(@)"}
checkpoints/codi3b_a1.0_b1.0_g1.0_ls1/wandb/offline-run-20260619_215302-txanrr5l/logs/debug-internal.log ADDED
@@ -0,0 +1,11 @@
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"time":"2026-06-19T21:53:03.291282981+02:00","level":"INFO","msg":"wandb-core"}
2
+ {"time":"2026-06-19T21:53:03.291300227+02:00","level":"INFO","msg":"stream: starting","core version":"0.27.2"}
3
+ {"time":"2026-06-19T21:53:03.416887341+02:00","level":"WARN","msg":"featurechecker: GraphQL client is nil, skipping feature loading"}
4
+ {"time":"2026-06-19T21:53:03.416909079+02:00","level":"WARN","msg":"featurechecker: GraphQL client is nil, skipping feature loading"}
5
+ {"time":"2026-06-19T21:53:03.416925111+02:00","level":"INFO","msg":"stream: created new stream","id":"txanrr5l"}
6
+ {"time":"2026-06-19T21:53:03.417036212+02:00","level":"INFO","msg":"handler: started"}
7
+ {"time":"2026-06-19T21:53:03.418690716+02:00","level":"INFO","msg":"stream: started"}
8
+ {"time":"2026-06-19T21:53:03.418814769+02:00","level":"INFO","msg":"writer: started","stream_id":"txanrr5l"}
9
+ {"time":"2026-06-19T21:53:03.418845474+02:00","level":"INFO","msg":"sender: started"}
10
+ {"time":"2026-06-19T21:53:03.439760344+02:00","level":"WARN","msg":"featurechecker: GraphQL client is nil, skipping feature loading"}
11
+ {"time":"2026-06-19T21:53:03.439773104+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/codi3b_a1.0_b1.0_g1.0_ls1/wandb/offline-run-20260619_215302-txanrr5l/logs/debug.log ADDED
@@ -0,0 +1,20 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ 2026-06-19 21:53:02,842 INFO MainThread:2910938 [wandb_setup.py:_flush():81] Current SDK version is 0.27.2
2
+ 2026-06-19 21:53:02,842 INFO MainThread:2910938 [wandb_setup.py:_flush():81] Configure stats pid to 2910938
3
+ 2026-06-19 21:53:02,842 INFO MainThread:2910938 [wandb_setup.py:_flush():81] Loading settings from environment variables
4
+ 2026-06-19 21:53:02,842 INFO MainThread:2910938 [wandb_init.py:setup_run_log_directory():723] Logging user logs to model_weights/codi-multi-3b/wandb/offline-run-20260619_215302-txanrr5l/logs/debug.log
5
+ 2026-06-19 21:53:02,843 INFO MainThread:2910938 [wandb_init.py:setup_run_log_directory():724] Logging internal logs to model_weights/codi-multi-3b/wandb/offline-run-20260619_215302-txanrr5l/logs/debug-internal.log
6
+ 2026-06-19 21:53:02,843 INFO MainThread:2910938 [wandb_init.py:init():766] calling init triggers
7
+ 2026-06-19 21:53:02,843 INFO MainThread:2910938 [wandb_init.py:init():771] wandb.init called with sweep_config: {}
8
+ config: {'model': 'model_weights/sft-coder-3b/checkpoint-3000', 'output_dir': 'model_weights/codi-multi-3b', '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': 1.0, 'beta': 1.0, 'gamma': 1.0, 'kd_layers': None, 'single_anchor': False, '_wandb': {}}
9
+ 2026-06-19 21:53:02,843 INFO MainThread:2910938 [wandb_init.py:init():814] starting backend
10
+ 2026-06-19 21:53:03,277 INFO MainThread:2910938 [wandb_init.py:init():829] sending inform_init request
11
+ 2026-06-19 21:53:03,423 INFO MainThread:2910938 [wandb_init.py:init():834] backend started and connected
12
+ 2026-06-19 21:53:03,425 INFO MainThread:2910938 [wandb_init.py:init():904] updated telemetry
13
+ 2026-06-19 21:53:03,438 INFO MainThread:2910938 [wandb_init.py:init():927] communicating run to backend with 90.0 second timeout
14
+ 2026-06-19 21:53:03,440 INFO MainThread:2910938 [wandb_init.py:init():972] starting run threads in backend
15
+ 2026-06-19 21:53:03,864 INFO MainThread:2910938 [wandb_run.py:_console_start():2523] atexit reg
16
+ 2026-06-19 21:53:03,864 INFO MainThread:2910938 [wandb_run.py:_redirect():2373] redirect: wrap_raw
17
+ 2026-06-19 21:53:03,864 INFO MainThread:2910938 [wandb_run.py:_redirect():2442] Wrapping output streams.
18
+ 2026-06-19 21:53:03,864 INFO MainThread:2910938 [wandb_run.py:_redirect():2465] Redirects installed.
19
+ 2026-06-19 21:53:03,871 INFO MainThread:2910938 [wandb_init.py:init():1010] run started, returning control to user process
20
+ 2026-06-19 21:53:09,123 INFO MainThread:2910938 [wandb_run.py:_config_callback():1346] config_cb None None {'output_dir': 'model_weights/codi-multi-3b', '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-multi-3b/runs/Jun19_21-53-03_alvis4-40', '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/codi3b_a1.0_b1.0_g1.0_ls1/wandb/offline-run-20260619_215302-txanrr5l/run-txanrr5l.wandb ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2d292398dd22e5b3734caacc80e0c7d9c1ea302c4262948b4f63543bc9653b60
3
+ size 7897088