Hugging Face's logo

ASSERT-KTH
/
codi-trace

Safetensors
Model card Files
xet
 Community
sirui6011 commited on
Commit
3b13b3f
·
verified ·
1 Parent(s): 9887261

add checkpoints/codi3b_a1.0_b1.0_g1.0_ls2

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. .gitattributes +6 -0
  2. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-1000/added_tokens.json +35 -0
  3. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-1000/chat_template.jinja +54 -0
  4. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-1000/config.json +66 -0
  5. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-1000/merges.txt +0 -0
  6. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-1000/pytorch_model.bin +3 -0
  7. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-1000/special_tokens_map.json +31 -0
  8. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-1000/thought_projector.pt +3 -0
  9. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-1000/tokenizer.json +3 -0
  10. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-1000/tokenizer_config.json +295 -0
  11. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-1000/trainer_state.json +2046 -0
  12. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-1000/vocab.json +0 -0
  13. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-200/added_tokens.json +35 -0
  14. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-200/chat_template.jinja +54 -0
  15. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-200/config.json +66 -0
  16. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-200/merges.txt +0 -0
  17. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-200/pytorch_model.bin +3 -0
  18. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-200/special_tokens_map.json +31 -0
  19. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-200/thought_projector.pt +3 -0
  20. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-200/tokenizer.json +3 -0
  21. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-200/tokenizer_config.json +295 -0
  22. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-200/trainer_state.json +434 -0
  23. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-200/vocab.json +0 -0
  24. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-400/added_tokens.json +35 -0
  25. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-400/chat_template.jinja +54 -0
  26. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-400/config.json +66 -0
  27. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-400/merges.txt +0 -0
  28. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-400/pytorch_model.bin +3 -0
  29. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-400/special_tokens_map.json +31 -0
  30. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-400/thought_projector.pt +3 -0
  31. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-400/tokenizer.json +3 -0
  32. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-400/tokenizer_config.json +295 -0
  33. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-400/trainer_state.json +834 -0
  34. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-400/vocab.json +0 -0
  35. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-600/added_tokens.json +35 -0
  36. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-600/chat_template.jinja +54 -0
  37. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-600/config.json +66 -0
  38. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-600/merges.txt +0 -0
  39. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-600/pytorch_model.bin +3 -0
  40. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-600/special_tokens_map.json +31 -0
  41. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-600/thought_projector.pt +3 -0
  42. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-600/tokenizer.json +3 -0
  43. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-600/tokenizer_config.json +295 -0
  44. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-600/trainer_state.json +1234 -0
  45. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-600/vocab.json +0 -0
  46. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-800/added_tokens.json +35 -0
  47. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-800/chat_template.jinja +54 -0
  48. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-800/config.json +66 -0
  49. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-800/merges.txt +0 -0
  50. checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-800/pytorch_model.bin +3 -0
.gitattributes CHANGED
@@ -84,3 +84,9 @@ checkpoints/codi3b_a1.0_b1.0_g1.0_ls1_ss0.50/checkpoint-2000/tokenizer.json filt
84
  checkpoints/codi3b_a1.0_b1.0_g1.0_ls1_ss0.50/checkpoint-2500/tokenizer.json filter=lfs diff=lfs merge=lfs -text
85
  checkpoints/codi3b_a1.0_b1.0_g1.0_ls1_ss0.50/checkpoint-500/tokenizer.json filter=lfs diff=lfs merge=lfs -text
86
  checkpoints/codi3b_a1.0_b1.0_g1.0_ls1_ss0.50/wandb/offline-run-20260622_173630-kv8jgb2s/run-kv8jgb2s.wandb filter=lfs diff=lfs merge=lfs -text
 
 
 
 
 
 
 
84
  checkpoints/codi3b_a1.0_b1.0_g1.0_ls1_ss0.50/checkpoint-2500/tokenizer.json filter=lfs diff=lfs merge=lfs -text
85
  checkpoints/codi3b_a1.0_b1.0_g1.0_ls1_ss0.50/checkpoint-500/tokenizer.json filter=lfs diff=lfs merge=lfs -text
86
  checkpoints/codi3b_a1.0_b1.0_g1.0_ls1_ss0.50/wandb/offline-run-20260622_173630-kv8jgb2s/run-kv8jgb2s.wandb filter=lfs diff=lfs merge=lfs -text
87
+ checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-1000/tokenizer.json filter=lfs diff=lfs merge=lfs -text
88
+ checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-200/tokenizer.json filter=lfs diff=lfs merge=lfs -text
89
+ checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-400/tokenizer.json filter=lfs diff=lfs merge=lfs -text
90
+ checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-600/tokenizer.json filter=lfs diff=lfs merge=lfs -text
91
+ checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-800/tokenizer.json filter=lfs diff=lfs merge=lfs -text
92
+ checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/wandb/offline-run-20260620_202751-dlf846se/run-dlf846se.wandb filter=lfs diff=lfs merge=lfs -text
checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/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_ls2/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_ls2/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_ls2/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_ls2/checkpoint-1000/pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:50fdb3d76be7c23bc973170e13a63f0398cae743bc0cd725362d34fc03a548c7
3
+ size 6187858991
checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/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_ls2/checkpoint-1000/thought_projector.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0ba9cb76f4b5b3c245d1e63b529d455d53eb6953fbff842155d57b749a664a78
3
+ size 16788033
checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/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_ls2/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_ls2/checkpoint-1000/trainer_state.json ADDED
@@ -0,0 +1,2046 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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": 77.0,
15
+ "kd_loss": 0.2900531589984894,
16
+ "learning_rate": 1.3333333333333334e-06,
17
+ "loss": 0.5715,
18
+ "step": 5,
19
+ "student_loss": 0.11784028261899948,
20
+ "teacher_loss": 0.001998942345380783
21
+ },
22
+ {
23
+ "epoch": 0.0028835063437139563,
24
+ "grad_norm": 26.875,
25
+ "kd_loss": 0.277434378862381,
26
+ "learning_rate": 3e-06,
27
+ "loss": 0.4692,
28
+ "step": 10,
29
+ "student_loss": 0.11281580477952957,
30
+ "teacher_loss": 0.005315451417118311
31
+ },
32
+ {
33
+ "epoch": 0.004325259515570935,
34
+ "grad_norm": 18.5,
35
+ "kd_loss": 0.21174456179141998,
36
+ "learning_rate": 4.666666666666667e-06,
37
+ "loss": 0.3373,
38
+ "step": 15,
39
+ "student_loss": 0.007377153728157282,
40
+ "teacher_loss": 0.0011863313848152757
41
+ },
42
+ {
43
+ "epoch": 0.0057670126874279125,
44
+ "grad_norm": 8.0625,
45
+ "kd_loss": 0.13275651633739471,
46
+ "learning_rate": 6.333333333333333e-06,
47
+ "loss": 0.2555,
48
+ "step": 20,
49
+ "student_loss": 0.04130968824028969,
50
+ "teacher_loss": 0.0005545365856960416
51
+ },
52
+ {
53
+ "epoch": 0.00720876585928489,
54
+ "grad_norm": 4.6875,
55
+ "kd_loss": 0.10908422619104385,
56
+ "learning_rate": 8.000000000000001e-06,
57
+ "loss": 0.2065,
58
+ "step": 25,
59
+ "student_loss": 0.0035145599395036697,
60
+ "teacher_loss": 0.0009682206436991692
61
+ },
62
+ {
63
+ "epoch": 0.00865051903114187,
64
+ "grad_norm": 2.328125,
65
+ "kd_loss": 0.06466298550367355,
66
+ "learning_rate": 9.666666666666667e-06,
67
+ "loss": 0.1499,
68
+ "step": 30,
69
+ "student_loss": 0.0010085003450512886,
70
+ "teacher_loss": 0.0005369208520278335
71
+ },
72
+ {
73
+ "epoch": 0.010092272202998846,
74
+ "grad_norm": 2.609375,
75
+ "kd_loss": 0.05857780575752258,
76
+ "learning_rate": 9.99958042442916e-06,
77
+ "loss": 0.1223,
78
+ "step": 35,
79
+ "student_loss": 0.0010954203316941857,
80
+ "teacher_loss": 0.0006188539555296302
81
+ },
82
+ {
83
+ "epoch": 0.011534025374855825,
84
+ "grad_norm": 1.8359375,
85
+ "kd_loss": 0.057602375745773315,
86
+ "learning_rate": 9.997876019358083e-06,
87
+ "loss": 0.1064,
88
+ "step": 40,
89
+ "student_loss": 0.032847270369529724,
90
+ "teacher_loss": 0.028142843395471573
91
+ },
92
+ {
93
+ "epoch": 0.012975778546712802,
94
+ "grad_norm": 1.578125,
95
+ "kd_loss": 0.08787942677736282,
96
+ "learning_rate": 9.99486100792044e-06,
97
+ "loss": 0.0811,
98
+ "step": 45,
99
+ "student_loss": 0.07371404767036438,
100
+ "teacher_loss": 0.002992093563079834
101
+ },
102
+ {
103
+ "epoch": 0.01441753171856978,
104
+ "grad_norm": 2.078125,
105
+ "kd_loss": 0.037376269698143005,
106
+ "learning_rate": 9.990536180750724e-06,
107
+ "loss": 0.0866,
108
+ "step": 50,
109
+ "student_loss": 0.027633950114250183,
110
+ "teacher_loss": 0.009523550048470497
111
+ },
112
+ {
113
+ "epoch": 0.015859284890426758,
114
+ "grad_norm": 1.8203125,
115
+ "kd_loss": 0.029546089470386505,
116
+ "learning_rate": 9.984902671959911e-06,
117
+ "loss": 0.0757,
118
+ "step": 55,
119
+ "student_loss": 0.0010951841250061989,
120
+ "teacher_loss": 0.0006982200429774821
121
+ },
122
+ {
123
+ "epoch": 0.01730103806228374,
124
+ "grad_norm": 2.21875,
125
+ "kd_loss": 0.029508251696825027,
126
+ "learning_rate": 9.97796195883804e-06,
127
+ "loss": 0.0683,
128
+ "step": 60,
129
+ "student_loss": 0.002835189923644066,
130
+ "teacher_loss": 0.001713551115244627
131
+ },
132
+ {
133
+ "epoch": 0.018742791234140715,
134
+ "grad_norm": 2.234375,
135
+ "kd_loss": 0.030333109200000763,
136
+ "learning_rate": 9.969715861466839e-06,
137
+ "loss": 0.0686,
138
+ "step": 65,
139
+ "student_loss": 0.03012149967253208,
140
+ "teacher_loss": 0.03870478644967079
141
+ },
142
+ {
143
+ "epoch": 0.020184544405997693,
144
+ "grad_norm": 1.265625,
145
+ "kd_loss": 0.05904582515358925,
146
+ "learning_rate": 9.96016654224243e-06,
147
+ "loss": 0.0737,
148
+ "step": 70,
149
+ "student_loss": 0.01526976004242897,
150
+ "teacher_loss": 0.0018924300093203783
151
+ },
152
+ {
153
+ "epoch": 0.02162629757785467,
154
+ "grad_norm": 1.7578125,
155
+ "kd_loss": 0.024645471945405006,
156
+ "learning_rate": 9.94931650530827e-06,
157
+ "loss": 0.0659,
158
+ "step": 75,
159
+ "student_loss": 0.002360731828957796,
160
+ "teacher_loss": 0.0017862127861008048
161
+ },
162
+ {
163
+ "epoch": 0.02306805074971165,
164
+ "grad_norm": 0.96484375,
165
+ "kd_loss": 0.02402629889547825,
166
+ "learning_rate": 9.93716859589851e-06,
167
+ "loss": 0.0587,
168
+ "step": 80,
169
+ "student_loss": 0.0011838233331218362,
170
+ "teacher_loss": 0.0007806048961356282
171
+ },
172
+ {
173
+ "epoch": 0.024509803921568627,
174
+ "grad_norm": 1.09375,
175
+ "kd_loss": 0.025694765150547028,
176
+ "learning_rate": 9.923725999591846e-06,
177
+ "loss": 0.061,
178
+ "step": 85,
179
+ "student_loss": 0.00045988388592377305,
180
+ "teacher_loss": 0.0004276020044926554
181
+ },
182
+ {
183
+ "epoch": 0.025951557093425604,
184
+ "grad_norm": 1.2421875,
185
+ "kd_loss": 0.0665941834449768,
186
+ "learning_rate": 9.908992241476189e-06,
187
+ "loss": 0.062,
188
+ "step": 90,
189
+ "student_loss": 0.018955115228891373,
190
+ "teacher_loss": 0.003640417940914631
191
+ },
192
+ {
193
+ "epoch": 0.027393310265282585,
194
+ "grad_norm": 1.625,
195
+ "kd_loss": 0.03899763152003288,
196
+ "learning_rate": 9.892971185224244e-06,
197
+ "loss": 0.0583,
198
+ "step": 95,
199
+ "student_loss": 0.008055699057877064,
200
+ "teacher_loss": 0.005101984832435846
201
+ },
202
+ {
203
+ "epoch": 0.02883506343713956,
204
+ "grad_norm": 1.3828125,
205
+ "kd_loss": 0.022396311163902283,
206
+ "learning_rate": 9.875667032080354e-06,
207
+ "loss": 0.0519,
208
+ "step": 100,
209
+ "student_loss": 0.0006799453985877335,
210
+ "teacher_loss": 0.0005385762779042125
211
+ },
212
+ {
213
+ "epoch": 0.03027681660899654,
214
+ "grad_norm": 1.609375,
215
+ "kd_loss": 0.0338844433426857,
216
+ "learning_rate": 9.857084319758772e-06,
217
+ "loss": 0.055,
218
+ "step": 105,
219
+ "student_loss": 0.004358346574008465,
220
+ "teacher_loss": 0.0003296585055068135
221
+ },
222
+ {
223
+ "epoch": 0.031718569780853516,
224
+ "grad_norm": 1.234375,
225
+ "kd_loss": 0.033958468586206436,
226
+ "learning_rate": 9.837227921253747e-06,
227
+ "loss": 0.0588,
228
+ "step": 110,
229
+ "student_loss": 0.031565263867378235,
230
+ "teacher_loss": 0.01906467415392399
231
+ },
232
+ {
233
+ "epoch": 0.03316032295271049,
234
+ "grad_norm": 1.4765625,
235
+ "kd_loss": 0.036942481994628906,
236
+ "learning_rate": 9.816103043561648e-06,
237
+ "loss": 0.0579,
238
+ "step": 115,
239
+ "student_loss": 0.018612468615174294,
240
+ "teacher_loss": 0.012586154974997044
241
+ },
242
+ {
243
+ "epoch": 0.03460207612456748,
244
+ "grad_norm": 1.359375,
245
+ "kd_loss": 0.02566700056195259,
246
+ "learning_rate": 9.79371522631553e-06,
247
+ "loss": 0.0494,
248
+ "step": 120,
249
+ "student_loss": 0.001015476998873055,
250
+ "teacher_loss": 0.0008012109901756048
251
+ },
252
+ {
253
+ "epoch": 0.036043829296424454,
254
+ "grad_norm": 1.234375,
255
+ "kd_loss": 0.0527762807905674,
256
+ "learning_rate": 9.770070340332457e-06,
257
+ "loss": 0.0544,
258
+ "step": 125,
259
+ "student_loss": 0.03289179503917694,
260
+ "teacher_loss": 0.00024821600527502596
261
+ },
262
+ {
263
+ "epoch": 0.03748558246828143,
264
+ "grad_norm": 1.4375,
265
+ "kd_loss": 0.020570307970046997,
266
+ "learning_rate": 9.745174586073982e-06,
267
+ "loss": 0.0531,
268
+ "step": 130,
269
+ "student_loss": 0.0006593615980818868,
270
+ "teacher_loss": 0.0006894106627441943
271
+ },
272
+ {
273
+ "epoch": 0.03892733564013841,
274
+ "grad_norm": 1.1953125,
275
+ "kd_loss": 0.03941934183239937,
276
+ "learning_rate": 9.719034492020183e-06,
277
+ "loss": 0.0584,
278
+ "step": 135,
279
+ "student_loss": 0.001388320466503501,
280
+ "teacher_loss": 0.00066422886447981
281
+ },
282
+ {
283
+ "epoch": 0.040369088811995385,
284
+ "grad_norm": 1.125,
285
+ "kd_loss": 0.023267682641744614,
286
+ "learning_rate": 9.691656912957686e-06,
287
+ "loss": 0.0554,
288
+ "step": 140,
289
+ "student_loss": 0.0006728945882059634,
290
+ "teacher_loss": 0.00048348584095947444
291
+ },
292
+ {
293
+ "epoch": 0.04181084198385236,
294
+ "grad_norm": 0.9140625,
295
+ "kd_loss": 0.06453745812177658,
296
+ "learning_rate": 9.663049028182112e-06,
297
+ "loss": 0.0511,
298
+ "step": 145,
299
+ "student_loss": 0.04263001307845116,
300
+ "teacher_loss": 0.006518054287880659
301
+ },
302
+ {
303
+ "epoch": 0.04325259515570934,
304
+ "grad_norm": 1.1640625,
305
+ "kd_loss": 0.014925969764590263,
306
+ "learning_rate": 9.633218339615433e-06,
307
+ "loss": 0.0518,
308
+ "step": 150,
309
+ "student_loss": 0.0011302934726700187,
310
+ "teacher_loss": 0.0012746027205139399
311
+ },
312
+ {
313
+ "epoch": 0.04469434832756632,
314
+ "grad_norm": 1.59375,
315
+ "kd_loss": 0.030081020668148994,
316
+ "learning_rate": 9.602172669838721e-06,
317
+ "loss": 0.0591,
318
+ "step": 155,
319
+ "student_loss": 0.0018543315818533301,
320
+ "teacher_loss": 0.0007238256512209773
321
+ },
322
+ {
323
+ "epoch": 0.0461361014994233,
324
+ "grad_norm": 2.15625,
325
+ "kd_loss": 0.026653118431568146,
326
+ "learning_rate": 9.569920160040815e-06,
327
+ "loss": 0.0574,
328
+ "step": 160,
329
+ "student_loss": 0.08247562497854233,
330
+ "teacher_loss": 0.04557507857680321
331
+ },
332
+ {
333
+ "epoch": 0.04757785467128028,
334
+ "grad_norm": 0.91015625,
335
+ "kd_loss": 0.030856940895318985,
336
+ "learning_rate": 9.536469267883432e-06,
337
+ "loss": 0.0493,
338
+ "step": 165,
339
+ "student_loss": 0.0005265406798571348,
340
+ "teacher_loss": 0.004277692176401615
341
+ },
342
+ {
343
+ "epoch": 0.049019607843137254,
344
+ "grad_norm": 2.875,
345
+ "kd_loss": 0.016948891803622246,
346
+ "learning_rate": 9.501828765283295e-06,
347
+ "loss": 0.0524,
348
+ "step": 170,
349
+ "student_loss": 0.0006478180293925107,
350
+ "teacher_loss": 0.000518353539519012
351
+ },
352
+ {
353
+ "epoch": 0.05046136101499423,
354
+ "grad_norm": 2.390625,
355
+ "kd_loss": 0.026514360681176186,
356
+ "learning_rate": 9.466007736111846e-06,
357
+ "loss": 0.0512,
358
+ "step": 175,
359
+ "student_loss": 0.005281130783259869,
360
+ "teacher_loss": 0.0008272691047750413
361
+ },
362
+ {
363
+ "epoch": 0.05190311418685121,
364
+ "grad_norm": 1.1015625,
365
+ "kd_loss": 0.036696940660476685,
366
+ "learning_rate": 9.429015573813163e-06,
367
+ "loss": 0.048,
368
+ "step": 180,
369
+ "student_loss": 0.0007505607791244984,
370
+ "teacher_loss": 0.0003059351583942771
371
+ },
372
+ {
373
+ "epoch": 0.05334486735870819,
374
+ "grad_norm": 1.421875,
375
+ "kd_loss": 0.035748232156038284,
376
+ "learning_rate": 9.390861978940687e-06,
377
+ "loss": 0.0538,
378
+ "step": 185,
379
+ "student_loss": 0.038561683148145676,
380
+ "teacher_loss": 0.00023784062068443745
381
+ },
382
+ {
383
+ "epoch": 0.05478662053056517,
384
+ "grad_norm": 2.171875,
385
+ "kd_loss": 0.051317859441041946,
386
+ "learning_rate": 9.351556956613423e-06,
387
+ "loss": 0.0494,
388
+ "step": 190,
389
+ "student_loss": 0.008348710834980011,
390
+ "teacher_loss": 0.0009534004493616521
391
+ },
392
+ {
393
+ "epoch": 0.056228373702422146,
394
+ "grad_norm": 3.3125,
395
+ "kd_loss": 0.04666966199874878,
396
+ "learning_rate": 9.31111081389227e-06,
397
+ "loss": 0.0497,
398
+ "step": 195,
399
+ "student_loss": 0.008153470233082771,
400
+ "teacher_loss": 0.0002795102773234248
401
+ },
402
+ {
403
+ "epoch": 0.05767012687427912,
404
+ "grad_norm": 2.328125,
405
+ "kd_loss": 0.03317176550626755,
406
+ "learning_rate": 9.269534157077177e-06,
407
+ "loss": 0.0596,
408
+ "step": 200,
409
+ "student_loss": 0.0004428599204402417,
410
+ "teacher_loss": 0.00018954814004246145
411
+ },
412
+ {
413
+ "epoch": 0.0591118800461361,
414
+ "grad_norm": 1.703125,
415
+ "kd_loss": 0.02631596103310585,
416
+ "learning_rate": 9.226837888925813e-06,
417
+ "loss": 0.0551,
418
+ "step": 205,
419
+ "student_loss": 0.03620046377182007,
420
+ "teacher_loss": 0.02737441286444664
421
+ },
422
+ {
423
+ "epoch": 0.06055363321799308,
424
+ "grad_norm": 1.9765625,
425
+ "kd_loss": 0.03267158567905426,
426
+ "learning_rate": 9.183033205794525e-06,
427
+ "loss": 0.0505,
428
+ "step": 210,
429
+ "student_loss": 0.0009676365298219025,
430
+ "teacher_loss": 0.0004164037818554789
431
+ },
432
+ {
433
+ "epoch": 0.061995386389850055,
434
+ "grad_norm": 1.4921875,
435
+ "kd_loss": 0.037410128861665726,
436
+ "learning_rate": 9.13813159470227e-06,
437
+ "loss": 0.0539,
438
+ "step": 215,
439
+ "student_loss": 0.00046113875578157604,
440
+ "teacher_loss": 0.00020874453184660524
441
+ },
442
+ {
443
+ "epoch": 0.06343713956170703,
444
+ "grad_norm": 1.6328125,
445
+ "kd_loss": 0.04395928606390953,
446
+ "learning_rate": 9.092144830318357e-06,
447
+ "loss": 0.0508,
448
+ "step": 220,
449
+ "student_loss": 0.030493613332509995,
450
+ "teacher_loss": 0.010191974230110645
451
+ },
452
+ {
453
+ "epoch": 0.06487889273356401,
454
+ "grad_norm": 1.359375,
455
+ "kd_loss": 0.022184843197464943,
456
+ "learning_rate": 9.045084971874738e-06,
457
+ "loss": 0.0603,
458
+ "step": 225,
459
+ "student_loss": 0.04274655878543854,
460
+ "teacher_loss": 0.001591266831383109
461
+ },
462
+ {
463
+ "epoch": 0.06632064590542099,
464
+ "grad_norm": 1.953125,
465
+ "kd_loss": 0.043670717626810074,
466
+ "learning_rate": 8.99696436000368e-06,
467
+ "loss": 0.0491,
468
+ "step": 230,
469
+ "student_loss": 0.022309448570013046,
470
+ "teacher_loss": 0.006051686126738787
471
+ },
472
+ {
473
+ "epoch": 0.06776239907727798,
474
+ "grad_norm": 1.3203125,
475
+ "kd_loss": 0.07893647253513336,
476
+ "learning_rate": 8.947795613501658e-06,
477
+ "loss": 0.0516,
478
+ "step": 235,
479
+ "student_loss": 0.012356958352029324,
480
+ "teacher_loss": 0.0008457938674837351
481
+ },
482
+ {
483
+ "epoch": 0.06920415224913495,
484
+ "grad_norm": 0.95703125,
485
+ "kd_loss": 0.030352523550391197,
486
+ "learning_rate": 8.897591626020284e-06,
487
+ "loss": 0.0553,
488
+ "step": 240,
489
+ "student_loss": 0.0036721325013786554,
490
+ "teacher_loss": 0.00047237545368261635
491
+ },
492
+ {
493
+ "epoch": 0.07064590542099193,
494
+ "grad_norm": 1.3125,
495
+ "kd_loss": 0.014727318659424782,
496
+ "learning_rate": 8.846365562685178e-06,
497
+ "loss": 0.0472,
498
+ "step": 245,
499
+ "student_loss": 0.002595798345282674,
500
+ "teacher_loss": 0.0030349844601005316
501
+ },
502
+ {
503
+ "epoch": 0.07208765859284891,
504
+ "grad_norm": 2.015625,
505
+ "kd_loss": 0.022538436576724052,
506
+ "learning_rate": 8.794130856643635e-06,
507
+ "loss": 0.0531,
508
+ "step": 250,
509
+ "student_loss": 0.0009441633592359722,
510
+ "teacher_loss": 0.0006299451342783868
511
+ },
512
+ {
513
+ "epoch": 0.07352941176470588,
514
+ "grad_norm": 1.3515625,
515
+ "kd_loss": 0.018597841262817383,
516
+ "learning_rate": 8.74090120554202e-06,
517
+ "loss": 0.0517,
518
+ "step": 255,
519
+ "student_loss": 0.0004831044061575085,
520
+ "teacher_loss": 0.0005120306159369648
521
+ },
522
+ {
523
+ "epoch": 0.07497116493656286,
524
+ "grad_norm": 1.65625,
525
+ "kd_loss": 0.027878236025571823,
526
+ "learning_rate": 8.686690567933803e-06,
527
+ "loss": 0.0558,
528
+ "step": 260,
529
+ "student_loss": 0.04166676476597786,
530
+ "teacher_loss": 0.032882433384656906
531
+ },
532
+ {
533
+ "epoch": 0.07641291810841984,
534
+ "grad_norm": 1.359375,
535
+ "kd_loss": 0.018291378393769264,
536
+ "learning_rate": 8.63151315961915e-06,
537
+ "loss": 0.0532,
538
+ "step": 265,
539
+ "student_loss": 0.0004272922524251044,
540
+ "teacher_loss": 0.000323377171298489
541
+ },
542
+ {
543
+ "epoch": 0.07785467128027682,
544
+ "grad_norm": 1.40625,
545
+ "kd_loss": 0.020099647343158722,
546
+ "learning_rate": 8.575383449917103e-06,
547
+ "loss": 0.0567,
548
+ "step": 270,
549
+ "student_loss": 0.001266423612833023,
550
+ "teacher_loss": 0.0014009763253852725
551
+ },
552
+ {
553
+ "epoch": 0.07929642445213379,
554
+ "grad_norm": 1.5390625,
555
+ "kd_loss": 0.022994602099061012,
556
+ "learning_rate": 8.518316157871232e-06,
557
+ "loss": 0.0493,
558
+ "step": 275,
559
+ "student_loss": 0.04728766158223152,
560
+ "teacher_loss": 0.034376151859760284
561
+ },
562
+ {
563
+ "epoch": 0.08073817762399077,
564
+ "grad_norm": 1.2421875,
565
+ "kd_loss": 0.03356819972395897,
566
+ "learning_rate": 8.460326248389825e-06,
567
+ "loss": 0.0483,
568
+ "step": 280,
569
+ "student_loss": 0.0006493565160781145,
570
+ "teacher_loss": 0.00017707925871945918
571
+ },
572
+ {
573
+ "epoch": 0.08217993079584775,
574
+ "grad_norm": 1.515625,
575
+ "kd_loss": 0.01917041279375553,
576
+ "learning_rate": 8.401428928321607e-06,
577
+ "loss": 0.0539,
578
+ "step": 285,
579
+ "student_loss": 0.0010354992700740695,
580
+ "teacher_loss": 0.0007419176981784403
581
+ },
582
+ {
583
+ "epoch": 0.08362168396770472,
584
+ "grad_norm": 1.4140625,
585
+ "kd_loss": 0.014347212389111519,
586
+ "learning_rate": 8.341639642468002e-06,
587
+ "loss": 0.0547,
588
+ "step": 290,
589
+ "student_loss": 0.01613372191786766,
590
+ "teacher_loss": 0.018294816836714745
591
+ },
592
+ {
593
+ "epoch": 0.0850634371395617,
594
+ "grad_norm": 1.7265625,
595
+ "kd_loss": 0.0206157136708498,
596
+ "learning_rate": 8.280974069532999e-06,
597
+ "loss": 0.0525,
598
+ "step": 295,
599
+ "student_loss": 0.0008157998672686517,
600
+ "teacher_loss": 0.0007138358778320253
601
+ },
602
+ {
603
+ "epoch": 0.08650519031141868,
604
+ "grad_norm": 1.3984375,
605
+ "kd_loss": 0.04587658867239952,
606
+ "learning_rate": 8.219448118011687e-06,
607
+ "loss": 0.0496,
608
+ "step": 300,
609
+ "student_loss": 0.015501612797379494,
610
+ "teacher_loss": 0.0018786232685670257
611
+ },
612
+ {
613
+ "epoch": 0.08794694348327567,
614
+ "grad_norm": 2.203125,
615
+ "kd_loss": 0.02352512627840042,
616
+ "learning_rate": 8.157077922018537e-06,
617
+ "loss": 0.0475,
618
+ "step": 305,
619
+ "student_loss": 0.008210180327296257,
620
+ "teacher_loss": 0.0028180002700537443
621
+ },
622
+ {
623
+ "epoch": 0.08938869665513265,
624
+ "grad_norm": 1.6484375,
625
+ "kd_loss": 0.01950286515057087,
626
+ "learning_rate": 8.093879837056486e-06,
627
+ "loss": 0.05,
628
+ "step": 310,
629
+ "student_loss": 0.0005316836177371442,
630
+ "teacher_loss": 0.00044056857586838305
631
+ },
632
+ {
633
+ "epoch": 0.09083044982698962,
634
+ "grad_norm": 1.9609375,
635
+ "kd_loss": 0.04243648052215576,
636
+ "learning_rate": 8.029870435728018e-06,
637
+ "loss": 0.049,
638
+ "step": 315,
639
+ "student_loss": 0.01187934447079897,
640
+ "teacher_loss": 0.0004662107094191015
641
+ },
642
+ {
643
+ "epoch": 0.0922722029988466,
644
+ "grad_norm": 1.7734375,
645
+ "kd_loss": 0.022677697241306305,
646
+ "learning_rate": 7.965066503389264e-06,
647
+ "loss": 0.0501,
648
+ "step": 320,
649
+ "student_loss": 0.0005311672575771809,
650
+ "teacher_loss": 0.0002600560837890953
651
+ },
652
+ {
653
+ "epoch": 0.09371395617070358,
654
+ "grad_norm": 1.03125,
655
+ "kd_loss": 0.06592908501625061,
656
+ "learning_rate": 7.89948503374835e-06,
657
+ "loss": 0.0465,
658
+ "step": 325,
659
+ "student_loss": 0.005857955198734999,
660
+ "teacher_loss": 0.02006148360669613
661
+ },
662
+ {
663
+ "epoch": 0.09515570934256055,
664
+ "grad_norm": 1.3125,
665
+ "kd_loss": 0.04595617577433586,
666
+ "learning_rate": 7.833143224409076e-06,
667
+ "loss": 0.0501,
668
+ "step": 330,
669
+ "student_loss": 0.01813417300581932,
670
+ "teacher_loss": 0.0033948116470128298
671
+ },
672
+ {
673
+ "epoch": 0.09659746251441753,
674
+ "grad_norm": 1.390625,
675
+ "kd_loss": 0.012639202177524567,
676
+ "learning_rate": 7.766058472361154e-06,
677
+ "loss": 0.0461,
678
+ "step": 335,
679
+ "student_loss": 0.0008777976036071777,
680
+ "teacher_loss": 0.0008721367339603603
681
+ },
682
+ {
683
+ "epoch": 0.09803921568627451,
684
+ "grad_norm": 1.0625,
685
+ "kd_loss": 0.026686903089284897,
686
+ "learning_rate": 7.698248369418146e-06,
687
+ "loss": 0.0515,
688
+ "step": 340,
689
+ "student_loss": 0.010171014815568924,
690
+ "teacher_loss": 0.007541630417108536
691
+ },
692
+ {
693
+ "epoch": 0.09948096885813149,
694
+ "grad_norm": 1.3203125,
695
+ "kd_loss": 0.03351679816842079,
696
+ "learning_rate": 7.629730697604314e-06,
697
+ "loss": 0.0516,
698
+ "step": 345,
699
+ "student_loss": 0.008214120753109455,
700
+ "teacher_loss": 0.0004564044938888401
701
+ },
702
+ {
703
+ "epoch": 0.10092272202998846,
704
+ "grad_norm": 1.71875,
705
+ "kd_loss": 0.05568505823612213,
706
+ "learning_rate": 7.560523424491595e-06,
707
+ "loss": 0.0457,
708
+ "step": 350,
709
+ "student_loss": 0.010019722394645214,
710
+ "teacher_loss": 0.003242993028834462
711
+ },
712
+ {
713
+ "epoch": 0.10236447520184544,
714
+ "grad_norm": 1.8515625,
715
+ "kd_loss": 0.016680920496582985,
716
+ "learning_rate": 7.490644698487909e-06,
717
+ "loss": 0.0531,
718
+ "step": 355,
719
+ "student_loss": 0.0006857452099211514,
720
+ "teacher_loss": 0.0006745306309312582
721
+ },
722
+ {
723
+ "epoch": 0.10380622837370242,
724
+ "grad_norm": 1.2109375,
725
+ "kd_loss": 0.01588292233645916,
726
+ "learning_rate": 7.420112844078066e-06,
727
+ "loss": 0.0509,
728
+ "step": 360,
729
+ "student_loss": 0.001193460077047348,
730
+ "teacher_loss": 0.001017728471197188
731
+ },
732
+ {
733
+ "epoch": 0.1052479815455594,
734
+ "grad_norm": 2.0625,
735
+ "kd_loss": 0.02928476221859455,
736
+ "learning_rate": 7.348946357018479e-06,
737
+ "loss": 0.053,
738
+ "step": 365,
739
+ "student_loss": 0.043851789087057114,
740
+ "teacher_loss": 0.0283951573073864
741
+ },
742
+ {
743
+ "epoch": 0.10668973471741638,
744
+ "grad_norm": 1.21875,
745
+ "kd_loss": 0.04628458246588707,
746
+ "learning_rate": 7.277163899486975e-06,
747
+ "loss": 0.0663,
748
+ "step": 370,
749
+ "student_loss": 0.023408163338899612,
750
+ "teacher_loss": 0.0004174449131824076
751
+ },
752
+ {
753
+ "epoch": 0.10813148788927336,
754
+ "grad_norm": 1.71875,
755
+ "kd_loss": 0.03154220059514046,
756
+ "learning_rate": 7.204784295188959e-06,
757
+ "loss": 0.05,
758
+ "step": 375,
759
+ "student_loss": 0.0324401780962944,
760
+ "teacher_loss": 0.020961124449968338
761
+ },
762
+ {
763
+ "epoch": 0.10957324106113034,
764
+ "grad_norm": 1.046875,
765
+ "kd_loss": 0.01982559822499752,
766
+ "learning_rate": 7.1318265244212305e-06,
767
+ "loss": 0.0493,
768
+ "step": 380,
769
+ "student_loss": 0.0006248670979402959,
770
+ "teacher_loss": 0.000449157232651487
771
+ },
772
+ {
773
+ "epoch": 0.11101499423298732,
774
+ "grad_norm": 1.390625,
775
+ "kd_loss": 0.0306891817599535,
776
+ "learning_rate": 7.05830971909472e-06,
777
+ "loss": 0.0447,
778
+ "step": 385,
779
+ "student_loss": 0.005440755281597376,
780
+ "teacher_loss": 0.00408168276771903
781
+ },
782
+ {
783
+ "epoch": 0.11245674740484429,
784
+ "grad_norm": 1.71875,
785
+ "kd_loss": 0.023891452699899673,
786
+ "learning_rate": 6.9842531577174865e-06,
787
+ "loss": 0.0458,
788
+ "step": 390,
789
+ "student_loss": 0.0003995662846136838,
790
+ "teacher_loss": 0.00039677225868217647
791
+ },
792
+ {
793
+ "epoch": 0.11389850057670127,
794
+ "grad_norm": 1.1328125,
795
+ "kd_loss": 0.018030904233455658,
796
+ "learning_rate": 6.9096762603392595e-06,
797
+ "loss": 0.0489,
798
+ "step": 395,
799
+ "student_loss": 0.0008614315302111208,
800
+ "teacher_loss": 0.0006612912402488291
801
+ },
802
+ {
803
+ "epoch": 0.11534025374855825,
804
+ "grad_norm": 1.703125,
805
+ "kd_loss": 0.02574356086552143,
806
+ "learning_rate": 6.834598583458862e-06,
807
+ "loss": 0.0432,
808
+ "step": 400,
809
+ "student_loss": 0.0009613529546186328,
810
+ "teacher_loss": 0.00023952587798703462
811
+ },
812
+ {
813
+ "epoch": 0.11678200692041522,
814
+ "grad_norm": 1.015625,
815
+ "kd_loss": 0.023935899138450623,
816
+ "learning_rate": 6.7590398148958625e-06,
817
+ "loss": 0.053,
818
+ "step": 405,
819
+ "student_loss": 0.011236889287829399,
820
+ "teacher_loss": 0.00042038856190629303
821
+ },
822
+ {
823
+ "epoch": 0.1182237600922722,
824
+ "grad_norm": 2.078125,
825
+ "kd_loss": 0.03511889651417732,
826
+ "learning_rate": 6.6830197686277945e-06,
827
+ "loss": 0.0539,
828
+ "step": 410,
829
+ "student_loss": 0.014045324176549911,
830
+ "teacher_loss": 0.008123157545924187
831
+ },
832
+ {
833
+ "epoch": 0.11966551326412918,
834
+ "grad_norm": 1.3828125,
835
+ "kd_loss": 0.025273097679018974,
836
+ "learning_rate": 6.6065583795942625e-06,
837
+ "loss": 0.0524,
838
+ "step": 415,
839
+ "student_loss": 0.004831122234463692,
840
+ "teacher_loss": 0.0038736157584935427
841
+ },
842
+ {
843
+ "epoch": 0.12110726643598616,
844
+ "grad_norm": 1.421875,
845
+ "kd_loss": 0.039099402725696564,
846
+ "learning_rate": 6.52967569846937e-06,
847
+ "loss": 0.0521,
848
+ "step": 420,
849
+ "student_loss": 0.04692080616950989,
850
+ "teacher_loss": 0.023907316848635674
851
+ },
852
+ {
853
+ "epoch": 0.12254901960784313,
854
+ "grad_norm": 1.4765625,
855
+ "kd_loss": 0.016112999990582466,
856
+ "learning_rate": 6.452391886403767e-06,
857
+ "loss": 0.0481,
858
+ "step": 425,
859
+ "student_loss": 0.002084826584905386,
860
+ "teacher_loss": 0.0022890178952366114
861
+ },
862
+ {
863
+ "epoch": 0.12399077277970011,
864
+ "grad_norm": 6.375,
865
+ "kd_loss": 0.05122941732406616,
866
+ "learning_rate": 6.374727209737743e-06,
867
+ "loss": 0.0585,
868
+ "step": 430,
869
+ "student_loss": 0.05258077755570412,
870
+ "teacher_loss": 0.06060388311743736
871
+ },
872
+ {
873
+ "epoch": 0.1254325259515571,
874
+ "grad_norm": 1.765625,
875
+ "kd_loss": 0.0188989769667387,
876
+ "learning_rate": 6.296702034686726e-06,
877
+ "loss": 0.0474,
878
+ "step": 435,
879
+ "student_loss": 0.001467574737034738,
880
+ "teacher_loss": 0.001621720613911748
881
+ },
882
+ {
883
+ "epoch": 0.12687427912341406,
884
+ "grad_norm": 0.9375,
885
+ "kd_loss": 0.03803834691643715,
886
+ "learning_rate": 6.218336822000598e-06,
887
+ "loss": 0.051,
888
+ "step": 440,
889
+ "student_loss": 0.012564362958073616,
890
+ "teacher_loss": 0.007415092084556818
891
+ },
892
+ {
893
+ "epoch": 0.12831603229527105,
894
+ "grad_norm": 1.265625,
895
+ "kd_loss": 0.028861815109848976,
896
+ "learning_rate": 6.139652121598219e-06,
897
+ "loss": 0.0508,
898
+ "step": 445,
899
+ "student_loss": 0.03139723464846611,
900
+ "teacher_loss": 0.018937230110168457
901
+ },
902
+ {
903
+ "epoch": 0.12975778546712802,
904
+ "grad_norm": 1.1015625,
905
+ "kd_loss": 0.03469434380531311,
906
+ "learning_rate": 6.060668567178561e-06,
907
+ "loss": 0.0506,
908
+ "step": 450,
909
+ "student_loss": 0.0018778342055156827,
910
+ "teacher_loss": 0.002150396816432476
911
+ },
912
+ {
913
+ "epoch": 0.131199538638985,
914
+ "grad_norm": 1.4375,
915
+ "kd_loss": 0.02303108386695385,
916
+ "learning_rate": 5.981406870809889e-06,
917
+ "loss": 0.0519,
918
+ "step": 455,
919
+ "student_loss": 0.0010092268930748105,
920
+ "teacher_loss": 0.0005768880364485085
921
+ },
922
+ {
923
+ "epoch": 0.13264129181084197,
924
+ "grad_norm": 1.6953125,
925
+ "kd_loss": 0.0384206622838974,
926
+ "learning_rate": 5.9018878174983674e-06,
927
+ "loss": 0.0457,
928
+ "step": 460,
929
+ "student_loss": 0.012533986009657383,
930
+ "teacher_loss": 0.0006231010193005204
931
+ },
932
+ {
933
+ "epoch": 0.13408304498269896,
934
+ "grad_norm": 1.453125,
935
+ "kd_loss": 0.027368323877453804,
936
+ "learning_rate": 5.822132259737565e-06,
937
+ "loss": 0.0517,
938
+ "step": 465,
939
+ "student_loss": 0.0006824060692451894,
940
+ "teacher_loss": 0.0005230730166658759
941
+ },
942
+ {
943
+ "epoch": 0.13552479815455595,
944
+ "grad_norm": 1.3359375,
945
+ "kd_loss": 0.025800921022892,
946
+ "learning_rate": 5.742161112040237e-06,
947
+ "loss": 0.0465,
948
+ "step": 470,
949
+ "student_loss": 0.00049140042392537,
950
+ "teacher_loss": 0.00034405241603963077
951
+ },
952
+ {
953
+ "epoch": 0.13696655132641292,
954
+ "grad_norm": 1.203125,
955
+ "kd_loss": 0.026706764474511147,
956
+ "learning_rate": 5.661995345453867e-06,
957
+ "loss": 0.0431,
958
+ "step": 475,
959
+ "student_loss": 0.0008521017152816057,
960
+ "teacher_loss": 0.0013468376128003001
961
+ },
962
+ {
963
+ "epoch": 0.1384083044982699,
964
+ "grad_norm": 1.4375,
965
+ "kd_loss": 0.030722549185156822,
966
+ "learning_rate": 5.581655982061367e-06,
967
+ "loss": 0.0595,
968
+ "step": 480,
969
+ "student_loss": 0.0362899973988533,
970
+ "teacher_loss": 0.03146218881011009
971
+ },
972
+ {
973
+ "epoch": 0.13985005767012687,
974
+ "grad_norm": 1.4765625,
975
+ "kd_loss": 0.019584491848945618,
976
+ "learning_rate": 5.501164089468406e-06,
977
+ "loss": 0.0469,
978
+ "step": 485,
979
+ "student_loss": 0.0011596218682825565,
980
+ "teacher_loss": 0.0012266576522961259
981
+ },
982
+ {
983
+ "epoch": 0.14129181084198386,
984
+ "grad_norm": 1.890625,
985
+ "kd_loss": 0.02283553034067154,
986
+ "learning_rate": 5.4205407752787884e-06,
987
+ "loss": 0.0567,
988
+ "step": 490,
989
+ "student_loss": 0.00145314063411206,
990
+ "teacher_loss": 0.0009193072910420597
991
+ },
992
+ {
993
+ "epoch": 0.14273356401384082,
994
+ "grad_norm": 1.671875,
995
+ "kd_loss": 0.034809987992048264,
996
+ "learning_rate": 5.339807181559359e-06,
997
+ "loss": 0.0612,
998
+ "step": 495,
999
+ "student_loss": 0.007140801288187504,
1000
+ "teacher_loss": 0.001571109751239419
1001
+ },
1002
+ {
1003
+ "epoch": 0.14417531718569782,
1004
+ "grad_norm": 1.53125,
1005
+ "kd_loss": 0.02246963419020176,
1006
+ "learning_rate": 5.258984479295853e-06,
1007
+ "loss": 0.0475,
1008
+ "step": 500,
1009
+ "student_loss": 0.000618905876763165,
1010
+ "teacher_loss": 0.0004656173987314105
1011
+ },
1012
+ {
1013
+ "epoch": 0.14561707035755478,
1014
+ "grad_norm": 1.5390625,
1015
+ "kd_loss": 0.02328832447528839,
1016
+ "learning_rate": 5.1780938628411795e-06,
1017
+ "loss": 0.0578,
1018
+ "step": 505,
1019
+ "student_loss": 0.0006587600219063461,
1020
+ "teacher_loss": 0.0005737307365052402
1021
+ },
1022
+ {
1023
+ "epoch": 0.14705882352941177,
1024
+ "grad_norm": 2.40625,
1025
+ "kd_loss": 0.02127482183277607,
1026
+ "learning_rate": 5.097156544357567e-06,
1027
+ "loss": 0.0491,
1028
+ "step": 510,
1029
+ "student_loss": 0.0005110295605845749,
1030
+ "teacher_loss": 0.00043354221270419657
1031
+ },
1032
+ {
1033
+ "epoch": 0.14850057670126873,
1034
+ "grad_norm": 1.9140625,
1035
+ "kd_loss": 0.025465093553066254,
1036
+ "learning_rate": 5.016193748254045e-06,
1037
+ "loss": 0.0459,
1038
+ "step": 515,
1039
+ "student_loss": 0.014592062681913376,
1040
+ "teacher_loss": 0.00036142065073363483
1041
+ },
1042
+ {
1043
+ "epoch": 0.14994232987312572,
1044
+ "grad_norm": 1.75,
1045
+ "kd_loss": 0.06094611808657646,
1046
+ "learning_rate": 4.935226705620699e-06,
1047
+ "loss": 0.0525,
1048
+ "step": 520,
1049
+ "student_loss": 0.04277863726019859,
1050
+ "teacher_loss": 0.0119513850659132
1051
+ },
1052
+ {
1053
+ "epoch": 0.1513840830449827,
1054
+ "grad_norm": 2.1875,
1055
+ "kd_loss": 0.02507575787603855,
1056
+ "learning_rate": 4.8542766486612035e-06,
1057
+ "loss": 0.0525,
1058
+ "step": 525,
1059
+ "student_loss": 0.0006827444885857403,
1060
+ "teacher_loss": 0.0003920310118701309
1061
+ },
1062
+ {
1063
+ "epoch": 0.15282583621683968,
1064
+ "grad_norm": 2.4375,
1065
+ "kd_loss": 0.027806738391518593,
1066
+ "learning_rate": 4.773364805125025e-06,
1067
+ "loss": 0.0489,
1068
+ "step": 530,
1069
+ "student_loss": 0.0006790340994484723,
1070
+ "teacher_loss": 0.0006053142133168876
1071
+ },
1072
+ {
1073
+ "epoch": 0.15426758938869667,
1074
+ "grad_norm": 2.859375,
1075
+ "kd_loss": 0.029496297240257263,
1076
+ "learning_rate": 4.6925123927408265e-06,
1077
+ "loss": 0.0469,
1078
+ "step": 535,
1079
+ "student_loss": 0.0009063734323717654,
1080
+ "teacher_loss": 0.0007879663608036935
1081
+ },
1082
+ {
1083
+ "epoch": 0.15570934256055363,
1084
+ "grad_norm": 1.96875,
1085
+ "kd_loss": 0.058626189827919006,
1086
+ "learning_rate": 4.611740613652485e-06,
1087
+ "loss": 0.0455,
1088
+ "step": 540,
1089
+ "student_loss": 0.021025465801358223,
1090
+ "teacher_loss": 0.0006112148403190076
1091
+ },
1092
+ {
1093
+ "epoch": 0.15715109573241062,
1094
+ "grad_norm": 1.9296875,
1095
+ "kd_loss": 0.020126501098275185,
1096
+ "learning_rate": 4.531070648859186e-06,
1097
+ "loss": 0.0436,
1098
+ "step": 545,
1099
+ "student_loss": 0.0014909773599356413,
1100
+ "teacher_loss": 0.0013666743179783225
1101
+ },
1102
+ {
1103
+ "epoch": 0.15859284890426759,
1104
+ "grad_norm": 2.25,
1105
+ "kd_loss": 0.045972585678100586,
1106
+ "learning_rate": 4.450523652661086e-06,
1107
+ "loss": 0.042,
1108
+ "step": 550,
1109
+ "student_loss": 0.000331137707689777,
1110
+ "teacher_loss": 0.0043593221344053745
1111
+ },
1112
+ {
1113
+ "epoch": 0.16003460207612458,
1114
+ "grad_norm": 1.6484375,
1115
+ "kd_loss": 0.0674143135547638,
1116
+ "learning_rate": 4.370120747111956e-06,
1117
+ "loss": 0.0488,
1118
+ "step": 555,
1119
+ "student_loss": 0.049931082874536514,
1120
+ "teacher_loss": 0.002049551112577319
1121
+ },
1122
+ {
1123
+ "epoch": 0.16147635524798154,
1124
+ "grad_norm": 1.78125,
1125
+ "kd_loss": 0.030732685700058937,
1126
+ "learning_rate": 4.289883016480291e-06,
1127
+ "loss": 0.052,
1128
+ "step": 560,
1129
+ "student_loss": 0.0015958240255713463,
1130
+ "teacher_loss": 0.0004530187579803169
1131
+ },
1132
+ {
1133
+ "epoch": 0.16291810841983853,
1134
+ "grad_norm": 1.71875,
1135
+ "kd_loss": 0.03263721987605095,
1136
+ "learning_rate": 4.209831501720328e-06,
1137
+ "loss": 0.0492,
1138
+ "step": 565,
1139
+ "student_loss": 0.027923284098505974,
1140
+ "teacher_loss": 0.02007977105677128
1141
+ },
1142
+ {
1143
+ "epoch": 0.1643598615916955,
1144
+ "grad_norm": 1.7109375,
1145
+ "kd_loss": 0.030005840584635735,
1146
+ "learning_rate": 4.129987194954421e-06,
1147
+ "loss": 0.0521,
1148
+ "step": 570,
1149
+ "student_loss": 0.010670044459402561,
1150
+ "teacher_loss": 0.007075584959238768
1151
+ },
1152
+ {
1153
+ "epoch": 0.16580161476355249,
1154
+ "grad_norm": 1.7421875,
1155
+ "kd_loss": 0.022702785208821297,
1156
+ "learning_rate": 4.050371033968216e-06,
1157
+ "loss": 0.0499,
1158
+ "step": 575,
1159
+ "student_loss": 0.0008753272122703493,
1160
+ "teacher_loss": 0.0006742495461367071
1161
+ },
1162
+ {
1163
+ "epoch": 0.16724336793540945,
1164
+ "grad_norm": 0.94140625,
1165
+ "kd_loss": 0.024699244648218155,
1166
+ "learning_rate": 3.9710038967200825e-06,
1167
+ "loss": 0.044,
1168
+ "step": 580,
1169
+ "student_loss": 0.003035952802747488,
1170
+ "teacher_loss": 0.0022496841847896576
1171
+ },
1172
+ {
1173
+ "epoch": 0.16868512110726644,
1174
+ "grad_norm": 1.2734375,
1175
+ "kd_loss": 0.02395116351544857,
1176
+ "learning_rate": 3.89190659586623e-06,
1177
+ "loss": 0.043,
1178
+ "step": 585,
1179
+ "student_loss": 0.0006281615933403373,
1180
+ "teacher_loss": 0.00044049997813999653
1181
+ },
1182
+ {
1183
+ "epoch": 0.1701268742791234,
1184
+ "grad_norm": 1.25,
1185
+ "kd_loss": 0.027902590110898018,
1186
+ "learning_rate": 3.8130998733029517e-06,
1187
+ "loss": 0.0462,
1188
+ "step": 590,
1189
+ "student_loss": 0.000795386906247586,
1190
+ "teacher_loss": 0.0014348695985972881
1191
+ },
1192
+ {
1193
+ "epoch": 0.1715686274509804,
1194
+ "grad_norm": 1.7890625,
1195
+ "kd_loss": 0.02104509435594082,
1196
+ "learning_rate": 3.734604394727419e-06,
1197
+ "loss": 0.0489,
1198
+ "step": 595,
1199
+ "student_loss": 0.0007530418806709349,
1200
+ "teacher_loss": 0.000717821647413075
1201
+ },
1202
+ {
1203
+ "epoch": 0.17301038062283736,
1204
+ "grad_norm": 1.7578125,
1205
+ "kd_loss": 0.033956341445446014,
1206
+ "learning_rate": 3.656440744218464e-06,
1207
+ "loss": 0.0514,
1208
+ "step": 600,
1209
+ "student_loss": 0.04278191551566124,
1210
+ "teacher_loss": 0.0042961700819432735
1211
+ },
1212
+ {
1213
+ "epoch": 0.17445213379469435,
1214
+ "grad_norm": 1.765625,
1215
+ "kd_loss": 0.043022528290748596,
1216
+ "learning_rate": 3.578629418838757e-06,
1217
+ "loss": 0.0503,
1218
+ "step": 605,
1219
+ "student_loss": 0.09618600457906723,
1220
+ "teacher_loss": 0.0364975742995739
1221
+ },
1222
+ {
1223
+ "epoch": 0.17589388696655134,
1224
+ "grad_norm": 1.53125,
1225
+ "kd_loss": 0.035916365683078766,
1226
+ "learning_rate": 3.5011908232598124e-06,
1227
+ "loss": 0.0467,
1228
+ "step": 610,
1229
+ "student_loss": 0.011542106047272682,
1230
+ "teacher_loss": 0.0006803967989981174
1231
+ },
1232
+ {
1233
+ "epoch": 0.1773356401384083,
1234
+ "grad_norm": 2.25,
1235
+ "kd_loss": 0.019939018413424492,
1236
+ "learning_rate": 3.4241452644112085e-06,
1237
+ "loss": 0.0427,
1238
+ "step": 615,
1239
+ "student_loss": 0.00031424107146449387,
1240
+ "teacher_loss": 0.0003367433091625571
1241
+ },
1242
+ {
1243
+ "epoch": 0.1787773933102653,
1244
+ "grad_norm": 1.25,
1245
+ "kd_loss": 0.02796308510005474,
1246
+ "learning_rate": 3.3475129461554567e-06,
1247
+ "loss": 0.0556,
1248
+ "step": 620,
1249
+ "student_loss": 0.011295602656900883,
1250
+ "teacher_loss": 0.007349366322159767
1251
+ },
1252
+ {
1253
+ "epoch": 0.18021914648212226,
1254
+ "grad_norm": 1.328125,
1255
+ "kd_loss": 0.02499685436487198,
1256
+ "learning_rate": 3.271313963989886e-06,
1257
+ "loss": 0.0437,
1258
+ "step": 625,
1259
+ "student_loss": 0.013104252517223358,
1260
+ "teacher_loss": 0.005263039842247963
1261
+ },
1262
+ {
1263
+ "epoch": 0.18166089965397925,
1264
+ "grad_norm": 1.2265625,
1265
+ "kd_loss": 0.03172106295824051,
1266
+ "learning_rate": 3.195568299776945e-06,
1267
+ "loss": 0.0456,
1268
+ "step": 630,
1269
+ "student_loss": 0.0007382084731943905,
1270
+ "teacher_loss": 0.0011308534303680062
1271
+ },
1272
+ {
1273
+ "epoch": 0.1831026528258362,
1274
+ "grad_norm": 1.0859375,
1275
+ "kd_loss": 0.026842625811696053,
1276
+ "learning_rate": 3.1202958165043053e-06,
1277
+ "loss": 0.0683,
1278
+ "step": 635,
1279
+ "student_loss": 0.0013574030017480254,
1280
+ "teacher_loss": 0.0005100882262922823
1281
+ },
1282
+ {
1283
+ "epoch": 0.1845444059976932,
1284
+ "grad_norm": 1.21875,
1285
+ "kd_loss": 0.028472959995269775,
1286
+ "learning_rate": 3.045516253076137e-06,
1287
+ "loss": 0.0432,
1288
+ "step": 640,
1289
+ "student_loss": 0.0005839015357196331,
1290
+ "teacher_loss": 0.00044177830568514764
1291
+ },
1292
+ {
1293
+ "epoch": 0.18598615916955016,
1294
+ "grad_norm": 1.5546875,
1295
+ "kd_loss": 0.0288334172219038,
1296
+ "learning_rate": 2.9712492191369245e-06,
1297
+ "loss": 0.0431,
1298
+ "step": 645,
1299
+ "student_loss": 0.0008110016933642328,
1300
+ "teacher_loss": 0.0007220477564260364
1301
+ },
1302
+ {
1303
+ "epoch": 0.18742791234140715,
1304
+ "grad_norm": 1.4765625,
1305
+ "kd_loss": 0.02081391215324402,
1306
+ "learning_rate": 2.8975141899291777e-06,
1307
+ "loss": 0.0435,
1308
+ "step": 650,
1309
+ "student_loss": 0.001152709242887795,
1310
+ "teacher_loss": 0.0009690260048955679
1311
+ },
1312
+ {
1313
+ "epoch": 0.18886966551326412,
1314
+ "grad_norm": 1.203125,
1315
+ "kd_loss": 0.033997293561697006,
1316
+ "learning_rate": 2.8243305011863843e-06,
1317
+ "loss": 0.0434,
1318
+ "step": 655,
1319
+ "student_loss": 0.000694808317348361,
1320
+ "teacher_loss": 0.0004087509005330503
1321
+ },
1322
+ {
1323
+ "epoch": 0.1903114186851211,
1324
+ "grad_norm": 2.265625,
1325
+ "kd_loss": 0.020430469885468483,
1326
+ "learning_rate": 2.751717344062552e-06,
1327
+ "loss": 0.0495,
1328
+ "step": 660,
1329
+ "student_loss": 0.0010103078093379736,
1330
+ "teacher_loss": 0.0016366998897865415
1331
+ },
1332
+ {
1333
+ "epoch": 0.19175317185697807,
1334
+ "grad_norm": 1.6640625,
1335
+ "kd_loss": 0.05377822369337082,
1336
+ "learning_rate": 2.6796937600996587e-06,
1337
+ "loss": 0.0528,
1338
+ "step": 665,
1339
+ "student_loss": 0.005403114948421717,
1340
+ "teacher_loss": 0.0009035084513016045
1341
+ },
1342
+ {
1343
+ "epoch": 0.19319492502883506,
1344
+ "grad_norm": 2.59375,
1345
+ "kd_loss": 0.026235708966851234,
1346
+ "learning_rate": 2.6082786362343377e-06,
1347
+ "loss": 0.0493,
1348
+ "step": 670,
1349
+ "student_loss": 0.003308928105980158,
1350
+ "teacher_loss": 0.0013608451699838042
1351
+ },
1352
+ {
1353
+ "epoch": 0.19463667820069205,
1354
+ "grad_norm": 1.265625,
1355
+ "kd_loss": 0.01900545135140419,
1356
+ "learning_rate": 2.5374906998451094e-06,
1357
+ "loss": 0.0492,
1358
+ "step": 675,
1359
+ "student_loss": 0.0011508408933877945,
1360
+ "teacher_loss": 0.001155366189777851
1361
+ },
1362
+ {
1363
+ "epoch": 0.19607843137254902,
1364
+ "grad_norm": 1.7890625,
1365
+ "kd_loss": 0.061810869723558426,
1366
+ "learning_rate": 2.467348513841447e-06,
1367
+ "loss": 0.0432,
1368
+ "step": 680,
1369
+ "student_loss": 0.12517237663269043,
1370
+ "teacher_loss": 0.044221408665180206
1371
+ },
1372
+ {
1373
+ "epoch": 0.197520184544406,
1374
+ "grad_norm": 1.2890625,
1375
+ "kd_loss": 0.024152066558599472,
1376
+ "learning_rate": 2.3978704717959777e-06,
1377
+ "loss": 0.0466,
1378
+ "step": 685,
1379
+ "student_loss": 0.03903283551335335,
1380
+ "teacher_loss": 0.031057028099894524
1381
+ },
1382
+ {
1383
+ "epoch": 0.19896193771626297,
1384
+ "grad_norm": 1.34375,
1385
+ "kd_loss": 0.04370317608118057,
1386
+ "learning_rate": 2.329074793121085e-06,
1387
+ "loss": 0.0486,
1388
+ "step": 690,
1389
+ "student_loss": 0.007788515649735928,
1390
+ "teacher_loss": 0.012201469391584396
1391
+ },
1392
+ {
1393
+ "epoch": 0.20040369088811996,
1394
+ "grad_norm": 1.3515625,
1395
+ "kd_loss": 0.02880249172449112,
1396
+ "learning_rate": 2.260979518291186e-06,
1397
+ "loss": 0.0515,
1398
+ "step": 695,
1399
+ "student_loss": 0.0037542004138231277,
1400
+ "teacher_loss": 0.006665734574198723
1401
+ },
1402
+ {
1403
+ "epoch": 0.20184544405997693,
1404
+ "grad_norm": 1.8828125,
1405
+ "kd_loss": 0.021196264773607254,
1406
+ "learning_rate": 2.1936025041119268e-06,
1407
+ "loss": 0.0468,
1408
+ "step": 700,
1409
+ "student_loss": 0.0005752092110924423,
1410
+ "teacher_loss": 0.0004531157319433987
1411
+ },
1412
+ {
1413
+ "epoch": 0.20328719723183392,
1414
+ "grad_norm": 2.171875,
1415
+ "kd_loss": 0.020803187042474747,
1416
+ "learning_rate": 2.1269614190375477e-06,
1417
+ "loss": 0.0466,
1418
+ "step": 705,
1419
+ "student_loss": 0.0011499158572405577,
1420
+ "teacher_loss": 0.0012183921644464135
1421
+ },
1422
+ {
1423
+ "epoch": 0.20472895040369088,
1424
+ "grad_norm": 1.5,
1425
+ "kd_loss": 0.049379028379917145,
1426
+ "learning_rate": 2.061073738537635e-06,
1427
+ "loss": 0.0499,
1428
+ "step": 710,
1429
+ "student_loss": 0.008894918486475945,
1430
+ "teacher_loss": 0.005386671982705593
1431
+ },
1432
+ {
1433
+ "epoch": 0.20617070357554787,
1434
+ "grad_norm": 1.15625,
1435
+ "kd_loss": 0.023851389065384865,
1436
+ "learning_rate": 1.9959567405144825e-06,
1437
+ "loss": 0.0567,
1438
+ "step": 715,
1439
+ "student_loss": 0.002494252985343337,
1440
+ "teacher_loss": 0.0045929416082799435
1441
+ },
1442
+ {
1443
+ "epoch": 0.20761245674740483,
1444
+ "grad_norm": 1.4453125,
1445
+ "kd_loss": 0.021184897050261497,
1446
+ "learning_rate": 1.931627500772263e-06,
1447
+ "loss": 0.049,
1448
+ "step": 720,
1449
+ "student_loss": 0.0017233651597052813,
1450
+ "teacher_loss": 0.0023725153878331184
1451
+ },
1452
+ {
1453
+ "epoch": 0.20905420991926182,
1454
+ "grad_norm": 1.1171875,
1455
+ "kd_loss": 0.02000688575208187,
1456
+ "learning_rate": 1.8681028885391905e-06,
1457
+ "loss": 0.0409,
1458
+ "step": 725,
1459
+ "student_loss": 0.0005524588632397354,
1460
+ "teacher_loss": 0.000574704899918288
1461
+ },
1462
+ {
1463
+ "epoch": 0.2104959630911188,
1464
+ "grad_norm": 2.484375,
1465
+ "kd_loss": 0.03425552695989609,
1466
+ "learning_rate": 1.8053995620438625e-06,
1467
+ "loss": 0.0464,
1468
+ "step": 730,
1469
+ "student_loss": 0.003905985504388809,
1470
+ "teacher_loss": 0.0014549260959029198
1471
+ },
1472
+ {
1473
+ "epoch": 0.21193771626297578,
1474
+ "grad_norm": 1.625,
1475
+ "kd_loss": 0.02555866912007332,
1476
+ "learning_rate": 1.743533964146924e-06,
1477
+ "loss": 0.0538,
1478
+ "step": 735,
1479
+ "student_loss": 0.001144330482929945,
1480
+ "teacher_loss": 0.00026118988171219826
1481
+ },
1482
+ {
1483
+ "epoch": 0.21337946943483277,
1484
+ "grad_norm": 1.9140625,
1485
+ "kd_loss": 0.029069842770695686,
1486
+ "learning_rate": 1.6825223180292138e-06,
1487
+ "loss": 0.0448,
1488
+ "step": 740,
1489
+ "student_loss": 0.0027026126626878977,
1490
+ "teacher_loss": 0.0003717290237545967
1491
+ },
1492
+ {
1493
+ "epoch": 0.21482122260668973,
1494
+ "grad_norm": 1.390625,
1495
+ "kd_loss": 0.026689402759075165,
1496
+ "learning_rate": 1.6223806229375182e-06,
1497
+ "loss": 0.0472,
1498
+ "step": 745,
1499
+ "student_loss": 0.043619196861982346,
1500
+ "teacher_loss": 0.01909303106367588
1501
+ },
1502
+ {
1503
+ "epoch": 0.21626297577854672,
1504
+ "grad_norm": 1.578125,
1505
+ "kd_loss": 0.04832395911216736,
1506
+ "learning_rate": 1.563124649989043e-06,
1507
+ "loss": 0.0463,
1508
+ "step": 750,
1509
+ "student_loss": 0.01544917095452547,
1510
+ "teacher_loss": 0.004783012438565493
1511
+ },
1512
+ {
1513
+ "epoch": 0.2177047289504037,
1514
+ "grad_norm": 2.21875,
1515
+ "kd_loss": 0.023376960307359695,
1516
+ "learning_rate": 1.5047699380357134e-06,
1517
+ "loss": 0.055,
1518
+ "step": 755,
1519
+ "student_loss": 0.03388846293091774,
1520
+ "teacher_loss": 0.008280983194708824
1521
+ },
1522
+ {
1523
+ "epoch": 0.21914648212226068,
1524
+ "grad_norm": 2.4375,
1525
+ "kd_loss": 0.06135249882936478,
1526
+ "learning_rate": 1.4473317895893773e-06,
1527
+ "loss": 0.0451,
1528
+ "step": 760,
1529
+ "student_loss": 0.06566105037927628,
1530
+ "teacher_loss": 0.0016040820628404617
1531
+ },
1532
+ {
1533
+ "epoch": 0.22058823529411764,
1534
+ "grad_norm": 2.296875,
1535
+ "kd_loss": 0.04429236426949501,
1536
+ "learning_rate": 1.39082526680899e-06,
1537
+ "loss": 0.0532,
1538
+ "step": 765,
1539
+ "student_loss": 0.003117841901257634,
1540
+ "teacher_loss": 0.0014330544508993626
1541
+ },
1542
+ {
1543
+ "epoch": 0.22202998846597463,
1544
+ "grad_norm": 1.109375,
1545
+ "kd_loss": 0.044245027005672455,
1546
+ "learning_rate": 1.3352651875508204e-06,
1547
+ "loss": 0.0475,
1548
+ "step": 770,
1549
+ "student_loss": 0.0025587521959096193,
1550
+ "teacher_loss": 0.0023581357672810555
1551
+ },
1552
+ {
1553
+ "epoch": 0.2234717416378316,
1554
+ "grad_norm": 1.203125,
1555
+ "kd_loss": 0.028671732172369957,
1556
+ "learning_rate": 1.2806661214827286e-06,
1557
+ "loss": 0.044,
1558
+ "step": 775,
1559
+ "student_loss": 0.0025355254765599966,
1560
+ "teacher_loss": 0.0005868234438821673
1561
+ },
1562
+ {
1563
+ "epoch": 0.22491349480968859,
1564
+ "grad_norm": 1.609375,
1565
+ "kd_loss": 0.025248337537050247,
1566
+ "learning_rate": 1.2270423862635188e-06,
1567
+ "loss": 0.0499,
1568
+ "step": 780,
1569
+ "student_loss": 0.0004664502339437604,
1570
+ "teacher_loss": 0.00047028783592395484
1571
+ },
1572
+ {
1573
+ "epoch": 0.22635524798154555,
1574
+ "grad_norm": 2.515625,
1575
+ "kd_loss": 0.020624032244086266,
1576
+ "learning_rate": 1.1744080437883859e-06,
1577
+ "loss": 0.0452,
1578
+ "step": 785,
1579
+ "student_loss": 0.0006135970470495522,
1580
+ "teacher_loss": 0.001232421607710421
1581
+ },
1582
+ {
1583
+ "epoch": 0.22779700115340254,
1584
+ "grad_norm": 2.125,
1585
+ "kd_loss": 0.017469381913542747,
1586
+ "learning_rate": 1.1227768965014246e-06,
1587
+ "loss": 0.043,
1588
+ "step": 790,
1589
+ "student_loss": 0.00502235721796751,
1590
+ "teacher_loss": 0.0049964492209255695
1591
+ },
1592
+ {
1593
+ "epoch": 0.2292387543252595,
1594
+ "grad_norm": 1.2421875,
1595
+ "kd_loss": 0.026635998860001564,
1596
+ "learning_rate": 1.0721624837761768e-06,
1597
+ "loss": 0.0489,
1598
+ "step": 795,
1599
+ "student_loss": 0.0006510152597911656,
1600
+ "teacher_loss": 0.0010630956385284662
1601
+ },
1602
+ {
1603
+ "epoch": 0.2306805074971165,
1604
+ "grad_norm": 1.5078125,
1605
+ "kd_loss": 0.042707618325948715,
1606
+ "learning_rate": 1.0225780783651689e-06,
1607
+ "loss": 0.0537,
1608
+ "step": 800,
1609
+ "student_loss": 0.06411539763212204,
1610
+ "teacher_loss": 0.03272087126970291
1611
+ },
1612
+ {
1613
+ "epoch": 0.23212226066897348,
1614
+ "grad_norm": 1.6796875,
1615
+ "kd_loss": 0.021439047530293465,
1616
+ "learning_rate": 9.740366829193587e-07,
1617
+ "loss": 0.0504,
1618
+ "step": 805,
1619
+ "student_loss": 0.0006748085143044591,
1620
+ "teacher_loss": 0.0005472498596645892
1621
+ },
1622
+ {
1623
+ "epoch": 0.23356401384083045,
1624
+ "grad_norm": 31.5,
1625
+ "kd_loss": 0.03162664547562599,
1626
+ "learning_rate": 9.265510265784189e-07,
1627
+ "loss": 0.0547,
1628
+ "step": 810,
1629
+ "student_loss": 0.0007046000100672245,
1630
+ "teacher_loss": 0.000402910663979128
1631
+ },
1632
+ {
1633
+ "epoch": 0.23500576701268744,
1634
+ "grad_norm": 1.9609375,
1635
+ "kd_loss": 0.022925056517124176,
1636
+ "learning_rate": 8.801335616327378e-07,
1637
+ "loss": 0.0465,
1638
+ "step": 815,
1639
+ "student_loss": 0.005376690998673439,
1640
+ "teacher_loss": 0.01123568881303072
1641
+ },
1642
+ {
1643
+ "epoch": 0.2364475201845444,
1644
+ "grad_norm": 1.890625,
1645
+ "kd_loss": 0.030163748189806938,
1646
+ "learning_rate": 8.347964602580245e-07,
1647
+ "loss": 0.0473,
1648
+ "step": 820,
1649
+ "student_loss": 0.0091484934091568,
1650
+ "teacher_loss": 0.0007933934102766216
1651
+ },
1652
+ {
1653
+ "epoch": 0.2378892733564014,
1654
+ "grad_norm": 2.109375,
1655
+ "kd_loss": 0.03393648937344551,
1656
+ "learning_rate": 7.905516113233652e-07,
1657
+ "loss": 0.0454,
1658
+ "step": 825,
1659
+ "student_loss": 0.0011236692080274224,
1660
+ "teacher_loss": 0.01654369756579399
1661
+ },
1662
+ {
1663
+ "epoch": 0.23933102652825836,
1664
+ "grad_norm": 1.421875,
1665
+ "kd_loss": 0.029087577015161514,
1666
+ "learning_rate": 7.474106172735746e-07,
1667
+ "loss": 0.0471,
1668
+ "step": 830,
1669
+ "student_loss": 0.015236456878483295,
1670
+ "teacher_loss": 0.007140653673559427
1671
+ },
1672
+ {
1673
+ "epoch": 0.24077277970011535,
1674
+ "grad_norm": 1.796875,
1675
+ "kd_loss": 0.024434193968772888,
1676
+ "learning_rate": 7.053847910866513e-07,
1677
+ "loss": 0.0437,
1678
+ "step": 835,
1679
+ "student_loss": 0.012794037349522114,
1680
+ "teacher_loss": 0.006644380744546652
1681
+ },
1682
+ {
1683
+ "epoch": 0.2422145328719723,
1684
+ "grad_norm": 1.21875,
1685
+ "kd_loss": 0.03438033163547516,
1686
+ "learning_rate": 6.644851533071556e-07,
1687
+ "loss": 0.0463,
1688
+ "step": 840,
1689
+ "student_loss": 0.0005363849923014641,
1690
+ "teacher_loss": 0.000252477650064975
1691
+ },
1692
+ {
1693
+ "epoch": 0.2436562860438293,
1694
+ "grad_norm": 1.625,
1695
+ "kd_loss": 0.024818820878863335,
1696
+ "learning_rate": 6.24722429156251e-07,
1697
+ "loss": 0.0645,
1698
+ "step": 845,
1699
+ "student_loss": 0.0007495523896068335,
1700
+ "teacher_loss": 0.0006931237294338644
1701
+ },
1702
+ {
1703
+ "epoch": 0.24509803921568626,
1704
+ "grad_norm": 2.15625,
1705
+ "kd_loss": 0.03822651132941246,
1706
+ "learning_rate": 5.861070457192081e-07,
1707
+ "loss": 0.0495,
1708
+ "step": 850,
1709
+ "student_loss": 0.016144895926117897,
1710
+ "teacher_loss": 0.0005814587930217385
1711
+ },
1712
+ {
1713
+ "epoch": 0.24653979238754326,
1714
+ "grad_norm": 1.578125,
1715
+ "kd_loss": 0.02833644114434719,
1716
+ "learning_rate": 5.486491292110796e-07,
1717
+ "loss": 0.0474,
1718
+ "step": 855,
1719
+ "student_loss": 0.0006240977672860026,
1720
+ "teacher_loss": 0.00048064070870168507
1721
+ },
1722
+ {
1723
+ "epoch": 0.24798154555940022,
1724
+ "grad_norm": 1.6640625,
1725
+ "kd_loss": 0.029966477304697037,
1726
+ "learning_rate": 5.123585023212785e-07,
1727
+ "loss": 0.0534,
1728
+ "step": 860,
1729
+ "student_loss": 0.0015769872115924954,
1730
+ "teacher_loss": 0.0009914073161780834
1731
+ },
1732
+ {
1733
+ "epoch": 0.2494232987312572,
1734
+ "grad_norm": 1.890625,
1735
+ "kd_loss": 0.034011196345090866,
1736
+ "learning_rate": 4.772446816377408e-07,
1737
+ "loss": 0.0493,
1738
+ "step": 865,
1739
+ "student_loss": 0.00044154218630865216,
1740
+ "teacher_loss": 0.0006335440557450056
1741
+ },
1742
+ {
1743
+ "epoch": 0.2508650519031142,
1744
+ "grad_norm": 1.8046875,
1745
+ "kd_loss": 0.0377616211771965,
1746
+ "learning_rate": 4.4331687515137614e-07,
1747
+ "loss": 0.0507,
1748
+ "step": 870,
1749
+ "student_loss": 0.03097929246723652,
1750
+ "teacher_loss": 0.0020129515323787928
1751
+ },
1752
+ {
1753
+ "epoch": 0.25230680507497116,
1754
+ "grad_norm": 1.9609375,
1755
+ "kd_loss": 0.027150675654411316,
1756
+ "learning_rate": 4.1058397984142405e-07,
1757
+ "loss": 0.0478,
1758
+ "step": 875,
1759
+ "student_loss": 0.00037838422576896846,
1760
+ "teacher_loss": 0.0002966223983094096
1761
+ },
1762
+ {
1763
+ "epoch": 0.2537485582468281,
1764
+ "grad_norm": 2.34375,
1765
+ "kd_loss": 0.02994398958981037,
1766
+ "learning_rate": 3.790545793423761e-07,
1767
+ "loss": 0.0508,
1768
+ "step": 880,
1769
+ "student_loss": 0.001047378871589899,
1770
+ "teacher_loss": 0.0005506668239831924
1771
+ },
1772
+ {
1773
+ "epoch": 0.25519031141868515,
1774
+ "grad_norm": 1.7421875,
1775
+ "kd_loss": 0.060813091695308685,
1776
+ "learning_rate": 3.4873694169306915e-07,
1777
+ "loss": 0.0423,
1778
+ "step": 885,
1779
+ "student_loss": 0.009220650419592857,
1780
+ "teacher_loss": 0.04958159476518631
1781
+ },
1782
+ {
1783
+ "epoch": 0.2566320645905421,
1784
+ "grad_norm": 1.703125,
1785
+ "kd_loss": 0.021925147622823715,
1786
+ "learning_rate": 3.196390171685343e-07,
1787
+ "loss": 0.0504,
1788
+ "step": 890,
1789
+ "student_loss": 0.0006026665214449167,
1790
+ "teacher_loss": 0.0006558591849170625
1791
+ },
1792
+ {
1793
+ "epoch": 0.25807381776239907,
1794
+ "grad_norm": 1.6328125,
1795
+ "kd_loss": 0.0360531285405159,
1796
+ "learning_rate": 2.917684361951728e-07,
1797
+ "loss": 0.0483,
1798
+ "step": 895,
1799
+ "student_loss": 0.01280931569635868,
1800
+ "teacher_loss": 0.006571149919182062
1801
+ },
1802
+ {
1803
+ "epoch": 0.25951557093425603,
1804
+ "grad_norm": 1.546875,
1805
+ "kd_loss": 0.03128186613321304,
1806
+ "learning_rate": 2.65132507349814e-07,
1807
+ "loss": 0.0517,
1808
+ "step": 900,
1809
+ "student_loss": 0.029152795672416687,
1810
+ "teacher_loss": 0.021574612706899643
1811
+ },
1812
+ {
1813
+ "epoch": 0.26095732410611305,
1814
+ "grad_norm": 1.2890625,
1815
+ "kd_loss": 0.022932501509785652,
1816
+ "learning_rate": 2.397382154431621e-07,
1817
+ "loss": 0.0462,
1818
+ "step": 905,
1819
+ "student_loss": 0.0031357663683593273,
1820
+ "teacher_loss": 0.0018265560502186418
1821
+ },
1822
+ {
1823
+ "epoch": 0.26239907727797,
1824
+ "grad_norm": 1.953125,
1825
+ "kd_loss": 0.02169475331902504,
1826
+ "learning_rate": 2.1559221968815547e-07,
1827
+ "loss": 0.0449,
1828
+ "step": 910,
1829
+ "student_loss": 0.000516595202498138,
1830
+ "teacher_loss": 0.0005862133693881333
1831
+ },
1832
+ {
1833
+ "epoch": 0.263840830449827,
1834
+ "grad_norm": 1.8359375,
1835
+ "kd_loss": 0.03421870246529579,
1836
+ "learning_rate": 1.9270085195370048e-07,
1837
+ "loss": 0.0447,
1838
+ "step": 915,
1839
+ "student_loss": 0.0019407154759392142,
1840
+ "teacher_loss": 0.0011921236291527748
1841
+ },
1842
+ {
1843
+ "epoch": 0.26528258362168394,
1844
+ "grad_norm": 1.546875,
1845
+ "kd_loss": 0.023406820371747017,
1846
+ "learning_rate": 1.7107011510424766e-07,
1847
+ "loss": 0.048,
1848
+ "step": 920,
1849
+ "student_loss": 0.014788295142352581,
1850
+ "teacher_loss": 0.012664122506976128
1851
+ },
1852
+ {
1853
+ "epoch": 0.26672433679354096,
1854
+ "grad_norm": 1.2578125,
1855
+ "kd_loss": 0.023140916600823402,
1856
+ "learning_rate": 1.5070568142564912e-07,
1857
+ "loss": 0.0451,
1858
+ "step": 925,
1859
+ "student_loss": 0.0005074160289950669,
1860
+ "teacher_loss": 0.0004883141373284161
1861
+ },
1862
+ {
1863
+ "epoch": 0.2681660899653979,
1864
+ "grad_norm": 1.765625,
1865
+ "kd_loss": 0.02654758095741272,
1866
+ "learning_rate": 1.3161289113769405e-07,
1867
+ "loss": 0.0464,
1868
+ "step": 930,
1869
+ "student_loss": 0.022677138447761536,
1870
+ "teacher_loss": 0.0006314730853773654
1871
+ },
1872
+ {
1873
+ "epoch": 0.2696078431372549,
1874
+ "grad_norm": 1.15625,
1875
+ "kd_loss": 0.02347838319838047,
1876
+ "learning_rate": 1.1379675099373489e-07,
1877
+ "loss": 0.045,
1878
+ "step": 935,
1879
+ "student_loss": 0.0015375192742794752,
1880
+ "teacher_loss": 0.0018927000928670168
1881
+ },
1882
+ {
1883
+ "epoch": 0.2710495963091119,
1884
+ "grad_norm": 1.3046875,
1885
+ "kd_loss": 0.04565216600894928,
1886
+ "learning_rate": 9.726193296774767e-08,
1887
+ "loss": 0.0502,
1888
+ "step": 940,
1889
+ "student_loss": 0.007527703884989023,
1890
+ "teacher_loss": 0.0015018682461231947
1891
+ },
1892
+ {
1893
+ "epoch": 0.27249134948096887,
1894
+ "grad_norm": 1.3203125,
1895
+ "kd_loss": 0.02283584326505661,
1896
+ "learning_rate": 8.201277302919086e-08,
1897
+ "loss": 0.051,
1898
+ "step": 945,
1899
+ "student_loss": 0.013407625257968903,
1900
+ "teacher_loss": 0.002004990354180336
1901
+ },
1902
+ {
1903
+ "epoch": 0.27393310265282583,
1904
+ "grad_norm": 1.6640625,
1905
+ "kd_loss": 0.030072957277297974,
1906
+ "learning_rate": 6.805327000596995e-08,
1907
+ "loss": 0.0453,
1908
+ "step": 950,
1909
+ "student_loss": 0.0004709431959781796,
1910
+ "teacher_loss": 0.0020170097704976797
1911
+ },
1912
+ {
1913
+ "epoch": 0.2753748558246828,
1914
+ "grad_norm": 1.296875,
1915
+ "kd_loss": 0.026053113862872124,
1916
+ "learning_rate": 5.538708453581787e-08,
1917
+ "loss": 0.0472,
1918
+ "step": 955,
1919
+ "student_loss": 0.001193143310956657,
1920
+ "teacher_loss": 0.0008201745804399252
1921
+ },
1922
+ {
1923
+ "epoch": 0.2768166089965398,
1924
+ "grad_norm": 2.15625,
1925
+ "kd_loss": 0.03029107116162777,
1926
+ "learning_rate": 4.40175381063529e-08,
1927
+ "loss": 0.0503,
1928
+ "step": 960,
1929
+ "student_loss": 0.0016266013262793422,
1930
+ "teacher_loss": 0.0011345352977514267
1931
+ },
1932
+ {
1933
+ "epoch": 0.2782583621683968,
1934
+ "grad_norm": 1.5546875,
1935
+ "kd_loss": 0.03487255796790123,
1936
+ "learning_rate": 3.394761218407705e-08,
1937
+ "loss": 0.0439,
1938
+ "step": 965,
1939
+ "student_loss": 0.010903997346758842,
1940
+ "teacher_loss": 0.0069407010450959206
1941
+ },
1942
+ {
1943
+ "epoch": 0.27970011534025374,
1944
+ "grad_norm": 2.234375,
1945
+ "kd_loss": 0.02785918489098549,
1946
+ "learning_rate": 2.5179947432540376e-08,
1947
+ "loss": 0.0466,
1948
+ "step": 970,
1949
+ "student_loss": 0.0006625642417930067,
1950
+ "teacher_loss": 0.0003359924885444343
1951
+ },
1952
+ {
1953
+ "epoch": 0.2811418685121107,
1954
+ "grad_norm": 1.671875,
1955
+ "kd_loss": 0.024715285748243332,
1956
+ "learning_rate": 1.7716843019867646e-08,
1957
+ "loss": 0.0433,
1958
+ "step": 975,
1959
+ "student_loss": 0.015628498047590256,
1960
+ "teacher_loss": 0.0008623237372376025
1961
+ },
1962
+ {
1963
+ "epoch": 0.2825836216839677,
1964
+ "grad_norm": 1.9765625,
1965
+ "kd_loss": 0.030769847333431244,
1966
+ "learning_rate": 1.156025601584676e-08,
1967
+ "loss": 0.0485,
1968
+ "step": 980,
1969
+ "student_loss": 0.000937971577513963,
1970
+ "teacher_loss": 0.00045465424773283303
1971
+ },
1972
+ {
1973
+ "epoch": 0.2840253748558247,
1974
+ "grad_norm": 3.546875,
1975
+ "kd_loss": 0.024554533883929253,
1976
+ "learning_rate": 6.711800878718144e-09,
1977
+ "loss": 0.0488,
1978
+ "step": 985,
1979
+ "student_loss": 0.00034612062154337764,
1980
+ "teacher_loss": 0.0003728184674400836
1981
+ },
1982
+ {
1983
+ "epoch": 0.28546712802768165,
1984
+ "grad_norm": 2.09375,
1985
+ "kd_loss": 0.02750871330499649,
1986
+ "learning_rate": 3.1727490318111953e-09,
1987
+ "loss": 0.0439,
1988
+ "step": 990,
1989
+ "student_loss": 0.0019528822740539908,
1990
+ "teacher_loss": 0.0003536756557878107
1991
+ },
1992
+ {
1993
+ "epoch": 0.2869088811995386,
1994
+ "grad_norm": 1.5546875,
1995
+ "kd_loss": 0.03954733535647392,
1996
+ "learning_rate": 9.440285301370865e-10,
1997
+ "loss": 0.0521,
1998
+ "step": 995,
1999
+ "student_loss": 0.002362085971981287,
2000
+ "teacher_loss": 0.001596570946276188
2001
+ },
2002
+ {
2003
+ "epoch": 0.28835063437139563,
2004
+ "grad_norm": 1.375,
2005
+ "kd_loss": 0.029186027124524117,
2006
+ "learning_rate": 2.622381702066523e-11,
2007
+ "loss": 0.0462,
2008
+ "step": 1000,
2009
+ "student_loss": 0.00639536464586854,
2010
+ "teacher_loss": 0.00032160498085431755
2011
+ },
2012
+ {
2013
+ "epoch": 0.28835063437139563,
2014
+ "kd_loss": 0.029186027124524117,
2015
+ "step": 1000,
2016
+ "student_loss": 0.00639536464586854,
2017
+ "teacher_loss": 0.00032160498085431755,
2018
+ "total_flos": 0.0,
2019
+ "train_loss": 0.06003223599493503,
2020
+ "train_runtime": 43009.204,
2021
+ "train_samples_per_second": 0.372,
2022
+ "train_steps_per_second": 0.023
2023
+ }
2024
+ ],
2025
+ "logging_steps": 5,
2026
+ "max_steps": 1000,
2027
+ "num_input_tokens_seen": 0,
2028
+ "num_train_epochs": 1,
2029
+ "save_steps": 200,
2030
+ "stateful_callbacks": {
2031
+ "TrainerControl": {
2032
+ "args": {
2033
+ "should_epoch_stop": false,
2034
+ "should_evaluate": false,
2035
+ "should_log": false,
2036
+ "should_save": false,
2037
+ "should_training_stop": true
2038
+ },
2039
+ "attributes": {}
2040
+ }
2041
+ },
2042
+ "total_flos": 0.0,
2043
+ "train_batch_size": 1,
2044
+ "trial_name": null,
2045
+ "trial_params": null
2046
+ }
checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/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_ls2/checkpoint-200/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_ls2/checkpoint-200/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_ls2/checkpoint-200/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_ls2/checkpoint-200/merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-200/pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d99d36b12cb7be3a1742c657588cccb973bd79d8cadda940632a9fe9b232be20
3
+ size 6187858991
checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-200/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_ls2/checkpoint-200/thought_projector.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f950ef56c72661d0836d9ea90f83a87bd7eb52e10a90b02e8b150250a37fe94c
3
+ size 16788033
checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-200/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_ls2/checkpoint-200/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_ls2/checkpoint-200/trainer_state.json ADDED
@@ -0,0 +1,434 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 0.05767012687427912,
6
+ "eval_steps": 500,
7
+ "global_step": 200,
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": 77.0,
15
+ "kd_loss": 0.2900531589984894,
16
+ "learning_rate": 1.3333333333333334e-06,
17
+ "loss": 0.5715,
18
+ "step": 5,
19
+ "student_loss": 0.11784028261899948,
20
+ "teacher_loss": 0.001998942345380783
21
+ },
22
+ {
23
+ "epoch": 0.0028835063437139563,
24
+ "grad_norm": 26.875,
25
+ "kd_loss": 0.277434378862381,
26
+ "learning_rate": 3e-06,
27
+ "loss": 0.4692,
28
+ "step": 10,
29
+ "student_loss": 0.11281580477952957,
30
+ "teacher_loss": 0.005315451417118311
31
+ },
32
+ {
33
+ "epoch": 0.004325259515570935,
34
+ "grad_norm": 18.5,
35
+ "kd_loss": 0.21174456179141998,
36
+ "learning_rate": 4.666666666666667e-06,
37
+ "loss": 0.3373,
38
+ "step": 15,
39
+ "student_loss": 0.007377153728157282,
40
+ "teacher_loss": 0.0011863313848152757
41
+ },
42
+ {
43
+ "epoch": 0.0057670126874279125,
44
+ "grad_norm": 8.0625,
45
+ "kd_loss": 0.13275651633739471,
46
+ "learning_rate": 6.333333333333333e-06,
47
+ "loss": 0.2555,
48
+ "step": 20,
49
+ "student_loss": 0.04130968824028969,
50
+ "teacher_loss": 0.0005545365856960416
51
+ },
52
+ {
53
+ "epoch": 0.00720876585928489,
54
+ "grad_norm": 4.6875,
55
+ "kd_loss": 0.10908422619104385,
56
+ "learning_rate": 8.000000000000001e-06,
57
+ "loss": 0.2065,
58
+ "step": 25,
59
+ "student_loss": 0.0035145599395036697,
60
+ "teacher_loss": 0.0009682206436991692
61
+ },
62
+ {
63
+ "epoch": 0.00865051903114187,
64
+ "grad_norm": 2.328125,
65
+ "kd_loss": 0.06466298550367355,
66
+ "learning_rate": 9.666666666666667e-06,
67
+ "loss": 0.1499,
68
+ "step": 30,
69
+ "student_loss": 0.0010085003450512886,
70
+ "teacher_loss": 0.0005369208520278335
71
+ },
72
+ {
73
+ "epoch": 0.010092272202998846,
74
+ "grad_norm": 2.609375,
75
+ "kd_loss": 0.05857780575752258,
76
+ "learning_rate": 9.99958042442916e-06,
77
+ "loss": 0.1223,
78
+ "step": 35,
79
+ "student_loss": 0.0010954203316941857,
80
+ "teacher_loss": 0.0006188539555296302
81
+ },
82
+ {
83
+ "epoch": 0.011534025374855825,
84
+ "grad_norm": 1.8359375,
85
+ "kd_loss": 0.057602375745773315,
86
+ "learning_rate": 9.997876019358083e-06,
87
+ "loss": 0.1064,
88
+ "step": 40,
89
+ "student_loss": 0.032847270369529724,
90
+ "teacher_loss": 0.028142843395471573
91
+ },
92
+ {
93
+ "epoch": 0.012975778546712802,
94
+ "grad_norm": 1.578125,
95
+ "kd_loss": 0.08787942677736282,
96
+ "learning_rate": 9.99486100792044e-06,
97
+ "loss": 0.0811,
98
+ "step": 45,
99
+ "student_loss": 0.07371404767036438,
100
+ "teacher_loss": 0.002992093563079834
101
+ },
102
+ {
103
+ "epoch": 0.01441753171856978,
104
+ "grad_norm": 2.078125,
105
+ "kd_loss": 0.037376269698143005,
106
+ "learning_rate": 9.990536180750724e-06,
107
+ "loss": 0.0866,
108
+ "step": 50,
109
+ "student_loss": 0.027633950114250183,
110
+ "teacher_loss": 0.009523550048470497
111
+ },
112
+ {
113
+ "epoch": 0.015859284890426758,
114
+ "grad_norm": 1.8203125,
115
+ "kd_loss": 0.029546089470386505,
116
+ "learning_rate": 9.984902671959911e-06,
117
+ "loss": 0.0757,
118
+ "step": 55,
119
+ "student_loss": 0.0010951841250061989,
120
+ "teacher_loss": 0.0006982200429774821
121
+ },
122
+ {
123
+ "epoch": 0.01730103806228374,
124
+ "grad_norm": 2.21875,
125
+ "kd_loss": 0.029508251696825027,
126
+ "learning_rate": 9.97796195883804e-06,
127
+ "loss": 0.0683,
128
+ "step": 60,
129
+ "student_loss": 0.002835189923644066,
130
+ "teacher_loss": 0.001713551115244627
131
+ },
132
+ {
133
+ "epoch": 0.018742791234140715,
134
+ "grad_norm": 2.234375,
135
+ "kd_loss": 0.030333109200000763,
136
+ "learning_rate": 9.969715861466839e-06,
137
+ "loss": 0.0686,
138
+ "step": 65,
139
+ "student_loss": 0.03012149967253208,
140
+ "teacher_loss": 0.03870478644967079
141
+ },
142
+ {
143
+ "epoch": 0.020184544405997693,
144
+ "grad_norm": 1.265625,
145
+ "kd_loss": 0.05904582515358925,
146
+ "learning_rate": 9.96016654224243e-06,
147
+ "loss": 0.0737,
148
+ "step": 70,
149
+ "student_loss": 0.01526976004242897,
150
+ "teacher_loss": 0.0018924300093203783
151
+ },
152
+ {
153
+ "epoch": 0.02162629757785467,
154
+ "grad_norm": 1.7578125,
155
+ "kd_loss": 0.024645471945405006,
156
+ "learning_rate": 9.94931650530827e-06,
157
+ "loss": 0.0659,
158
+ "step": 75,
159
+ "student_loss": 0.002360731828957796,
160
+ "teacher_loss": 0.0017862127861008048
161
+ },
162
+ {
163
+ "epoch": 0.02306805074971165,
164
+ "grad_norm": 0.96484375,
165
+ "kd_loss": 0.02402629889547825,
166
+ "learning_rate": 9.93716859589851e-06,
167
+ "loss": 0.0587,
168
+ "step": 80,
169
+ "student_loss": 0.0011838233331218362,
170
+ "teacher_loss": 0.0007806048961356282
171
+ },
172
+ {
173
+ "epoch": 0.024509803921568627,
174
+ "grad_norm": 1.09375,
175
+ "kd_loss": 0.025694765150547028,
176
+ "learning_rate": 9.923725999591846e-06,
177
+ "loss": 0.061,
178
+ "step": 85,
179
+ "student_loss": 0.00045988388592377305,
180
+ "teacher_loss": 0.0004276020044926554
181
+ },
182
+ {
183
+ "epoch": 0.025951557093425604,
184
+ "grad_norm": 1.2421875,
185
+ "kd_loss": 0.0665941834449768,
186
+ "learning_rate": 9.908992241476189e-06,
187
+ "loss": 0.062,
188
+ "step": 90,
189
+ "student_loss": 0.018955115228891373,
190
+ "teacher_loss": 0.003640417940914631
191
+ },
192
+ {
193
+ "epoch": 0.027393310265282585,
194
+ "grad_norm": 1.625,
195
+ "kd_loss": 0.03899763152003288,
196
+ "learning_rate": 9.892971185224244e-06,
197
+ "loss": 0.0583,
198
+ "step": 95,
199
+ "student_loss": 0.008055699057877064,
200
+ "teacher_loss": 0.005101984832435846
201
+ },
202
+ {
203
+ "epoch": 0.02883506343713956,
204
+ "grad_norm": 1.3828125,
205
+ "kd_loss": 0.022396311163902283,
206
+ "learning_rate": 9.875667032080354e-06,
207
+ "loss": 0.0519,
208
+ "step": 100,
209
+ "student_loss": 0.0006799453985877335,
210
+ "teacher_loss": 0.0005385762779042125
211
+ },
212
+ {
213
+ "epoch": 0.03027681660899654,
214
+ "grad_norm": 1.609375,
215
+ "kd_loss": 0.0338844433426857,
216
+ "learning_rate": 9.857084319758772e-06,
217
+ "loss": 0.055,
218
+ "step": 105,
219
+ "student_loss": 0.004358346574008465,
220
+ "teacher_loss": 0.0003296585055068135
221
+ },
222
+ {
223
+ "epoch": 0.031718569780853516,
224
+ "grad_norm": 1.234375,
225
+ "kd_loss": 0.033958468586206436,
226
+ "learning_rate": 9.837227921253747e-06,
227
+ "loss": 0.0588,
228
+ "step": 110,
229
+ "student_loss": 0.031565263867378235,
230
+ "teacher_loss": 0.01906467415392399
231
+ },
232
+ {
233
+ "epoch": 0.03316032295271049,
234
+ "grad_norm": 1.4765625,
235
+ "kd_loss": 0.036942481994628906,
236
+ "learning_rate": 9.816103043561648e-06,
237
+ "loss": 0.0579,
238
+ "step": 115,
239
+ "student_loss": 0.018612468615174294,
240
+ "teacher_loss": 0.012586154974997044
241
+ },
242
+ {
243
+ "epoch": 0.03460207612456748,
244
+ "grad_norm": 1.359375,
245
+ "kd_loss": 0.02566700056195259,
246
+ "learning_rate": 9.79371522631553e-06,
247
+ "loss": 0.0494,
248
+ "step": 120,
249
+ "student_loss": 0.001015476998873055,
250
+ "teacher_loss": 0.0008012109901756048
251
+ },
252
+ {
253
+ "epoch": 0.036043829296424454,
254
+ "grad_norm": 1.234375,
255
+ "kd_loss": 0.0527762807905674,
256
+ "learning_rate": 9.770070340332457e-06,
257
+ "loss": 0.0544,
258
+ "step": 125,
259
+ "student_loss": 0.03289179503917694,
260
+ "teacher_loss": 0.00024821600527502596
261
+ },
262
+ {
263
+ "epoch": 0.03748558246828143,
264
+ "grad_norm": 1.4375,
265
+ "kd_loss": 0.020570307970046997,
266
+ "learning_rate": 9.745174586073982e-06,
267
+ "loss": 0.0531,
268
+ "step": 130,
269
+ "student_loss": 0.0006593615980818868,
270
+ "teacher_loss": 0.0006894106627441943
271
+ },
272
+ {
273
+ "epoch": 0.03892733564013841,
274
+ "grad_norm": 1.1953125,
275
+ "kd_loss": 0.03941934183239937,
276
+ "learning_rate": 9.719034492020183e-06,
277
+ "loss": 0.0584,
278
+ "step": 135,
279
+ "student_loss": 0.001388320466503501,
280
+ "teacher_loss": 0.00066422886447981
281
+ },
282
+ {
283
+ "epoch": 0.040369088811995385,
284
+ "grad_norm": 1.125,
285
+ "kd_loss": 0.023267682641744614,
286
+ "learning_rate": 9.691656912957686e-06,
287
+ "loss": 0.0554,
288
+ "step": 140,
289
+ "student_loss": 0.0006728945882059634,
290
+ "teacher_loss": 0.00048348584095947444
291
+ },
292
+ {
293
+ "epoch": 0.04181084198385236,
294
+ "grad_norm": 0.9140625,
295
+ "kd_loss": 0.06453745812177658,
296
+ "learning_rate": 9.663049028182112e-06,
297
+ "loss": 0.0511,
298
+ "step": 145,
299
+ "student_loss": 0.04263001307845116,
300
+ "teacher_loss": 0.006518054287880659
301
+ },
302
+ {
303
+ "epoch": 0.04325259515570934,
304
+ "grad_norm": 1.1640625,
305
+ "kd_loss": 0.014925969764590263,
306
+ "learning_rate": 9.633218339615433e-06,
307
+ "loss": 0.0518,
308
+ "step": 150,
309
+ "student_loss": 0.0011302934726700187,
310
+ "teacher_loss": 0.0012746027205139399
311
+ },
312
+ {
313
+ "epoch": 0.04469434832756632,
314
+ "grad_norm": 1.59375,
315
+ "kd_loss": 0.030081020668148994,
316
+ "learning_rate": 9.602172669838721e-06,
317
+ "loss": 0.0591,
318
+ "step": 155,
319
+ "student_loss": 0.0018543315818533301,
320
+ "teacher_loss": 0.0007238256512209773
321
+ },
322
+ {
323
+ "epoch": 0.0461361014994233,
324
+ "grad_norm": 2.15625,
325
+ "kd_loss": 0.026653118431568146,
326
+ "learning_rate": 9.569920160040815e-06,
327
+ "loss": 0.0574,
328
+ "step": 160,
329
+ "student_loss": 0.08247562497854233,
330
+ "teacher_loss": 0.04557507857680321
331
+ },
332
+ {
333
+ "epoch": 0.04757785467128028,
334
+ "grad_norm": 0.91015625,
335
+ "kd_loss": 0.030856940895318985,
336
+ "learning_rate": 9.536469267883432e-06,
337
+ "loss": 0.0493,
338
+ "step": 165,
339
+ "student_loss": 0.0005265406798571348,
340
+ "teacher_loss": 0.004277692176401615
341
+ },
342
+ {
343
+ "epoch": 0.049019607843137254,
344
+ "grad_norm": 2.875,
345
+ "kd_loss": 0.016948891803622246,
346
+ "learning_rate": 9.501828765283295e-06,
347
+ "loss": 0.0524,
348
+ "step": 170,
349
+ "student_loss": 0.0006478180293925107,
350
+ "teacher_loss": 0.000518353539519012
351
+ },
352
+ {
353
+ "epoch": 0.05046136101499423,
354
+ "grad_norm": 2.390625,
355
+ "kd_loss": 0.026514360681176186,
356
+ "learning_rate": 9.466007736111846e-06,
357
+ "loss": 0.0512,
358
+ "step": 175,
359
+ "student_loss": 0.005281130783259869,
360
+ "teacher_loss": 0.0008272691047750413
361
+ },
362
+ {
363
+ "epoch": 0.05190311418685121,
364
+ "grad_norm": 1.1015625,
365
+ "kd_loss": 0.036696940660476685,
366
+ "learning_rate": 9.429015573813163e-06,
367
+ "loss": 0.048,
368
+ "step": 180,
369
+ "student_loss": 0.0007505607791244984,
370
+ "teacher_loss": 0.0003059351583942771
371
+ },
372
+ {
373
+ "epoch": 0.05334486735870819,
374
+ "grad_norm": 1.421875,
375
+ "kd_loss": 0.035748232156038284,
376
+ "learning_rate": 9.390861978940687e-06,
377
+ "loss": 0.0538,
378
+ "step": 185,
379
+ "student_loss": 0.038561683148145676,
380
+ "teacher_loss": 0.00023784062068443745
381
+ },
382
+ {
383
+ "epoch": 0.05478662053056517,
384
+ "grad_norm": 2.171875,
385
+ "kd_loss": 0.051317859441041946,
386
+ "learning_rate": 9.351556956613423e-06,
387
+ "loss": 0.0494,
388
+ "step": 190,
389
+ "student_loss": 0.008348710834980011,
390
+ "teacher_loss": 0.0009534004493616521
391
+ },
392
+ {
393
+ "epoch": 0.056228373702422146,
394
+ "grad_norm": 3.3125,
395
+ "kd_loss": 0.04666966199874878,
396
+ "learning_rate": 9.31111081389227e-06,
397
+ "loss": 0.0497,
398
+ "step": 195,
399
+ "student_loss": 0.008153470233082771,
400
+ "teacher_loss": 0.0002795102773234248
401
+ },
402
+ {
403
+ "epoch": 0.05767012687427912,
404
+ "grad_norm": 2.328125,
405
+ "kd_loss": 0.03317176550626755,
406
+ "learning_rate": 9.269534157077177e-06,
407
+ "loss": 0.0596,
408
+ "step": 200,
409
+ "student_loss": 0.0004428599204402417,
410
+ "teacher_loss": 0.00018954814004246145
411
+ }
412
+ ],
413
+ "logging_steps": 5,
414
+ "max_steps": 1000,
415
+ "num_input_tokens_seen": 0,
416
+ "num_train_epochs": 1,
417
+ "save_steps": 200,
418
+ "stateful_callbacks": {
419
+ "TrainerControl": {
420
+ "args": {
421
+ "should_epoch_stop": false,
422
+ "should_evaluate": false,
423
+ "should_log": false,
424
+ "should_save": true,
425
+ "should_training_stop": false
426
+ },
427
+ "attributes": {}
428
+ }
429
+ },
430
+ "total_flos": 0.0,
431
+ "train_batch_size": 1,
432
+ "trial_name": null,
433
+ "trial_params": null
434
+ }
checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-200/vocab.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-400/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_ls2/checkpoint-400/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_ls2/checkpoint-400/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_ls2/checkpoint-400/merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-400/pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4b1c606a61677aa1816c3586bd19a810db09076c5c7a67b808931b719821684b
3
+ size 6187858991
checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-400/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_ls2/checkpoint-400/thought_projector.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ede142b7a07ef2b3f7447fc76f05f591789e04ebf5bc4508783064adc212ed6a
3
+ size 16788033
checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-400/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_ls2/checkpoint-400/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_ls2/checkpoint-400/trainer_state.json ADDED
@@ -0,0 +1,834 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 0.11534025374855825,
6
+ "eval_steps": 500,
7
+ "global_step": 400,
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": 77.0,
15
+ "kd_loss": 0.2900531589984894,
16
+ "learning_rate": 1.3333333333333334e-06,
17
+ "loss": 0.5715,
18
+ "step": 5,
19
+ "student_loss": 0.11784028261899948,
20
+ "teacher_loss": 0.001998942345380783
21
+ },
22
+ {
23
+ "epoch": 0.0028835063437139563,
24
+ "grad_norm": 26.875,
25
+ "kd_loss": 0.277434378862381,
26
+ "learning_rate": 3e-06,
27
+ "loss": 0.4692,
28
+ "step": 10,
29
+ "student_loss": 0.11281580477952957,
30
+ "teacher_loss": 0.005315451417118311
31
+ },
32
+ {
33
+ "epoch": 0.004325259515570935,
34
+ "grad_norm": 18.5,
35
+ "kd_loss": 0.21174456179141998,
36
+ "learning_rate": 4.666666666666667e-06,
37
+ "loss": 0.3373,
38
+ "step": 15,
39
+ "student_loss": 0.007377153728157282,
40
+ "teacher_loss": 0.0011863313848152757
41
+ },
42
+ {
43
+ "epoch": 0.0057670126874279125,
44
+ "grad_norm": 8.0625,
45
+ "kd_loss": 0.13275651633739471,
46
+ "learning_rate": 6.333333333333333e-06,
47
+ "loss": 0.2555,
48
+ "step": 20,
49
+ "student_loss": 0.04130968824028969,
50
+ "teacher_loss": 0.0005545365856960416
51
+ },
52
+ {
53
+ "epoch": 0.00720876585928489,
54
+ "grad_norm": 4.6875,
55
+ "kd_loss": 0.10908422619104385,
56
+ "learning_rate": 8.000000000000001e-06,
57
+ "loss": 0.2065,
58
+ "step": 25,
59
+ "student_loss": 0.0035145599395036697,
60
+ "teacher_loss": 0.0009682206436991692
61
+ },
62
+ {
63
+ "epoch": 0.00865051903114187,
64
+ "grad_norm": 2.328125,
65
+ "kd_loss": 0.06466298550367355,
66
+ "learning_rate": 9.666666666666667e-06,
67
+ "loss": 0.1499,
68
+ "step": 30,
69
+ "student_loss": 0.0010085003450512886,
70
+ "teacher_loss": 0.0005369208520278335
71
+ },
72
+ {
73
+ "epoch": 0.010092272202998846,
74
+ "grad_norm": 2.609375,
75
+ "kd_loss": 0.05857780575752258,
76
+ "learning_rate": 9.99958042442916e-06,
77
+ "loss": 0.1223,
78
+ "step": 35,
79
+ "student_loss": 0.0010954203316941857,
80
+ "teacher_loss": 0.0006188539555296302
81
+ },
82
+ {
83
+ "epoch": 0.011534025374855825,
84
+ "grad_norm": 1.8359375,
85
+ "kd_loss": 0.057602375745773315,
86
+ "learning_rate": 9.997876019358083e-06,
87
+ "loss": 0.1064,
88
+ "step": 40,
89
+ "student_loss": 0.032847270369529724,
90
+ "teacher_loss": 0.028142843395471573
91
+ },
92
+ {
93
+ "epoch": 0.012975778546712802,
94
+ "grad_norm": 1.578125,
95
+ "kd_loss": 0.08787942677736282,
96
+ "learning_rate": 9.99486100792044e-06,
97
+ "loss": 0.0811,
98
+ "step": 45,
99
+ "student_loss": 0.07371404767036438,
100
+ "teacher_loss": 0.002992093563079834
101
+ },
102
+ {
103
+ "epoch": 0.01441753171856978,
104
+ "grad_norm": 2.078125,
105
+ "kd_loss": 0.037376269698143005,
106
+ "learning_rate": 9.990536180750724e-06,
107
+ "loss": 0.0866,
108
+ "step": 50,
109
+ "student_loss": 0.027633950114250183,
110
+ "teacher_loss": 0.009523550048470497
111
+ },
112
+ {
113
+ "epoch": 0.015859284890426758,
114
+ "grad_norm": 1.8203125,
115
+ "kd_loss": 0.029546089470386505,
116
+ "learning_rate": 9.984902671959911e-06,
117
+ "loss": 0.0757,
118
+ "step": 55,
119
+ "student_loss": 0.0010951841250061989,
120
+ "teacher_loss": 0.0006982200429774821
121
+ },
122
+ {
123
+ "epoch": 0.01730103806228374,
124
+ "grad_norm": 2.21875,
125
+ "kd_loss": 0.029508251696825027,
126
+ "learning_rate": 9.97796195883804e-06,
127
+ "loss": 0.0683,
128
+ "step": 60,
129
+ "student_loss": 0.002835189923644066,
130
+ "teacher_loss": 0.001713551115244627
131
+ },
132
+ {
133
+ "epoch": 0.018742791234140715,
134
+ "grad_norm": 2.234375,
135
+ "kd_loss": 0.030333109200000763,
136
+ "learning_rate": 9.969715861466839e-06,
137
+ "loss": 0.0686,
138
+ "step": 65,
139
+ "student_loss": 0.03012149967253208,
140
+ "teacher_loss": 0.03870478644967079
141
+ },
142
+ {
143
+ "epoch": 0.020184544405997693,
144
+ "grad_norm": 1.265625,
145
+ "kd_loss": 0.05904582515358925,
146
+ "learning_rate": 9.96016654224243e-06,
147
+ "loss": 0.0737,
148
+ "step": 70,
149
+ "student_loss": 0.01526976004242897,
150
+ "teacher_loss": 0.0018924300093203783
151
+ },
152
+ {
153
+ "epoch": 0.02162629757785467,
154
+ "grad_norm": 1.7578125,
155
+ "kd_loss": 0.024645471945405006,
156
+ "learning_rate": 9.94931650530827e-06,
157
+ "loss": 0.0659,
158
+ "step": 75,
159
+ "student_loss": 0.002360731828957796,
160
+ "teacher_loss": 0.0017862127861008048
161
+ },
162
+ {
163
+ "epoch": 0.02306805074971165,
164
+ "grad_norm": 0.96484375,
165
+ "kd_loss": 0.02402629889547825,
166
+ "learning_rate": 9.93716859589851e-06,
167
+ "loss": 0.0587,
168
+ "step": 80,
169
+ "student_loss": 0.0011838233331218362,
170
+ "teacher_loss": 0.0007806048961356282
171
+ },
172
+ {
173
+ "epoch": 0.024509803921568627,
174
+ "grad_norm": 1.09375,
175
+ "kd_loss": 0.025694765150547028,
176
+ "learning_rate": 9.923725999591846e-06,
177
+ "loss": 0.061,
178
+ "step": 85,
179
+ "student_loss": 0.00045988388592377305,
180
+ "teacher_loss": 0.0004276020044926554
181
+ },
182
+ {
183
+ "epoch": 0.025951557093425604,
184
+ "grad_norm": 1.2421875,
185
+ "kd_loss": 0.0665941834449768,
186
+ "learning_rate": 9.908992241476189e-06,
187
+ "loss": 0.062,
188
+ "step": 90,
189
+ "student_loss": 0.018955115228891373,
190
+ "teacher_loss": 0.003640417940914631
191
+ },
192
+ {
193
+ "epoch": 0.027393310265282585,
194
+ "grad_norm": 1.625,
195
+ "kd_loss": 0.03899763152003288,
196
+ "learning_rate": 9.892971185224244e-06,
197
+ "loss": 0.0583,
198
+ "step": 95,
199
+ "student_loss": 0.008055699057877064,
200
+ "teacher_loss": 0.005101984832435846
201
+ },
202
+ {
203
+ "epoch": 0.02883506343713956,
204
+ "grad_norm": 1.3828125,
205
+ "kd_loss": 0.022396311163902283,
206
+ "learning_rate": 9.875667032080354e-06,
207
+ "loss": 0.0519,
208
+ "step": 100,
209
+ "student_loss": 0.0006799453985877335,
210
+ "teacher_loss": 0.0005385762779042125
211
+ },
212
+ {
213
+ "epoch": 0.03027681660899654,
214
+ "grad_norm": 1.609375,
215
+ "kd_loss": 0.0338844433426857,
216
+ "learning_rate": 9.857084319758772e-06,
217
+ "loss": 0.055,
218
+ "step": 105,
219
+ "student_loss": 0.004358346574008465,
220
+ "teacher_loss": 0.0003296585055068135
221
+ },
222
+ {
223
+ "epoch": 0.031718569780853516,
224
+ "grad_norm": 1.234375,
225
+ "kd_loss": 0.033958468586206436,
226
+ "learning_rate": 9.837227921253747e-06,
227
+ "loss": 0.0588,
228
+ "step": 110,
229
+ "student_loss": 0.031565263867378235,
230
+ "teacher_loss": 0.01906467415392399
231
+ },
232
+ {
233
+ "epoch": 0.03316032295271049,
234
+ "grad_norm": 1.4765625,
235
+ "kd_loss": 0.036942481994628906,
236
+ "learning_rate": 9.816103043561648e-06,
237
+ "loss": 0.0579,
238
+ "step": 115,
239
+ "student_loss": 0.018612468615174294,
240
+ "teacher_loss": 0.012586154974997044
241
+ },
242
+ {
243
+ "epoch": 0.03460207612456748,
244
+ "grad_norm": 1.359375,
245
+ "kd_loss": 0.02566700056195259,
246
+ "learning_rate": 9.79371522631553e-06,
247
+ "loss": 0.0494,
248
+ "step": 120,
249
+ "student_loss": 0.001015476998873055,
250
+ "teacher_loss": 0.0008012109901756048
251
+ },
252
+ {
253
+ "epoch": 0.036043829296424454,
254
+ "grad_norm": 1.234375,
255
+ "kd_loss": 0.0527762807905674,
256
+ "learning_rate": 9.770070340332457e-06,
257
+ "loss": 0.0544,
258
+ "step": 125,
259
+ "student_loss": 0.03289179503917694,
260
+ "teacher_loss": 0.00024821600527502596
261
+ },
262
+ {
263
+ "epoch": 0.03748558246828143,
264
+ "grad_norm": 1.4375,
265
+ "kd_loss": 0.020570307970046997,
266
+ "learning_rate": 9.745174586073982e-06,
267
+ "loss": 0.0531,
268
+ "step": 130,
269
+ "student_loss": 0.0006593615980818868,
270
+ "teacher_loss": 0.0006894106627441943
271
+ },
272
+ {
273
+ "epoch": 0.03892733564013841,
274
+ "grad_norm": 1.1953125,
275
+ "kd_loss": 0.03941934183239937,
276
+ "learning_rate": 9.719034492020183e-06,
277
+ "loss": 0.0584,
278
+ "step": 135,
279
+ "student_loss": 0.001388320466503501,
280
+ "teacher_loss": 0.00066422886447981
281
+ },
282
+ {
283
+ "epoch": 0.040369088811995385,
284
+ "grad_norm": 1.125,
285
+ "kd_loss": 0.023267682641744614,
286
+ "learning_rate": 9.691656912957686e-06,
287
+ "loss": 0.0554,
288
+ "step": 140,
289
+ "student_loss": 0.0006728945882059634,
290
+ "teacher_loss": 0.00048348584095947444
291
+ },
292
+ {
293
+ "epoch": 0.04181084198385236,
294
+ "grad_norm": 0.9140625,
295
+ "kd_loss": 0.06453745812177658,
296
+ "learning_rate": 9.663049028182112e-06,
297
+ "loss": 0.0511,
298
+ "step": 145,
299
+ "student_loss": 0.04263001307845116,
300
+ "teacher_loss": 0.006518054287880659
301
+ },
302
+ {
303
+ "epoch": 0.04325259515570934,
304
+ "grad_norm": 1.1640625,
305
+ "kd_loss": 0.014925969764590263,
306
+ "learning_rate": 9.633218339615433e-06,
307
+ "loss": 0.0518,
308
+ "step": 150,
309
+ "student_loss": 0.0011302934726700187,
310
+ "teacher_loss": 0.0012746027205139399
311
+ },
312
+ {
313
+ "epoch": 0.04469434832756632,
314
+ "grad_norm": 1.59375,
315
+ "kd_loss": 0.030081020668148994,
316
+ "learning_rate": 9.602172669838721e-06,
317
+ "loss": 0.0591,
318
+ "step": 155,
319
+ "student_loss": 0.0018543315818533301,
320
+ "teacher_loss": 0.0007238256512209773
321
+ },
322
+ {
323
+ "epoch": 0.0461361014994233,
324
+ "grad_norm": 2.15625,
325
+ "kd_loss": 0.026653118431568146,
326
+ "learning_rate": 9.569920160040815e-06,
327
+ "loss": 0.0574,
328
+ "step": 160,
329
+ "student_loss": 0.08247562497854233,
330
+ "teacher_loss": 0.04557507857680321
331
+ },
332
+ {
333
+ "epoch": 0.04757785467128028,
334
+ "grad_norm": 0.91015625,
335
+ "kd_loss": 0.030856940895318985,
336
+ "learning_rate": 9.536469267883432e-06,
337
+ "loss": 0.0493,
338
+ "step": 165,
339
+ "student_loss": 0.0005265406798571348,
340
+ "teacher_loss": 0.004277692176401615
341
+ },
342
+ {
343
+ "epoch": 0.049019607843137254,
344
+ "grad_norm": 2.875,
345
+ "kd_loss": 0.016948891803622246,
346
+ "learning_rate": 9.501828765283295e-06,
347
+ "loss": 0.0524,
348
+ "step": 170,
349
+ "student_loss": 0.0006478180293925107,
350
+ "teacher_loss": 0.000518353539519012
351
+ },
352
+ {
353
+ "epoch": 0.05046136101499423,
354
+ "grad_norm": 2.390625,
355
+ "kd_loss": 0.026514360681176186,
356
+ "learning_rate": 9.466007736111846e-06,
357
+ "loss": 0.0512,
358
+ "step": 175,
359
+ "student_loss": 0.005281130783259869,
360
+ "teacher_loss": 0.0008272691047750413
361
+ },
362
+ {
363
+ "epoch": 0.05190311418685121,
364
+ "grad_norm": 1.1015625,
365
+ "kd_loss": 0.036696940660476685,
366
+ "learning_rate": 9.429015573813163e-06,
367
+ "loss": 0.048,
368
+ "step": 180,
369
+ "student_loss": 0.0007505607791244984,
370
+ "teacher_loss": 0.0003059351583942771
371
+ },
372
+ {
373
+ "epoch": 0.05334486735870819,
374
+ "grad_norm": 1.421875,
375
+ "kd_loss": 0.035748232156038284,
376
+ "learning_rate": 9.390861978940687e-06,
377
+ "loss": 0.0538,
378
+ "step": 185,
379
+ "student_loss": 0.038561683148145676,
380
+ "teacher_loss": 0.00023784062068443745
381
+ },
382
+ {
383
+ "epoch": 0.05478662053056517,
384
+ "grad_norm": 2.171875,
385
+ "kd_loss": 0.051317859441041946,
386
+ "learning_rate": 9.351556956613423e-06,
387
+ "loss": 0.0494,
388
+ "step": 190,
389
+ "student_loss": 0.008348710834980011,
390
+ "teacher_loss": 0.0009534004493616521
391
+ },
392
+ {
393
+ "epoch": 0.056228373702422146,
394
+ "grad_norm": 3.3125,
395
+ "kd_loss": 0.04666966199874878,
396
+ "learning_rate": 9.31111081389227e-06,
397
+ "loss": 0.0497,
398
+ "step": 195,
399
+ "student_loss": 0.008153470233082771,
400
+ "teacher_loss": 0.0002795102773234248
401
+ },
402
+ {
403
+ "epoch": 0.05767012687427912,
404
+ "grad_norm": 2.328125,
405
+ "kd_loss": 0.03317176550626755,
406
+ "learning_rate": 9.269534157077177e-06,
407
+ "loss": 0.0596,
408
+ "step": 200,
409
+ "student_loss": 0.0004428599204402417,
410
+ "teacher_loss": 0.00018954814004246145
411
+ },
412
+ {
413
+ "epoch": 0.0591118800461361,
414
+ "grad_norm": 1.703125,
415
+ "kd_loss": 0.02631596103310585,
416
+ "learning_rate": 9.226837888925813e-06,
417
+ "loss": 0.0551,
418
+ "step": 205,
419
+ "student_loss": 0.03620046377182007,
420
+ "teacher_loss": 0.02737441286444664
421
+ },
422
+ {
423
+ "epoch": 0.06055363321799308,
424
+ "grad_norm": 1.9765625,
425
+ "kd_loss": 0.03267158567905426,
426
+ "learning_rate": 9.183033205794525e-06,
427
+ "loss": 0.0505,
428
+ "step": 210,
429
+ "student_loss": 0.0009676365298219025,
430
+ "teacher_loss": 0.0004164037818554789
431
+ },
432
+ {
433
+ "epoch": 0.061995386389850055,
434
+ "grad_norm": 1.4921875,
435
+ "kd_loss": 0.037410128861665726,
436
+ "learning_rate": 9.13813159470227e-06,
437
+ "loss": 0.0539,
438
+ "step": 215,
439
+ "student_loss": 0.00046113875578157604,
440
+ "teacher_loss": 0.00020874453184660524
441
+ },
442
+ {
443
+ "epoch": 0.06343713956170703,
444
+ "grad_norm": 1.6328125,
445
+ "kd_loss": 0.04395928606390953,
446
+ "learning_rate": 9.092144830318357e-06,
447
+ "loss": 0.0508,
448
+ "step": 220,
449
+ "student_loss": 0.030493613332509995,
450
+ "teacher_loss": 0.010191974230110645
451
+ },
452
+ {
453
+ "epoch": 0.06487889273356401,
454
+ "grad_norm": 1.359375,
455
+ "kd_loss": 0.022184843197464943,
456
+ "learning_rate": 9.045084971874738e-06,
457
+ "loss": 0.0603,
458
+ "step": 225,
459
+ "student_loss": 0.04274655878543854,
460
+ "teacher_loss": 0.001591266831383109
461
+ },
462
+ {
463
+ "epoch": 0.06632064590542099,
464
+ "grad_norm": 1.953125,
465
+ "kd_loss": 0.043670717626810074,
466
+ "learning_rate": 8.99696436000368e-06,
467
+ "loss": 0.0491,
468
+ "step": 230,
469
+ "student_loss": 0.022309448570013046,
470
+ "teacher_loss": 0.006051686126738787
471
+ },
472
+ {
473
+ "epoch": 0.06776239907727798,
474
+ "grad_norm": 1.3203125,
475
+ "kd_loss": 0.07893647253513336,
476
+ "learning_rate": 8.947795613501658e-06,
477
+ "loss": 0.0516,
478
+ "step": 235,
479
+ "student_loss": 0.012356958352029324,
480
+ "teacher_loss": 0.0008457938674837351
481
+ },
482
+ {
483
+ "epoch": 0.06920415224913495,
484
+ "grad_norm": 0.95703125,
485
+ "kd_loss": 0.030352523550391197,
486
+ "learning_rate": 8.897591626020284e-06,
487
+ "loss": 0.0553,
488
+ "step": 240,
489
+ "student_loss": 0.0036721325013786554,
490
+ "teacher_loss": 0.00047237545368261635
491
+ },
492
+ {
493
+ "epoch": 0.07064590542099193,
494
+ "grad_norm": 1.3125,
495
+ "kd_loss": 0.014727318659424782,
496
+ "learning_rate": 8.846365562685178e-06,
497
+ "loss": 0.0472,
498
+ "step": 245,
499
+ "student_loss": 0.002595798345282674,
500
+ "teacher_loss": 0.0030349844601005316
501
+ },
502
+ {
503
+ "epoch": 0.07208765859284891,
504
+ "grad_norm": 2.015625,
505
+ "kd_loss": 0.022538436576724052,
506
+ "learning_rate": 8.794130856643635e-06,
507
+ "loss": 0.0531,
508
+ "step": 250,
509
+ "student_loss": 0.0009441633592359722,
510
+ "teacher_loss": 0.0006299451342783868
511
+ },
512
+ {
513
+ "epoch": 0.07352941176470588,
514
+ "grad_norm": 1.3515625,
515
+ "kd_loss": 0.018597841262817383,
516
+ "learning_rate": 8.74090120554202e-06,
517
+ "loss": 0.0517,
518
+ "step": 255,
519
+ "student_loss": 0.0004831044061575085,
520
+ "teacher_loss": 0.0005120306159369648
521
+ },
522
+ {
523
+ "epoch": 0.07497116493656286,
524
+ "grad_norm": 1.65625,
525
+ "kd_loss": 0.027878236025571823,
526
+ "learning_rate": 8.686690567933803e-06,
527
+ "loss": 0.0558,
528
+ "step": 260,
529
+ "student_loss": 0.04166676476597786,
530
+ "teacher_loss": 0.032882433384656906
531
+ },
532
+ {
533
+ "epoch": 0.07641291810841984,
534
+ "grad_norm": 1.359375,
535
+ "kd_loss": 0.018291378393769264,
536
+ "learning_rate": 8.63151315961915e-06,
537
+ "loss": 0.0532,
538
+ "step": 265,
539
+ "student_loss": 0.0004272922524251044,
540
+ "teacher_loss": 0.000323377171298489
541
+ },
542
+ {
543
+ "epoch": 0.07785467128027682,
544
+ "grad_norm": 1.40625,
545
+ "kd_loss": 0.020099647343158722,
546
+ "learning_rate": 8.575383449917103e-06,
547
+ "loss": 0.0567,
548
+ "step": 270,
549
+ "student_loss": 0.001266423612833023,
550
+ "teacher_loss": 0.0014009763253852725
551
+ },
552
+ {
553
+ "epoch": 0.07929642445213379,
554
+ "grad_norm": 1.5390625,
555
+ "kd_loss": 0.022994602099061012,
556
+ "learning_rate": 8.518316157871232e-06,
557
+ "loss": 0.0493,
558
+ "step": 275,
559
+ "student_loss": 0.04728766158223152,
560
+ "teacher_loss": 0.034376151859760284
561
+ },
562
+ {
563
+ "epoch": 0.08073817762399077,
564
+ "grad_norm": 1.2421875,
565
+ "kd_loss": 0.03356819972395897,
566
+ "learning_rate": 8.460326248389825e-06,
567
+ "loss": 0.0483,
568
+ "step": 280,
569
+ "student_loss": 0.0006493565160781145,
570
+ "teacher_loss": 0.00017707925871945918
571
+ },
572
+ {
573
+ "epoch": 0.08217993079584775,
574
+ "grad_norm": 1.515625,
575
+ "kd_loss": 0.01917041279375553,
576
+ "learning_rate": 8.401428928321607e-06,
577
+ "loss": 0.0539,
578
+ "step": 285,
579
+ "student_loss": 0.0010354992700740695,
580
+ "teacher_loss": 0.0007419176981784403
581
+ },
582
+ {
583
+ "epoch": 0.08362168396770472,
584
+ "grad_norm": 1.4140625,
585
+ "kd_loss": 0.014347212389111519,
586
+ "learning_rate": 8.341639642468002e-06,
587
+ "loss": 0.0547,
588
+ "step": 290,
589
+ "student_loss": 0.01613372191786766,
590
+ "teacher_loss": 0.018294816836714745
591
+ },
592
+ {
593
+ "epoch": 0.0850634371395617,
594
+ "grad_norm": 1.7265625,
595
+ "kd_loss": 0.0206157136708498,
596
+ "learning_rate": 8.280974069532999e-06,
597
+ "loss": 0.0525,
598
+ "step": 295,
599
+ "student_loss": 0.0008157998672686517,
600
+ "teacher_loss": 0.0007138358778320253
601
+ },
602
+ {
603
+ "epoch": 0.08650519031141868,
604
+ "grad_norm": 1.3984375,
605
+ "kd_loss": 0.04587658867239952,
606
+ "learning_rate": 8.219448118011687e-06,
607
+ "loss": 0.0496,
608
+ "step": 300,
609
+ "student_loss": 0.015501612797379494,
610
+ "teacher_loss": 0.0018786232685670257
611
+ },
612
+ {
613
+ "epoch": 0.08794694348327567,
614
+ "grad_norm": 2.203125,
615
+ "kd_loss": 0.02352512627840042,
616
+ "learning_rate": 8.157077922018537e-06,
617
+ "loss": 0.0475,
618
+ "step": 305,
619
+ "student_loss": 0.008210180327296257,
620
+ "teacher_loss": 0.0028180002700537443
621
+ },
622
+ {
623
+ "epoch": 0.08938869665513265,
624
+ "grad_norm": 1.6484375,
625
+ "kd_loss": 0.01950286515057087,
626
+ "learning_rate": 8.093879837056486e-06,
627
+ "loss": 0.05,
628
+ "step": 310,
629
+ "student_loss": 0.0005316836177371442,
630
+ "teacher_loss": 0.00044056857586838305
631
+ },
632
+ {
633
+ "epoch": 0.09083044982698962,
634
+ "grad_norm": 1.9609375,
635
+ "kd_loss": 0.04243648052215576,
636
+ "learning_rate": 8.029870435728018e-06,
637
+ "loss": 0.049,
638
+ "step": 315,
639
+ "student_loss": 0.01187934447079897,
640
+ "teacher_loss": 0.0004662107094191015
641
+ },
642
+ {
643
+ "epoch": 0.0922722029988466,
644
+ "grad_norm": 1.7734375,
645
+ "kd_loss": 0.022677697241306305,
646
+ "learning_rate": 7.965066503389264e-06,
647
+ "loss": 0.0501,
648
+ "step": 320,
649
+ "student_loss": 0.0005311672575771809,
650
+ "teacher_loss": 0.0002600560837890953
651
+ },
652
+ {
653
+ "epoch": 0.09371395617070358,
654
+ "grad_norm": 1.03125,
655
+ "kd_loss": 0.06592908501625061,
656
+ "learning_rate": 7.89948503374835e-06,
657
+ "loss": 0.0465,
658
+ "step": 325,
659
+ "student_loss": 0.005857955198734999,
660
+ "teacher_loss": 0.02006148360669613
661
+ },
662
+ {
663
+ "epoch": 0.09515570934256055,
664
+ "grad_norm": 1.3125,
665
+ "kd_loss": 0.04595617577433586,
666
+ "learning_rate": 7.833143224409076e-06,
667
+ "loss": 0.0501,
668
+ "step": 330,
669
+ "student_loss": 0.01813417300581932,
670
+ "teacher_loss": 0.0033948116470128298
671
+ },
672
+ {
673
+ "epoch": 0.09659746251441753,
674
+ "grad_norm": 1.390625,
675
+ "kd_loss": 0.012639202177524567,
676
+ "learning_rate": 7.766058472361154e-06,
677
+ "loss": 0.0461,
678
+ "step": 335,
679
+ "student_loss": 0.0008777976036071777,
680
+ "teacher_loss": 0.0008721367339603603
681
+ },
682
+ {
683
+ "epoch": 0.09803921568627451,
684
+ "grad_norm": 1.0625,
685
+ "kd_loss": 0.026686903089284897,
686
+ "learning_rate": 7.698248369418146e-06,
687
+ "loss": 0.0515,
688
+ "step": 340,
689
+ "student_loss": 0.010171014815568924,
690
+ "teacher_loss": 0.007541630417108536
691
+ },
692
+ {
693
+ "epoch": 0.09948096885813149,
694
+ "grad_norm": 1.3203125,
695
+ "kd_loss": 0.03351679816842079,
696
+ "learning_rate": 7.629730697604314e-06,
697
+ "loss": 0.0516,
698
+ "step": 345,
699
+ "student_loss": 0.008214120753109455,
700
+ "teacher_loss": 0.0004564044938888401
701
+ },
702
+ {
703
+ "epoch": 0.10092272202998846,
704
+ "grad_norm": 1.71875,
705
+ "kd_loss": 0.05568505823612213,
706
+ "learning_rate": 7.560523424491595e-06,
707
+ "loss": 0.0457,
708
+ "step": 350,
709
+ "student_loss": 0.010019722394645214,
710
+ "teacher_loss": 0.003242993028834462
711
+ },
712
+ {
713
+ "epoch": 0.10236447520184544,
714
+ "grad_norm": 1.8515625,
715
+ "kd_loss": 0.016680920496582985,
716
+ "learning_rate": 7.490644698487909e-06,
717
+ "loss": 0.0531,
718
+ "step": 355,
719
+ "student_loss": 0.0006857452099211514,
720
+ "teacher_loss": 0.0006745306309312582
721
+ },
722
+ {
723
+ "epoch": 0.10380622837370242,
724
+ "grad_norm": 1.2109375,
725
+ "kd_loss": 0.01588292233645916,
726
+ "learning_rate": 7.420112844078066e-06,
727
+ "loss": 0.0509,
728
+ "step": 360,
729
+ "student_loss": 0.001193460077047348,
730
+ "teacher_loss": 0.001017728471197188
731
+ },
732
+ {
733
+ "epoch": 0.1052479815455594,
734
+ "grad_norm": 2.0625,
735
+ "kd_loss": 0.02928476221859455,
736
+ "learning_rate": 7.348946357018479e-06,
737
+ "loss": 0.053,
738
+ "step": 365,
739
+ "student_loss": 0.043851789087057114,
740
+ "teacher_loss": 0.0283951573073864
741
+ },
742
+ {
743
+ "epoch": 0.10668973471741638,
744
+ "grad_norm": 1.21875,
745
+ "kd_loss": 0.04628458246588707,
746
+ "learning_rate": 7.277163899486975e-06,
747
+ "loss": 0.0663,
748
+ "step": 370,
749
+ "student_loss": 0.023408163338899612,
750
+ "teacher_loss": 0.0004174449131824076
751
+ },
752
+ {
753
+ "epoch": 0.10813148788927336,
754
+ "grad_norm": 1.71875,
755
+ "kd_loss": 0.03154220059514046,
756
+ "learning_rate": 7.204784295188959e-06,
757
+ "loss": 0.05,
758
+ "step": 375,
759
+ "student_loss": 0.0324401780962944,
760
+ "teacher_loss": 0.020961124449968338
761
+ },
762
+ {
763
+ "epoch": 0.10957324106113034,
764
+ "grad_norm": 1.046875,
765
+ "kd_loss": 0.01982559822499752,
766
+ "learning_rate": 7.1318265244212305e-06,
767
+ "loss": 0.0493,
768
+ "step": 380,
769
+ "student_loss": 0.0006248670979402959,
770
+ "teacher_loss": 0.000449157232651487
771
+ },
772
+ {
773
+ "epoch": 0.11101499423298732,
774
+ "grad_norm": 1.390625,
775
+ "kd_loss": 0.0306891817599535,
776
+ "learning_rate": 7.05830971909472e-06,
777
+ "loss": 0.0447,
778
+ "step": 385,
779
+ "student_loss": 0.005440755281597376,
780
+ "teacher_loss": 0.00408168276771903
781
+ },
782
+ {
783
+ "epoch": 0.11245674740484429,
784
+ "grad_norm": 1.71875,
785
+ "kd_loss": 0.023891452699899673,
786
+ "learning_rate": 6.9842531577174865e-06,
787
+ "loss": 0.0458,
788
+ "step": 390,
789
+ "student_loss": 0.0003995662846136838,
790
+ "teacher_loss": 0.00039677225868217647
791
+ },
792
+ {
793
+ "epoch": 0.11389850057670127,
794
+ "grad_norm": 1.1328125,
795
+ "kd_loss": 0.018030904233455658,
796
+ "learning_rate": 6.9096762603392595e-06,
797
+ "loss": 0.0489,
798
+ "step": 395,
799
+ "student_loss": 0.0008614315302111208,
800
+ "teacher_loss": 0.0006612912402488291
801
+ },
802
+ {
803
+ "epoch": 0.11534025374855825,
804
+ "grad_norm": 1.703125,
805
+ "kd_loss": 0.02574356086552143,
806
+ "learning_rate": 6.834598583458862e-06,
807
+ "loss": 0.0432,
808
+ "step": 400,
809
+ "student_loss": 0.0009613529546186328,
810
+ "teacher_loss": 0.00023952587798703462
811
+ }
812
+ ],
813
+ "logging_steps": 5,
814
+ "max_steps": 1000,
815
+ "num_input_tokens_seen": 0,
816
+ "num_train_epochs": 1,
817
+ "save_steps": 200,
818
+ "stateful_callbacks": {
819
+ "TrainerControl": {
820
+ "args": {
821
+ "should_epoch_stop": false,
822
+ "should_evaluate": false,
823
+ "should_log": false,
824
+ "should_save": true,
825
+ "should_training_stop": false
826
+ },
827
+ "attributes": {}
828
+ }
829
+ },
830
+ "total_flos": 0.0,
831
+ "train_batch_size": 1,
832
+ "trial_name": null,
833
+ "trial_params": null
834
+ }
checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-400/vocab.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-600/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_ls2/checkpoint-600/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_ls2/checkpoint-600/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_ls2/checkpoint-600/merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-600/pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d9c1e32964034f19d2eea6200ecfac0e046141e729303cc2d10efaf8c2be33cd
3
+ size 6187858991
checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-600/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_ls2/checkpoint-600/thought_projector.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7edc988914cb4cd8bc98f5f4784f82b2b3856fbce315fda7cec1535b925fe3c8
3
+ size 16788033
checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-600/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_ls2/checkpoint-600/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_ls2/checkpoint-600/trainer_state.json ADDED
@@ -0,0 +1,1234 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 0.17301038062283736,
6
+ "eval_steps": 500,
7
+ "global_step": 600,
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": 77.0,
15
+ "kd_loss": 0.2900531589984894,
16
+ "learning_rate": 1.3333333333333334e-06,
17
+ "loss": 0.5715,
18
+ "step": 5,
19
+ "student_loss": 0.11784028261899948,
20
+ "teacher_loss": 0.001998942345380783
21
+ },
22
+ {
23
+ "epoch": 0.0028835063437139563,
24
+ "grad_norm": 26.875,
25
+ "kd_loss": 0.277434378862381,
26
+ "learning_rate": 3e-06,
27
+ "loss": 0.4692,
28
+ "step": 10,
29
+ "student_loss": 0.11281580477952957,
30
+ "teacher_loss": 0.005315451417118311
31
+ },
32
+ {
33
+ "epoch": 0.004325259515570935,
34
+ "grad_norm": 18.5,
35
+ "kd_loss": 0.21174456179141998,
36
+ "learning_rate": 4.666666666666667e-06,
37
+ "loss": 0.3373,
38
+ "step": 15,
39
+ "student_loss": 0.007377153728157282,
40
+ "teacher_loss": 0.0011863313848152757
41
+ },
42
+ {
43
+ "epoch": 0.0057670126874279125,
44
+ "grad_norm": 8.0625,
45
+ "kd_loss": 0.13275651633739471,
46
+ "learning_rate": 6.333333333333333e-06,
47
+ "loss": 0.2555,
48
+ "step": 20,
49
+ "student_loss": 0.04130968824028969,
50
+ "teacher_loss": 0.0005545365856960416
51
+ },
52
+ {
53
+ "epoch": 0.00720876585928489,
54
+ "grad_norm": 4.6875,
55
+ "kd_loss": 0.10908422619104385,
56
+ "learning_rate": 8.000000000000001e-06,
57
+ "loss": 0.2065,
58
+ "step": 25,
59
+ "student_loss": 0.0035145599395036697,
60
+ "teacher_loss": 0.0009682206436991692
61
+ },
62
+ {
63
+ "epoch": 0.00865051903114187,
64
+ "grad_norm": 2.328125,
65
+ "kd_loss": 0.06466298550367355,
66
+ "learning_rate": 9.666666666666667e-06,
67
+ "loss": 0.1499,
68
+ "step": 30,
69
+ "student_loss": 0.0010085003450512886,
70
+ "teacher_loss": 0.0005369208520278335
71
+ },
72
+ {
73
+ "epoch": 0.010092272202998846,
74
+ "grad_norm": 2.609375,
75
+ "kd_loss": 0.05857780575752258,
76
+ "learning_rate": 9.99958042442916e-06,
77
+ "loss": 0.1223,
78
+ "step": 35,
79
+ "student_loss": 0.0010954203316941857,
80
+ "teacher_loss": 0.0006188539555296302
81
+ },
82
+ {
83
+ "epoch": 0.011534025374855825,
84
+ "grad_norm": 1.8359375,
85
+ "kd_loss": 0.057602375745773315,
86
+ "learning_rate": 9.997876019358083e-06,
87
+ "loss": 0.1064,
88
+ "step": 40,
89
+ "student_loss": 0.032847270369529724,
90
+ "teacher_loss": 0.028142843395471573
91
+ },
92
+ {
93
+ "epoch": 0.012975778546712802,
94
+ "grad_norm": 1.578125,
95
+ "kd_loss": 0.08787942677736282,
96
+ "learning_rate": 9.99486100792044e-06,
97
+ "loss": 0.0811,
98
+ "step": 45,
99
+ "student_loss": 0.07371404767036438,
100
+ "teacher_loss": 0.002992093563079834
101
+ },
102
+ {
103
+ "epoch": 0.01441753171856978,
104
+ "grad_norm": 2.078125,
105
+ "kd_loss": 0.037376269698143005,
106
+ "learning_rate": 9.990536180750724e-06,
107
+ "loss": 0.0866,
108
+ "step": 50,
109
+ "student_loss": 0.027633950114250183,
110
+ "teacher_loss": 0.009523550048470497
111
+ },
112
+ {
113
+ "epoch": 0.015859284890426758,
114
+ "grad_norm": 1.8203125,
115
+ "kd_loss": 0.029546089470386505,
116
+ "learning_rate": 9.984902671959911e-06,
117
+ "loss": 0.0757,
118
+ "step": 55,
119
+ "student_loss": 0.0010951841250061989,
120
+ "teacher_loss": 0.0006982200429774821
121
+ },
122
+ {
123
+ "epoch": 0.01730103806228374,
124
+ "grad_norm": 2.21875,
125
+ "kd_loss": 0.029508251696825027,
126
+ "learning_rate": 9.97796195883804e-06,
127
+ "loss": 0.0683,
128
+ "step": 60,
129
+ "student_loss": 0.002835189923644066,
130
+ "teacher_loss": 0.001713551115244627
131
+ },
132
+ {
133
+ "epoch": 0.018742791234140715,
134
+ "grad_norm": 2.234375,
135
+ "kd_loss": 0.030333109200000763,
136
+ "learning_rate": 9.969715861466839e-06,
137
+ "loss": 0.0686,
138
+ "step": 65,
139
+ "student_loss": 0.03012149967253208,
140
+ "teacher_loss": 0.03870478644967079
141
+ },
142
+ {
143
+ "epoch": 0.020184544405997693,
144
+ "grad_norm": 1.265625,
145
+ "kd_loss": 0.05904582515358925,
146
+ "learning_rate": 9.96016654224243e-06,
147
+ "loss": 0.0737,
148
+ "step": 70,
149
+ "student_loss": 0.01526976004242897,
150
+ "teacher_loss": 0.0018924300093203783
151
+ },
152
+ {
153
+ "epoch": 0.02162629757785467,
154
+ "grad_norm": 1.7578125,
155
+ "kd_loss": 0.024645471945405006,
156
+ "learning_rate": 9.94931650530827e-06,
157
+ "loss": 0.0659,
158
+ "step": 75,
159
+ "student_loss": 0.002360731828957796,
160
+ "teacher_loss": 0.0017862127861008048
161
+ },
162
+ {
163
+ "epoch": 0.02306805074971165,
164
+ "grad_norm": 0.96484375,
165
+ "kd_loss": 0.02402629889547825,
166
+ "learning_rate": 9.93716859589851e-06,
167
+ "loss": 0.0587,
168
+ "step": 80,
169
+ "student_loss": 0.0011838233331218362,
170
+ "teacher_loss": 0.0007806048961356282
171
+ },
172
+ {
173
+ "epoch": 0.024509803921568627,
174
+ "grad_norm": 1.09375,
175
+ "kd_loss": 0.025694765150547028,
176
+ "learning_rate": 9.923725999591846e-06,
177
+ "loss": 0.061,
178
+ "step": 85,
179
+ "student_loss": 0.00045988388592377305,
180
+ "teacher_loss": 0.0004276020044926554
181
+ },
182
+ {
183
+ "epoch": 0.025951557093425604,
184
+ "grad_norm": 1.2421875,
185
+ "kd_loss": 0.0665941834449768,
186
+ "learning_rate": 9.908992241476189e-06,
187
+ "loss": 0.062,
188
+ "step": 90,
189
+ "student_loss": 0.018955115228891373,
190
+ "teacher_loss": 0.003640417940914631
191
+ },
192
+ {
193
+ "epoch": 0.027393310265282585,
194
+ "grad_norm": 1.625,
195
+ "kd_loss": 0.03899763152003288,
196
+ "learning_rate": 9.892971185224244e-06,
197
+ "loss": 0.0583,
198
+ "step": 95,
199
+ "student_loss": 0.008055699057877064,
200
+ "teacher_loss": 0.005101984832435846
201
+ },
202
+ {
203
+ "epoch": 0.02883506343713956,
204
+ "grad_norm": 1.3828125,
205
+ "kd_loss": 0.022396311163902283,
206
+ "learning_rate": 9.875667032080354e-06,
207
+ "loss": 0.0519,
208
+ "step": 100,
209
+ "student_loss": 0.0006799453985877335,
210
+ "teacher_loss": 0.0005385762779042125
211
+ },
212
+ {
213
+ "epoch": 0.03027681660899654,
214
+ "grad_norm": 1.609375,
215
+ "kd_loss": 0.0338844433426857,
216
+ "learning_rate": 9.857084319758772e-06,
217
+ "loss": 0.055,
218
+ "step": 105,
219
+ "student_loss": 0.004358346574008465,
220
+ "teacher_loss": 0.0003296585055068135
221
+ },
222
+ {
223
+ "epoch": 0.031718569780853516,
224
+ "grad_norm": 1.234375,
225
+ "kd_loss": 0.033958468586206436,
226
+ "learning_rate": 9.837227921253747e-06,
227
+ "loss": 0.0588,
228
+ "step": 110,
229
+ "student_loss": 0.031565263867378235,
230
+ "teacher_loss": 0.01906467415392399
231
+ },
232
+ {
233
+ "epoch": 0.03316032295271049,
234
+ "grad_norm": 1.4765625,
235
+ "kd_loss": 0.036942481994628906,
236
+ "learning_rate": 9.816103043561648e-06,
237
+ "loss": 0.0579,
238
+ "step": 115,
239
+ "student_loss": 0.018612468615174294,
240
+ "teacher_loss": 0.012586154974997044
241
+ },
242
+ {
243
+ "epoch": 0.03460207612456748,
244
+ "grad_norm": 1.359375,
245
+ "kd_loss": 0.02566700056195259,
246
+ "learning_rate": 9.79371522631553e-06,
247
+ "loss": 0.0494,
248
+ "step": 120,
249
+ "student_loss": 0.001015476998873055,
250
+ "teacher_loss": 0.0008012109901756048
251
+ },
252
+ {
253
+ "epoch": 0.036043829296424454,
254
+ "grad_norm": 1.234375,
255
+ "kd_loss": 0.0527762807905674,
256
+ "learning_rate": 9.770070340332457e-06,
257
+ "loss": 0.0544,
258
+ "step": 125,
259
+ "student_loss": 0.03289179503917694,
260
+ "teacher_loss": 0.00024821600527502596
261
+ },
262
+ {
263
+ "epoch": 0.03748558246828143,
264
+ "grad_norm": 1.4375,
265
+ "kd_loss": 0.020570307970046997,
266
+ "learning_rate": 9.745174586073982e-06,
267
+ "loss": 0.0531,
268
+ "step": 130,
269
+ "student_loss": 0.0006593615980818868,
270
+ "teacher_loss": 0.0006894106627441943
271
+ },
272
+ {
273
+ "epoch": 0.03892733564013841,
274
+ "grad_norm": 1.1953125,
275
+ "kd_loss": 0.03941934183239937,
276
+ "learning_rate": 9.719034492020183e-06,
277
+ "loss": 0.0584,
278
+ "step": 135,
279
+ "student_loss": 0.001388320466503501,
280
+ "teacher_loss": 0.00066422886447981
281
+ },
282
+ {
283
+ "epoch": 0.040369088811995385,
284
+ "grad_norm": 1.125,
285
+ "kd_loss": 0.023267682641744614,
286
+ "learning_rate": 9.691656912957686e-06,
287
+ "loss": 0.0554,
288
+ "step": 140,
289
+ "student_loss": 0.0006728945882059634,
290
+ "teacher_loss": 0.00048348584095947444
291
+ },
292
+ {
293
+ "epoch": 0.04181084198385236,
294
+ "grad_norm": 0.9140625,
295
+ "kd_loss": 0.06453745812177658,
296
+ "learning_rate": 9.663049028182112e-06,
297
+ "loss": 0.0511,
298
+ "step": 145,
299
+ "student_loss": 0.04263001307845116,
300
+ "teacher_loss": 0.006518054287880659
301
+ },
302
+ {
303
+ "epoch": 0.04325259515570934,
304
+ "grad_norm": 1.1640625,
305
+ "kd_loss": 0.014925969764590263,
306
+ "learning_rate": 9.633218339615433e-06,
307
+ "loss": 0.0518,
308
+ "step": 150,
309
+ "student_loss": 0.0011302934726700187,
310
+ "teacher_loss": 0.0012746027205139399
311
+ },
312
+ {
313
+ "epoch": 0.04469434832756632,
314
+ "grad_norm": 1.59375,
315
+ "kd_loss": 0.030081020668148994,
316
+ "learning_rate": 9.602172669838721e-06,
317
+ "loss": 0.0591,
318
+ "step": 155,
319
+ "student_loss": 0.0018543315818533301,
320
+ "teacher_loss": 0.0007238256512209773
321
+ },
322
+ {
323
+ "epoch": 0.0461361014994233,
324
+ "grad_norm": 2.15625,
325
+ "kd_loss": 0.026653118431568146,
326
+ "learning_rate": 9.569920160040815e-06,
327
+ "loss": 0.0574,
328
+ "step": 160,
329
+ "student_loss": 0.08247562497854233,
330
+ "teacher_loss": 0.04557507857680321
331
+ },
332
+ {
333
+ "epoch": 0.04757785467128028,
334
+ "grad_norm": 0.91015625,
335
+ "kd_loss": 0.030856940895318985,
336
+ "learning_rate": 9.536469267883432e-06,
337
+ "loss": 0.0493,
338
+ "step": 165,
339
+ "student_loss": 0.0005265406798571348,
340
+ "teacher_loss": 0.004277692176401615
341
+ },
342
+ {
343
+ "epoch": 0.049019607843137254,
344
+ "grad_norm": 2.875,
345
+ "kd_loss": 0.016948891803622246,
346
+ "learning_rate": 9.501828765283295e-06,
347
+ "loss": 0.0524,
348
+ "step": 170,
349
+ "student_loss": 0.0006478180293925107,
350
+ "teacher_loss": 0.000518353539519012
351
+ },
352
+ {
353
+ "epoch": 0.05046136101499423,
354
+ "grad_norm": 2.390625,
355
+ "kd_loss": 0.026514360681176186,
356
+ "learning_rate": 9.466007736111846e-06,
357
+ "loss": 0.0512,
358
+ "step": 175,
359
+ "student_loss": 0.005281130783259869,
360
+ "teacher_loss": 0.0008272691047750413
361
+ },
362
+ {
363
+ "epoch": 0.05190311418685121,
364
+ "grad_norm": 1.1015625,
365
+ "kd_loss": 0.036696940660476685,
366
+ "learning_rate": 9.429015573813163e-06,
367
+ "loss": 0.048,
368
+ "step": 180,
369
+ "student_loss": 0.0007505607791244984,
370
+ "teacher_loss": 0.0003059351583942771
371
+ },
372
+ {
373
+ "epoch": 0.05334486735870819,
374
+ "grad_norm": 1.421875,
375
+ "kd_loss": 0.035748232156038284,
376
+ "learning_rate": 9.390861978940687e-06,
377
+ "loss": 0.0538,
378
+ "step": 185,
379
+ "student_loss": 0.038561683148145676,
380
+ "teacher_loss": 0.00023784062068443745
381
+ },
382
+ {
383
+ "epoch": 0.05478662053056517,
384
+ "grad_norm": 2.171875,
385
+ "kd_loss": 0.051317859441041946,
386
+ "learning_rate": 9.351556956613423e-06,
387
+ "loss": 0.0494,
388
+ "step": 190,
389
+ "student_loss": 0.008348710834980011,
390
+ "teacher_loss": 0.0009534004493616521
391
+ },
392
+ {
393
+ "epoch": 0.056228373702422146,
394
+ "grad_norm": 3.3125,
395
+ "kd_loss": 0.04666966199874878,
396
+ "learning_rate": 9.31111081389227e-06,
397
+ "loss": 0.0497,
398
+ "step": 195,
399
+ "student_loss": 0.008153470233082771,
400
+ "teacher_loss": 0.0002795102773234248
401
+ },
402
+ {
403
+ "epoch": 0.05767012687427912,
404
+ "grad_norm": 2.328125,
405
+ "kd_loss": 0.03317176550626755,
406
+ "learning_rate": 9.269534157077177e-06,
407
+ "loss": 0.0596,
408
+ "step": 200,
409
+ "student_loss": 0.0004428599204402417,
410
+ "teacher_loss": 0.00018954814004246145
411
+ },
412
+ {
413
+ "epoch": 0.0591118800461361,
414
+ "grad_norm": 1.703125,
415
+ "kd_loss": 0.02631596103310585,
416
+ "learning_rate": 9.226837888925813e-06,
417
+ "loss": 0.0551,
418
+ "step": 205,
419
+ "student_loss": 0.03620046377182007,
420
+ "teacher_loss": 0.02737441286444664
421
+ },
422
+ {
423
+ "epoch": 0.06055363321799308,
424
+ "grad_norm": 1.9765625,
425
+ "kd_loss": 0.03267158567905426,
426
+ "learning_rate": 9.183033205794525e-06,
427
+ "loss": 0.0505,
428
+ "step": 210,
429
+ "student_loss": 0.0009676365298219025,
430
+ "teacher_loss": 0.0004164037818554789
431
+ },
432
+ {
433
+ "epoch": 0.061995386389850055,
434
+ "grad_norm": 1.4921875,
435
+ "kd_loss": 0.037410128861665726,
436
+ "learning_rate": 9.13813159470227e-06,
437
+ "loss": 0.0539,
438
+ "step": 215,
439
+ "student_loss": 0.00046113875578157604,
440
+ "teacher_loss": 0.00020874453184660524
441
+ },
442
+ {
443
+ "epoch": 0.06343713956170703,
444
+ "grad_norm": 1.6328125,
445
+ "kd_loss": 0.04395928606390953,
446
+ "learning_rate": 9.092144830318357e-06,
447
+ "loss": 0.0508,
448
+ "step": 220,
449
+ "student_loss": 0.030493613332509995,
450
+ "teacher_loss": 0.010191974230110645
451
+ },
452
+ {
453
+ "epoch": 0.06487889273356401,
454
+ "grad_norm": 1.359375,
455
+ "kd_loss": 0.022184843197464943,
456
+ "learning_rate": 9.045084971874738e-06,
457
+ "loss": 0.0603,
458
+ "step": 225,
459
+ "student_loss": 0.04274655878543854,
460
+ "teacher_loss": 0.001591266831383109
461
+ },
462
+ {
463
+ "epoch": 0.06632064590542099,
464
+ "grad_norm": 1.953125,
465
+ "kd_loss": 0.043670717626810074,
466
+ "learning_rate": 8.99696436000368e-06,
467
+ "loss": 0.0491,
468
+ "step": 230,
469
+ "student_loss": 0.022309448570013046,
470
+ "teacher_loss": 0.006051686126738787
471
+ },
472
+ {
473
+ "epoch": 0.06776239907727798,
474
+ "grad_norm": 1.3203125,
475
+ "kd_loss": 0.07893647253513336,
476
+ "learning_rate": 8.947795613501658e-06,
477
+ "loss": 0.0516,
478
+ "step": 235,
479
+ "student_loss": 0.012356958352029324,
480
+ "teacher_loss": 0.0008457938674837351
481
+ },
482
+ {
483
+ "epoch": 0.06920415224913495,
484
+ "grad_norm": 0.95703125,
485
+ "kd_loss": 0.030352523550391197,
486
+ "learning_rate": 8.897591626020284e-06,
487
+ "loss": 0.0553,
488
+ "step": 240,
489
+ "student_loss": 0.0036721325013786554,
490
+ "teacher_loss": 0.00047237545368261635
491
+ },
492
+ {
493
+ "epoch": 0.07064590542099193,
494
+ "grad_norm": 1.3125,
495
+ "kd_loss": 0.014727318659424782,
496
+ "learning_rate": 8.846365562685178e-06,
497
+ "loss": 0.0472,
498
+ "step": 245,
499
+ "student_loss": 0.002595798345282674,
500
+ "teacher_loss": 0.0030349844601005316
501
+ },
502
+ {
503
+ "epoch": 0.07208765859284891,
504
+ "grad_norm": 2.015625,
505
+ "kd_loss": 0.022538436576724052,
506
+ "learning_rate": 8.794130856643635e-06,
507
+ "loss": 0.0531,
508
+ "step": 250,
509
+ "student_loss": 0.0009441633592359722,
510
+ "teacher_loss": 0.0006299451342783868
511
+ },
512
+ {
513
+ "epoch": 0.07352941176470588,
514
+ "grad_norm": 1.3515625,
515
+ "kd_loss": 0.018597841262817383,
516
+ "learning_rate": 8.74090120554202e-06,
517
+ "loss": 0.0517,
518
+ "step": 255,
519
+ "student_loss": 0.0004831044061575085,
520
+ "teacher_loss": 0.0005120306159369648
521
+ },
522
+ {
523
+ "epoch": 0.07497116493656286,
524
+ "grad_norm": 1.65625,
525
+ "kd_loss": 0.027878236025571823,
526
+ "learning_rate": 8.686690567933803e-06,
527
+ "loss": 0.0558,
528
+ "step": 260,
529
+ "student_loss": 0.04166676476597786,
530
+ "teacher_loss": 0.032882433384656906
531
+ },
532
+ {
533
+ "epoch": 0.07641291810841984,
534
+ "grad_norm": 1.359375,
535
+ "kd_loss": 0.018291378393769264,
536
+ "learning_rate": 8.63151315961915e-06,
537
+ "loss": 0.0532,
538
+ "step": 265,
539
+ "student_loss": 0.0004272922524251044,
540
+ "teacher_loss": 0.000323377171298489
541
+ },
542
+ {
543
+ "epoch": 0.07785467128027682,
544
+ "grad_norm": 1.40625,
545
+ "kd_loss": 0.020099647343158722,
546
+ "learning_rate": 8.575383449917103e-06,
547
+ "loss": 0.0567,
548
+ "step": 270,
549
+ "student_loss": 0.001266423612833023,
550
+ "teacher_loss": 0.0014009763253852725
551
+ },
552
+ {
553
+ "epoch": 0.07929642445213379,
554
+ "grad_norm": 1.5390625,
555
+ "kd_loss": 0.022994602099061012,
556
+ "learning_rate": 8.518316157871232e-06,
557
+ "loss": 0.0493,
558
+ "step": 275,
559
+ "student_loss": 0.04728766158223152,
560
+ "teacher_loss": 0.034376151859760284
561
+ },
562
+ {
563
+ "epoch": 0.08073817762399077,
564
+ "grad_norm": 1.2421875,
565
+ "kd_loss": 0.03356819972395897,
566
+ "learning_rate": 8.460326248389825e-06,
567
+ "loss": 0.0483,
568
+ "step": 280,
569
+ "student_loss": 0.0006493565160781145,
570
+ "teacher_loss": 0.00017707925871945918
571
+ },
572
+ {
573
+ "epoch": 0.08217993079584775,
574
+ "grad_norm": 1.515625,
575
+ "kd_loss": 0.01917041279375553,
576
+ "learning_rate": 8.401428928321607e-06,
577
+ "loss": 0.0539,
578
+ "step": 285,
579
+ "student_loss": 0.0010354992700740695,
580
+ "teacher_loss": 0.0007419176981784403
581
+ },
582
+ {
583
+ "epoch": 0.08362168396770472,
584
+ "grad_norm": 1.4140625,
585
+ "kd_loss": 0.014347212389111519,
586
+ "learning_rate": 8.341639642468002e-06,
587
+ "loss": 0.0547,
588
+ "step": 290,
589
+ "student_loss": 0.01613372191786766,
590
+ "teacher_loss": 0.018294816836714745
591
+ },
592
+ {
593
+ "epoch": 0.0850634371395617,
594
+ "grad_norm": 1.7265625,
595
+ "kd_loss": 0.0206157136708498,
596
+ "learning_rate": 8.280974069532999e-06,
597
+ "loss": 0.0525,
598
+ "step": 295,
599
+ "student_loss": 0.0008157998672686517,
600
+ "teacher_loss": 0.0007138358778320253
601
+ },
602
+ {
603
+ "epoch": 0.08650519031141868,
604
+ "grad_norm": 1.3984375,
605
+ "kd_loss": 0.04587658867239952,
606
+ "learning_rate": 8.219448118011687e-06,
607
+ "loss": 0.0496,
608
+ "step": 300,
609
+ "student_loss": 0.015501612797379494,
610
+ "teacher_loss": 0.0018786232685670257
611
+ },
612
+ {
613
+ "epoch": 0.08794694348327567,
614
+ "grad_norm": 2.203125,
615
+ "kd_loss": 0.02352512627840042,
616
+ "learning_rate": 8.157077922018537e-06,
617
+ "loss": 0.0475,
618
+ "step": 305,
619
+ "student_loss": 0.008210180327296257,
620
+ "teacher_loss": 0.0028180002700537443
621
+ },
622
+ {
623
+ "epoch": 0.08938869665513265,
624
+ "grad_norm": 1.6484375,
625
+ "kd_loss": 0.01950286515057087,
626
+ "learning_rate": 8.093879837056486e-06,
627
+ "loss": 0.05,
628
+ "step": 310,
629
+ "student_loss": 0.0005316836177371442,
630
+ "teacher_loss": 0.00044056857586838305
631
+ },
632
+ {
633
+ "epoch": 0.09083044982698962,
634
+ "grad_norm": 1.9609375,
635
+ "kd_loss": 0.04243648052215576,
636
+ "learning_rate": 8.029870435728018e-06,
637
+ "loss": 0.049,
638
+ "step": 315,
639
+ "student_loss": 0.01187934447079897,
640
+ "teacher_loss": 0.0004662107094191015
641
+ },
642
+ {
643
+ "epoch": 0.0922722029988466,
644
+ "grad_norm": 1.7734375,
645
+ "kd_loss": 0.022677697241306305,
646
+ "learning_rate": 7.965066503389264e-06,
647
+ "loss": 0.0501,
648
+ "step": 320,
649
+ "student_loss": 0.0005311672575771809,
650
+ "teacher_loss": 0.0002600560837890953
651
+ },
652
+ {
653
+ "epoch": 0.09371395617070358,
654
+ "grad_norm": 1.03125,
655
+ "kd_loss": 0.06592908501625061,
656
+ "learning_rate": 7.89948503374835e-06,
657
+ "loss": 0.0465,
658
+ "step": 325,
659
+ "student_loss": 0.005857955198734999,
660
+ "teacher_loss": 0.02006148360669613
661
+ },
662
+ {
663
+ "epoch": 0.09515570934256055,
664
+ "grad_norm": 1.3125,
665
+ "kd_loss": 0.04595617577433586,
666
+ "learning_rate": 7.833143224409076e-06,
667
+ "loss": 0.0501,
668
+ "step": 330,
669
+ "student_loss": 0.01813417300581932,
670
+ "teacher_loss": 0.0033948116470128298
671
+ },
672
+ {
673
+ "epoch": 0.09659746251441753,
674
+ "grad_norm": 1.390625,
675
+ "kd_loss": 0.012639202177524567,
676
+ "learning_rate": 7.766058472361154e-06,
677
+ "loss": 0.0461,
678
+ "step": 335,
679
+ "student_loss": 0.0008777976036071777,
680
+ "teacher_loss": 0.0008721367339603603
681
+ },
682
+ {
683
+ "epoch": 0.09803921568627451,
684
+ "grad_norm": 1.0625,
685
+ "kd_loss": 0.026686903089284897,
686
+ "learning_rate": 7.698248369418146e-06,
687
+ "loss": 0.0515,
688
+ "step": 340,
689
+ "student_loss": 0.010171014815568924,
690
+ "teacher_loss": 0.007541630417108536
691
+ },
692
+ {
693
+ "epoch": 0.09948096885813149,
694
+ "grad_norm": 1.3203125,
695
+ "kd_loss": 0.03351679816842079,
696
+ "learning_rate": 7.629730697604314e-06,
697
+ "loss": 0.0516,
698
+ "step": 345,
699
+ "student_loss": 0.008214120753109455,
700
+ "teacher_loss": 0.0004564044938888401
701
+ },
702
+ {
703
+ "epoch": 0.10092272202998846,
704
+ "grad_norm": 1.71875,
705
+ "kd_loss": 0.05568505823612213,
706
+ "learning_rate": 7.560523424491595e-06,
707
+ "loss": 0.0457,
708
+ "step": 350,
709
+ "student_loss": 0.010019722394645214,
710
+ "teacher_loss": 0.003242993028834462
711
+ },
712
+ {
713
+ "epoch": 0.10236447520184544,
714
+ "grad_norm": 1.8515625,
715
+ "kd_loss": 0.016680920496582985,
716
+ "learning_rate": 7.490644698487909e-06,
717
+ "loss": 0.0531,
718
+ "step": 355,
719
+ "student_loss": 0.0006857452099211514,
720
+ "teacher_loss": 0.0006745306309312582
721
+ },
722
+ {
723
+ "epoch": 0.10380622837370242,
724
+ "grad_norm": 1.2109375,
725
+ "kd_loss": 0.01588292233645916,
726
+ "learning_rate": 7.420112844078066e-06,
727
+ "loss": 0.0509,
728
+ "step": 360,
729
+ "student_loss": 0.001193460077047348,
730
+ "teacher_loss": 0.001017728471197188
731
+ },
732
+ {
733
+ "epoch": 0.1052479815455594,
734
+ "grad_norm": 2.0625,
735
+ "kd_loss": 0.02928476221859455,
736
+ "learning_rate": 7.348946357018479e-06,
737
+ "loss": 0.053,
738
+ "step": 365,
739
+ "student_loss": 0.043851789087057114,
740
+ "teacher_loss": 0.0283951573073864
741
+ },
742
+ {
743
+ "epoch": 0.10668973471741638,
744
+ "grad_norm": 1.21875,
745
+ "kd_loss": 0.04628458246588707,
746
+ "learning_rate": 7.277163899486975e-06,
747
+ "loss": 0.0663,
748
+ "step": 370,
749
+ "student_loss": 0.023408163338899612,
750
+ "teacher_loss": 0.0004174449131824076
751
+ },
752
+ {
753
+ "epoch": 0.10813148788927336,
754
+ "grad_norm": 1.71875,
755
+ "kd_loss": 0.03154220059514046,
756
+ "learning_rate": 7.204784295188959e-06,
757
+ "loss": 0.05,
758
+ "step": 375,
759
+ "student_loss": 0.0324401780962944,
760
+ "teacher_loss": 0.020961124449968338
761
+ },
762
+ {
763
+ "epoch": 0.10957324106113034,
764
+ "grad_norm": 1.046875,
765
+ "kd_loss": 0.01982559822499752,
766
+ "learning_rate": 7.1318265244212305e-06,
767
+ "loss": 0.0493,
768
+ "step": 380,
769
+ "student_loss": 0.0006248670979402959,
770
+ "teacher_loss": 0.000449157232651487
771
+ },
772
+ {
773
+ "epoch": 0.11101499423298732,
774
+ "grad_norm": 1.390625,
775
+ "kd_loss": 0.0306891817599535,
776
+ "learning_rate": 7.05830971909472e-06,
777
+ "loss": 0.0447,
778
+ "step": 385,
779
+ "student_loss": 0.005440755281597376,
780
+ "teacher_loss": 0.00408168276771903
781
+ },
782
+ {
783
+ "epoch": 0.11245674740484429,
784
+ "grad_norm": 1.71875,
785
+ "kd_loss": 0.023891452699899673,
786
+ "learning_rate": 6.9842531577174865e-06,
787
+ "loss": 0.0458,
788
+ "step": 390,
789
+ "student_loss": 0.0003995662846136838,
790
+ "teacher_loss": 0.00039677225868217647
791
+ },
792
+ {
793
+ "epoch": 0.11389850057670127,
794
+ "grad_norm": 1.1328125,
795
+ "kd_loss": 0.018030904233455658,
796
+ "learning_rate": 6.9096762603392595e-06,
797
+ "loss": 0.0489,
798
+ "step": 395,
799
+ "student_loss": 0.0008614315302111208,
800
+ "teacher_loss": 0.0006612912402488291
801
+ },
802
+ {
803
+ "epoch": 0.11534025374855825,
804
+ "grad_norm": 1.703125,
805
+ "kd_loss": 0.02574356086552143,
806
+ "learning_rate": 6.834598583458862e-06,
807
+ "loss": 0.0432,
808
+ "step": 400,
809
+ "student_loss": 0.0009613529546186328,
810
+ "teacher_loss": 0.00023952587798703462
811
+ },
812
+ {
813
+ "epoch": 0.11678200692041522,
814
+ "grad_norm": 1.015625,
815
+ "kd_loss": 0.023935899138450623,
816
+ "learning_rate": 6.7590398148958625e-06,
817
+ "loss": 0.053,
818
+ "step": 405,
819
+ "student_loss": 0.011236889287829399,
820
+ "teacher_loss": 0.00042038856190629303
821
+ },
822
+ {
823
+ "epoch": 0.1182237600922722,
824
+ "grad_norm": 2.078125,
825
+ "kd_loss": 0.03511889651417732,
826
+ "learning_rate": 6.6830197686277945e-06,
827
+ "loss": 0.0539,
828
+ "step": 410,
829
+ "student_loss": 0.014045324176549911,
830
+ "teacher_loss": 0.008123157545924187
831
+ },
832
+ {
833
+ "epoch": 0.11966551326412918,
834
+ "grad_norm": 1.3828125,
835
+ "kd_loss": 0.025273097679018974,
836
+ "learning_rate": 6.6065583795942625e-06,
837
+ "loss": 0.0524,
838
+ "step": 415,
839
+ "student_loss": 0.004831122234463692,
840
+ "teacher_loss": 0.0038736157584935427
841
+ },
842
+ {
843
+ "epoch": 0.12110726643598616,
844
+ "grad_norm": 1.421875,
845
+ "kd_loss": 0.039099402725696564,
846
+ "learning_rate": 6.52967569846937e-06,
847
+ "loss": 0.0521,
848
+ "step": 420,
849
+ "student_loss": 0.04692080616950989,
850
+ "teacher_loss": 0.023907316848635674
851
+ },
852
+ {
853
+ "epoch": 0.12254901960784313,
854
+ "grad_norm": 1.4765625,
855
+ "kd_loss": 0.016112999990582466,
856
+ "learning_rate": 6.452391886403767e-06,
857
+ "loss": 0.0481,
858
+ "step": 425,
859
+ "student_loss": 0.002084826584905386,
860
+ "teacher_loss": 0.0022890178952366114
861
+ },
862
+ {
863
+ "epoch": 0.12399077277970011,
864
+ "grad_norm": 6.375,
865
+ "kd_loss": 0.05122941732406616,
866
+ "learning_rate": 6.374727209737743e-06,
867
+ "loss": 0.0585,
868
+ "step": 430,
869
+ "student_loss": 0.05258077755570412,
870
+ "teacher_loss": 0.06060388311743736
871
+ },
872
+ {
873
+ "epoch": 0.1254325259515571,
874
+ "grad_norm": 1.765625,
875
+ "kd_loss": 0.0188989769667387,
876
+ "learning_rate": 6.296702034686726e-06,
877
+ "loss": 0.0474,
878
+ "step": 435,
879
+ "student_loss": 0.001467574737034738,
880
+ "teacher_loss": 0.001621720613911748
881
+ },
882
+ {
883
+ "epoch": 0.12687427912341406,
884
+ "grad_norm": 0.9375,
885
+ "kd_loss": 0.03803834691643715,
886
+ "learning_rate": 6.218336822000598e-06,
887
+ "loss": 0.051,
888
+ "step": 440,
889
+ "student_loss": 0.012564362958073616,
890
+ "teacher_loss": 0.007415092084556818
891
+ },
892
+ {
893
+ "epoch": 0.12831603229527105,
894
+ "grad_norm": 1.265625,
895
+ "kd_loss": 0.028861815109848976,
896
+ "learning_rate": 6.139652121598219e-06,
897
+ "loss": 0.0508,
898
+ "step": 445,
899
+ "student_loss": 0.03139723464846611,
900
+ "teacher_loss": 0.018937230110168457
901
+ },
902
+ {
903
+ "epoch": 0.12975778546712802,
904
+ "grad_norm": 1.1015625,
905
+ "kd_loss": 0.03469434380531311,
906
+ "learning_rate": 6.060668567178561e-06,
907
+ "loss": 0.0506,
908
+ "step": 450,
909
+ "student_loss": 0.0018778342055156827,
910
+ "teacher_loss": 0.002150396816432476
911
+ },
912
+ {
913
+ "epoch": 0.131199538638985,
914
+ "grad_norm": 1.4375,
915
+ "kd_loss": 0.02303108386695385,
916
+ "learning_rate": 5.981406870809889e-06,
917
+ "loss": 0.0519,
918
+ "step": 455,
919
+ "student_loss": 0.0010092268930748105,
920
+ "teacher_loss": 0.0005768880364485085
921
+ },
922
+ {
923
+ "epoch": 0.13264129181084197,
924
+ "grad_norm": 1.6953125,
925
+ "kd_loss": 0.0384206622838974,
926
+ "learning_rate": 5.9018878174983674e-06,
927
+ "loss": 0.0457,
928
+ "step": 460,
929
+ "student_loss": 0.012533986009657383,
930
+ "teacher_loss": 0.0006231010193005204
931
+ },
932
+ {
933
+ "epoch": 0.13408304498269896,
934
+ "grad_norm": 1.453125,
935
+ "kd_loss": 0.027368323877453804,
936
+ "learning_rate": 5.822132259737565e-06,
937
+ "loss": 0.0517,
938
+ "step": 465,
939
+ "student_loss": 0.0006824060692451894,
940
+ "teacher_loss": 0.0005230730166658759
941
+ },
942
+ {
943
+ "epoch": 0.13552479815455595,
944
+ "grad_norm": 1.3359375,
945
+ "kd_loss": 0.025800921022892,
946
+ "learning_rate": 5.742161112040237e-06,
947
+ "loss": 0.0465,
948
+ "step": 470,
949
+ "student_loss": 0.00049140042392537,
950
+ "teacher_loss": 0.00034405241603963077
951
+ },
952
+ {
953
+ "epoch": 0.13696655132641292,
954
+ "grad_norm": 1.203125,
955
+ "kd_loss": 0.026706764474511147,
956
+ "learning_rate": 5.661995345453867e-06,
957
+ "loss": 0.0431,
958
+ "step": 475,
959
+ "student_loss": 0.0008521017152816057,
960
+ "teacher_loss": 0.0013468376128003001
961
+ },
962
+ {
963
+ "epoch": 0.1384083044982699,
964
+ "grad_norm": 1.4375,
965
+ "kd_loss": 0.030722549185156822,
966
+ "learning_rate": 5.581655982061367e-06,
967
+ "loss": 0.0595,
968
+ "step": 480,
969
+ "student_loss": 0.0362899973988533,
970
+ "teacher_loss": 0.03146218881011009
971
+ },
972
+ {
973
+ "epoch": 0.13985005767012687,
974
+ "grad_norm": 1.4765625,
975
+ "kd_loss": 0.019584491848945618,
976
+ "learning_rate": 5.501164089468406e-06,
977
+ "loss": 0.0469,
978
+ "step": 485,
979
+ "student_loss": 0.0011596218682825565,
980
+ "teacher_loss": 0.0012266576522961259
981
+ },
982
+ {
983
+ "epoch": 0.14129181084198386,
984
+ "grad_norm": 1.890625,
985
+ "kd_loss": 0.02283553034067154,
986
+ "learning_rate": 5.4205407752787884e-06,
987
+ "loss": 0.0567,
988
+ "step": 490,
989
+ "student_loss": 0.00145314063411206,
990
+ "teacher_loss": 0.0009193072910420597
991
+ },
992
+ {
993
+ "epoch": 0.14273356401384082,
994
+ "grad_norm": 1.671875,
995
+ "kd_loss": 0.034809987992048264,
996
+ "learning_rate": 5.339807181559359e-06,
997
+ "loss": 0.0612,
998
+ "step": 495,
999
+ "student_loss": 0.007140801288187504,
1000
+ "teacher_loss": 0.001571109751239419
1001
+ },
1002
+ {
1003
+ "epoch": 0.14417531718569782,
1004
+ "grad_norm": 1.53125,
1005
+ "kd_loss": 0.02246963419020176,
1006
+ "learning_rate": 5.258984479295853e-06,
1007
+ "loss": 0.0475,
1008
+ "step": 500,
1009
+ "student_loss": 0.000618905876763165,
1010
+ "teacher_loss": 0.0004656173987314105
1011
+ },
1012
+ {
1013
+ "epoch": 0.14561707035755478,
1014
+ "grad_norm": 1.5390625,
1015
+ "kd_loss": 0.02328832447528839,
1016
+ "learning_rate": 5.1780938628411795e-06,
1017
+ "loss": 0.0578,
1018
+ "step": 505,
1019
+ "student_loss": 0.0006587600219063461,
1020
+ "teacher_loss": 0.0005737307365052402
1021
+ },
1022
+ {
1023
+ "epoch": 0.14705882352941177,
1024
+ "grad_norm": 2.40625,
1025
+ "kd_loss": 0.02127482183277607,
1026
+ "learning_rate": 5.097156544357567e-06,
1027
+ "loss": 0.0491,
1028
+ "step": 510,
1029
+ "student_loss": 0.0005110295605845749,
1030
+ "teacher_loss": 0.00043354221270419657
1031
+ },
1032
+ {
1033
+ "epoch": 0.14850057670126873,
1034
+ "grad_norm": 1.9140625,
1035
+ "kd_loss": 0.025465093553066254,
1036
+ "learning_rate": 5.016193748254045e-06,
1037
+ "loss": 0.0459,
1038
+ "step": 515,
1039
+ "student_loss": 0.014592062681913376,
1040
+ "teacher_loss": 0.00036142065073363483
1041
+ },
1042
+ {
1043
+ "epoch": 0.14994232987312572,
1044
+ "grad_norm": 1.75,
1045
+ "kd_loss": 0.06094611808657646,
1046
+ "learning_rate": 4.935226705620699e-06,
1047
+ "loss": 0.0525,
1048
+ "step": 520,
1049
+ "student_loss": 0.04277863726019859,
1050
+ "teacher_loss": 0.0119513850659132
1051
+ },
1052
+ {
1053
+ "epoch": 0.1513840830449827,
1054
+ "grad_norm": 2.1875,
1055
+ "kd_loss": 0.02507575787603855,
1056
+ "learning_rate": 4.8542766486612035e-06,
1057
+ "loss": 0.0525,
1058
+ "step": 525,
1059
+ "student_loss": 0.0006827444885857403,
1060
+ "teacher_loss": 0.0003920310118701309
1061
+ },
1062
+ {
1063
+ "epoch": 0.15282583621683968,
1064
+ "grad_norm": 2.4375,
1065
+ "kd_loss": 0.027806738391518593,
1066
+ "learning_rate": 4.773364805125025e-06,
1067
+ "loss": 0.0489,
1068
+ "step": 530,
1069
+ "student_loss": 0.0006790340994484723,
1070
+ "teacher_loss": 0.0006053142133168876
1071
+ },
1072
+ {
1073
+ "epoch": 0.15426758938869667,
1074
+ "grad_norm": 2.859375,
1075
+ "kd_loss": 0.029496297240257263,
1076
+ "learning_rate": 4.6925123927408265e-06,
1077
+ "loss": 0.0469,
1078
+ "step": 535,
1079
+ "student_loss": 0.0009063734323717654,
1080
+ "teacher_loss": 0.0007879663608036935
1081
+ },
1082
+ {
1083
+ "epoch": 0.15570934256055363,
1084
+ "grad_norm": 1.96875,
1085
+ "kd_loss": 0.058626189827919006,
1086
+ "learning_rate": 4.611740613652485e-06,
1087
+ "loss": 0.0455,
1088
+ "step": 540,
1089
+ "student_loss": 0.021025465801358223,
1090
+ "teacher_loss": 0.0006112148403190076
1091
+ },
1092
+ {
1093
+ "epoch": 0.15715109573241062,
1094
+ "grad_norm": 1.9296875,
1095
+ "kd_loss": 0.020126501098275185,
1096
+ "learning_rate": 4.531070648859186e-06,
1097
+ "loss": 0.0436,
1098
+ "step": 545,
1099
+ "student_loss": 0.0014909773599356413,
1100
+ "teacher_loss": 0.0013666743179783225
1101
+ },
1102
+ {
1103
+ "epoch": 0.15859284890426759,
1104
+ "grad_norm": 2.25,
1105
+ "kd_loss": 0.045972585678100586,
1106
+ "learning_rate": 4.450523652661086e-06,
1107
+ "loss": 0.042,
1108
+ "step": 550,
1109
+ "student_loss": 0.000331137707689777,
1110
+ "teacher_loss": 0.0043593221344053745
1111
+ },
1112
+ {
1113
+ "epoch": 0.16003460207612458,
1114
+ "grad_norm": 1.6484375,
1115
+ "kd_loss": 0.0674143135547638,
1116
+ "learning_rate": 4.370120747111956e-06,
1117
+ "loss": 0.0488,
1118
+ "step": 555,
1119
+ "student_loss": 0.049931082874536514,
1120
+ "teacher_loss": 0.002049551112577319
1121
+ },
1122
+ {
1123
+ "epoch": 0.16147635524798154,
1124
+ "grad_norm": 1.78125,
1125
+ "kd_loss": 0.030732685700058937,
1126
+ "learning_rate": 4.289883016480291e-06,
1127
+ "loss": 0.052,
1128
+ "step": 560,
1129
+ "student_loss": 0.0015958240255713463,
1130
+ "teacher_loss": 0.0004530187579803169
1131
+ },
1132
+ {
1133
+ "epoch": 0.16291810841983853,
1134
+ "grad_norm": 1.71875,
1135
+ "kd_loss": 0.03263721987605095,
1136
+ "learning_rate": 4.209831501720328e-06,
1137
+ "loss": 0.0492,
1138
+ "step": 565,
1139
+ "student_loss": 0.027923284098505974,
1140
+ "teacher_loss": 0.02007977105677128
1141
+ },
1142
+ {
1143
+ "epoch": 0.1643598615916955,
1144
+ "grad_norm": 1.7109375,
1145
+ "kd_loss": 0.030005840584635735,
1146
+ "learning_rate": 4.129987194954421e-06,
1147
+ "loss": 0.0521,
1148
+ "step": 570,
1149
+ "student_loss": 0.010670044459402561,
1150
+ "teacher_loss": 0.007075584959238768
1151
+ },
1152
+ {
1153
+ "epoch": 0.16580161476355249,
1154
+ "grad_norm": 1.7421875,
1155
+ "kd_loss": 0.022702785208821297,
1156
+ "learning_rate": 4.050371033968216e-06,
1157
+ "loss": 0.0499,
1158
+ "step": 575,
1159
+ "student_loss": 0.0008753272122703493,
1160
+ "teacher_loss": 0.0006742495461367071
1161
+ },
1162
+ {
1163
+ "epoch": 0.16724336793540945,
1164
+ "grad_norm": 0.94140625,
1165
+ "kd_loss": 0.024699244648218155,
1166
+ "learning_rate": 3.9710038967200825e-06,
1167
+ "loss": 0.044,
1168
+ "step": 580,
1169
+ "student_loss": 0.003035952802747488,
1170
+ "teacher_loss": 0.0022496841847896576
1171
+ },
1172
+ {
1173
+ "epoch": 0.16868512110726644,
1174
+ "grad_norm": 1.2734375,
1175
+ "kd_loss": 0.02395116351544857,
1176
+ "learning_rate": 3.89190659586623e-06,
1177
+ "loss": 0.043,
1178
+ "step": 585,
1179
+ "student_loss": 0.0006281615933403373,
1180
+ "teacher_loss": 0.00044049997813999653
1181
+ },
1182
+ {
1183
+ "epoch": 0.1701268742791234,
1184
+ "grad_norm": 1.25,
1185
+ "kd_loss": 0.027902590110898018,
1186
+ "learning_rate": 3.8130998733029517e-06,
1187
+ "loss": 0.0462,
1188
+ "step": 590,
1189
+ "student_loss": 0.000795386906247586,
1190
+ "teacher_loss": 0.0014348695985972881
1191
+ },
1192
+ {
1193
+ "epoch": 0.1715686274509804,
1194
+ "grad_norm": 1.7890625,
1195
+ "kd_loss": 0.02104509435594082,
1196
+ "learning_rate": 3.734604394727419e-06,
1197
+ "loss": 0.0489,
1198
+ "step": 595,
1199
+ "student_loss": 0.0007530418806709349,
1200
+ "teacher_loss": 0.000717821647413075
1201
+ },
1202
+ {
1203
+ "epoch": 0.17301038062283736,
1204
+ "grad_norm": 1.7578125,
1205
+ "kd_loss": 0.033956341445446014,
1206
+ "learning_rate": 3.656440744218464e-06,
1207
+ "loss": 0.0514,
1208
+ "step": 600,
1209
+ "student_loss": 0.04278191551566124,
1210
+ "teacher_loss": 0.0042961700819432735
1211
+ }
1212
+ ],
1213
+ "logging_steps": 5,
1214
+ "max_steps": 1000,
1215
+ "num_input_tokens_seen": 0,
1216
+ "num_train_epochs": 1,
1217
+ "save_steps": 200,
1218
+ "stateful_callbacks": {
1219
+ "TrainerControl": {
1220
+ "args": {
1221
+ "should_epoch_stop": false,
1222
+ "should_evaluate": false,
1223
+ "should_log": false,
1224
+ "should_save": true,
1225
+ "should_training_stop": false
1226
+ },
1227
+ "attributes": {}
1228
+ }
1229
+ },
1230
+ "total_flos": 0.0,
1231
+ "train_batch_size": 1,
1232
+ "trial_name": null,
1233
+ "trial_params": null
1234
+ }
checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-600/vocab.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-800/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_ls2/checkpoint-800/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_ls2/checkpoint-800/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_ls2/checkpoint-800/merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
checkpoints/codi3b_a1.0_b1.0_g1.0_ls2/checkpoint-800/pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1064fd6c634f76911f294801c7dae7f6e820d14da7a02fa71cc24bbd26d9a151
3
+ size 6187858991