Hugging Face's logo

ASSERT-KTH
/
codi-trace

Safetensors
Model card Files
xet
 Community
sirui6011 commited on
Commit
dc871d8
·
verified ·
1 Parent(s): 24d9a63

add checkpoints/codi3b_a0.5_b1.0_g0.5_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_a0.5_b1.0_g0.5_ls2/checkpoint-1000/added_tokens.json +35 -0
  3. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-1000/chat_template.jinja +54 -0
  4. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-1000/config.json +66 -0
  5. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-1000/merges.txt +0 -0
  6. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-1000/pytorch_model.bin +3 -0
  7. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-1000/special_tokens_map.json +31 -0
  8. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-1000/thought_projector.pt +3 -0
  9. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-1000/tokenizer.json +3 -0
  10. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-1000/tokenizer_config.json +295 -0
  11. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-1000/trainer_state.json +2046 -0
  12. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-1000/vocab.json +0 -0
  13. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-200/added_tokens.json +35 -0
  14. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-200/chat_template.jinja +54 -0
  15. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-200/config.json +66 -0
  16. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-200/merges.txt +0 -0
  17. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-200/pytorch_model.bin +3 -0
  18. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-200/special_tokens_map.json +31 -0
  19. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-200/thought_projector.pt +3 -0
  20. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-200/tokenizer.json +3 -0
  21. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-200/tokenizer_config.json +295 -0
  22. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-200/trainer_state.json +434 -0
  23. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-200/vocab.json +0 -0
  24. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-400/added_tokens.json +35 -0
  25. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-400/chat_template.jinja +54 -0
  26. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-400/config.json +66 -0
  27. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-400/merges.txt +0 -0
  28. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-400/pytorch_model.bin +3 -0
  29. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-400/special_tokens_map.json +31 -0
  30. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-400/thought_projector.pt +3 -0
  31. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-400/tokenizer.json +3 -0
  32. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-400/tokenizer_config.json +295 -0
  33. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-400/trainer_state.json +834 -0
  34. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-400/vocab.json +0 -0
  35. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-600/added_tokens.json +35 -0
  36. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-600/chat_template.jinja +54 -0
  37. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-600/config.json +66 -0
  38. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-600/merges.txt +0 -0
  39. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-600/pytorch_model.bin +3 -0
  40. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-600/special_tokens_map.json +31 -0
  41. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-600/thought_projector.pt +3 -0
  42. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-600/tokenizer.json +3 -0
  43. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-600/tokenizer_config.json +295 -0
  44. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-600/trainer_state.json +1234 -0
  45. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-600/vocab.json +0 -0
  46. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-800/added_tokens.json +35 -0
  47. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-800/chat_template.jinja +54 -0
  48. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-800/config.json +66 -0
  49. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-800/merges.txt +0 -0
  50. checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-800/pytorch_model.bin +3 -0
.gitattributes CHANGED
@@ -62,3 +62,9 @@ checkpoints/codi3b_a0.5_b1.0_g0.5_ls1/checkpoint-400/tokenizer.json filter=lfs d
62
  checkpoints/codi3b_a0.5_b1.0_g0.5_ls1/checkpoint-600/tokenizer.json filter=lfs diff=lfs merge=lfs -text
63
  checkpoints/codi3b_a0.5_b1.0_g0.5_ls1/checkpoint-800/tokenizer.json filter=lfs diff=lfs merge=lfs -text
64
  checkpoints/codi3b_a0.5_b1.0_g0.5_ls1/wandb/offline-run-20260620_190723-v54ju1cd/run-v54ju1cd.wandb filter=lfs diff=lfs merge=lfs -text
 
 
 
 
 
 
 
62
  checkpoints/codi3b_a0.5_b1.0_g0.5_ls1/checkpoint-600/tokenizer.json filter=lfs diff=lfs merge=lfs -text
63
  checkpoints/codi3b_a0.5_b1.0_g0.5_ls1/checkpoint-800/tokenizer.json filter=lfs diff=lfs merge=lfs -text
64
  checkpoints/codi3b_a0.5_b1.0_g0.5_ls1/wandb/offline-run-20260620_190723-v54ju1cd/run-v54ju1cd.wandb filter=lfs diff=lfs merge=lfs -text
65
+ checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-1000/tokenizer.json filter=lfs diff=lfs merge=lfs -text
66
+ checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-200/tokenizer.json filter=lfs diff=lfs merge=lfs -text
67
+ checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-400/tokenizer.json filter=lfs diff=lfs merge=lfs -text
68
+ checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-600/tokenizer.json filter=lfs diff=lfs merge=lfs -text
69
+ checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-800/tokenizer.json filter=lfs diff=lfs merge=lfs -text
70
+ checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/wandb/offline-run-20260620_205602-mgxj3qvh/run-mgxj3qvh.wandb filter=lfs diff=lfs merge=lfs -text
checkpoints/codi3b_a0.5_b1.0_g0.5_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_a0.5_b1.0_g0.5_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_a0.5_b1.0_g0.5_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_a0.5_b1.0_g0.5_ls2/checkpoint-1000/merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-1000/pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7fab2179217b40bd1d9699c0150feea4a61964b67f6e47b1faf4137925dee730
3
+ size 6187858991
checkpoints/codi3b_a0.5_b1.0_g0.5_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_a0.5_b1.0_g0.5_ls2/checkpoint-1000/thought_projector.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:80bed173a3b415528e742bdd7eb367f051e29649d3850f528c258ccd893ef46b
3
+ size 16788033
checkpoints/codi3b_a0.5_b1.0_g0.5_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_a0.5_b1.0_g0.5_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_a0.5_b1.0_g0.5_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": 64.5,
15
+ "kd_loss": 0.2920137941837311,
16
+ "learning_rate": 1.3333333333333334e-06,
17
+ "loss": 0.4591,
18
+ "step": 5,
19
+ "student_loss": 0.18811793625354767,
20
+ "teacher_loss": 0.0020189557690173388
21
+ },
22
+ {
23
+ "epoch": 0.0028835063437139563,
24
+ "grad_norm": 37.25,
25
+ "kd_loss": 0.27189651131629944,
26
+ "learning_rate": 3e-06,
27
+ "loss": 0.3445,
28
+ "step": 10,
29
+ "student_loss": 0.15471063554286957,
30
+ "teacher_loss": 0.00530141731724143
31
+ },
32
+ {
33
+ "epoch": 0.004325259515570935,
34
+ "grad_norm": 7.1875,
35
+ "kd_loss": 0.22213532030582428,
36
+ "learning_rate": 4.666666666666667e-06,
37
+ "loss": 0.2034,
38
+ "step": 15,
39
+ "student_loss": 0.007600904442369938,
40
+ "teacher_loss": 0.0011894693598151207
41
+ },
42
+ {
43
+ "epoch": 0.0057670126874279125,
44
+ "grad_norm": 7.21875,
45
+ "kd_loss": 0.14968645572662354,
46
+ "learning_rate": 6.333333333333333e-06,
47
+ "loss": 0.1464,
48
+ "step": 20,
49
+ "student_loss": 0.026719728484749794,
50
+ "teacher_loss": 0.0005942200659774244
51
+ },
52
+ {
53
+ "epoch": 0.00720876585928489,
54
+ "grad_norm": 2.703125,
55
+ "kd_loss": 0.11552157998085022,
56
+ "learning_rate": 8.000000000000001e-06,
57
+ "loss": 0.1171,
58
+ "step": 25,
59
+ "student_loss": 0.002865401329472661,
60
+ "teacher_loss": 0.0010151522001251578
61
+ },
62
+ {
63
+ "epoch": 0.00865051903114187,
64
+ "grad_norm": 1.484375,
65
+ "kd_loss": 0.07486728578805923,
66
+ "learning_rate": 9.666666666666667e-06,
67
+ "loss": 0.0901,
68
+ "step": 30,
69
+ "student_loss": 0.0011324587976559997,
70
+ "teacher_loss": 0.0005539655685424805
71
+ },
72
+ {
73
+ "epoch": 0.010092272202998846,
74
+ "grad_norm": 2.140625,
75
+ "kd_loss": 0.06898625195026398,
76
+ "learning_rate": 9.99958042442916e-06,
77
+ "loss": 0.0757,
78
+ "step": 35,
79
+ "student_loss": 0.0009480358567088842,
80
+ "teacher_loss": 0.0005732738063670695
81
+ },
82
+ {
83
+ "epoch": 0.011534025374855825,
84
+ "grad_norm": 1.296875,
85
+ "kd_loss": 0.0642521008849144,
86
+ "learning_rate": 9.997876019358083e-06,
87
+ "loss": 0.0685,
88
+ "step": 40,
89
+ "student_loss": 0.031368859112262726,
90
+ "teacher_loss": 0.027499590069055557
91
+ },
92
+ {
93
+ "epoch": 0.012975778546712802,
94
+ "grad_norm": 1.3359375,
95
+ "kd_loss": 0.11454806476831436,
96
+ "learning_rate": 9.99486100792044e-06,
97
+ "loss": 0.0541,
98
+ "step": 45,
99
+ "student_loss": 0.06097811087965965,
100
+ "teacher_loss": 0.0027094183024019003
101
+ },
102
+ {
103
+ "epoch": 0.01441753171856978,
104
+ "grad_norm": 1.5859375,
105
+ "kd_loss": 0.03876315429806709,
106
+ "learning_rate": 9.990536180750724e-06,
107
+ "loss": 0.0563,
108
+ "step": 50,
109
+ "student_loss": 0.02282995544373989,
110
+ "teacher_loss": 0.008940276689827442
111
+ },
112
+ {
113
+ "epoch": 0.015859284890426758,
114
+ "grad_norm": 1.125,
115
+ "kd_loss": 0.033022914081811905,
116
+ "learning_rate": 9.984902671959911e-06,
117
+ "loss": 0.0481,
118
+ "step": 55,
119
+ "student_loss": 0.0011594295501708984,
120
+ "teacher_loss": 0.0007985035772435367
121
+ },
122
+ {
123
+ "epoch": 0.01730103806228374,
124
+ "grad_norm": 0.8828125,
125
+ "kd_loss": 0.030243180692195892,
126
+ "learning_rate": 9.97796195883804e-06,
127
+ "loss": 0.0429,
128
+ "step": 60,
129
+ "student_loss": 0.0034187885466963053,
130
+ "teacher_loss": 0.0017242392059415579
131
+ },
132
+ {
133
+ "epoch": 0.018742791234140715,
134
+ "grad_norm": 1.4453125,
135
+ "kd_loss": 0.032282304018735886,
136
+ "learning_rate": 9.969715861466839e-06,
137
+ "loss": 0.0446,
138
+ "step": 65,
139
+ "student_loss": 0.029871758073568344,
140
+ "teacher_loss": 0.039595428854227066
141
+ },
142
+ {
143
+ "epoch": 0.020184544405997693,
144
+ "grad_norm": 1.1953125,
145
+ "kd_loss": 0.06090804189443588,
146
+ "learning_rate": 9.96016654224243e-06,
147
+ "loss": 0.0477,
148
+ "step": 70,
149
+ "student_loss": 0.017709577456116676,
150
+ "teacher_loss": 0.0022253147326409817
151
+ },
152
+ {
153
+ "epoch": 0.02162629757785467,
154
+ "grad_norm": 0.84765625,
155
+ "kd_loss": 0.024834414944052696,
156
+ "learning_rate": 9.94931650530827e-06,
157
+ "loss": 0.0418,
158
+ "step": 75,
159
+ "student_loss": 0.0012920524459332228,
160
+ "teacher_loss": 0.0008862126851454377
161
+ },
162
+ {
163
+ "epoch": 0.02306805074971165,
164
+ "grad_norm": 0.60546875,
165
+ "kd_loss": 0.025287121534347534,
166
+ "learning_rate": 9.93716859589851e-06,
167
+ "loss": 0.0365,
168
+ "step": 80,
169
+ "student_loss": 0.0013494148151949048,
170
+ "teacher_loss": 0.0008630359079688787
171
+ },
172
+ {
173
+ "epoch": 0.024509803921568627,
174
+ "grad_norm": 0.86328125,
175
+ "kd_loss": 0.024378223344683647,
176
+ "learning_rate": 9.923725999591846e-06,
177
+ "loss": 0.0395,
178
+ "step": 85,
179
+ "student_loss": 0.0005228935624472797,
180
+ "teacher_loss": 0.0004561410460155457
181
+ },
182
+ {
183
+ "epoch": 0.025951557093425604,
184
+ "grad_norm": 0.70703125,
185
+ "kd_loss": 0.07289917767047882,
186
+ "learning_rate": 9.908992241476189e-06,
187
+ "loss": 0.0394,
188
+ "step": 90,
189
+ "student_loss": 0.020422162488102913,
190
+ "teacher_loss": 0.0035499960649758577
191
+ },
192
+ {
193
+ "epoch": 0.027393310265282585,
194
+ "grad_norm": 1.125,
195
+ "kd_loss": 0.044955912977457047,
196
+ "learning_rate": 9.892971185224244e-06,
197
+ "loss": 0.0351,
198
+ "step": 95,
199
+ "student_loss": 0.008261171169579029,
200
+ "teacher_loss": 0.005078152287751436
201
+ },
202
+ {
203
+ "epoch": 0.02883506343713956,
204
+ "grad_norm": 0.83203125,
205
+ "kd_loss": 0.02227398194372654,
206
+ "learning_rate": 9.875667032080354e-06,
207
+ "loss": 0.0326,
208
+ "step": 100,
209
+ "student_loss": 0.0006025677430443466,
210
+ "teacher_loss": 0.00046476206625811756
211
+ },
212
+ {
213
+ "epoch": 0.03027681660899654,
214
+ "grad_norm": 1.2734375,
215
+ "kd_loss": 0.03532887250185013,
216
+ "learning_rate": 9.857084319758772e-06,
217
+ "loss": 0.036,
218
+ "step": 105,
219
+ "student_loss": 0.0034369053319096565,
220
+ "teacher_loss": 0.00029834595625288785
221
+ },
222
+ {
223
+ "epoch": 0.031718569780853516,
224
+ "grad_norm": 1.1640625,
225
+ "kd_loss": 0.04033924266695976,
226
+ "learning_rate": 9.837227921253747e-06,
227
+ "loss": 0.0375,
228
+ "step": 110,
229
+ "student_loss": 0.03080393560230732,
230
+ "teacher_loss": 0.018437745049595833
231
+ },
232
+ {
233
+ "epoch": 0.03316032295271049,
234
+ "grad_norm": 0.703125,
235
+ "kd_loss": 0.04197424277663231,
236
+ "learning_rate": 9.816103043561648e-06,
237
+ "loss": 0.0347,
238
+ "step": 115,
239
+ "student_loss": 0.0021668823901563883,
240
+ "teacher_loss": 0.00045062918798066676
241
+ },
242
+ {
243
+ "epoch": 0.03460207612456748,
244
+ "grad_norm": 0.98828125,
245
+ "kd_loss": 0.027563175186514854,
246
+ "learning_rate": 9.79371522631553e-06,
247
+ "loss": 0.032,
248
+ "step": 120,
249
+ "student_loss": 0.0016319500282406807,
250
+ "teacher_loss": 0.0008567498298361897
251
+ },
252
+ {
253
+ "epoch": 0.036043829296424454,
254
+ "grad_norm": 0.92578125,
255
+ "kd_loss": 0.06173818185925484,
256
+ "learning_rate": 9.770070340332457e-06,
257
+ "loss": 0.0364,
258
+ "step": 125,
259
+ "student_loss": 0.02385914884507656,
260
+ "teacher_loss": 0.00027849775506183505
261
+ },
262
+ {
263
+ "epoch": 0.03748558246828143,
264
+ "grad_norm": 1.0859375,
265
+ "kd_loss": 0.029248492792248726,
266
+ "learning_rate": 9.745174586073982e-06,
267
+ "loss": 0.0346,
268
+ "step": 130,
269
+ "student_loss": 0.0005455865757539868,
270
+ "teacher_loss": 0.0004959598300047219
271
+ },
272
+ {
273
+ "epoch": 0.03892733564013841,
274
+ "grad_norm": 0.953125,
275
+ "kd_loss": 0.0406946986913681,
276
+ "learning_rate": 9.719034492020183e-06,
277
+ "loss": 0.0377,
278
+ "step": 135,
279
+ "student_loss": 0.0013323032762855291,
280
+ "teacher_loss": 0.0005948346224613488
281
+ },
282
+ {
283
+ "epoch": 0.040369088811995385,
284
+ "grad_norm": 1.0390625,
285
+ "kd_loss": 0.023682042956352234,
286
+ "learning_rate": 9.691656912957686e-06,
287
+ "loss": 0.036,
288
+ "step": 140,
289
+ "student_loss": 0.0005881476681679487,
290
+ "teacher_loss": 0.0003679130459204316
291
+ },
292
+ {
293
+ "epoch": 0.04181084198385236,
294
+ "grad_norm": 0.5859375,
295
+ "kd_loss": 0.07271980494260788,
296
+ "learning_rate": 9.663049028182112e-06,
297
+ "loss": 0.0325,
298
+ "step": 145,
299
+ "student_loss": 0.028793470934033394,
300
+ "teacher_loss": 0.005983584560453892
301
+ },
302
+ {
303
+ "epoch": 0.04325259515570934,
304
+ "grad_norm": 0.609375,
305
+ "kd_loss": 0.01843745820224285,
306
+ "learning_rate": 9.633218339615433e-06,
307
+ "loss": 0.0316,
308
+ "step": 150,
309
+ "student_loss": 0.001051027444191277,
310
+ "teacher_loss": 0.000913174357265234
311
+ },
312
+ {
313
+ "epoch": 0.04469434832756632,
314
+ "grad_norm": 1.1171875,
315
+ "kd_loss": 0.023381218314170837,
316
+ "learning_rate": 9.602172669838721e-06,
317
+ "loss": 0.0381,
318
+ "step": 155,
319
+ "student_loss": 0.002775231609120965,
320
+ "teacher_loss": 0.0007182428380474448
321
+ },
322
+ {
323
+ "epoch": 0.0461361014994233,
324
+ "grad_norm": 1.6875,
325
+ "kd_loss": 0.03144950047135353,
326
+ "learning_rate": 9.569920160040815e-06,
327
+ "loss": 0.0344,
328
+ "step": 160,
329
+ "student_loss": 0.0778423622250557,
330
+ "teacher_loss": 0.044956743717193604
331
+ },
332
+ {
333
+ "epoch": 0.04757785467128028,
334
+ "grad_norm": 0.75,
335
+ "kd_loss": 0.03335012122988701,
336
+ "learning_rate": 9.536469267883432e-06,
337
+ "loss": 0.0311,
338
+ "step": 165,
339
+ "student_loss": 0.0012603362556546926,
340
+ "teacher_loss": 0.00417186226695776
341
+ },
342
+ {
343
+ "epoch": 0.049019607843137254,
344
+ "grad_norm": 1.8828125,
345
+ "kd_loss": 0.018793689087033272,
346
+ "learning_rate": 9.501828765283295e-06,
347
+ "loss": 0.0355,
348
+ "step": 170,
349
+ "student_loss": 0.0005546013708226383,
350
+ "teacher_loss": 0.0003494401171337813
351
+ },
352
+ {
353
+ "epoch": 0.05046136101499423,
354
+ "grad_norm": 1.09375,
355
+ "kd_loss": 0.027096513658761978,
356
+ "learning_rate": 9.466007736111846e-06,
357
+ "loss": 0.0322,
358
+ "step": 175,
359
+ "student_loss": 0.0030481775756925344,
360
+ "teacher_loss": 0.0003126203373540193
361
+ },
362
+ {
363
+ "epoch": 0.05190311418685121,
364
+ "grad_norm": 0.69140625,
365
+ "kd_loss": 0.03634097799658775,
366
+ "learning_rate": 9.429015573813163e-06,
367
+ "loss": 0.0302,
368
+ "step": 180,
369
+ "student_loss": 0.0012228424893692136,
370
+ "teacher_loss": 0.00034309024340473115
371
+ },
372
+ {
373
+ "epoch": 0.05334486735870819,
374
+ "grad_norm": 1.34375,
375
+ "kd_loss": 0.03704367205500603,
376
+ "learning_rate": 9.390861978940687e-06,
377
+ "loss": 0.0363,
378
+ "step": 185,
379
+ "student_loss": 0.035696618258953094,
380
+ "teacher_loss": 0.00034108557156287134
381
+ },
382
+ {
383
+ "epoch": 0.05478662053056517,
384
+ "grad_norm": 1.3046875,
385
+ "kd_loss": 0.052902791649103165,
386
+ "learning_rate": 9.351556956613423e-06,
387
+ "loss": 0.0321,
388
+ "step": 190,
389
+ "student_loss": 0.009327664040029049,
390
+ "teacher_loss": 0.0015260990476235747
391
+ },
392
+ {
393
+ "epoch": 0.056228373702422146,
394
+ "grad_norm": 2.828125,
395
+ "kd_loss": 0.05315268039703369,
396
+ "learning_rate": 9.31111081389227e-06,
397
+ "loss": 0.0327,
398
+ "step": 195,
399
+ "student_loss": 0.011893535032868385,
400
+ "teacher_loss": 0.0007797812577337027
401
+ },
402
+ {
403
+ "epoch": 0.05767012687427912,
404
+ "grad_norm": 0.8125,
405
+ "kd_loss": 0.03527738153934479,
406
+ "learning_rate": 9.269534157077177e-06,
407
+ "loss": 0.0394,
408
+ "step": 200,
409
+ "student_loss": 0.0005083663854748011,
410
+ "teacher_loss": 0.0002656931465025991
411
+ },
412
+ {
413
+ "epoch": 0.0591118800461361,
414
+ "grad_norm": 0.9921875,
415
+ "kd_loss": 0.02736036665737629,
416
+ "learning_rate": 9.226837888925813e-06,
417
+ "loss": 0.0341,
418
+ "step": 205,
419
+ "student_loss": 0.03568984195590019,
420
+ "teacher_loss": 0.027640890330076218
421
+ },
422
+ {
423
+ "epoch": 0.06055363321799308,
424
+ "grad_norm": 1.78125,
425
+ "kd_loss": 0.030339844524860382,
426
+ "learning_rate": 9.183033205794525e-06,
427
+ "loss": 0.0302,
428
+ "step": 210,
429
+ "student_loss": 0.000763049116358161,
430
+ "teacher_loss": 0.0003548153617884964
431
+ },
432
+ {
433
+ "epoch": 0.061995386389850055,
434
+ "grad_norm": 1.1484375,
435
+ "kd_loss": 0.03865697979927063,
436
+ "learning_rate": 9.13813159470227e-06,
437
+ "loss": 0.0326,
438
+ "step": 215,
439
+ "student_loss": 0.0005148217896930873,
440
+ "teacher_loss": 0.00020245747873559594
441
+ },
442
+ {
443
+ "epoch": 0.06343713956170703,
444
+ "grad_norm": 1.3671875,
445
+ "kd_loss": 0.04260854423046112,
446
+ "learning_rate": 9.092144830318357e-06,
447
+ "loss": 0.0316,
448
+ "step": 220,
449
+ "student_loss": 0.030991079285740852,
450
+ "teacher_loss": 0.00993641559034586
451
+ },
452
+ {
453
+ "epoch": 0.06487889273356401,
454
+ "grad_norm": 1.203125,
455
+ "kd_loss": 0.023958567529916763,
456
+ "learning_rate": 9.045084971874738e-06,
457
+ "loss": 0.0376,
458
+ "step": 225,
459
+ "student_loss": 0.043197184801101685,
460
+ "teacher_loss": 0.00216495874337852
461
+ },
462
+ {
463
+ "epoch": 0.06632064590542099,
464
+ "grad_norm": 1.7578125,
465
+ "kd_loss": 0.04370651766657829,
466
+ "learning_rate": 8.99696436000368e-06,
467
+ "loss": 0.0299,
468
+ "step": 230,
469
+ "student_loss": 0.005392159800976515,
470
+ "teacher_loss": 0.003661371534690261
471
+ },
472
+ {
473
+ "epoch": 0.06776239907727798,
474
+ "grad_norm": 0.8671875,
475
+ "kd_loss": 0.07619086652994156,
476
+ "learning_rate": 8.947795613501658e-06,
477
+ "loss": 0.0314,
478
+ "step": 235,
479
+ "student_loss": 0.008586333133280277,
480
+ "teacher_loss": 0.0005631999811157584
481
+ },
482
+ {
483
+ "epoch": 0.06920415224913495,
484
+ "grad_norm": 1.0625,
485
+ "kd_loss": 0.03111647628247738,
486
+ "learning_rate": 8.897591626020284e-06,
487
+ "loss": 0.034,
488
+ "step": 240,
489
+ "student_loss": 0.003906027879565954,
490
+ "teacher_loss": 0.0004787310608662665
491
+ },
492
+ {
493
+ "epoch": 0.07064590542099193,
494
+ "grad_norm": 1.453125,
495
+ "kd_loss": 0.017389042302966118,
496
+ "learning_rate": 8.846365562685178e-06,
497
+ "loss": 0.0279,
498
+ "step": 245,
499
+ "student_loss": 0.013969292864203453,
500
+ "teacher_loss": 0.006873726844787598
501
+ },
502
+ {
503
+ "epoch": 0.07208765859284891,
504
+ "grad_norm": 1.6015625,
505
+ "kd_loss": 0.0229345690459013,
506
+ "learning_rate": 8.794130856643635e-06,
507
+ "loss": 0.0311,
508
+ "step": 250,
509
+ "student_loss": 0.0008334179292432964,
510
+ "teacher_loss": 0.0003169570700265467
511
+ },
512
+ {
513
+ "epoch": 0.07352941176470588,
514
+ "grad_norm": 0.83203125,
515
+ "kd_loss": 0.01808979921042919,
516
+ "learning_rate": 8.74090120554202e-06,
517
+ "loss": 0.0312,
518
+ "step": 255,
519
+ "student_loss": 0.0003083710907958448,
520
+ "teacher_loss": 0.00030115401023067534
521
+ },
522
+ {
523
+ "epoch": 0.07497116493656286,
524
+ "grad_norm": 0.88671875,
525
+ "kd_loss": 0.026108454912900925,
526
+ "learning_rate": 8.686690567933803e-06,
527
+ "loss": 0.0333,
528
+ "step": 260,
529
+ "student_loss": 0.042571116238832474,
530
+ "teacher_loss": 0.03388316184282303
531
+ },
532
+ {
533
+ "epoch": 0.07641291810841984,
534
+ "grad_norm": 0.80078125,
535
+ "kd_loss": 0.016656002029776573,
536
+ "learning_rate": 8.63151315961915e-06,
537
+ "loss": 0.0317,
538
+ "step": 265,
539
+ "student_loss": 0.0003702428948599845,
540
+ "teacher_loss": 0.0002555136161390692
541
+ },
542
+ {
543
+ "epoch": 0.07785467128027682,
544
+ "grad_norm": 1.046875,
545
+ "kd_loss": 0.019304102286696434,
546
+ "learning_rate": 8.575383449917103e-06,
547
+ "loss": 0.0342,
548
+ "step": 270,
549
+ "student_loss": 0.001813149661757052,
550
+ "teacher_loss": 0.0011580288410186768
551
+ },
552
+ {
553
+ "epoch": 0.07929642445213379,
554
+ "grad_norm": 1.1796875,
555
+ "kd_loss": 0.023480774834752083,
556
+ "learning_rate": 8.518316157871232e-06,
557
+ "loss": 0.029,
558
+ "step": 275,
559
+ "student_loss": 0.04764978215098381,
560
+ "teacher_loss": 0.03439468517899513
561
+ },
562
+ {
563
+ "epoch": 0.08073817762399077,
564
+ "grad_norm": 1.015625,
565
+ "kd_loss": 0.03275206685066223,
566
+ "learning_rate": 8.460326248389825e-06,
567
+ "loss": 0.0289,
568
+ "step": 280,
569
+ "student_loss": 0.0005029537715017796,
570
+ "teacher_loss": 0.00019533037266228348
571
+ },
572
+ {
573
+ "epoch": 0.08217993079584775,
574
+ "grad_norm": 1.140625,
575
+ "kd_loss": 0.019457675516605377,
576
+ "learning_rate": 8.401428928321607e-06,
577
+ "loss": 0.0322,
578
+ "step": 285,
579
+ "student_loss": 0.0007758038118481636,
580
+ "teacher_loss": 0.000900130660738796
581
+ },
582
+ {
583
+ "epoch": 0.08362168396770472,
584
+ "grad_norm": 1.3125,
585
+ "kd_loss": 0.017802242189645767,
586
+ "learning_rate": 8.341639642468002e-06,
587
+ "loss": 0.0348,
588
+ "step": 290,
589
+ "student_loss": 0.023405462503433228,
590
+ "teacher_loss": 0.021540865302085876
591
+ },
592
+ {
593
+ "epoch": 0.0850634371395617,
594
+ "grad_norm": 1.03125,
595
+ "kd_loss": 0.01647804118692875,
596
+ "learning_rate": 8.280974069532999e-06,
597
+ "loss": 0.0328,
598
+ "step": 295,
599
+ "student_loss": 0.0006198033224791288,
600
+ "teacher_loss": 0.000554366793949157
601
+ },
602
+ {
603
+ "epoch": 0.08650519031141868,
604
+ "grad_norm": 0.9296875,
605
+ "kd_loss": 0.04784730449318886,
606
+ "learning_rate": 8.219448118011687e-06,
607
+ "loss": 0.0308,
608
+ "step": 300,
609
+ "student_loss": 0.015613794326782227,
610
+ "teacher_loss": 0.001725711626932025
611
+ },
612
+ {
613
+ "epoch": 0.08794694348327567,
614
+ "grad_norm": 1.1953125,
615
+ "kd_loss": 0.019965235143899918,
616
+ "learning_rate": 8.157077922018537e-06,
617
+ "loss": 0.0289,
618
+ "step": 305,
619
+ "student_loss": 0.006098510231822729,
620
+ "teacher_loss": 0.002777156652882695
621
+ },
622
+ {
623
+ "epoch": 0.08938869665513265,
624
+ "grad_norm": 1.0390625,
625
+ "kd_loss": 0.020748196169734,
626
+ "learning_rate": 8.093879837056486e-06,
627
+ "loss": 0.0309,
628
+ "step": 310,
629
+ "student_loss": 0.000493990199174732,
630
+ "teacher_loss": 0.00039537265547551215
631
+ },
632
+ {
633
+ "epoch": 0.09083044982698962,
634
+ "grad_norm": 1.4765625,
635
+ "kd_loss": 0.03892743960022926,
636
+ "learning_rate": 8.029870435728018e-06,
637
+ "loss": 0.0294,
638
+ "step": 315,
639
+ "student_loss": 0.0073717073537409306,
640
+ "teacher_loss": 0.00021180440671741962
641
+ },
642
+ {
643
+ "epoch": 0.0922722029988466,
644
+ "grad_norm": 1.3828125,
645
+ "kd_loss": 0.02363528683781624,
646
+ "learning_rate": 7.965066503389264e-06,
647
+ "loss": 0.0313,
648
+ "step": 320,
649
+ "student_loss": 0.0004119524674024433,
650
+ "teacher_loss": 0.00022927633835934103
651
+ },
652
+ {
653
+ "epoch": 0.09371395617070358,
654
+ "grad_norm": 1.7421875,
655
+ "kd_loss": 0.06230268254876137,
656
+ "learning_rate": 7.89948503374835e-06,
657
+ "loss": 0.0284,
658
+ "step": 325,
659
+ "student_loss": 0.009024329483509064,
660
+ "teacher_loss": 0.01963256485760212
661
+ },
662
+ {
663
+ "epoch": 0.09515570934256055,
664
+ "grad_norm": 0.90625,
665
+ "kd_loss": 0.04712303727865219,
666
+ "learning_rate": 7.833143224409076e-06,
667
+ "loss": 0.0302,
668
+ "step": 330,
669
+ "student_loss": 0.008720812387764454,
670
+ "teacher_loss": 0.0013501221546903253
671
+ },
672
+ {
673
+ "epoch": 0.09659746251441753,
674
+ "grad_norm": 0.9140625,
675
+ "kd_loss": 0.018766457214951515,
676
+ "learning_rate": 7.766058472361154e-06,
677
+ "loss": 0.0283,
678
+ "step": 335,
679
+ "student_loss": 0.001242243917658925,
680
+ "teacher_loss": 0.0023425817489624023
681
+ },
682
+ {
683
+ "epoch": 0.09803921568627451,
684
+ "grad_norm": 2.078125,
685
+ "kd_loss": 0.02911657840013504,
686
+ "learning_rate": 7.698248369418146e-06,
687
+ "loss": 0.0326,
688
+ "step": 340,
689
+ "student_loss": 0.009861858561635017,
690
+ "teacher_loss": 0.006796200294047594
691
+ },
692
+ {
693
+ "epoch": 0.09948096885813149,
694
+ "grad_norm": 1.2265625,
695
+ "kd_loss": 0.032899159938097,
696
+ "learning_rate": 7.629730697604314e-06,
697
+ "loss": 0.0335,
698
+ "step": 345,
699
+ "student_loss": 0.00739182299003005,
700
+ "teacher_loss": 0.0002512071805540472
701
+ },
702
+ {
703
+ "epoch": 0.10092272202998846,
704
+ "grad_norm": 1.0859375,
705
+ "kd_loss": 0.06120900437235832,
706
+ "learning_rate": 7.560523424491595e-06,
707
+ "loss": 0.0282,
708
+ "step": 350,
709
+ "student_loss": 0.003599822986871004,
710
+ "teacher_loss": 0.0007063632365316153
711
+ },
712
+ {
713
+ "epoch": 0.10236447520184544,
714
+ "grad_norm": 2.296875,
715
+ "kd_loss": 0.02087453007698059,
716
+ "learning_rate": 7.490644698487909e-06,
717
+ "loss": 0.035,
718
+ "step": 355,
719
+ "student_loss": 0.001389339566230774,
720
+ "teacher_loss": 0.0017551770433783531
721
+ },
722
+ {
723
+ "epoch": 0.10380622837370242,
724
+ "grad_norm": 0.90234375,
725
+ "kd_loss": 0.019557664170861244,
726
+ "learning_rate": 7.420112844078066e-06,
727
+ "loss": 0.0321,
728
+ "step": 360,
729
+ "student_loss": 0.0006360138650052249,
730
+ "teacher_loss": 0.0005147532792761922
731
+ },
732
+ {
733
+ "epoch": 0.1052479815455594,
734
+ "grad_norm": 1.3203125,
735
+ "kd_loss": 0.030097220093011856,
736
+ "learning_rate": 7.348946357018479e-06,
737
+ "loss": 0.0333,
738
+ "step": 365,
739
+ "student_loss": 0.04143820330500603,
740
+ "teacher_loss": 0.027219083160161972
741
+ },
742
+ {
743
+ "epoch": 0.10668973471741638,
744
+ "grad_norm": 0.9453125,
745
+ "kd_loss": 0.05344080179929733,
746
+ "learning_rate": 7.277163899486975e-06,
747
+ "loss": 0.0351,
748
+ "step": 370,
749
+ "student_loss": 0.021593965590000153,
750
+ "teacher_loss": 0.0005673202103935182
751
+ },
752
+ {
753
+ "epoch": 0.10813148788927336,
754
+ "grad_norm": 0.7265625,
755
+ "kd_loss": 0.03127056360244751,
756
+ "learning_rate": 7.204784295188959e-06,
757
+ "loss": 0.0287,
758
+ "step": 375,
759
+ "student_loss": 0.03163963928818703,
760
+ "teacher_loss": 0.020587248727679253
761
+ },
762
+ {
763
+ "epoch": 0.10957324106113034,
764
+ "grad_norm": 0.84375,
765
+ "kd_loss": 0.01984175480902195,
766
+ "learning_rate": 7.1318265244212305e-06,
767
+ "loss": 0.0311,
768
+ "step": 380,
769
+ "student_loss": 0.0005506337620317936,
770
+ "teacher_loss": 0.00038444914389401674
771
+ },
772
+ {
773
+ "epoch": 0.11101499423298732,
774
+ "grad_norm": 0.59765625,
775
+ "kd_loss": 0.02916550263762474,
776
+ "learning_rate": 7.05830971909472e-06,
777
+ "loss": 0.0294,
778
+ "step": 385,
779
+ "student_loss": 0.00045756620238535106,
780
+ "teacher_loss": 0.0003828653716482222
781
+ },
782
+ {
783
+ "epoch": 0.11245674740484429,
784
+ "grad_norm": 1.171875,
785
+ "kd_loss": 0.021802764385938644,
786
+ "learning_rate": 6.9842531577174865e-06,
787
+ "loss": 0.0271,
788
+ "step": 390,
789
+ "student_loss": 0.00026892355526797473,
790
+ "teacher_loss": 0.0002502185816410929
791
+ },
792
+ {
793
+ "epoch": 0.11389850057670127,
794
+ "grad_norm": 0.94140625,
795
+ "kd_loss": 0.020051907747983932,
796
+ "learning_rate": 6.9096762603392595e-06,
797
+ "loss": 0.0307,
798
+ "step": 395,
799
+ "student_loss": 0.0004568768781609833,
800
+ "teacher_loss": 0.00039133013342507184
801
+ },
802
+ {
803
+ "epoch": 0.11534025374855825,
804
+ "grad_norm": 0.78125,
805
+ "kd_loss": 0.03054620325565338,
806
+ "learning_rate": 6.834598583458862e-06,
807
+ "loss": 0.0275,
808
+ "step": 400,
809
+ "student_loss": 0.0005328103434294462,
810
+ "teacher_loss": 0.00015425821766257286
811
+ },
812
+ {
813
+ "epoch": 0.11678200692041522,
814
+ "grad_norm": 0.640625,
815
+ "kd_loss": 0.02604236640036106,
816
+ "learning_rate": 6.7590398148958625e-06,
817
+ "loss": 0.0335,
818
+ "step": 405,
819
+ "student_loss": 0.011330639012157917,
820
+ "teacher_loss": 0.00021900788124185055
821
+ },
822
+ {
823
+ "epoch": 0.1182237600922722,
824
+ "grad_norm": 0.8046875,
825
+ "kd_loss": 0.03826223686337471,
826
+ "learning_rate": 6.6830197686277945e-06,
827
+ "loss": 0.0366,
828
+ "step": 410,
829
+ "student_loss": 0.014453927055001259,
830
+ "teacher_loss": 0.00844341516494751
831
+ },
832
+ {
833
+ "epoch": 0.11966551326412918,
834
+ "grad_norm": 1.203125,
835
+ "kd_loss": 0.021850954741239548,
836
+ "learning_rate": 6.6065583795942625e-06,
837
+ "loss": 0.0347,
838
+ "step": 415,
839
+ "student_loss": 0.004759110510349274,
840
+ "teacher_loss": 0.0034685542341321707
841
+ },
842
+ {
843
+ "epoch": 0.12110726643598616,
844
+ "grad_norm": 0.93359375,
845
+ "kd_loss": 0.045137468725442886,
846
+ "learning_rate": 6.52967569846937e-06,
847
+ "loss": 0.0331,
848
+ "step": 420,
849
+ "student_loss": 0.04586087912321091,
850
+ "teacher_loss": 0.021448055282235146
851
+ },
852
+ {
853
+ "epoch": 0.12254901960784313,
854
+ "grad_norm": 1.4453125,
855
+ "kd_loss": 0.014929288066923618,
856
+ "learning_rate": 6.452391886403767e-06,
857
+ "loss": 0.0299,
858
+ "step": 425,
859
+ "student_loss": 0.002178685739636421,
860
+ "teacher_loss": 0.0021052202209830284
861
+ },
862
+ {
863
+ "epoch": 0.12399077277970011,
864
+ "grad_norm": 1.3671875,
865
+ "kd_loss": 0.0547032505273819,
866
+ "learning_rate": 6.374727209737743e-06,
867
+ "loss": 0.0368,
868
+ "step": 430,
869
+ "student_loss": 0.05219801887869835,
870
+ "teacher_loss": 0.060560259968042374
871
+ },
872
+ {
873
+ "epoch": 0.1254325259515571,
874
+ "grad_norm": 1.453125,
875
+ "kd_loss": 0.01706705428659916,
876
+ "learning_rate": 6.296702034686726e-06,
877
+ "loss": 0.0301,
878
+ "step": 435,
879
+ "student_loss": 0.0008379022474400699,
880
+ "teacher_loss": 0.0006224109092727304
881
+ },
882
+ {
883
+ "epoch": 0.12687427912341406,
884
+ "grad_norm": 1.0,
885
+ "kd_loss": 0.03890637308359146,
886
+ "learning_rate": 6.218336822000598e-06,
887
+ "loss": 0.0314,
888
+ "step": 440,
889
+ "student_loss": 0.012142423540353775,
890
+ "teacher_loss": 0.007434291299432516
891
+ },
892
+ {
893
+ "epoch": 0.12831603229527105,
894
+ "grad_norm": 0.890625,
895
+ "kd_loss": 0.03179040178656578,
896
+ "learning_rate": 6.139652121598219e-06,
897
+ "loss": 0.0313,
898
+ "step": 445,
899
+ "student_loss": 0.03341586887836456,
900
+ "teacher_loss": 0.018860887736082077
901
+ },
902
+ {
903
+ "epoch": 0.12975778546712802,
904
+ "grad_norm": 0.69140625,
905
+ "kd_loss": 0.03629063814878464,
906
+ "learning_rate": 6.060668567178561e-06,
907
+ "loss": 0.0329,
908
+ "step": 450,
909
+ "student_loss": 0.0010012147249653935,
910
+ "teacher_loss": 0.0014839700888842344
911
+ },
912
+ {
913
+ "epoch": 0.131199538638985,
914
+ "grad_norm": 0.8828125,
915
+ "kd_loss": 0.02263510413467884,
916
+ "learning_rate": 5.981406870809889e-06,
917
+ "loss": 0.0326,
918
+ "step": 455,
919
+ "student_loss": 0.0007265008171088994,
920
+ "teacher_loss": 0.0003751025360543281
921
+ },
922
+ {
923
+ "epoch": 0.13264129181084197,
924
+ "grad_norm": 1.2109375,
925
+ "kd_loss": 0.0492476262152195,
926
+ "learning_rate": 5.9018878174983674e-06,
927
+ "loss": 0.0295,
928
+ "step": 460,
929
+ "student_loss": 0.005101657006889582,
930
+ "teacher_loss": 0.0002927044697571546
931
+ },
932
+ {
933
+ "epoch": 0.13408304498269896,
934
+ "grad_norm": 1.2578125,
935
+ "kd_loss": 0.031007954850792885,
936
+ "learning_rate": 5.822132259737565e-06,
937
+ "loss": 0.034,
938
+ "step": 465,
939
+ "student_loss": 0.00047535731573589146,
940
+ "teacher_loss": 0.00024468303308822215
941
+ },
942
+ {
943
+ "epoch": 0.13552479815455595,
944
+ "grad_norm": 1.7265625,
945
+ "kd_loss": 0.02650834433734417,
946
+ "learning_rate": 5.742161112040237e-06,
947
+ "loss": 0.0313,
948
+ "step": 470,
949
+ "student_loss": 0.000378905504476279,
950
+ "teacher_loss": 0.0002708406245801598
951
+ },
952
+ {
953
+ "epoch": 0.13696655132641292,
954
+ "grad_norm": 1.171875,
955
+ "kd_loss": 0.03383399918675423,
956
+ "learning_rate": 5.661995345453867e-06,
957
+ "loss": 0.0289,
958
+ "step": 475,
959
+ "student_loss": 0.0004789994563907385,
960
+ "teacher_loss": 0.0002899342798627913
961
+ },
962
+ {
963
+ "epoch": 0.1384083044982699,
964
+ "grad_norm": 1.1953125,
965
+ "kd_loss": 0.03584924340248108,
966
+ "learning_rate": 5.581655982061367e-06,
967
+ "loss": 0.0422,
968
+ "step": 480,
969
+ "student_loss": 0.03862505778670311,
970
+ "teacher_loss": 0.036061737686395645
971
+ },
972
+ {
973
+ "epoch": 0.13985005767012687,
974
+ "grad_norm": 1.203125,
975
+ "kd_loss": 0.020711667835712433,
976
+ "learning_rate": 5.501164089468406e-06,
977
+ "loss": 0.0313,
978
+ "step": 485,
979
+ "student_loss": 0.0004981390666216612,
980
+ "teacher_loss": 0.0005692155100405216
981
+ },
982
+ {
983
+ "epoch": 0.14129181084198386,
984
+ "grad_norm": 1.3984375,
985
+ "kd_loss": 0.027012551203370094,
986
+ "learning_rate": 5.4205407752787884e-06,
987
+ "loss": 0.0367,
988
+ "step": 490,
989
+ "student_loss": 0.0008920110412873328,
990
+ "teacher_loss": 0.0007303191814571619
991
+ },
992
+ {
993
+ "epoch": 0.14273356401384082,
994
+ "grad_norm": 1.3671875,
995
+ "kd_loss": 0.040799129754304886,
996
+ "learning_rate": 5.339807181559359e-06,
997
+ "loss": 0.0399,
998
+ "step": 495,
999
+ "student_loss": 0.0009312813053838909,
1000
+ "teacher_loss": 0.0005571586079895496
1001
+ },
1002
+ {
1003
+ "epoch": 0.14417531718569782,
1004
+ "grad_norm": 1.6875,
1005
+ "kd_loss": 0.03058801032602787,
1006
+ "learning_rate": 5.258984479295853e-06,
1007
+ "loss": 0.0316,
1008
+ "step": 500,
1009
+ "student_loss": 0.00041255459655076265,
1010
+ "teacher_loss": 0.00031951998244039714
1011
+ },
1012
+ {
1013
+ "epoch": 0.14561707035755478,
1014
+ "grad_norm": 1.734375,
1015
+ "kd_loss": 0.02832438424229622,
1016
+ "learning_rate": 5.1780938628411795e-06,
1017
+ "loss": 0.0402,
1018
+ "step": 505,
1019
+ "student_loss": 0.0005547971813939512,
1020
+ "teacher_loss": 0.0004580595705192536
1021
+ },
1022
+ {
1023
+ "epoch": 0.14705882352941177,
1024
+ "grad_norm": 2.28125,
1025
+ "kd_loss": 0.019691068679094315,
1026
+ "learning_rate": 5.097156544357567e-06,
1027
+ "loss": 0.032,
1028
+ "step": 510,
1029
+ "student_loss": 0.00035549805033952,
1030
+ "teacher_loss": 0.00029006152180954814
1031
+ },
1032
+ {
1033
+ "epoch": 0.14850057670126873,
1034
+ "grad_norm": 1.4609375,
1035
+ "kd_loss": 0.028027402237057686,
1036
+ "learning_rate": 5.016193748254045e-06,
1037
+ "loss": 0.0299,
1038
+ "step": 515,
1039
+ "student_loss": 0.016550345346331596,
1040
+ "teacher_loss": 0.00025012606056407094
1041
+ },
1042
+ {
1043
+ "epoch": 0.14994232987312572,
1044
+ "grad_norm": 1.7734375,
1045
+ "kd_loss": 0.06320768594741821,
1046
+ "learning_rate": 4.935226705620699e-06,
1047
+ "loss": 0.0344,
1048
+ "step": 520,
1049
+ "student_loss": 0.04100845754146576,
1050
+ "teacher_loss": 0.011567573063075542
1051
+ },
1052
+ {
1053
+ "epoch": 0.1513840830449827,
1054
+ "grad_norm": 0.859375,
1055
+ "kd_loss": 0.032060496509075165,
1056
+ "learning_rate": 4.8542766486612035e-06,
1057
+ "loss": 0.033,
1058
+ "step": 525,
1059
+ "student_loss": 0.0004020243068225682,
1060
+ "teacher_loss": 0.00020839735225308686
1061
+ },
1062
+ {
1063
+ "epoch": 0.15282583621683968,
1064
+ "grad_norm": 1.203125,
1065
+ "kd_loss": 0.024977991357445717,
1066
+ "learning_rate": 4.773364805125025e-06,
1067
+ "loss": 0.0309,
1068
+ "step": 530,
1069
+ "student_loss": 0.0005168431089259684,
1070
+ "teacher_loss": 0.00039579602889716625
1071
+ },
1072
+ {
1073
+ "epoch": 0.15426758938869667,
1074
+ "grad_norm": 2.046875,
1075
+ "kd_loss": 0.03925255313515663,
1076
+ "learning_rate": 4.6925123927408265e-06,
1077
+ "loss": 0.0308,
1078
+ "step": 535,
1079
+ "student_loss": 0.0007517871563322842,
1080
+ "teacher_loss": 0.00039392782491631806
1081
+ },
1082
+ {
1083
+ "epoch": 0.15570934256055363,
1084
+ "grad_norm": 1.8203125,
1085
+ "kd_loss": 0.05933701992034912,
1086
+ "learning_rate": 4.611740613652485e-06,
1087
+ "loss": 0.0286,
1088
+ "step": 540,
1089
+ "student_loss": 0.0131508968770504,
1090
+ "teacher_loss": 0.0005503469728864729
1091
+ },
1092
+ {
1093
+ "epoch": 0.15715109573241062,
1094
+ "grad_norm": 1.0390625,
1095
+ "kd_loss": 0.023665662854909897,
1096
+ "learning_rate": 4.531070648859186e-06,
1097
+ "loss": 0.0282,
1098
+ "step": 545,
1099
+ "student_loss": 0.0027271448634564877,
1100
+ "teacher_loss": 0.0027394567150622606
1101
+ },
1102
+ {
1103
+ "epoch": 0.15859284890426759,
1104
+ "grad_norm": 2.046875,
1105
+ "kd_loss": 0.05410351976752281,
1106
+ "learning_rate": 4.450523652661086e-06,
1107
+ "loss": 0.0276,
1108
+ "step": 550,
1109
+ "student_loss": 0.00023584096925333142,
1110
+ "teacher_loss": 0.0036266453098505735
1111
+ },
1112
+ {
1113
+ "epoch": 0.16003460207612458,
1114
+ "grad_norm": 1.1796875,
1115
+ "kd_loss": 0.08308840543031693,
1116
+ "learning_rate": 4.370120747111956e-06,
1117
+ "loss": 0.0327,
1118
+ "step": 555,
1119
+ "student_loss": 0.046959906816482544,
1120
+ "teacher_loss": 0.0017155115492641926
1121
+ },
1122
+ {
1123
+ "epoch": 0.16147635524798154,
1124
+ "grad_norm": 2.4375,
1125
+ "kd_loss": 0.0380968302488327,
1126
+ "learning_rate": 4.289883016480291e-06,
1127
+ "loss": 0.0349,
1128
+ "step": 560,
1129
+ "student_loss": 0.001130731194280088,
1130
+ "teacher_loss": 0.00036855213693343103
1131
+ },
1132
+ {
1133
+ "epoch": 0.16291810841983853,
1134
+ "grad_norm": 1.1328125,
1135
+ "kd_loss": 0.03884750232100487,
1136
+ "learning_rate": 4.209831501720328e-06,
1137
+ "loss": 0.0325,
1138
+ "step": 565,
1139
+ "student_loss": 0.02741910330951214,
1140
+ "teacher_loss": 0.019537584856152534
1141
+ },
1142
+ {
1143
+ "epoch": 0.1643598615916955,
1144
+ "grad_norm": 5.1875,
1145
+ "kd_loss": 0.0330473892390728,
1146
+ "learning_rate": 4.129987194954421e-06,
1147
+ "loss": 0.0356,
1148
+ "step": 570,
1149
+ "student_loss": 0.010371256619691849,
1150
+ "teacher_loss": 0.0070776245556771755
1151
+ },
1152
+ {
1153
+ "epoch": 0.16580161476355249,
1154
+ "grad_norm": 1.1015625,
1155
+ "kd_loss": 0.028591087087988853,
1156
+ "learning_rate": 4.050371033968216e-06,
1157
+ "loss": 0.0328,
1158
+ "step": 575,
1159
+ "student_loss": 0.000467249978100881,
1160
+ "teacher_loss": 0.0002494181098882109
1161
+ },
1162
+ {
1163
+ "epoch": 0.16724336793540945,
1164
+ "grad_norm": 0.80078125,
1165
+ "kd_loss": 0.02640584297478199,
1166
+ "learning_rate": 3.9710038967200825e-06,
1167
+ "loss": 0.0284,
1168
+ "step": 580,
1169
+ "student_loss": 0.0017712963744997978,
1170
+ "teacher_loss": 0.0019773358944803476
1171
+ },
1172
+ {
1173
+ "epoch": 0.16868512110726644,
1174
+ "grad_norm": 0.85546875,
1175
+ "kd_loss": 0.030399348586797714,
1176
+ "learning_rate": 3.89190659586623e-06,
1177
+ "loss": 0.0281,
1178
+ "step": 585,
1179
+ "student_loss": 0.0004607080190908164,
1180
+ "teacher_loss": 0.0002647584769874811
1181
+ },
1182
+ {
1183
+ "epoch": 0.1701268742791234,
1184
+ "grad_norm": 1.1328125,
1185
+ "kd_loss": 0.03226058557629585,
1186
+ "learning_rate": 3.8130998733029517e-06,
1187
+ "loss": 0.0302,
1188
+ "step": 590,
1189
+ "student_loss": 0.0005413145408965647,
1190
+ "teacher_loss": 0.0011012081522494555
1191
+ },
1192
+ {
1193
+ "epoch": 0.1715686274509804,
1194
+ "grad_norm": 1.8203125,
1195
+ "kd_loss": 0.023172084242105484,
1196
+ "learning_rate": 3.734604394727419e-06,
1197
+ "loss": 0.0325,
1198
+ "step": 595,
1199
+ "student_loss": 0.00059064372908324,
1200
+ "teacher_loss": 0.0005090903723612428
1201
+ },
1202
+ {
1203
+ "epoch": 0.17301038062283736,
1204
+ "grad_norm": 1.1640625,
1205
+ "kd_loss": 0.038484539836645126,
1206
+ "learning_rate": 3.656440744218464e-06,
1207
+ "loss": 0.0346,
1208
+ "step": 600,
1209
+ "student_loss": 0.041177455335855484,
1210
+ "teacher_loss": 0.004930450581014156
1211
+ },
1212
+ {
1213
+ "epoch": 0.17445213379469435,
1214
+ "grad_norm": 2.515625,
1215
+ "kd_loss": 0.04977564886212349,
1216
+ "learning_rate": 3.578629418838757e-06,
1217
+ "loss": 0.0334,
1218
+ "step": 605,
1219
+ "student_loss": 0.07967430353164673,
1220
+ "teacher_loss": 0.03648905083537102
1221
+ },
1222
+ {
1223
+ "epoch": 0.17589388696655134,
1224
+ "grad_norm": 1.34375,
1225
+ "kd_loss": 0.04043799638748169,
1226
+ "learning_rate": 3.5011908232598124e-06,
1227
+ "loss": 0.0305,
1228
+ "step": 610,
1229
+ "student_loss": 0.00654568150639534,
1230
+ "teacher_loss": 0.00039838344673626125
1231
+ },
1232
+ {
1233
+ "epoch": 0.1773356401384083,
1234
+ "grad_norm": 1.109375,
1235
+ "kd_loss": 0.023836029693484306,
1236
+ "learning_rate": 3.4241452644112085e-06,
1237
+ "loss": 0.0266,
1238
+ "step": 615,
1239
+ "student_loss": 0.0002936046803370118,
1240
+ "teacher_loss": 0.0003476935962680727
1241
+ },
1242
+ {
1243
+ "epoch": 0.1787773933102653,
1244
+ "grad_norm": 1.1640625,
1245
+ "kd_loss": 0.033266592770814896,
1246
+ "learning_rate": 3.3475129461554567e-06,
1247
+ "loss": 0.0375,
1248
+ "step": 620,
1249
+ "student_loss": 0.01093846745789051,
1250
+ "teacher_loss": 0.007155448663979769
1251
+ },
1252
+ {
1253
+ "epoch": 0.18021914648212226,
1254
+ "grad_norm": 1.4609375,
1255
+ "kd_loss": 0.02975156530737877,
1256
+ "learning_rate": 3.271313963989886e-06,
1257
+ "loss": 0.0286,
1258
+ "step": 625,
1259
+ "student_loss": 0.014491051435470581,
1260
+ "teacher_loss": 0.004247736185789108
1261
+ },
1262
+ {
1263
+ "epoch": 0.18166089965397925,
1264
+ "grad_norm": 0.98046875,
1265
+ "kd_loss": 0.04514408856630325,
1266
+ "learning_rate": 3.195568299776945e-06,
1267
+ "loss": 0.0303,
1268
+ "step": 630,
1269
+ "student_loss": 0.0005301684141159058,
1270
+ "teacher_loss": 0.0009141535847447813
1271
+ },
1272
+ {
1273
+ "epoch": 0.1831026528258362,
1274
+ "grad_norm": 0.82421875,
1275
+ "kd_loss": 0.03230506554245949,
1276
+ "learning_rate": 3.1202958165043053e-06,
1277
+ "loss": 0.0417,
1278
+ "step": 635,
1279
+ "student_loss": 0.0006954488926567137,
1280
+ "teacher_loss": 0.0002712230198085308
1281
+ },
1282
+ {
1283
+ "epoch": 0.1845444059976932,
1284
+ "grad_norm": 1.46875,
1285
+ "kd_loss": 0.02938377857208252,
1286
+ "learning_rate": 3.045516253076137e-06,
1287
+ "loss": 0.0276,
1288
+ "step": 640,
1289
+ "student_loss": 0.0007398583693429828,
1290
+ "teacher_loss": 0.00036641399492509663
1291
+ },
1292
+ {
1293
+ "epoch": 0.18598615916955016,
1294
+ "grad_norm": 0.90234375,
1295
+ "kd_loss": 0.03232554718852043,
1296
+ "learning_rate": 2.9712492191369245e-06,
1297
+ "loss": 0.0279,
1298
+ "step": 645,
1299
+ "student_loss": 0.0006595580489374697,
1300
+ "teacher_loss": 0.0004537871282082051
1301
+ },
1302
+ {
1303
+ "epoch": 0.18742791234140715,
1304
+ "grad_norm": 0.72265625,
1305
+ "kd_loss": 0.02090008556842804,
1306
+ "learning_rate": 2.8975141899291777e-06,
1307
+ "loss": 0.0271,
1308
+ "step": 650,
1309
+ "student_loss": 0.000823685375507921,
1310
+ "teacher_loss": 0.0008547761826775968
1311
+ },
1312
+ {
1313
+ "epoch": 0.18886966551326412,
1314
+ "grad_norm": 1.0546875,
1315
+ "kd_loss": 0.04636139050126076,
1316
+ "learning_rate": 2.8243305011863843e-06,
1317
+ "loss": 0.0281,
1318
+ "step": 655,
1319
+ "student_loss": 0.0005695814033970237,
1320
+ "teacher_loss": 0.00031179824145510793
1321
+ },
1322
+ {
1323
+ "epoch": 0.1903114186851211,
1324
+ "grad_norm": 1.6875,
1325
+ "kd_loss": 0.024737229570746422,
1326
+ "learning_rate": 2.751717344062552e-06,
1327
+ "loss": 0.0316,
1328
+ "step": 660,
1329
+ "student_loss": 0.0006874947575852275,
1330
+ "teacher_loss": 0.002697322051972151
1331
+ },
1332
+ {
1333
+ "epoch": 0.19175317185697807,
1334
+ "grad_norm": 0.953125,
1335
+ "kd_loss": 0.06534231454133987,
1336
+ "learning_rate": 2.6796937600996587e-06,
1337
+ "loss": 0.0336,
1338
+ "step": 665,
1339
+ "student_loss": 0.0013298896374180913,
1340
+ "teacher_loss": 0.0003429361677262932
1341
+ },
1342
+ {
1343
+ "epoch": 0.19319492502883506,
1344
+ "grad_norm": 1.3046875,
1345
+ "kd_loss": 0.03078162856400013,
1346
+ "learning_rate": 2.6082786362343377e-06,
1347
+ "loss": 0.035,
1348
+ "step": 670,
1349
+ "student_loss": 0.0029936288483440876,
1350
+ "teacher_loss": 0.0007150344317778945
1351
+ },
1352
+ {
1353
+ "epoch": 0.19463667820069205,
1354
+ "grad_norm": 0.94140625,
1355
+ "kd_loss": 0.019795667380094528,
1356
+ "learning_rate": 2.5374906998451094e-06,
1357
+ "loss": 0.0333,
1358
+ "step": 675,
1359
+ "student_loss": 0.0008592633530497551,
1360
+ "teacher_loss": 0.0008608284406363964
1361
+ },
1362
+ {
1363
+ "epoch": 0.19607843137254902,
1364
+ "grad_norm": 1.34375,
1365
+ "kd_loss": 0.061516523361206055,
1366
+ "learning_rate": 2.467348513841447e-06,
1367
+ "loss": 0.0279,
1368
+ "step": 680,
1369
+ "student_loss": 0.11378592997789383,
1370
+ "teacher_loss": 0.04401650279760361
1371
+ },
1372
+ {
1373
+ "epoch": 0.197520184544406,
1374
+ "grad_norm": 1.1953125,
1375
+ "kd_loss": 0.02606314793229103,
1376
+ "learning_rate": 2.3978704717959777e-06,
1377
+ "loss": 0.0311,
1378
+ "step": 685,
1379
+ "student_loss": 0.037850309163331985,
1380
+ "teacher_loss": 0.0306412260979414
1381
+ },
1382
+ {
1383
+ "epoch": 0.19896193771626297,
1384
+ "grad_norm": 2.078125,
1385
+ "kd_loss": 0.05581173300743103,
1386
+ "learning_rate": 2.329074793121085e-06,
1387
+ "loss": 0.0321,
1388
+ "step": 690,
1389
+ "student_loss": 0.007378917653113604,
1390
+ "teacher_loss": 0.01118262205272913
1391
+ },
1392
+ {
1393
+ "epoch": 0.20040369088811996,
1394
+ "grad_norm": 0.984375,
1395
+ "kd_loss": 0.035412125289440155,
1396
+ "learning_rate": 2.260979518291186e-06,
1397
+ "loss": 0.0349,
1398
+ "step": 695,
1399
+ "student_loss": 0.0029192951042205095,
1400
+ "teacher_loss": 0.007233661133795977
1401
+ },
1402
+ {
1403
+ "epoch": 0.20184544405997693,
1404
+ "grad_norm": 1.3359375,
1405
+ "kd_loss": 0.024152226746082306,
1406
+ "learning_rate": 2.1936025041119268e-06,
1407
+ "loss": 0.031,
1408
+ "step": 700,
1409
+ "student_loss": 0.0007802582113072276,
1410
+ "teacher_loss": 0.0007531556766480207
1411
+ },
1412
+ {
1413
+ "epoch": 0.20328719723183392,
1414
+ "grad_norm": 1.796875,
1415
+ "kd_loss": 0.02112882025539875,
1416
+ "learning_rate": 2.1269614190375477e-06,
1417
+ "loss": 0.0307,
1418
+ "step": 705,
1419
+ "student_loss": 0.001995673170313239,
1420
+ "teacher_loss": 0.0013640234246850014
1421
+ },
1422
+ {
1423
+ "epoch": 0.20472895040369088,
1424
+ "grad_norm": 0.76171875,
1425
+ "kd_loss": 0.060643333941698074,
1426
+ "learning_rate": 2.061073738537635e-06,
1427
+ "loss": 0.0335,
1428
+ "step": 710,
1429
+ "student_loss": 0.005931171588599682,
1430
+ "teacher_loss": 0.00436427490785718
1431
+ },
1432
+ {
1433
+ "epoch": 0.20617070357554787,
1434
+ "grad_norm": 0.9453125,
1435
+ "kd_loss": 0.025081951171159744,
1436
+ "learning_rate": 1.9959567405144825e-06,
1437
+ "loss": 0.0378,
1438
+ "step": 715,
1439
+ "student_loss": 0.0024764742702245712,
1440
+ "teacher_loss": 0.005250995047390461
1441
+ },
1442
+ {
1443
+ "epoch": 0.20761245674740483,
1444
+ "grad_norm": 0.9765625,
1445
+ "kd_loss": 0.023644844070076942,
1446
+ "learning_rate": 1.931627500772263e-06,
1447
+ "loss": 0.0328,
1448
+ "step": 720,
1449
+ "student_loss": 0.0015447784680873156,
1450
+ "teacher_loss": 0.0032466305419802666
1451
+ },
1452
+ {
1453
+ "epoch": 0.20905420991926182,
1454
+ "grad_norm": 0.828125,
1455
+ "kd_loss": 0.02127092145383358,
1456
+ "learning_rate": 1.8681028885391905e-06,
1457
+ "loss": 0.0261,
1458
+ "step": 725,
1459
+ "student_loss": 0.00040122531936503947,
1460
+ "teacher_loss": 0.00032715691486373544
1461
+ },
1462
+ {
1463
+ "epoch": 0.2104959630911188,
1464
+ "grad_norm": 1.21875,
1465
+ "kd_loss": 0.03333493322134018,
1466
+ "learning_rate": 1.8053995620438625e-06,
1467
+ "loss": 0.0308,
1468
+ "step": 730,
1469
+ "student_loss": 0.0026169014163315296,
1470
+ "teacher_loss": 0.0015284213004633784
1471
+ },
1472
+ {
1473
+ "epoch": 0.21193771626297578,
1474
+ "grad_norm": 1.2578125,
1475
+ "kd_loss": 0.03426002338528633,
1476
+ "learning_rate": 1.743533964146924e-06,
1477
+ "loss": 0.0366,
1478
+ "step": 735,
1479
+ "student_loss": 0.0005074241198599339,
1480
+ "teacher_loss": 0.0002774895401671529
1481
+ },
1482
+ {
1483
+ "epoch": 0.21337946943483277,
1484
+ "grad_norm": 1.0703125,
1485
+ "kd_loss": 0.03785933926701546,
1486
+ "learning_rate": 1.6825223180292138e-06,
1487
+ "loss": 0.0288,
1488
+ "step": 740,
1489
+ "student_loss": 0.0019946948159486055,
1490
+ "teacher_loss": 0.00017226416093762964
1491
+ },
1492
+ {
1493
+ "epoch": 0.21482122260668973,
1494
+ "grad_norm": 0.98828125,
1495
+ "kd_loss": 0.024361569434404373,
1496
+ "learning_rate": 1.6223806229375182e-06,
1497
+ "loss": 0.0301,
1498
+ "step": 745,
1499
+ "student_loss": 0.04029940441250801,
1500
+ "teacher_loss": 0.018720334395766258
1501
+ },
1502
+ {
1503
+ "epoch": 0.21626297577854672,
1504
+ "grad_norm": 1.671875,
1505
+ "kd_loss": 0.06201966851949692,
1506
+ "learning_rate": 1.563124649989043e-06,
1507
+ "loss": 0.0294,
1508
+ "step": 750,
1509
+ "student_loss": 0.01399194449186325,
1510
+ "teacher_loss": 0.004409614019095898
1511
+ },
1512
+ {
1513
+ "epoch": 0.2177047289504037,
1514
+ "grad_norm": 1.7421875,
1515
+ "kd_loss": 0.024494940415024757,
1516
+ "learning_rate": 1.5047699380357134e-06,
1517
+ "loss": 0.035,
1518
+ "step": 755,
1519
+ "student_loss": 0.0298053789883852,
1520
+ "teacher_loss": 0.006942529231309891
1521
+ },
1522
+ {
1523
+ "epoch": 0.21914648212226068,
1524
+ "grad_norm": 1.5234375,
1525
+ "kd_loss": 0.06555049866437912,
1526
+ "learning_rate": 1.4473317895893773e-06,
1527
+ "loss": 0.0292,
1528
+ "step": 760,
1529
+ "student_loss": 0.054234448820352554,
1530
+ "teacher_loss": 0.0010094160679727793
1531
+ },
1532
+ {
1533
+ "epoch": 0.22058823529411764,
1534
+ "grad_norm": 1.1015625,
1535
+ "kd_loss": 0.050752874463796616,
1536
+ "learning_rate": 1.39082526680899e-06,
1537
+ "loss": 0.0358,
1538
+ "step": 765,
1539
+ "student_loss": 0.0037405432667583227,
1540
+ "teacher_loss": 0.0012883292511105537
1541
+ },
1542
+ {
1543
+ "epoch": 0.22202998846597463,
1544
+ "grad_norm": 1.4609375,
1545
+ "kd_loss": 0.05563059076666832,
1546
+ "learning_rate": 1.3352651875508204e-06,
1547
+ "loss": 0.0312,
1548
+ "step": 770,
1549
+ "student_loss": 0.0009953544940799475,
1550
+ "teacher_loss": 0.0020596326794475317
1551
+ },
1552
+ {
1553
+ "epoch": 0.2234717416378316,
1554
+ "grad_norm": 1.625,
1555
+ "kd_loss": 0.03560180589556694,
1556
+ "learning_rate": 1.2806661214827286e-06,
1557
+ "loss": 0.0281,
1558
+ "step": 775,
1559
+ "student_loss": 0.002246091840788722,
1560
+ "teacher_loss": 0.0005796651821583509
1561
+ },
1562
+ {
1563
+ "epoch": 0.22491349480968859,
1564
+ "grad_norm": 1.2421875,
1565
+ "kd_loss": 0.029239855706691742,
1566
+ "learning_rate": 1.2270423862635188e-06,
1567
+ "loss": 0.0322,
1568
+ "step": 780,
1569
+ "student_loss": 0.0003545498475432396,
1570
+ "teacher_loss": 0.00024268838751595467
1571
+ },
1572
+ {
1573
+ "epoch": 0.22635524798154555,
1574
+ "grad_norm": 1.53125,
1575
+ "kd_loss": 0.019190076738595963,
1576
+ "learning_rate": 1.1744080437883859e-06,
1577
+ "loss": 0.0288,
1578
+ "step": 785,
1579
+ "student_loss": 0.0004118382930755615,
1580
+ "teacher_loss": 0.0013559224316850305
1581
+ },
1582
+ {
1583
+ "epoch": 0.22779700115340254,
1584
+ "grad_norm": 1.71875,
1585
+ "kd_loss": 0.01850035786628723,
1586
+ "learning_rate": 1.1227768965014246e-06,
1587
+ "loss": 0.0274,
1588
+ "step": 790,
1589
+ "student_loss": 0.006309983320534229,
1590
+ "teacher_loss": 0.004283738788217306
1591
+ },
1592
+ {
1593
+ "epoch": 0.2292387543252595,
1594
+ "grad_norm": 1.1796875,
1595
+ "kd_loss": 0.029416479170322418,
1596
+ "learning_rate": 1.0721624837761768e-06,
1597
+ "loss": 0.0311,
1598
+ "step": 795,
1599
+ "student_loss": 0.0004396810254547745,
1600
+ "teacher_loss": 0.0006457470590248704
1601
+ },
1602
+ {
1603
+ "epoch": 0.2306805074971165,
1604
+ "grad_norm": 1.203125,
1605
+ "kd_loss": 0.04670260474085808,
1606
+ "learning_rate": 1.0225780783651689e-06,
1607
+ "loss": 0.0362,
1608
+ "step": 800,
1609
+ "student_loss": 0.0542101114988327,
1610
+ "teacher_loss": 0.0266594085842371
1611
+ },
1612
+ {
1613
+ "epoch": 0.23212226066897348,
1614
+ "grad_norm": 1.25,
1615
+ "kd_loss": 0.019798053428530693,
1616
+ "learning_rate": 9.740366829193587e-07,
1617
+ "loss": 0.0339,
1618
+ "step": 805,
1619
+ "student_loss": 0.00048328418051823974,
1620
+ "teacher_loss": 0.00044713987153954804
1621
+ },
1622
+ {
1623
+ "epoch": 0.23356401384083045,
1624
+ "grad_norm": 6.125,
1625
+ "kd_loss": 0.03847261890769005,
1626
+ "learning_rate": 9.265510265784189e-07,
1627
+ "loss": 0.0292,
1628
+ "step": 810,
1629
+ "student_loss": 0.0004956105258315802,
1630
+ "teacher_loss": 0.00024124003539327532
1631
+ },
1632
+ {
1633
+ "epoch": 0.23500576701268744,
1634
+ "grad_norm": 1.8671875,
1635
+ "kd_loss": 0.02178817056119442,
1636
+ "learning_rate": 8.801335616327378e-07,
1637
+ "loss": 0.0305,
1638
+ "step": 815,
1639
+ "student_loss": 0.0008903697016648948,
1640
+ "teacher_loss": 0.01014482881873846
1641
+ },
1642
+ {
1643
+ "epoch": 0.2364475201845444,
1644
+ "grad_norm": 1.6328125,
1645
+ "kd_loss": 0.036839839071035385,
1646
+ "learning_rate": 8.347964602580245e-07,
1647
+ "loss": 0.0308,
1648
+ "step": 820,
1649
+ "student_loss": 0.008828239515423775,
1650
+ "teacher_loss": 0.0019727114122360945
1651
+ },
1652
+ {
1653
+ "epoch": 0.2378892733564014,
1654
+ "grad_norm": 1.828125,
1655
+ "kd_loss": 0.03977164626121521,
1656
+ "learning_rate": 7.905516113233652e-07,
1657
+ "loss": 0.0289,
1658
+ "step": 825,
1659
+ "student_loss": 0.0008772791479714215,
1660
+ "teacher_loss": 0.016588712111115456
1661
+ },
1662
+ {
1663
+ "epoch": 0.23933102652825836,
1664
+ "grad_norm": 0.8359375,
1665
+ "kd_loss": 0.03716457635164261,
1666
+ "learning_rate": 7.474106172735746e-07,
1667
+ "loss": 0.0306,
1668
+ "step": 830,
1669
+ "student_loss": 0.007541711442172527,
1670
+ "teacher_loss": 0.004410990513861179
1671
+ },
1672
+ {
1673
+ "epoch": 0.24077277970011535,
1674
+ "grad_norm": 1.2890625,
1675
+ "kd_loss": 0.028623711317777634,
1676
+ "learning_rate": 7.053847910866513e-07,
1677
+ "loss": 0.0283,
1678
+ "step": 835,
1679
+ "student_loss": 0.01320332009345293,
1680
+ "teacher_loss": 0.004977709148079157
1681
+ },
1682
+ {
1683
+ "epoch": 0.2422145328719723,
1684
+ "grad_norm": 1.2265625,
1685
+ "kd_loss": 0.038853954523801804,
1686
+ "learning_rate": 6.644851533071556e-07,
1687
+ "loss": 0.0303,
1688
+ "step": 840,
1689
+ "student_loss": 0.0004099447396583855,
1690
+ "teacher_loss": 0.00023886460985522717
1691
+ },
1692
+ {
1693
+ "epoch": 0.2436562860438293,
1694
+ "grad_norm": 1.59375,
1695
+ "kd_loss": 0.026872437447309494,
1696
+ "learning_rate": 6.24722429156251e-07,
1697
+ "loss": 0.0455,
1698
+ "step": 845,
1699
+ "student_loss": 0.0005929277976974845,
1700
+ "teacher_loss": 0.0004925825633108616
1701
+ },
1702
+ {
1703
+ "epoch": 0.24509803921568626,
1704
+ "grad_norm": 1.953125,
1705
+ "kd_loss": 0.045193299651145935,
1706
+ "learning_rate": 5.861070457192081e-07,
1707
+ "loss": 0.0324,
1708
+ "step": 850,
1709
+ "student_loss": 0.01363430917263031,
1710
+ "teacher_loss": 0.00036361668026074767
1711
+ },
1712
+ {
1713
+ "epoch": 0.24653979238754326,
1714
+ "grad_norm": 0.83203125,
1715
+ "kd_loss": 0.029308617115020752,
1716
+ "learning_rate": 5.486491292110796e-07,
1717
+ "loss": 0.0307,
1718
+ "step": 855,
1719
+ "student_loss": 0.00042216500150971115,
1720
+ "teacher_loss": 0.00046156966709531844
1721
+ },
1722
+ {
1723
+ "epoch": 0.24798154555940022,
1724
+ "grad_norm": 1.5625,
1725
+ "kd_loss": 0.028431078419089317,
1726
+ "learning_rate": 5.123585023212785e-07,
1727
+ "loss": 0.0361,
1728
+ "step": 860,
1729
+ "student_loss": 0.0017358119366690516,
1730
+ "teacher_loss": 0.0008302436908707023
1731
+ },
1732
+ {
1733
+ "epoch": 0.2494232987312572,
1734
+ "grad_norm": 1.859375,
1735
+ "kd_loss": 0.03732982277870178,
1736
+ "learning_rate": 4.772446816377408e-07,
1737
+ "loss": 0.0322,
1738
+ "step": 865,
1739
+ "student_loss": 0.0003742146072909236,
1740
+ "teacher_loss": 0.0004992606700398028
1741
+ },
1742
+ {
1743
+ "epoch": 0.2508650519031142,
1744
+ "grad_norm": 1.2109375,
1745
+ "kd_loss": 0.045105963945388794,
1746
+ "learning_rate": 4.4331687515137614e-07,
1747
+ "loss": 0.032,
1748
+ "step": 870,
1749
+ "student_loss": 0.0020948813762515783,
1750
+ "teacher_loss": 0.0004395451978780329
1751
+ },
1752
+ {
1753
+ "epoch": 0.25230680507497116,
1754
+ "grad_norm": 1.1484375,
1755
+ "kd_loss": 0.028816962614655495,
1756
+ "learning_rate": 4.1058397984142405e-07,
1757
+ "loss": 0.0309,
1758
+ "step": 875,
1759
+ "student_loss": 0.00026272557443007827,
1760
+ "teacher_loss": 0.00027639055042527616
1761
+ },
1762
+ {
1763
+ "epoch": 0.2537485582468281,
1764
+ "grad_norm": 1.7265625,
1765
+ "kd_loss": 0.036270152777433395,
1766
+ "learning_rate": 3.790545793423761e-07,
1767
+ "loss": 0.0331,
1768
+ "step": 880,
1769
+ "student_loss": 0.0007760376902297139,
1770
+ "teacher_loss": 0.0004516106564551592
1771
+ },
1772
+ {
1773
+ "epoch": 0.25519031141868515,
1774
+ "grad_norm": 1.6875,
1775
+ "kd_loss": 0.07943305373191833,
1776
+ "learning_rate": 3.4873694169306915e-07,
1777
+ "loss": 0.0268,
1778
+ "step": 885,
1779
+ "student_loss": 0.014952230267226696,
1780
+ "teacher_loss": 0.05265064164996147
1781
+ },
1782
+ {
1783
+ "epoch": 0.2566320645905421,
1784
+ "grad_norm": 1.296875,
1785
+ "kd_loss": 0.022127894684672356,
1786
+ "learning_rate": 3.196390171685343e-07,
1787
+ "loss": 0.0328,
1788
+ "step": 890,
1789
+ "student_loss": 0.000466528203105554,
1790
+ "teacher_loss": 0.00047626433661207557
1791
+ },
1792
+ {
1793
+ "epoch": 0.25807381776239907,
1794
+ "grad_norm": 1.359375,
1795
+ "kd_loss": 0.044547051191329956,
1796
+ "learning_rate": 2.917684361951728e-07,
1797
+ "loss": 0.0313,
1798
+ "step": 895,
1799
+ "student_loss": 0.012537115253508091,
1800
+ "teacher_loss": 0.004908856004476547
1801
+ },
1802
+ {
1803
+ "epoch": 0.25951557093425603,
1804
+ "grad_norm": 1.2265625,
1805
+ "kd_loss": 0.030574947595596313,
1806
+ "learning_rate": 2.65132507349814e-07,
1807
+ "loss": 0.0349,
1808
+ "step": 900,
1809
+ "student_loss": 0.02907959371805191,
1810
+ "teacher_loss": 0.021374521777033806
1811
+ },
1812
+ {
1813
+ "epoch": 0.26095732410611305,
1814
+ "grad_norm": 0.87890625,
1815
+ "kd_loss": 0.02914801612496376,
1816
+ "learning_rate": 2.397382154431621e-07,
1817
+ "loss": 0.0301,
1818
+ "step": 905,
1819
+ "student_loss": 0.0034685099963098764,
1820
+ "teacher_loss": 0.002060187980532646
1821
+ },
1822
+ {
1823
+ "epoch": 0.26239907727797,
1824
+ "grad_norm": 1.984375,
1825
+ "kd_loss": 0.0164653267711401,
1826
+ "learning_rate": 2.1559221968815547e-07,
1827
+ "loss": 0.0285,
1828
+ "step": 910,
1829
+ "student_loss": 0.0004796285356860608,
1830
+ "teacher_loss": 0.0005832899478264153
1831
+ },
1832
+ {
1833
+ "epoch": 0.263840830449827,
1834
+ "grad_norm": 1.3046875,
1835
+ "kd_loss": 0.049093734472990036,
1836
+ "learning_rate": 1.9270085195370048e-07,
1837
+ "loss": 0.0291,
1838
+ "step": 915,
1839
+ "student_loss": 0.0019156670896336436,
1840
+ "teacher_loss": 0.0006669871509075165
1841
+ },
1842
+ {
1843
+ "epoch": 0.26528258362168394,
1844
+ "grad_norm": 0.77734375,
1845
+ "kd_loss": 0.024058707058429718,
1846
+ "learning_rate": 1.7107011510424766e-07,
1847
+ "loss": 0.0306,
1848
+ "step": 920,
1849
+ "student_loss": 0.014235829003155231,
1850
+ "teacher_loss": 0.012093200348317623
1851
+ },
1852
+ {
1853
+ "epoch": 0.26672433679354096,
1854
+ "grad_norm": 0.78515625,
1855
+ "kd_loss": 0.020182941108942032,
1856
+ "learning_rate": 1.5070568142564912e-07,
1857
+ "loss": 0.0286,
1858
+ "step": 925,
1859
+ "student_loss": 0.0004554464831016958,
1860
+ "teacher_loss": 0.00047793317935429513
1861
+ },
1862
+ {
1863
+ "epoch": 0.2681660899653979,
1864
+ "grad_norm": 1.1640625,
1865
+ "kd_loss": 0.030689207836985588,
1866
+ "learning_rate": 1.3161289113769405e-07,
1867
+ "loss": 0.0296,
1868
+ "step": 930,
1869
+ "student_loss": 0.015923185274004936,
1870
+ "teacher_loss": 0.0007836474105715752
1871
+ },
1872
+ {
1873
+ "epoch": 0.2696078431372549,
1874
+ "grad_norm": 1.1015625,
1875
+ "kd_loss": 0.026608861982822418,
1876
+ "learning_rate": 1.1379675099373489e-07,
1877
+ "loss": 0.0283,
1878
+ "step": 935,
1879
+ "student_loss": 0.0022123996168375015,
1880
+ "teacher_loss": 0.0022658726666122675
1881
+ },
1882
+ {
1883
+ "epoch": 0.2710495963091119,
1884
+ "grad_norm": 1.0625,
1885
+ "kd_loss": 0.04891452193260193,
1886
+ "learning_rate": 9.726193296774767e-08,
1887
+ "loss": 0.032,
1888
+ "step": 940,
1889
+ "student_loss": 0.007106420584022999,
1890
+ "teacher_loss": 0.001212292816489935
1891
+ },
1892
+ {
1893
+ "epoch": 0.27249134948096887,
1894
+ "grad_norm": 1.03125,
1895
+ "kd_loss": 0.02279752679169178,
1896
+ "learning_rate": 8.201277302919086e-08,
1897
+ "loss": 0.0335,
1898
+ "step": 945,
1899
+ "student_loss": 0.009904825128614902,
1900
+ "teacher_loss": 0.0015650882851332426
1901
+ },
1902
+ {
1903
+ "epoch": 0.27393310265282583,
1904
+ "grad_norm": 0.875,
1905
+ "kd_loss": 0.031176593154668808,
1906
+ "learning_rate": 6.805327000596995e-08,
1907
+ "loss": 0.0286,
1908
+ "step": 950,
1909
+ "student_loss": 0.0004509532009251416,
1910
+ "teacher_loss": 0.00026176305254921317
1911
+ },
1912
+ {
1913
+ "epoch": 0.2753748558246828,
1914
+ "grad_norm": 1.1875,
1915
+ "kd_loss": 0.029909612610936165,
1916
+ "learning_rate": 5.538708453581787e-08,
1917
+ "loss": 0.0302,
1918
+ "step": 955,
1919
+ "student_loss": 0.0009037847630679607,
1920
+ "teacher_loss": 0.0007386825745925307
1921
+ },
1922
+ {
1923
+ "epoch": 0.2768166089965398,
1924
+ "grad_norm": 1.8515625,
1925
+ "kd_loss": 0.03285350278019905,
1926
+ "learning_rate": 4.40175381063529e-08,
1927
+ "loss": 0.0321,
1928
+ "step": 960,
1929
+ "student_loss": 0.0016368315555155277,
1930
+ "teacher_loss": 0.0012363821733742952
1931
+ },
1932
+ {
1933
+ "epoch": 0.2782583621683968,
1934
+ "grad_norm": 1.3515625,
1935
+ "kd_loss": 0.038273654878139496,
1936
+ "learning_rate": 3.394761218407705e-08,
1937
+ "loss": 0.0286,
1938
+ "step": 965,
1939
+ "student_loss": 0.010774951428174973,
1940
+ "teacher_loss": 0.006840071175247431
1941
+ },
1942
+ {
1943
+ "epoch": 0.27970011534025374,
1944
+ "grad_norm": 1.1796875,
1945
+ "kd_loss": 0.033728428184986115,
1946
+ "learning_rate": 2.5179947432540376e-08,
1947
+ "loss": 0.0303,
1948
+ "step": 970,
1949
+ "student_loss": 0.0003708581207320094,
1950
+ "teacher_loss": 0.00017761511844582856
1951
+ },
1952
+ {
1953
+ "epoch": 0.2811418685121107,
1954
+ "grad_norm": 2.46875,
1955
+ "kd_loss": 0.025371067225933075,
1956
+ "learning_rate": 1.7716843019867646e-08,
1957
+ "loss": 0.0281,
1958
+ "step": 975,
1959
+ "student_loss": 0.013592376373708248,
1960
+ "teacher_loss": 0.0011630237568169832
1961
+ },
1962
+ {
1963
+ "epoch": 0.2825836216839677,
1964
+ "grad_norm": 0.90625,
1965
+ "kd_loss": 0.03571967035531998,
1966
+ "learning_rate": 1.156025601584676e-08,
1967
+ "loss": 0.0317,
1968
+ "step": 980,
1969
+ "student_loss": 0.00045868984307162464,
1970
+ "teacher_loss": 0.0002273312275065109
1971
+ },
1972
+ {
1973
+ "epoch": 0.2840253748558247,
1974
+ "grad_norm": 1.3046875,
1975
+ "kd_loss": 0.030186688527464867,
1976
+ "learning_rate": 6.711800878718144e-09,
1977
+ "loss": 0.0318,
1978
+ "step": 985,
1979
+ "student_loss": 0.0002502851712051779,
1980
+ "teacher_loss": 0.0002345747489016503
1981
+ },
1982
+ {
1983
+ "epoch": 0.28546712802768165,
1984
+ "grad_norm": 1.671875,
1985
+ "kd_loss": 0.0296288151293993,
1986
+ "learning_rate": 3.1727490318111953e-09,
1987
+ "loss": 0.0292,
1988
+ "step": 990,
1989
+ "student_loss": 0.0013771216617897153,
1990
+ "teacher_loss": 0.00027968850918114185
1991
+ },
1992
+ {
1993
+ "epoch": 0.2869088811995386,
1994
+ "grad_norm": 0.953125,
1995
+ "kd_loss": 0.04763549193739891,
1996
+ "learning_rate": 9.440285301370865e-10,
1997
+ "loss": 0.0333,
1998
+ "step": 995,
1999
+ "student_loss": 0.0019419729942455888,
2000
+ "teacher_loss": 0.0009000069694593549
2001
+ },
2002
+ {
2003
+ "epoch": 0.28835063437139563,
2004
+ "grad_norm": 1.1328125,
2005
+ "kd_loss": 0.03819262236356735,
2006
+ "learning_rate": 2.622381702066523e-11,
2007
+ "loss": 0.0293,
2008
+ "step": 1000,
2009
+ "student_loss": 0.006347914692014456,
2010
+ "teacher_loss": 0.00022976804757490754
2011
+ },
2012
+ {
2013
+ "epoch": 0.28835063437139563,
2014
+ "kd_loss": 0.03819262236356735,
2015
+ "step": 1000,
2016
+ "student_loss": 0.006347914692014456,
2017
+ "teacher_loss": 0.00022976804757490754,
2018
+ "total_flos": 0.0,
2019
+ "train_loss": 0.038887605339288714,
2020
+ "train_runtime": 42585.1685,
2021
+ "train_samples_per_second": 0.376,
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_a0.5_b1.0_g0.5_ls2/checkpoint-1000/vocab.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoints/codi3b_a0.5_b1.0_g0.5_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_a0.5_b1.0_g0.5_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_a0.5_b1.0_g0.5_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_a0.5_b1.0_g0.5_ls2/checkpoint-200/merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-200/pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e5a0fb2eb54ab86efa50b7fbd0472cd1a04e921f87f21a6abcd29b5f358429ed
3
+ size 6187858991
checkpoints/codi3b_a0.5_b1.0_g0.5_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_a0.5_b1.0_g0.5_ls2/checkpoint-200/thought_projector.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:523885c5ce55a07b08632c61501c45ac0817d35409b075d4e080b994be75d8ce
3
+ size 16788033
checkpoints/codi3b_a0.5_b1.0_g0.5_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_a0.5_b1.0_g0.5_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_a0.5_b1.0_g0.5_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": 64.5,
15
+ "kd_loss": 0.2920137941837311,
16
+ "learning_rate": 1.3333333333333334e-06,
17
+ "loss": 0.4591,
18
+ "step": 5,
19
+ "student_loss": 0.18811793625354767,
20
+ "teacher_loss": 0.0020189557690173388
21
+ },
22
+ {
23
+ "epoch": 0.0028835063437139563,
24
+ "grad_norm": 37.25,
25
+ "kd_loss": 0.27189651131629944,
26
+ "learning_rate": 3e-06,
27
+ "loss": 0.3445,
28
+ "step": 10,
29
+ "student_loss": 0.15471063554286957,
30
+ "teacher_loss": 0.00530141731724143
31
+ },
32
+ {
33
+ "epoch": 0.004325259515570935,
34
+ "grad_norm": 7.1875,
35
+ "kd_loss": 0.22213532030582428,
36
+ "learning_rate": 4.666666666666667e-06,
37
+ "loss": 0.2034,
38
+ "step": 15,
39
+ "student_loss": 0.007600904442369938,
40
+ "teacher_loss": 0.0011894693598151207
41
+ },
42
+ {
43
+ "epoch": 0.0057670126874279125,
44
+ "grad_norm": 7.21875,
45
+ "kd_loss": 0.14968645572662354,
46
+ "learning_rate": 6.333333333333333e-06,
47
+ "loss": 0.1464,
48
+ "step": 20,
49
+ "student_loss": 0.026719728484749794,
50
+ "teacher_loss": 0.0005942200659774244
51
+ },
52
+ {
53
+ "epoch": 0.00720876585928489,
54
+ "grad_norm": 2.703125,
55
+ "kd_loss": 0.11552157998085022,
56
+ "learning_rate": 8.000000000000001e-06,
57
+ "loss": 0.1171,
58
+ "step": 25,
59
+ "student_loss": 0.002865401329472661,
60
+ "teacher_loss": 0.0010151522001251578
61
+ },
62
+ {
63
+ "epoch": 0.00865051903114187,
64
+ "grad_norm": 1.484375,
65
+ "kd_loss": 0.07486728578805923,
66
+ "learning_rate": 9.666666666666667e-06,
67
+ "loss": 0.0901,
68
+ "step": 30,
69
+ "student_loss": 0.0011324587976559997,
70
+ "teacher_loss": 0.0005539655685424805
71
+ },
72
+ {
73
+ "epoch": 0.010092272202998846,
74
+ "grad_norm": 2.140625,
75
+ "kd_loss": 0.06898625195026398,
76
+ "learning_rate": 9.99958042442916e-06,
77
+ "loss": 0.0757,
78
+ "step": 35,
79
+ "student_loss": 0.0009480358567088842,
80
+ "teacher_loss": 0.0005732738063670695
81
+ },
82
+ {
83
+ "epoch": 0.011534025374855825,
84
+ "grad_norm": 1.296875,
85
+ "kd_loss": 0.0642521008849144,
86
+ "learning_rate": 9.997876019358083e-06,
87
+ "loss": 0.0685,
88
+ "step": 40,
89
+ "student_loss": 0.031368859112262726,
90
+ "teacher_loss": 0.027499590069055557
91
+ },
92
+ {
93
+ "epoch": 0.012975778546712802,
94
+ "grad_norm": 1.3359375,
95
+ "kd_loss": 0.11454806476831436,
96
+ "learning_rate": 9.99486100792044e-06,
97
+ "loss": 0.0541,
98
+ "step": 45,
99
+ "student_loss": 0.06097811087965965,
100
+ "teacher_loss": 0.0027094183024019003
101
+ },
102
+ {
103
+ "epoch": 0.01441753171856978,
104
+ "grad_norm": 1.5859375,
105
+ "kd_loss": 0.03876315429806709,
106
+ "learning_rate": 9.990536180750724e-06,
107
+ "loss": 0.0563,
108
+ "step": 50,
109
+ "student_loss": 0.02282995544373989,
110
+ "teacher_loss": 0.008940276689827442
111
+ },
112
+ {
113
+ "epoch": 0.015859284890426758,
114
+ "grad_norm": 1.125,
115
+ "kd_loss": 0.033022914081811905,
116
+ "learning_rate": 9.984902671959911e-06,
117
+ "loss": 0.0481,
118
+ "step": 55,
119
+ "student_loss": 0.0011594295501708984,
120
+ "teacher_loss": 0.0007985035772435367
121
+ },
122
+ {
123
+ "epoch": 0.01730103806228374,
124
+ "grad_norm": 0.8828125,
125
+ "kd_loss": 0.030243180692195892,
126
+ "learning_rate": 9.97796195883804e-06,
127
+ "loss": 0.0429,
128
+ "step": 60,
129
+ "student_loss": 0.0034187885466963053,
130
+ "teacher_loss": 0.0017242392059415579
131
+ },
132
+ {
133
+ "epoch": 0.018742791234140715,
134
+ "grad_norm": 1.4453125,
135
+ "kd_loss": 0.032282304018735886,
136
+ "learning_rate": 9.969715861466839e-06,
137
+ "loss": 0.0446,
138
+ "step": 65,
139
+ "student_loss": 0.029871758073568344,
140
+ "teacher_loss": 0.039595428854227066
141
+ },
142
+ {
143
+ "epoch": 0.020184544405997693,
144
+ "grad_norm": 1.1953125,
145
+ "kd_loss": 0.06090804189443588,
146
+ "learning_rate": 9.96016654224243e-06,
147
+ "loss": 0.0477,
148
+ "step": 70,
149
+ "student_loss": 0.017709577456116676,
150
+ "teacher_loss": 0.0022253147326409817
151
+ },
152
+ {
153
+ "epoch": 0.02162629757785467,
154
+ "grad_norm": 0.84765625,
155
+ "kd_loss": 0.024834414944052696,
156
+ "learning_rate": 9.94931650530827e-06,
157
+ "loss": 0.0418,
158
+ "step": 75,
159
+ "student_loss": 0.0012920524459332228,
160
+ "teacher_loss": 0.0008862126851454377
161
+ },
162
+ {
163
+ "epoch": 0.02306805074971165,
164
+ "grad_norm": 0.60546875,
165
+ "kd_loss": 0.025287121534347534,
166
+ "learning_rate": 9.93716859589851e-06,
167
+ "loss": 0.0365,
168
+ "step": 80,
169
+ "student_loss": 0.0013494148151949048,
170
+ "teacher_loss": 0.0008630359079688787
171
+ },
172
+ {
173
+ "epoch": 0.024509803921568627,
174
+ "grad_norm": 0.86328125,
175
+ "kd_loss": 0.024378223344683647,
176
+ "learning_rate": 9.923725999591846e-06,
177
+ "loss": 0.0395,
178
+ "step": 85,
179
+ "student_loss": 0.0005228935624472797,
180
+ "teacher_loss": 0.0004561410460155457
181
+ },
182
+ {
183
+ "epoch": 0.025951557093425604,
184
+ "grad_norm": 0.70703125,
185
+ "kd_loss": 0.07289917767047882,
186
+ "learning_rate": 9.908992241476189e-06,
187
+ "loss": 0.0394,
188
+ "step": 90,
189
+ "student_loss": 0.020422162488102913,
190
+ "teacher_loss": 0.0035499960649758577
191
+ },
192
+ {
193
+ "epoch": 0.027393310265282585,
194
+ "grad_norm": 1.125,
195
+ "kd_loss": 0.044955912977457047,
196
+ "learning_rate": 9.892971185224244e-06,
197
+ "loss": 0.0351,
198
+ "step": 95,
199
+ "student_loss": 0.008261171169579029,
200
+ "teacher_loss": 0.005078152287751436
201
+ },
202
+ {
203
+ "epoch": 0.02883506343713956,
204
+ "grad_norm": 0.83203125,
205
+ "kd_loss": 0.02227398194372654,
206
+ "learning_rate": 9.875667032080354e-06,
207
+ "loss": 0.0326,
208
+ "step": 100,
209
+ "student_loss": 0.0006025677430443466,
210
+ "teacher_loss": 0.00046476206625811756
211
+ },
212
+ {
213
+ "epoch": 0.03027681660899654,
214
+ "grad_norm": 1.2734375,
215
+ "kd_loss": 0.03532887250185013,
216
+ "learning_rate": 9.857084319758772e-06,
217
+ "loss": 0.036,
218
+ "step": 105,
219
+ "student_loss": 0.0034369053319096565,
220
+ "teacher_loss": 0.00029834595625288785
221
+ },
222
+ {
223
+ "epoch": 0.031718569780853516,
224
+ "grad_norm": 1.1640625,
225
+ "kd_loss": 0.04033924266695976,
226
+ "learning_rate": 9.837227921253747e-06,
227
+ "loss": 0.0375,
228
+ "step": 110,
229
+ "student_loss": 0.03080393560230732,
230
+ "teacher_loss": 0.018437745049595833
231
+ },
232
+ {
233
+ "epoch": 0.03316032295271049,
234
+ "grad_norm": 0.703125,
235
+ "kd_loss": 0.04197424277663231,
236
+ "learning_rate": 9.816103043561648e-06,
237
+ "loss": 0.0347,
238
+ "step": 115,
239
+ "student_loss": 0.0021668823901563883,
240
+ "teacher_loss": 0.00045062918798066676
241
+ },
242
+ {
243
+ "epoch": 0.03460207612456748,
244
+ "grad_norm": 0.98828125,
245
+ "kd_loss": 0.027563175186514854,
246
+ "learning_rate": 9.79371522631553e-06,
247
+ "loss": 0.032,
248
+ "step": 120,
249
+ "student_loss": 0.0016319500282406807,
250
+ "teacher_loss": 0.0008567498298361897
251
+ },
252
+ {
253
+ "epoch": 0.036043829296424454,
254
+ "grad_norm": 0.92578125,
255
+ "kd_loss": 0.06173818185925484,
256
+ "learning_rate": 9.770070340332457e-06,
257
+ "loss": 0.0364,
258
+ "step": 125,
259
+ "student_loss": 0.02385914884507656,
260
+ "teacher_loss": 0.00027849775506183505
261
+ },
262
+ {
263
+ "epoch": 0.03748558246828143,
264
+ "grad_norm": 1.0859375,
265
+ "kd_loss": 0.029248492792248726,
266
+ "learning_rate": 9.745174586073982e-06,
267
+ "loss": 0.0346,
268
+ "step": 130,
269
+ "student_loss": 0.0005455865757539868,
270
+ "teacher_loss": 0.0004959598300047219
271
+ },
272
+ {
273
+ "epoch": 0.03892733564013841,
274
+ "grad_norm": 0.953125,
275
+ "kd_loss": 0.0406946986913681,
276
+ "learning_rate": 9.719034492020183e-06,
277
+ "loss": 0.0377,
278
+ "step": 135,
279
+ "student_loss": 0.0013323032762855291,
280
+ "teacher_loss": 0.0005948346224613488
281
+ },
282
+ {
283
+ "epoch": 0.040369088811995385,
284
+ "grad_norm": 1.0390625,
285
+ "kd_loss": 0.023682042956352234,
286
+ "learning_rate": 9.691656912957686e-06,
287
+ "loss": 0.036,
288
+ "step": 140,
289
+ "student_loss": 0.0005881476681679487,
290
+ "teacher_loss": 0.0003679130459204316
291
+ },
292
+ {
293
+ "epoch": 0.04181084198385236,
294
+ "grad_norm": 0.5859375,
295
+ "kd_loss": 0.07271980494260788,
296
+ "learning_rate": 9.663049028182112e-06,
297
+ "loss": 0.0325,
298
+ "step": 145,
299
+ "student_loss": 0.028793470934033394,
300
+ "teacher_loss": 0.005983584560453892
301
+ },
302
+ {
303
+ "epoch": 0.04325259515570934,
304
+ "grad_norm": 0.609375,
305
+ "kd_loss": 0.01843745820224285,
306
+ "learning_rate": 9.633218339615433e-06,
307
+ "loss": 0.0316,
308
+ "step": 150,
309
+ "student_loss": 0.001051027444191277,
310
+ "teacher_loss": 0.000913174357265234
311
+ },
312
+ {
313
+ "epoch": 0.04469434832756632,
314
+ "grad_norm": 1.1171875,
315
+ "kd_loss": 0.023381218314170837,
316
+ "learning_rate": 9.602172669838721e-06,
317
+ "loss": 0.0381,
318
+ "step": 155,
319
+ "student_loss": 0.002775231609120965,
320
+ "teacher_loss": 0.0007182428380474448
321
+ },
322
+ {
323
+ "epoch": 0.0461361014994233,
324
+ "grad_norm": 1.6875,
325
+ "kd_loss": 0.03144950047135353,
326
+ "learning_rate": 9.569920160040815e-06,
327
+ "loss": 0.0344,
328
+ "step": 160,
329
+ "student_loss": 0.0778423622250557,
330
+ "teacher_loss": 0.044956743717193604
331
+ },
332
+ {
333
+ "epoch": 0.04757785467128028,
334
+ "grad_norm": 0.75,
335
+ "kd_loss": 0.03335012122988701,
336
+ "learning_rate": 9.536469267883432e-06,
337
+ "loss": 0.0311,
338
+ "step": 165,
339
+ "student_loss": 0.0012603362556546926,
340
+ "teacher_loss": 0.00417186226695776
341
+ },
342
+ {
343
+ "epoch": 0.049019607843137254,
344
+ "grad_norm": 1.8828125,
345
+ "kd_loss": 0.018793689087033272,
346
+ "learning_rate": 9.501828765283295e-06,
347
+ "loss": 0.0355,
348
+ "step": 170,
349
+ "student_loss": 0.0005546013708226383,
350
+ "teacher_loss": 0.0003494401171337813
351
+ },
352
+ {
353
+ "epoch": 0.05046136101499423,
354
+ "grad_norm": 1.09375,
355
+ "kd_loss": 0.027096513658761978,
356
+ "learning_rate": 9.466007736111846e-06,
357
+ "loss": 0.0322,
358
+ "step": 175,
359
+ "student_loss": 0.0030481775756925344,
360
+ "teacher_loss": 0.0003126203373540193
361
+ },
362
+ {
363
+ "epoch": 0.05190311418685121,
364
+ "grad_norm": 0.69140625,
365
+ "kd_loss": 0.03634097799658775,
366
+ "learning_rate": 9.429015573813163e-06,
367
+ "loss": 0.0302,
368
+ "step": 180,
369
+ "student_loss": 0.0012228424893692136,
370
+ "teacher_loss": 0.00034309024340473115
371
+ },
372
+ {
373
+ "epoch": 0.05334486735870819,
374
+ "grad_norm": 1.34375,
375
+ "kd_loss": 0.03704367205500603,
376
+ "learning_rate": 9.390861978940687e-06,
377
+ "loss": 0.0363,
378
+ "step": 185,
379
+ "student_loss": 0.035696618258953094,
380
+ "teacher_loss": 0.00034108557156287134
381
+ },
382
+ {
383
+ "epoch": 0.05478662053056517,
384
+ "grad_norm": 1.3046875,
385
+ "kd_loss": 0.052902791649103165,
386
+ "learning_rate": 9.351556956613423e-06,
387
+ "loss": 0.0321,
388
+ "step": 190,
389
+ "student_loss": 0.009327664040029049,
390
+ "teacher_loss": 0.0015260990476235747
391
+ },
392
+ {
393
+ "epoch": 0.056228373702422146,
394
+ "grad_norm": 2.828125,
395
+ "kd_loss": 0.05315268039703369,
396
+ "learning_rate": 9.31111081389227e-06,
397
+ "loss": 0.0327,
398
+ "step": 195,
399
+ "student_loss": 0.011893535032868385,
400
+ "teacher_loss": 0.0007797812577337027
401
+ },
402
+ {
403
+ "epoch": 0.05767012687427912,
404
+ "grad_norm": 0.8125,
405
+ "kd_loss": 0.03527738153934479,
406
+ "learning_rate": 9.269534157077177e-06,
407
+ "loss": 0.0394,
408
+ "step": 200,
409
+ "student_loss": 0.0005083663854748011,
410
+ "teacher_loss": 0.0002656931465025991
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_a0.5_b1.0_g0.5_ls2/checkpoint-200/vocab.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoints/codi3b_a0.5_b1.0_g0.5_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_a0.5_b1.0_g0.5_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_a0.5_b1.0_g0.5_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_a0.5_b1.0_g0.5_ls2/checkpoint-400/merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-400/pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a5b555de6ba696a39fd5cd9998487dd6e8c54b33d39ba83e48f21a287406f531
3
+ size 6187858991
checkpoints/codi3b_a0.5_b1.0_g0.5_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_a0.5_b1.0_g0.5_ls2/checkpoint-400/thought_projector.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:85684960a1d3635f643b98b163e29c8ad5fcdee2b4b15b6e9a2f5fd31f95c859
3
+ size 16788033
checkpoints/codi3b_a0.5_b1.0_g0.5_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_a0.5_b1.0_g0.5_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_a0.5_b1.0_g0.5_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": 64.5,
15
+ "kd_loss": 0.2920137941837311,
16
+ "learning_rate": 1.3333333333333334e-06,
17
+ "loss": 0.4591,
18
+ "step": 5,
19
+ "student_loss": 0.18811793625354767,
20
+ "teacher_loss": 0.0020189557690173388
21
+ },
22
+ {
23
+ "epoch": 0.0028835063437139563,
24
+ "grad_norm": 37.25,
25
+ "kd_loss": 0.27189651131629944,
26
+ "learning_rate": 3e-06,
27
+ "loss": 0.3445,
28
+ "step": 10,
29
+ "student_loss": 0.15471063554286957,
30
+ "teacher_loss": 0.00530141731724143
31
+ },
32
+ {
33
+ "epoch": 0.004325259515570935,
34
+ "grad_norm": 7.1875,
35
+ "kd_loss": 0.22213532030582428,
36
+ "learning_rate": 4.666666666666667e-06,
37
+ "loss": 0.2034,
38
+ "step": 15,
39
+ "student_loss": 0.007600904442369938,
40
+ "teacher_loss": 0.0011894693598151207
41
+ },
42
+ {
43
+ "epoch": 0.0057670126874279125,
44
+ "grad_norm": 7.21875,
45
+ "kd_loss": 0.14968645572662354,
46
+ "learning_rate": 6.333333333333333e-06,
47
+ "loss": 0.1464,
48
+ "step": 20,
49
+ "student_loss": 0.026719728484749794,
50
+ "teacher_loss": 0.0005942200659774244
51
+ },
52
+ {
53
+ "epoch": 0.00720876585928489,
54
+ "grad_norm": 2.703125,
55
+ "kd_loss": 0.11552157998085022,
56
+ "learning_rate": 8.000000000000001e-06,
57
+ "loss": 0.1171,
58
+ "step": 25,
59
+ "student_loss": 0.002865401329472661,
60
+ "teacher_loss": 0.0010151522001251578
61
+ },
62
+ {
63
+ "epoch": 0.00865051903114187,
64
+ "grad_norm": 1.484375,
65
+ "kd_loss": 0.07486728578805923,
66
+ "learning_rate": 9.666666666666667e-06,
67
+ "loss": 0.0901,
68
+ "step": 30,
69
+ "student_loss": 0.0011324587976559997,
70
+ "teacher_loss": 0.0005539655685424805
71
+ },
72
+ {
73
+ "epoch": 0.010092272202998846,
74
+ "grad_norm": 2.140625,
75
+ "kd_loss": 0.06898625195026398,
76
+ "learning_rate": 9.99958042442916e-06,
77
+ "loss": 0.0757,
78
+ "step": 35,
79
+ "student_loss": 0.0009480358567088842,
80
+ "teacher_loss": 0.0005732738063670695
81
+ },
82
+ {
83
+ "epoch": 0.011534025374855825,
84
+ "grad_norm": 1.296875,
85
+ "kd_loss": 0.0642521008849144,
86
+ "learning_rate": 9.997876019358083e-06,
87
+ "loss": 0.0685,
88
+ "step": 40,
89
+ "student_loss": 0.031368859112262726,
90
+ "teacher_loss": 0.027499590069055557
91
+ },
92
+ {
93
+ "epoch": 0.012975778546712802,
94
+ "grad_norm": 1.3359375,
95
+ "kd_loss": 0.11454806476831436,
96
+ "learning_rate": 9.99486100792044e-06,
97
+ "loss": 0.0541,
98
+ "step": 45,
99
+ "student_loss": 0.06097811087965965,
100
+ "teacher_loss": 0.0027094183024019003
101
+ },
102
+ {
103
+ "epoch": 0.01441753171856978,
104
+ "grad_norm": 1.5859375,
105
+ "kd_loss": 0.03876315429806709,
106
+ "learning_rate": 9.990536180750724e-06,
107
+ "loss": 0.0563,
108
+ "step": 50,
109
+ "student_loss": 0.02282995544373989,
110
+ "teacher_loss": 0.008940276689827442
111
+ },
112
+ {
113
+ "epoch": 0.015859284890426758,
114
+ "grad_norm": 1.125,
115
+ "kd_loss": 0.033022914081811905,
116
+ "learning_rate": 9.984902671959911e-06,
117
+ "loss": 0.0481,
118
+ "step": 55,
119
+ "student_loss": 0.0011594295501708984,
120
+ "teacher_loss": 0.0007985035772435367
121
+ },
122
+ {
123
+ "epoch": 0.01730103806228374,
124
+ "grad_norm": 0.8828125,
125
+ "kd_loss": 0.030243180692195892,
126
+ "learning_rate": 9.97796195883804e-06,
127
+ "loss": 0.0429,
128
+ "step": 60,
129
+ "student_loss": 0.0034187885466963053,
130
+ "teacher_loss": 0.0017242392059415579
131
+ },
132
+ {
133
+ "epoch": 0.018742791234140715,
134
+ "grad_norm": 1.4453125,
135
+ "kd_loss": 0.032282304018735886,
136
+ "learning_rate": 9.969715861466839e-06,
137
+ "loss": 0.0446,
138
+ "step": 65,
139
+ "student_loss": 0.029871758073568344,
140
+ "teacher_loss": 0.039595428854227066
141
+ },
142
+ {
143
+ "epoch": 0.020184544405997693,
144
+ "grad_norm": 1.1953125,
145
+ "kd_loss": 0.06090804189443588,
146
+ "learning_rate": 9.96016654224243e-06,
147
+ "loss": 0.0477,
148
+ "step": 70,
149
+ "student_loss": 0.017709577456116676,
150
+ "teacher_loss": 0.0022253147326409817
151
+ },
152
+ {
153
+ "epoch": 0.02162629757785467,
154
+ "grad_norm": 0.84765625,
155
+ "kd_loss": 0.024834414944052696,
156
+ "learning_rate": 9.94931650530827e-06,
157
+ "loss": 0.0418,
158
+ "step": 75,
159
+ "student_loss": 0.0012920524459332228,
160
+ "teacher_loss": 0.0008862126851454377
161
+ },
162
+ {
163
+ "epoch": 0.02306805074971165,
164
+ "grad_norm": 0.60546875,
165
+ "kd_loss": 0.025287121534347534,
166
+ "learning_rate": 9.93716859589851e-06,
167
+ "loss": 0.0365,
168
+ "step": 80,
169
+ "student_loss": 0.0013494148151949048,
170
+ "teacher_loss": 0.0008630359079688787
171
+ },
172
+ {
173
+ "epoch": 0.024509803921568627,
174
+ "grad_norm": 0.86328125,
175
+ "kd_loss": 0.024378223344683647,
176
+ "learning_rate": 9.923725999591846e-06,
177
+ "loss": 0.0395,
178
+ "step": 85,
179
+ "student_loss": 0.0005228935624472797,
180
+ "teacher_loss": 0.0004561410460155457
181
+ },
182
+ {
183
+ "epoch": 0.025951557093425604,
184
+ "grad_norm": 0.70703125,
185
+ "kd_loss": 0.07289917767047882,
186
+ "learning_rate": 9.908992241476189e-06,
187
+ "loss": 0.0394,
188
+ "step": 90,
189
+ "student_loss": 0.020422162488102913,
190
+ "teacher_loss": 0.0035499960649758577
191
+ },
192
+ {
193
+ "epoch": 0.027393310265282585,
194
+ "grad_norm": 1.125,
195
+ "kd_loss": 0.044955912977457047,
196
+ "learning_rate": 9.892971185224244e-06,
197
+ "loss": 0.0351,
198
+ "step": 95,
199
+ "student_loss": 0.008261171169579029,
200
+ "teacher_loss": 0.005078152287751436
201
+ },
202
+ {
203
+ "epoch": 0.02883506343713956,
204
+ "grad_norm": 0.83203125,
205
+ "kd_loss": 0.02227398194372654,
206
+ "learning_rate": 9.875667032080354e-06,
207
+ "loss": 0.0326,
208
+ "step": 100,
209
+ "student_loss": 0.0006025677430443466,
210
+ "teacher_loss": 0.00046476206625811756
211
+ },
212
+ {
213
+ "epoch": 0.03027681660899654,
214
+ "grad_norm": 1.2734375,
215
+ "kd_loss": 0.03532887250185013,
216
+ "learning_rate": 9.857084319758772e-06,
217
+ "loss": 0.036,
218
+ "step": 105,
219
+ "student_loss": 0.0034369053319096565,
220
+ "teacher_loss": 0.00029834595625288785
221
+ },
222
+ {
223
+ "epoch": 0.031718569780853516,
224
+ "grad_norm": 1.1640625,
225
+ "kd_loss": 0.04033924266695976,
226
+ "learning_rate": 9.837227921253747e-06,
227
+ "loss": 0.0375,
228
+ "step": 110,
229
+ "student_loss": 0.03080393560230732,
230
+ "teacher_loss": 0.018437745049595833
231
+ },
232
+ {
233
+ "epoch": 0.03316032295271049,
234
+ "grad_norm": 0.703125,
235
+ "kd_loss": 0.04197424277663231,
236
+ "learning_rate": 9.816103043561648e-06,
237
+ "loss": 0.0347,
238
+ "step": 115,
239
+ "student_loss": 0.0021668823901563883,
240
+ "teacher_loss": 0.00045062918798066676
241
+ },
242
+ {
243
+ "epoch": 0.03460207612456748,
244
+ "grad_norm": 0.98828125,
245
+ "kd_loss": 0.027563175186514854,
246
+ "learning_rate": 9.79371522631553e-06,
247
+ "loss": 0.032,
248
+ "step": 120,
249
+ "student_loss": 0.0016319500282406807,
250
+ "teacher_loss": 0.0008567498298361897
251
+ },
252
+ {
253
+ "epoch": 0.036043829296424454,
254
+ "grad_norm": 0.92578125,
255
+ "kd_loss": 0.06173818185925484,
256
+ "learning_rate": 9.770070340332457e-06,
257
+ "loss": 0.0364,
258
+ "step": 125,
259
+ "student_loss": 0.02385914884507656,
260
+ "teacher_loss": 0.00027849775506183505
261
+ },
262
+ {
263
+ "epoch": 0.03748558246828143,
264
+ "grad_norm": 1.0859375,
265
+ "kd_loss": 0.029248492792248726,
266
+ "learning_rate": 9.745174586073982e-06,
267
+ "loss": 0.0346,
268
+ "step": 130,
269
+ "student_loss": 0.0005455865757539868,
270
+ "teacher_loss": 0.0004959598300047219
271
+ },
272
+ {
273
+ "epoch": 0.03892733564013841,
274
+ "grad_norm": 0.953125,
275
+ "kd_loss": 0.0406946986913681,
276
+ "learning_rate": 9.719034492020183e-06,
277
+ "loss": 0.0377,
278
+ "step": 135,
279
+ "student_loss": 0.0013323032762855291,
280
+ "teacher_loss": 0.0005948346224613488
281
+ },
282
+ {
283
+ "epoch": 0.040369088811995385,
284
+ "grad_norm": 1.0390625,
285
+ "kd_loss": 0.023682042956352234,
286
+ "learning_rate": 9.691656912957686e-06,
287
+ "loss": 0.036,
288
+ "step": 140,
289
+ "student_loss": 0.0005881476681679487,
290
+ "teacher_loss": 0.0003679130459204316
291
+ },
292
+ {
293
+ "epoch": 0.04181084198385236,
294
+ "grad_norm": 0.5859375,
295
+ "kd_loss": 0.07271980494260788,
296
+ "learning_rate": 9.663049028182112e-06,
297
+ "loss": 0.0325,
298
+ "step": 145,
299
+ "student_loss": 0.028793470934033394,
300
+ "teacher_loss": 0.005983584560453892
301
+ },
302
+ {
303
+ "epoch": 0.04325259515570934,
304
+ "grad_norm": 0.609375,
305
+ "kd_loss": 0.01843745820224285,
306
+ "learning_rate": 9.633218339615433e-06,
307
+ "loss": 0.0316,
308
+ "step": 150,
309
+ "student_loss": 0.001051027444191277,
310
+ "teacher_loss": 0.000913174357265234
311
+ },
312
+ {
313
+ "epoch": 0.04469434832756632,
314
+ "grad_norm": 1.1171875,
315
+ "kd_loss": 0.023381218314170837,
316
+ "learning_rate": 9.602172669838721e-06,
317
+ "loss": 0.0381,
318
+ "step": 155,
319
+ "student_loss": 0.002775231609120965,
320
+ "teacher_loss": 0.0007182428380474448
321
+ },
322
+ {
323
+ "epoch": 0.0461361014994233,
324
+ "grad_norm": 1.6875,
325
+ "kd_loss": 0.03144950047135353,
326
+ "learning_rate": 9.569920160040815e-06,
327
+ "loss": 0.0344,
328
+ "step": 160,
329
+ "student_loss": 0.0778423622250557,
330
+ "teacher_loss": 0.044956743717193604
331
+ },
332
+ {
333
+ "epoch": 0.04757785467128028,
334
+ "grad_norm": 0.75,
335
+ "kd_loss": 0.03335012122988701,
336
+ "learning_rate": 9.536469267883432e-06,
337
+ "loss": 0.0311,
338
+ "step": 165,
339
+ "student_loss": 0.0012603362556546926,
340
+ "teacher_loss": 0.00417186226695776
341
+ },
342
+ {
343
+ "epoch": 0.049019607843137254,
344
+ "grad_norm": 1.8828125,
345
+ "kd_loss": 0.018793689087033272,
346
+ "learning_rate": 9.501828765283295e-06,
347
+ "loss": 0.0355,
348
+ "step": 170,
349
+ "student_loss": 0.0005546013708226383,
350
+ "teacher_loss": 0.0003494401171337813
351
+ },
352
+ {
353
+ "epoch": 0.05046136101499423,
354
+ "grad_norm": 1.09375,
355
+ "kd_loss": 0.027096513658761978,
356
+ "learning_rate": 9.466007736111846e-06,
357
+ "loss": 0.0322,
358
+ "step": 175,
359
+ "student_loss": 0.0030481775756925344,
360
+ "teacher_loss": 0.0003126203373540193
361
+ },
362
+ {
363
+ "epoch": 0.05190311418685121,
364
+ "grad_norm": 0.69140625,
365
+ "kd_loss": 0.03634097799658775,
366
+ "learning_rate": 9.429015573813163e-06,
367
+ "loss": 0.0302,
368
+ "step": 180,
369
+ "student_loss": 0.0012228424893692136,
370
+ "teacher_loss": 0.00034309024340473115
371
+ },
372
+ {
373
+ "epoch": 0.05334486735870819,
374
+ "grad_norm": 1.34375,
375
+ "kd_loss": 0.03704367205500603,
376
+ "learning_rate": 9.390861978940687e-06,
377
+ "loss": 0.0363,
378
+ "step": 185,
379
+ "student_loss": 0.035696618258953094,
380
+ "teacher_loss": 0.00034108557156287134
381
+ },
382
+ {
383
+ "epoch": 0.05478662053056517,
384
+ "grad_norm": 1.3046875,
385
+ "kd_loss": 0.052902791649103165,
386
+ "learning_rate": 9.351556956613423e-06,
387
+ "loss": 0.0321,
388
+ "step": 190,
389
+ "student_loss": 0.009327664040029049,
390
+ "teacher_loss": 0.0015260990476235747
391
+ },
392
+ {
393
+ "epoch": 0.056228373702422146,
394
+ "grad_norm": 2.828125,
395
+ "kd_loss": 0.05315268039703369,
396
+ "learning_rate": 9.31111081389227e-06,
397
+ "loss": 0.0327,
398
+ "step": 195,
399
+ "student_loss": 0.011893535032868385,
400
+ "teacher_loss": 0.0007797812577337027
401
+ },
402
+ {
403
+ "epoch": 0.05767012687427912,
404
+ "grad_norm": 0.8125,
405
+ "kd_loss": 0.03527738153934479,
406
+ "learning_rate": 9.269534157077177e-06,
407
+ "loss": 0.0394,
408
+ "step": 200,
409
+ "student_loss": 0.0005083663854748011,
410
+ "teacher_loss": 0.0002656931465025991
411
+ },
412
+ {
413
+ "epoch": 0.0591118800461361,
414
+ "grad_norm": 0.9921875,
415
+ "kd_loss": 0.02736036665737629,
416
+ "learning_rate": 9.226837888925813e-06,
417
+ "loss": 0.0341,
418
+ "step": 205,
419
+ "student_loss": 0.03568984195590019,
420
+ "teacher_loss": 0.027640890330076218
421
+ },
422
+ {
423
+ "epoch": 0.06055363321799308,
424
+ "grad_norm": 1.78125,
425
+ "kd_loss": 0.030339844524860382,
426
+ "learning_rate": 9.183033205794525e-06,
427
+ "loss": 0.0302,
428
+ "step": 210,
429
+ "student_loss": 0.000763049116358161,
430
+ "teacher_loss": 0.0003548153617884964
431
+ },
432
+ {
433
+ "epoch": 0.061995386389850055,
434
+ "grad_norm": 1.1484375,
435
+ "kd_loss": 0.03865697979927063,
436
+ "learning_rate": 9.13813159470227e-06,
437
+ "loss": 0.0326,
438
+ "step": 215,
439
+ "student_loss": 0.0005148217896930873,
440
+ "teacher_loss": 0.00020245747873559594
441
+ },
442
+ {
443
+ "epoch": 0.06343713956170703,
444
+ "grad_norm": 1.3671875,
445
+ "kd_loss": 0.04260854423046112,
446
+ "learning_rate": 9.092144830318357e-06,
447
+ "loss": 0.0316,
448
+ "step": 220,
449
+ "student_loss": 0.030991079285740852,
450
+ "teacher_loss": 0.00993641559034586
451
+ },
452
+ {
453
+ "epoch": 0.06487889273356401,
454
+ "grad_norm": 1.203125,
455
+ "kd_loss": 0.023958567529916763,
456
+ "learning_rate": 9.045084971874738e-06,
457
+ "loss": 0.0376,
458
+ "step": 225,
459
+ "student_loss": 0.043197184801101685,
460
+ "teacher_loss": 0.00216495874337852
461
+ },
462
+ {
463
+ "epoch": 0.06632064590542099,
464
+ "grad_norm": 1.7578125,
465
+ "kd_loss": 0.04370651766657829,
466
+ "learning_rate": 8.99696436000368e-06,
467
+ "loss": 0.0299,
468
+ "step": 230,
469
+ "student_loss": 0.005392159800976515,
470
+ "teacher_loss": 0.003661371534690261
471
+ },
472
+ {
473
+ "epoch": 0.06776239907727798,
474
+ "grad_norm": 0.8671875,
475
+ "kd_loss": 0.07619086652994156,
476
+ "learning_rate": 8.947795613501658e-06,
477
+ "loss": 0.0314,
478
+ "step": 235,
479
+ "student_loss": 0.008586333133280277,
480
+ "teacher_loss": 0.0005631999811157584
481
+ },
482
+ {
483
+ "epoch": 0.06920415224913495,
484
+ "grad_norm": 1.0625,
485
+ "kd_loss": 0.03111647628247738,
486
+ "learning_rate": 8.897591626020284e-06,
487
+ "loss": 0.034,
488
+ "step": 240,
489
+ "student_loss": 0.003906027879565954,
490
+ "teacher_loss": 0.0004787310608662665
491
+ },
492
+ {
493
+ "epoch": 0.07064590542099193,
494
+ "grad_norm": 1.453125,
495
+ "kd_loss": 0.017389042302966118,
496
+ "learning_rate": 8.846365562685178e-06,
497
+ "loss": 0.0279,
498
+ "step": 245,
499
+ "student_loss": 0.013969292864203453,
500
+ "teacher_loss": 0.006873726844787598
501
+ },
502
+ {
503
+ "epoch": 0.07208765859284891,
504
+ "grad_norm": 1.6015625,
505
+ "kd_loss": 0.0229345690459013,
506
+ "learning_rate": 8.794130856643635e-06,
507
+ "loss": 0.0311,
508
+ "step": 250,
509
+ "student_loss": 0.0008334179292432964,
510
+ "teacher_loss": 0.0003169570700265467
511
+ },
512
+ {
513
+ "epoch": 0.07352941176470588,
514
+ "grad_norm": 0.83203125,
515
+ "kd_loss": 0.01808979921042919,
516
+ "learning_rate": 8.74090120554202e-06,
517
+ "loss": 0.0312,
518
+ "step": 255,
519
+ "student_loss": 0.0003083710907958448,
520
+ "teacher_loss": 0.00030115401023067534
521
+ },
522
+ {
523
+ "epoch": 0.07497116493656286,
524
+ "grad_norm": 0.88671875,
525
+ "kd_loss": 0.026108454912900925,
526
+ "learning_rate": 8.686690567933803e-06,
527
+ "loss": 0.0333,
528
+ "step": 260,
529
+ "student_loss": 0.042571116238832474,
530
+ "teacher_loss": 0.03388316184282303
531
+ },
532
+ {
533
+ "epoch": 0.07641291810841984,
534
+ "grad_norm": 0.80078125,
535
+ "kd_loss": 0.016656002029776573,
536
+ "learning_rate": 8.63151315961915e-06,
537
+ "loss": 0.0317,
538
+ "step": 265,
539
+ "student_loss": 0.0003702428948599845,
540
+ "teacher_loss": 0.0002555136161390692
541
+ },
542
+ {
543
+ "epoch": 0.07785467128027682,
544
+ "grad_norm": 1.046875,
545
+ "kd_loss": 0.019304102286696434,
546
+ "learning_rate": 8.575383449917103e-06,
547
+ "loss": 0.0342,
548
+ "step": 270,
549
+ "student_loss": 0.001813149661757052,
550
+ "teacher_loss": 0.0011580288410186768
551
+ },
552
+ {
553
+ "epoch": 0.07929642445213379,
554
+ "grad_norm": 1.1796875,
555
+ "kd_loss": 0.023480774834752083,
556
+ "learning_rate": 8.518316157871232e-06,
557
+ "loss": 0.029,
558
+ "step": 275,
559
+ "student_loss": 0.04764978215098381,
560
+ "teacher_loss": 0.03439468517899513
561
+ },
562
+ {
563
+ "epoch": 0.08073817762399077,
564
+ "grad_norm": 1.015625,
565
+ "kd_loss": 0.03275206685066223,
566
+ "learning_rate": 8.460326248389825e-06,
567
+ "loss": 0.0289,
568
+ "step": 280,
569
+ "student_loss": 0.0005029537715017796,
570
+ "teacher_loss": 0.00019533037266228348
571
+ },
572
+ {
573
+ "epoch": 0.08217993079584775,
574
+ "grad_norm": 1.140625,
575
+ "kd_loss": 0.019457675516605377,
576
+ "learning_rate": 8.401428928321607e-06,
577
+ "loss": 0.0322,
578
+ "step": 285,
579
+ "student_loss": 0.0007758038118481636,
580
+ "teacher_loss": 0.000900130660738796
581
+ },
582
+ {
583
+ "epoch": 0.08362168396770472,
584
+ "grad_norm": 1.3125,
585
+ "kd_loss": 0.017802242189645767,
586
+ "learning_rate": 8.341639642468002e-06,
587
+ "loss": 0.0348,
588
+ "step": 290,
589
+ "student_loss": 0.023405462503433228,
590
+ "teacher_loss": 0.021540865302085876
591
+ },
592
+ {
593
+ "epoch": 0.0850634371395617,
594
+ "grad_norm": 1.03125,
595
+ "kd_loss": 0.01647804118692875,
596
+ "learning_rate": 8.280974069532999e-06,
597
+ "loss": 0.0328,
598
+ "step": 295,
599
+ "student_loss": 0.0006198033224791288,
600
+ "teacher_loss": 0.000554366793949157
601
+ },
602
+ {
603
+ "epoch": 0.08650519031141868,
604
+ "grad_norm": 0.9296875,
605
+ "kd_loss": 0.04784730449318886,
606
+ "learning_rate": 8.219448118011687e-06,
607
+ "loss": 0.0308,
608
+ "step": 300,
609
+ "student_loss": 0.015613794326782227,
610
+ "teacher_loss": 0.001725711626932025
611
+ },
612
+ {
613
+ "epoch": 0.08794694348327567,
614
+ "grad_norm": 1.1953125,
615
+ "kd_loss": 0.019965235143899918,
616
+ "learning_rate": 8.157077922018537e-06,
617
+ "loss": 0.0289,
618
+ "step": 305,
619
+ "student_loss": 0.006098510231822729,
620
+ "teacher_loss": 0.002777156652882695
621
+ },
622
+ {
623
+ "epoch": 0.08938869665513265,
624
+ "grad_norm": 1.0390625,
625
+ "kd_loss": 0.020748196169734,
626
+ "learning_rate": 8.093879837056486e-06,
627
+ "loss": 0.0309,
628
+ "step": 310,
629
+ "student_loss": 0.000493990199174732,
630
+ "teacher_loss": 0.00039537265547551215
631
+ },
632
+ {
633
+ "epoch": 0.09083044982698962,
634
+ "grad_norm": 1.4765625,
635
+ "kd_loss": 0.03892743960022926,
636
+ "learning_rate": 8.029870435728018e-06,
637
+ "loss": 0.0294,
638
+ "step": 315,
639
+ "student_loss": 0.0073717073537409306,
640
+ "teacher_loss": 0.00021180440671741962
641
+ },
642
+ {
643
+ "epoch": 0.0922722029988466,
644
+ "grad_norm": 1.3828125,
645
+ "kd_loss": 0.02363528683781624,
646
+ "learning_rate": 7.965066503389264e-06,
647
+ "loss": 0.0313,
648
+ "step": 320,
649
+ "student_loss": 0.0004119524674024433,
650
+ "teacher_loss": 0.00022927633835934103
651
+ },
652
+ {
653
+ "epoch": 0.09371395617070358,
654
+ "grad_norm": 1.7421875,
655
+ "kd_loss": 0.06230268254876137,
656
+ "learning_rate": 7.89948503374835e-06,
657
+ "loss": 0.0284,
658
+ "step": 325,
659
+ "student_loss": 0.009024329483509064,
660
+ "teacher_loss": 0.01963256485760212
661
+ },
662
+ {
663
+ "epoch": 0.09515570934256055,
664
+ "grad_norm": 0.90625,
665
+ "kd_loss": 0.04712303727865219,
666
+ "learning_rate": 7.833143224409076e-06,
667
+ "loss": 0.0302,
668
+ "step": 330,
669
+ "student_loss": 0.008720812387764454,
670
+ "teacher_loss": 0.0013501221546903253
671
+ },
672
+ {
673
+ "epoch": 0.09659746251441753,
674
+ "grad_norm": 0.9140625,
675
+ "kd_loss": 0.018766457214951515,
676
+ "learning_rate": 7.766058472361154e-06,
677
+ "loss": 0.0283,
678
+ "step": 335,
679
+ "student_loss": 0.001242243917658925,
680
+ "teacher_loss": 0.0023425817489624023
681
+ },
682
+ {
683
+ "epoch": 0.09803921568627451,
684
+ "grad_norm": 2.078125,
685
+ "kd_loss": 0.02911657840013504,
686
+ "learning_rate": 7.698248369418146e-06,
687
+ "loss": 0.0326,
688
+ "step": 340,
689
+ "student_loss": 0.009861858561635017,
690
+ "teacher_loss": 0.006796200294047594
691
+ },
692
+ {
693
+ "epoch": 0.09948096885813149,
694
+ "grad_norm": 1.2265625,
695
+ "kd_loss": 0.032899159938097,
696
+ "learning_rate": 7.629730697604314e-06,
697
+ "loss": 0.0335,
698
+ "step": 345,
699
+ "student_loss": 0.00739182299003005,
700
+ "teacher_loss": 0.0002512071805540472
701
+ },
702
+ {
703
+ "epoch": 0.10092272202998846,
704
+ "grad_norm": 1.0859375,
705
+ "kd_loss": 0.06120900437235832,
706
+ "learning_rate": 7.560523424491595e-06,
707
+ "loss": 0.0282,
708
+ "step": 350,
709
+ "student_loss": 0.003599822986871004,
710
+ "teacher_loss": 0.0007063632365316153
711
+ },
712
+ {
713
+ "epoch": 0.10236447520184544,
714
+ "grad_norm": 2.296875,
715
+ "kd_loss": 0.02087453007698059,
716
+ "learning_rate": 7.490644698487909e-06,
717
+ "loss": 0.035,
718
+ "step": 355,
719
+ "student_loss": 0.001389339566230774,
720
+ "teacher_loss": 0.0017551770433783531
721
+ },
722
+ {
723
+ "epoch": 0.10380622837370242,
724
+ "grad_norm": 0.90234375,
725
+ "kd_loss": 0.019557664170861244,
726
+ "learning_rate": 7.420112844078066e-06,
727
+ "loss": 0.0321,
728
+ "step": 360,
729
+ "student_loss": 0.0006360138650052249,
730
+ "teacher_loss": 0.0005147532792761922
731
+ },
732
+ {
733
+ "epoch": 0.1052479815455594,
734
+ "grad_norm": 1.3203125,
735
+ "kd_loss": 0.030097220093011856,
736
+ "learning_rate": 7.348946357018479e-06,
737
+ "loss": 0.0333,
738
+ "step": 365,
739
+ "student_loss": 0.04143820330500603,
740
+ "teacher_loss": 0.027219083160161972
741
+ },
742
+ {
743
+ "epoch": 0.10668973471741638,
744
+ "grad_norm": 0.9453125,
745
+ "kd_loss": 0.05344080179929733,
746
+ "learning_rate": 7.277163899486975e-06,
747
+ "loss": 0.0351,
748
+ "step": 370,
749
+ "student_loss": 0.021593965590000153,
750
+ "teacher_loss": 0.0005673202103935182
751
+ },
752
+ {
753
+ "epoch": 0.10813148788927336,
754
+ "grad_norm": 0.7265625,
755
+ "kd_loss": 0.03127056360244751,
756
+ "learning_rate": 7.204784295188959e-06,
757
+ "loss": 0.0287,
758
+ "step": 375,
759
+ "student_loss": 0.03163963928818703,
760
+ "teacher_loss": 0.020587248727679253
761
+ },
762
+ {
763
+ "epoch": 0.10957324106113034,
764
+ "grad_norm": 0.84375,
765
+ "kd_loss": 0.01984175480902195,
766
+ "learning_rate": 7.1318265244212305e-06,
767
+ "loss": 0.0311,
768
+ "step": 380,
769
+ "student_loss": 0.0005506337620317936,
770
+ "teacher_loss": 0.00038444914389401674
771
+ },
772
+ {
773
+ "epoch": 0.11101499423298732,
774
+ "grad_norm": 0.59765625,
775
+ "kd_loss": 0.02916550263762474,
776
+ "learning_rate": 7.05830971909472e-06,
777
+ "loss": 0.0294,
778
+ "step": 385,
779
+ "student_loss": 0.00045756620238535106,
780
+ "teacher_loss": 0.0003828653716482222
781
+ },
782
+ {
783
+ "epoch": 0.11245674740484429,
784
+ "grad_norm": 1.171875,
785
+ "kd_loss": 0.021802764385938644,
786
+ "learning_rate": 6.9842531577174865e-06,
787
+ "loss": 0.0271,
788
+ "step": 390,
789
+ "student_loss": 0.00026892355526797473,
790
+ "teacher_loss": 0.0002502185816410929
791
+ },
792
+ {
793
+ "epoch": 0.11389850057670127,
794
+ "grad_norm": 0.94140625,
795
+ "kd_loss": 0.020051907747983932,
796
+ "learning_rate": 6.9096762603392595e-06,
797
+ "loss": 0.0307,
798
+ "step": 395,
799
+ "student_loss": 0.0004568768781609833,
800
+ "teacher_loss": 0.00039133013342507184
801
+ },
802
+ {
803
+ "epoch": 0.11534025374855825,
804
+ "grad_norm": 0.78125,
805
+ "kd_loss": 0.03054620325565338,
806
+ "learning_rate": 6.834598583458862e-06,
807
+ "loss": 0.0275,
808
+ "step": 400,
809
+ "student_loss": 0.0005328103434294462,
810
+ "teacher_loss": 0.00015425821766257286
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_a0.5_b1.0_g0.5_ls2/checkpoint-400/vocab.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoints/codi3b_a0.5_b1.0_g0.5_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_a0.5_b1.0_g0.5_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_a0.5_b1.0_g0.5_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_a0.5_b1.0_g0.5_ls2/checkpoint-600/merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-600/pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6244ca33c3a3fef3be7e4a026d26ccc4173f9ba0ff320a9f37dcab8df189ad0b
3
+ size 6187858991
checkpoints/codi3b_a0.5_b1.0_g0.5_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_a0.5_b1.0_g0.5_ls2/checkpoint-600/thought_projector.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:80e5886a026245018a58a248884d160c0e06852fc283c5566d08bbc7cf5afe5d
3
+ size 16788033
checkpoints/codi3b_a0.5_b1.0_g0.5_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_a0.5_b1.0_g0.5_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_a0.5_b1.0_g0.5_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": 64.5,
15
+ "kd_loss": 0.2920137941837311,
16
+ "learning_rate": 1.3333333333333334e-06,
17
+ "loss": 0.4591,
18
+ "step": 5,
19
+ "student_loss": 0.18811793625354767,
20
+ "teacher_loss": 0.0020189557690173388
21
+ },
22
+ {
23
+ "epoch": 0.0028835063437139563,
24
+ "grad_norm": 37.25,
25
+ "kd_loss": 0.27189651131629944,
26
+ "learning_rate": 3e-06,
27
+ "loss": 0.3445,
28
+ "step": 10,
29
+ "student_loss": 0.15471063554286957,
30
+ "teacher_loss": 0.00530141731724143
31
+ },
32
+ {
33
+ "epoch": 0.004325259515570935,
34
+ "grad_norm": 7.1875,
35
+ "kd_loss": 0.22213532030582428,
36
+ "learning_rate": 4.666666666666667e-06,
37
+ "loss": 0.2034,
38
+ "step": 15,
39
+ "student_loss": 0.007600904442369938,
40
+ "teacher_loss": 0.0011894693598151207
41
+ },
42
+ {
43
+ "epoch": 0.0057670126874279125,
44
+ "grad_norm": 7.21875,
45
+ "kd_loss": 0.14968645572662354,
46
+ "learning_rate": 6.333333333333333e-06,
47
+ "loss": 0.1464,
48
+ "step": 20,
49
+ "student_loss": 0.026719728484749794,
50
+ "teacher_loss": 0.0005942200659774244
51
+ },
52
+ {
53
+ "epoch": 0.00720876585928489,
54
+ "grad_norm": 2.703125,
55
+ "kd_loss": 0.11552157998085022,
56
+ "learning_rate": 8.000000000000001e-06,
57
+ "loss": 0.1171,
58
+ "step": 25,
59
+ "student_loss": 0.002865401329472661,
60
+ "teacher_loss": 0.0010151522001251578
61
+ },
62
+ {
63
+ "epoch": 0.00865051903114187,
64
+ "grad_norm": 1.484375,
65
+ "kd_loss": 0.07486728578805923,
66
+ "learning_rate": 9.666666666666667e-06,
67
+ "loss": 0.0901,
68
+ "step": 30,
69
+ "student_loss": 0.0011324587976559997,
70
+ "teacher_loss": 0.0005539655685424805
71
+ },
72
+ {
73
+ "epoch": 0.010092272202998846,
74
+ "grad_norm": 2.140625,
75
+ "kd_loss": 0.06898625195026398,
76
+ "learning_rate": 9.99958042442916e-06,
77
+ "loss": 0.0757,
78
+ "step": 35,
79
+ "student_loss": 0.0009480358567088842,
80
+ "teacher_loss": 0.0005732738063670695
81
+ },
82
+ {
83
+ "epoch": 0.011534025374855825,
84
+ "grad_norm": 1.296875,
85
+ "kd_loss": 0.0642521008849144,
86
+ "learning_rate": 9.997876019358083e-06,
87
+ "loss": 0.0685,
88
+ "step": 40,
89
+ "student_loss": 0.031368859112262726,
90
+ "teacher_loss": 0.027499590069055557
91
+ },
92
+ {
93
+ "epoch": 0.012975778546712802,
94
+ "grad_norm": 1.3359375,
95
+ "kd_loss": 0.11454806476831436,
96
+ "learning_rate": 9.99486100792044e-06,
97
+ "loss": 0.0541,
98
+ "step": 45,
99
+ "student_loss": 0.06097811087965965,
100
+ "teacher_loss": 0.0027094183024019003
101
+ },
102
+ {
103
+ "epoch": 0.01441753171856978,
104
+ "grad_norm": 1.5859375,
105
+ "kd_loss": 0.03876315429806709,
106
+ "learning_rate": 9.990536180750724e-06,
107
+ "loss": 0.0563,
108
+ "step": 50,
109
+ "student_loss": 0.02282995544373989,
110
+ "teacher_loss": 0.008940276689827442
111
+ },
112
+ {
113
+ "epoch": 0.015859284890426758,
114
+ "grad_norm": 1.125,
115
+ "kd_loss": 0.033022914081811905,
116
+ "learning_rate": 9.984902671959911e-06,
117
+ "loss": 0.0481,
118
+ "step": 55,
119
+ "student_loss": 0.0011594295501708984,
120
+ "teacher_loss": 0.0007985035772435367
121
+ },
122
+ {
123
+ "epoch": 0.01730103806228374,
124
+ "grad_norm": 0.8828125,
125
+ "kd_loss": 0.030243180692195892,
126
+ "learning_rate": 9.97796195883804e-06,
127
+ "loss": 0.0429,
128
+ "step": 60,
129
+ "student_loss": 0.0034187885466963053,
130
+ "teacher_loss": 0.0017242392059415579
131
+ },
132
+ {
133
+ "epoch": 0.018742791234140715,
134
+ "grad_norm": 1.4453125,
135
+ "kd_loss": 0.032282304018735886,
136
+ "learning_rate": 9.969715861466839e-06,
137
+ "loss": 0.0446,
138
+ "step": 65,
139
+ "student_loss": 0.029871758073568344,
140
+ "teacher_loss": 0.039595428854227066
141
+ },
142
+ {
143
+ "epoch": 0.020184544405997693,
144
+ "grad_norm": 1.1953125,
145
+ "kd_loss": 0.06090804189443588,
146
+ "learning_rate": 9.96016654224243e-06,
147
+ "loss": 0.0477,
148
+ "step": 70,
149
+ "student_loss": 0.017709577456116676,
150
+ "teacher_loss": 0.0022253147326409817
151
+ },
152
+ {
153
+ "epoch": 0.02162629757785467,
154
+ "grad_norm": 0.84765625,
155
+ "kd_loss": 0.024834414944052696,
156
+ "learning_rate": 9.94931650530827e-06,
157
+ "loss": 0.0418,
158
+ "step": 75,
159
+ "student_loss": 0.0012920524459332228,
160
+ "teacher_loss": 0.0008862126851454377
161
+ },
162
+ {
163
+ "epoch": 0.02306805074971165,
164
+ "grad_norm": 0.60546875,
165
+ "kd_loss": 0.025287121534347534,
166
+ "learning_rate": 9.93716859589851e-06,
167
+ "loss": 0.0365,
168
+ "step": 80,
169
+ "student_loss": 0.0013494148151949048,
170
+ "teacher_loss": 0.0008630359079688787
171
+ },
172
+ {
173
+ "epoch": 0.024509803921568627,
174
+ "grad_norm": 0.86328125,
175
+ "kd_loss": 0.024378223344683647,
176
+ "learning_rate": 9.923725999591846e-06,
177
+ "loss": 0.0395,
178
+ "step": 85,
179
+ "student_loss": 0.0005228935624472797,
180
+ "teacher_loss": 0.0004561410460155457
181
+ },
182
+ {
183
+ "epoch": 0.025951557093425604,
184
+ "grad_norm": 0.70703125,
185
+ "kd_loss": 0.07289917767047882,
186
+ "learning_rate": 9.908992241476189e-06,
187
+ "loss": 0.0394,
188
+ "step": 90,
189
+ "student_loss": 0.020422162488102913,
190
+ "teacher_loss": 0.0035499960649758577
191
+ },
192
+ {
193
+ "epoch": 0.027393310265282585,
194
+ "grad_norm": 1.125,
195
+ "kd_loss": 0.044955912977457047,
196
+ "learning_rate": 9.892971185224244e-06,
197
+ "loss": 0.0351,
198
+ "step": 95,
199
+ "student_loss": 0.008261171169579029,
200
+ "teacher_loss": 0.005078152287751436
201
+ },
202
+ {
203
+ "epoch": 0.02883506343713956,
204
+ "grad_norm": 0.83203125,
205
+ "kd_loss": 0.02227398194372654,
206
+ "learning_rate": 9.875667032080354e-06,
207
+ "loss": 0.0326,
208
+ "step": 100,
209
+ "student_loss": 0.0006025677430443466,
210
+ "teacher_loss": 0.00046476206625811756
211
+ },
212
+ {
213
+ "epoch": 0.03027681660899654,
214
+ "grad_norm": 1.2734375,
215
+ "kd_loss": 0.03532887250185013,
216
+ "learning_rate": 9.857084319758772e-06,
217
+ "loss": 0.036,
218
+ "step": 105,
219
+ "student_loss": 0.0034369053319096565,
220
+ "teacher_loss": 0.00029834595625288785
221
+ },
222
+ {
223
+ "epoch": 0.031718569780853516,
224
+ "grad_norm": 1.1640625,
225
+ "kd_loss": 0.04033924266695976,
226
+ "learning_rate": 9.837227921253747e-06,
227
+ "loss": 0.0375,
228
+ "step": 110,
229
+ "student_loss": 0.03080393560230732,
230
+ "teacher_loss": 0.018437745049595833
231
+ },
232
+ {
233
+ "epoch": 0.03316032295271049,
234
+ "grad_norm": 0.703125,
235
+ "kd_loss": 0.04197424277663231,
236
+ "learning_rate": 9.816103043561648e-06,
237
+ "loss": 0.0347,
238
+ "step": 115,
239
+ "student_loss": 0.0021668823901563883,
240
+ "teacher_loss": 0.00045062918798066676
241
+ },
242
+ {
243
+ "epoch": 0.03460207612456748,
244
+ "grad_norm": 0.98828125,
245
+ "kd_loss": 0.027563175186514854,
246
+ "learning_rate": 9.79371522631553e-06,
247
+ "loss": 0.032,
248
+ "step": 120,
249
+ "student_loss": 0.0016319500282406807,
250
+ "teacher_loss": 0.0008567498298361897
251
+ },
252
+ {
253
+ "epoch": 0.036043829296424454,
254
+ "grad_norm": 0.92578125,
255
+ "kd_loss": 0.06173818185925484,
256
+ "learning_rate": 9.770070340332457e-06,
257
+ "loss": 0.0364,
258
+ "step": 125,
259
+ "student_loss": 0.02385914884507656,
260
+ "teacher_loss": 0.00027849775506183505
261
+ },
262
+ {
263
+ "epoch": 0.03748558246828143,
264
+ "grad_norm": 1.0859375,
265
+ "kd_loss": 0.029248492792248726,
266
+ "learning_rate": 9.745174586073982e-06,
267
+ "loss": 0.0346,
268
+ "step": 130,
269
+ "student_loss": 0.0005455865757539868,
270
+ "teacher_loss": 0.0004959598300047219
271
+ },
272
+ {
273
+ "epoch": 0.03892733564013841,
274
+ "grad_norm": 0.953125,
275
+ "kd_loss": 0.0406946986913681,
276
+ "learning_rate": 9.719034492020183e-06,
277
+ "loss": 0.0377,
278
+ "step": 135,
279
+ "student_loss": 0.0013323032762855291,
280
+ "teacher_loss": 0.0005948346224613488
281
+ },
282
+ {
283
+ "epoch": 0.040369088811995385,
284
+ "grad_norm": 1.0390625,
285
+ "kd_loss": 0.023682042956352234,
286
+ "learning_rate": 9.691656912957686e-06,
287
+ "loss": 0.036,
288
+ "step": 140,
289
+ "student_loss": 0.0005881476681679487,
290
+ "teacher_loss": 0.0003679130459204316
291
+ },
292
+ {
293
+ "epoch": 0.04181084198385236,
294
+ "grad_norm": 0.5859375,
295
+ "kd_loss": 0.07271980494260788,
296
+ "learning_rate": 9.663049028182112e-06,
297
+ "loss": 0.0325,
298
+ "step": 145,
299
+ "student_loss": 0.028793470934033394,
300
+ "teacher_loss": 0.005983584560453892
301
+ },
302
+ {
303
+ "epoch": 0.04325259515570934,
304
+ "grad_norm": 0.609375,
305
+ "kd_loss": 0.01843745820224285,
306
+ "learning_rate": 9.633218339615433e-06,
307
+ "loss": 0.0316,
308
+ "step": 150,
309
+ "student_loss": 0.001051027444191277,
310
+ "teacher_loss": 0.000913174357265234
311
+ },
312
+ {
313
+ "epoch": 0.04469434832756632,
314
+ "grad_norm": 1.1171875,
315
+ "kd_loss": 0.023381218314170837,
316
+ "learning_rate": 9.602172669838721e-06,
317
+ "loss": 0.0381,
318
+ "step": 155,
319
+ "student_loss": 0.002775231609120965,
320
+ "teacher_loss": 0.0007182428380474448
321
+ },
322
+ {
323
+ "epoch": 0.0461361014994233,
324
+ "grad_norm": 1.6875,
325
+ "kd_loss": 0.03144950047135353,
326
+ "learning_rate": 9.569920160040815e-06,
327
+ "loss": 0.0344,
328
+ "step": 160,
329
+ "student_loss": 0.0778423622250557,
330
+ "teacher_loss": 0.044956743717193604
331
+ },
332
+ {
333
+ "epoch": 0.04757785467128028,
334
+ "grad_norm": 0.75,
335
+ "kd_loss": 0.03335012122988701,
336
+ "learning_rate": 9.536469267883432e-06,
337
+ "loss": 0.0311,
338
+ "step": 165,
339
+ "student_loss": 0.0012603362556546926,
340
+ "teacher_loss": 0.00417186226695776
341
+ },
342
+ {
343
+ "epoch": 0.049019607843137254,
344
+ "grad_norm": 1.8828125,
345
+ "kd_loss": 0.018793689087033272,
346
+ "learning_rate": 9.501828765283295e-06,
347
+ "loss": 0.0355,
348
+ "step": 170,
349
+ "student_loss": 0.0005546013708226383,
350
+ "teacher_loss": 0.0003494401171337813
351
+ },
352
+ {
353
+ "epoch": 0.05046136101499423,
354
+ "grad_norm": 1.09375,
355
+ "kd_loss": 0.027096513658761978,
356
+ "learning_rate": 9.466007736111846e-06,
357
+ "loss": 0.0322,
358
+ "step": 175,
359
+ "student_loss": 0.0030481775756925344,
360
+ "teacher_loss": 0.0003126203373540193
361
+ },
362
+ {
363
+ "epoch": 0.05190311418685121,
364
+ "grad_norm": 0.69140625,
365
+ "kd_loss": 0.03634097799658775,
366
+ "learning_rate": 9.429015573813163e-06,
367
+ "loss": 0.0302,
368
+ "step": 180,
369
+ "student_loss": 0.0012228424893692136,
370
+ "teacher_loss": 0.00034309024340473115
371
+ },
372
+ {
373
+ "epoch": 0.05334486735870819,
374
+ "grad_norm": 1.34375,
375
+ "kd_loss": 0.03704367205500603,
376
+ "learning_rate": 9.390861978940687e-06,
377
+ "loss": 0.0363,
378
+ "step": 185,
379
+ "student_loss": 0.035696618258953094,
380
+ "teacher_loss": 0.00034108557156287134
381
+ },
382
+ {
383
+ "epoch": 0.05478662053056517,
384
+ "grad_norm": 1.3046875,
385
+ "kd_loss": 0.052902791649103165,
386
+ "learning_rate": 9.351556956613423e-06,
387
+ "loss": 0.0321,
388
+ "step": 190,
389
+ "student_loss": 0.009327664040029049,
390
+ "teacher_loss": 0.0015260990476235747
391
+ },
392
+ {
393
+ "epoch": 0.056228373702422146,
394
+ "grad_norm": 2.828125,
395
+ "kd_loss": 0.05315268039703369,
396
+ "learning_rate": 9.31111081389227e-06,
397
+ "loss": 0.0327,
398
+ "step": 195,
399
+ "student_loss": 0.011893535032868385,
400
+ "teacher_loss": 0.0007797812577337027
401
+ },
402
+ {
403
+ "epoch": 0.05767012687427912,
404
+ "grad_norm": 0.8125,
405
+ "kd_loss": 0.03527738153934479,
406
+ "learning_rate": 9.269534157077177e-06,
407
+ "loss": 0.0394,
408
+ "step": 200,
409
+ "student_loss": 0.0005083663854748011,
410
+ "teacher_loss": 0.0002656931465025991
411
+ },
412
+ {
413
+ "epoch": 0.0591118800461361,
414
+ "grad_norm": 0.9921875,
415
+ "kd_loss": 0.02736036665737629,
416
+ "learning_rate": 9.226837888925813e-06,
417
+ "loss": 0.0341,
418
+ "step": 205,
419
+ "student_loss": 0.03568984195590019,
420
+ "teacher_loss": 0.027640890330076218
421
+ },
422
+ {
423
+ "epoch": 0.06055363321799308,
424
+ "grad_norm": 1.78125,
425
+ "kd_loss": 0.030339844524860382,
426
+ "learning_rate": 9.183033205794525e-06,
427
+ "loss": 0.0302,
428
+ "step": 210,
429
+ "student_loss": 0.000763049116358161,
430
+ "teacher_loss": 0.0003548153617884964
431
+ },
432
+ {
433
+ "epoch": 0.061995386389850055,
434
+ "grad_norm": 1.1484375,
435
+ "kd_loss": 0.03865697979927063,
436
+ "learning_rate": 9.13813159470227e-06,
437
+ "loss": 0.0326,
438
+ "step": 215,
439
+ "student_loss": 0.0005148217896930873,
440
+ "teacher_loss": 0.00020245747873559594
441
+ },
442
+ {
443
+ "epoch": 0.06343713956170703,
444
+ "grad_norm": 1.3671875,
445
+ "kd_loss": 0.04260854423046112,
446
+ "learning_rate": 9.092144830318357e-06,
447
+ "loss": 0.0316,
448
+ "step": 220,
449
+ "student_loss": 0.030991079285740852,
450
+ "teacher_loss": 0.00993641559034586
451
+ },
452
+ {
453
+ "epoch": 0.06487889273356401,
454
+ "grad_norm": 1.203125,
455
+ "kd_loss": 0.023958567529916763,
456
+ "learning_rate": 9.045084971874738e-06,
457
+ "loss": 0.0376,
458
+ "step": 225,
459
+ "student_loss": 0.043197184801101685,
460
+ "teacher_loss": 0.00216495874337852
461
+ },
462
+ {
463
+ "epoch": 0.06632064590542099,
464
+ "grad_norm": 1.7578125,
465
+ "kd_loss": 0.04370651766657829,
466
+ "learning_rate": 8.99696436000368e-06,
467
+ "loss": 0.0299,
468
+ "step": 230,
469
+ "student_loss": 0.005392159800976515,
470
+ "teacher_loss": 0.003661371534690261
471
+ },
472
+ {
473
+ "epoch": 0.06776239907727798,
474
+ "grad_norm": 0.8671875,
475
+ "kd_loss": 0.07619086652994156,
476
+ "learning_rate": 8.947795613501658e-06,
477
+ "loss": 0.0314,
478
+ "step": 235,
479
+ "student_loss": 0.008586333133280277,
480
+ "teacher_loss": 0.0005631999811157584
481
+ },
482
+ {
483
+ "epoch": 0.06920415224913495,
484
+ "grad_norm": 1.0625,
485
+ "kd_loss": 0.03111647628247738,
486
+ "learning_rate": 8.897591626020284e-06,
487
+ "loss": 0.034,
488
+ "step": 240,
489
+ "student_loss": 0.003906027879565954,
490
+ "teacher_loss": 0.0004787310608662665
491
+ },
492
+ {
493
+ "epoch": 0.07064590542099193,
494
+ "grad_norm": 1.453125,
495
+ "kd_loss": 0.017389042302966118,
496
+ "learning_rate": 8.846365562685178e-06,
497
+ "loss": 0.0279,
498
+ "step": 245,
499
+ "student_loss": 0.013969292864203453,
500
+ "teacher_loss": 0.006873726844787598
501
+ },
502
+ {
503
+ "epoch": 0.07208765859284891,
504
+ "grad_norm": 1.6015625,
505
+ "kd_loss": 0.0229345690459013,
506
+ "learning_rate": 8.794130856643635e-06,
507
+ "loss": 0.0311,
508
+ "step": 250,
509
+ "student_loss": 0.0008334179292432964,
510
+ "teacher_loss": 0.0003169570700265467
511
+ },
512
+ {
513
+ "epoch": 0.07352941176470588,
514
+ "grad_norm": 0.83203125,
515
+ "kd_loss": 0.01808979921042919,
516
+ "learning_rate": 8.74090120554202e-06,
517
+ "loss": 0.0312,
518
+ "step": 255,
519
+ "student_loss": 0.0003083710907958448,
520
+ "teacher_loss": 0.00030115401023067534
521
+ },
522
+ {
523
+ "epoch": 0.07497116493656286,
524
+ "grad_norm": 0.88671875,
525
+ "kd_loss": 0.026108454912900925,
526
+ "learning_rate": 8.686690567933803e-06,
527
+ "loss": 0.0333,
528
+ "step": 260,
529
+ "student_loss": 0.042571116238832474,
530
+ "teacher_loss": 0.03388316184282303
531
+ },
532
+ {
533
+ "epoch": 0.07641291810841984,
534
+ "grad_norm": 0.80078125,
535
+ "kd_loss": 0.016656002029776573,
536
+ "learning_rate": 8.63151315961915e-06,
537
+ "loss": 0.0317,
538
+ "step": 265,
539
+ "student_loss": 0.0003702428948599845,
540
+ "teacher_loss": 0.0002555136161390692
541
+ },
542
+ {
543
+ "epoch": 0.07785467128027682,
544
+ "grad_norm": 1.046875,
545
+ "kd_loss": 0.019304102286696434,
546
+ "learning_rate": 8.575383449917103e-06,
547
+ "loss": 0.0342,
548
+ "step": 270,
549
+ "student_loss": 0.001813149661757052,
550
+ "teacher_loss": 0.0011580288410186768
551
+ },
552
+ {
553
+ "epoch": 0.07929642445213379,
554
+ "grad_norm": 1.1796875,
555
+ "kd_loss": 0.023480774834752083,
556
+ "learning_rate": 8.518316157871232e-06,
557
+ "loss": 0.029,
558
+ "step": 275,
559
+ "student_loss": 0.04764978215098381,
560
+ "teacher_loss": 0.03439468517899513
561
+ },
562
+ {
563
+ "epoch": 0.08073817762399077,
564
+ "grad_norm": 1.015625,
565
+ "kd_loss": 0.03275206685066223,
566
+ "learning_rate": 8.460326248389825e-06,
567
+ "loss": 0.0289,
568
+ "step": 280,
569
+ "student_loss": 0.0005029537715017796,
570
+ "teacher_loss": 0.00019533037266228348
571
+ },
572
+ {
573
+ "epoch": 0.08217993079584775,
574
+ "grad_norm": 1.140625,
575
+ "kd_loss": 0.019457675516605377,
576
+ "learning_rate": 8.401428928321607e-06,
577
+ "loss": 0.0322,
578
+ "step": 285,
579
+ "student_loss": 0.0007758038118481636,
580
+ "teacher_loss": 0.000900130660738796
581
+ },
582
+ {
583
+ "epoch": 0.08362168396770472,
584
+ "grad_norm": 1.3125,
585
+ "kd_loss": 0.017802242189645767,
586
+ "learning_rate": 8.341639642468002e-06,
587
+ "loss": 0.0348,
588
+ "step": 290,
589
+ "student_loss": 0.023405462503433228,
590
+ "teacher_loss": 0.021540865302085876
591
+ },
592
+ {
593
+ "epoch": 0.0850634371395617,
594
+ "grad_norm": 1.03125,
595
+ "kd_loss": 0.01647804118692875,
596
+ "learning_rate": 8.280974069532999e-06,
597
+ "loss": 0.0328,
598
+ "step": 295,
599
+ "student_loss": 0.0006198033224791288,
600
+ "teacher_loss": 0.000554366793949157
601
+ },
602
+ {
603
+ "epoch": 0.08650519031141868,
604
+ "grad_norm": 0.9296875,
605
+ "kd_loss": 0.04784730449318886,
606
+ "learning_rate": 8.219448118011687e-06,
607
+ "loss": 0.0308,
608
+ "step": 300,
609
+ "student_loss": 0.015613794326782227,
610
+ "teacher_loss": 0.001725711626932025
611
+ },
612
+ {
613
+ "epoch": 0.08794694348327567,
614
+ "grad_norm": 1.1953125,
615
+ "kd_loss": 0.019965235143899918,
616
+ "learning_rate": 8.157077922018537e-06,
617
+ "loss": 0.0289,
618
+ "step": 305,
619
+ "student_loss": 0.006098510231822729,
620
+ "teacher_loss": 0.002777156652882695
621
+ },
622
+ {
623
+ "epoch": 0.08938869665513265,
624
+ "grad_norm": 1.0390625,
625
+ "kd_loss": 0.020748196169734,
626
+ "learning_rate": 8.093879837056486e-06,
627
+ "loss": 0.0309,
628
+ "step": 310,
629
+ "student_loss": 0.000493990199174732,
630
+ "teacher_loss": 0.00039537265547551215
631
+ },
632
+ {
633
+ "epoch": 0.09083044982698962,
634
+ "grad_norm": 1.4765625,
635
+ "kd_loss": 0.03892743960022926,
636
+ "learning_rate": 8.029870435728018e-06,
637
+ "loss": 0.0294,
638
+ "step": 315,
639
+ "student_loss": 0.0073717073537409306,
640
+ "teacher_loss": 0.00021180440671741962
641
+ },
642
+ {
643
+ "epoch": 0.0922722029988466,
644
+ "grad_norm": 1.3828125,
645
+ "kd_loss": 0.02363528683781624,
646
+ "learning_rate": 7.965066503389264e-06,
647
+ "loss": 0.0313,
648
+ "step": 320,
649
+ "student_loss": 0.0004119524674024433,
650
+ "teacher_loss": 0.00022927633835934103
651
+ },
652
+ {
653
+ "epoch": 0.09371395617070358,
654
+ "grad_norm": 1.7421875,
655
+ "kd_loss": 0.06230268254876137,
656
+ "learning_rate": 7.89948503374835e-06,
657
+ "loss": 0.0284,
658
+ "step": 325,
659
+ "student_loss": 0.009024329483509064,
660
+ "teacher_loss": 0.01963256485760212
661
+ },
662
+ {
663
+ "epoch": 0.09515570934256055,
664
+ "grad_norm": 0.90625,
665
+ "kd_loss": 0.04712303727865219,
666
+ "learning_rate": 7.833143224409076e-06,
667
+ "loss": 0.0302,
668
+ "step": 330,
669
+ "student_loss": 0.008720812387764454,
670
+ "teacher_loss": 0.0013501221546903253
671
+ },
672
+ {
673
+ "epoch": 0.09659746251441753,
674
+ "grad_norm": 0.9140625,
675
+ "kd_loss": 0.018766457214951515,
676
+ "learning_rate": 7.766058472361154e-06,
677
+ "loss": 0.0283,
678
+ "step": 335,
679
+ "student_loss": 0.001242243917658925,
680
+ "teacher_loss": 0.0023425817489624023
681
+ },
682
+ {
683
+ "epoch": 0.09803921568627451,
684
+ "grad_norm": 2.078125,
685
+ "kd_loss": 0.02911657840013504,
686
+ "learning_rate": 7.698248369418146e-06,
687
+ "loss": 0.0326,
688
+ "step": 340,
689
+ "student_loss": 0.009861858561635017,
690
+ "teacher_loss": 0.006796200294047594
691
+ },
692
+ {
693
+ "epoch": 0.09948096885813149,
694
+ "grad_norm": 1.2265625,
695
+ "kd_loss": 0.032899159938097,
696
+ "learning_rate": 7.629730697604314e-06,
697
+ "loss": 0.0335,
698
+ "step": 345,
699
+ "student_loss": 0.00739182299003005,
700
+ "teacher_loss": 0.0002512071805540472
701
+ },
702
+ {
703
+ "epoch": 0.10092272202998846,
704
+ "grad_norm": 1.0859375,
705
+ "kd_loss": 0.06120900437235832,
706
+ "learning_rate": 7.560523424491595e-06,
707
+ "loss": 0.0282,
708
+ "step": 350,
709
+ "student_loss": 0.003599822986871004,
710
+ "teacher_loss": 0.0007063632365316153
711
+ },
712
+ {
713
+ "epoch": 0.10236447520184544,
714
+ "grad_norm": 2.296875,
715
+ "kd_loss": 0.02087453007698059,
716
+ "learning_rate": 7.490644698487909e-06,
717
+ "loss": 0.035,
718
+ "step": 355,
719
+ "student_loss": 0.001389339566230774,
720
+ "teacher_loss": 0.0017551770433783531
721
+ },
722
+ {
723
+ "epoch": 0.10380622837370242,
724
+ "grad_norm": 0.90234375,
725
+ "kd_loss": 0.019557664170861244,
726
+ "learning_rate": 7.420112844078066e-06,
727
+ "loss": 0.0321,
728
+ "step": 360,
729
+ "student_loss": 0.0006360138650052249,
730
+ "teacher_loss": 0.0005147532792761922
731
+ },
732
+ {
733
+ "epoch": 0.1052479815455594,
734
+ "grad_norm": 1.3203125,
735
+ "kd_loss": 0.030097220093011856,
736
+ "learning_rate": 7.348946357018479e-06,
737
+ "loss": 0.0333,
738
+ "step": 365,
739
+ "student_loss": 0.04143820330500603,
740
+ "teacher_loss": 0.027219083160161972
741
+ },
742
+ {
743
+ "epoch": 0.10668973471741638,
744
+ "grad_norm": 0.9453125,
745
+ "kd_loss": 0.05344080179929733,
746
+ "learning_rate": 7.277163899486975e-06,
747
+ "loss": 0.0351,
748
+ "step": 370,
749
+ "student_loss": 0.021593965590000153,
750
+ "teacher_loss": 0.0005673202103935182
751
+ },
752
+ {
753
+ "epoch": 0.10813148788927336,
754
+ "grad_norm": 0.7265625,
755
+ "kd_loss": 0.03127056360244751,
756
+ "learning_rate": 7.204784295188959e-06,
757
+ "loss": 0.0287,
758
+ "step": 375,
759
+ "student_loss": 0.03163963928818703,
760
+ "teacher_loss": 0.020587248727679253
761
+ },
762
+ {
763
+ "epoch": 0.10957324106113034,
764
+ "grad_norm": 0.84375,
765
+ "kd_loss": 0.01984175480902195,
766
+ "learning_rate": 7.1318265244212305e-06,
767
+ "loss": 0.0311,
768
+ "step": 380,
769
+ "student_loss": 0.0005506337620317936,
770
+ "teacher_loss": 0.00038444914389401674
771
+ },
772
+ {
773
+ "epoch": 0.11101499423298732,
774
+ "grad_norm": 0.59765625,
775
+ "kd_loss": 0.02916550263762474,
776
+ "learning_rate": 7.05830971909472e-06,
777
+ "loss": 0.0294,
778
+ "step": 385,
779
+ "student_loss": 0.00045756620238535106,
780
+ "teacher_loss": 0.0003828653716482222
781
+ },
782
+ {
783
+ "epoch": 0.11245674740484429,
784
+ "grad_norm": 1.171875,
785
+ "kd_loss": 0.021802764385938644,
786
+ "learning_rate": 6.9842531577174865e-06,
787
+ "loss": 0.0271,
788
+ "step": 390,
789
+ "student_loss": 0.00026892355526797473,
790
+ "teacher_loss": 0.0002502185816410929
791
+ },
792
+ {
793
+ "epoch": 0.11389850057670127,
794
+ "grad_norm": 0.94140625,
795
+ "kd_loss": 0.020051907747983932,
796
+ "learning_rate": 6.9096762603392595e-06,
797
+ "loss": 0.0307,
798
+ "step": 395,
799
+ "student_loss": 0.0004568768781609833,
800
+ "teacher_loss": 0.00039133013342507184
801
+ },
802
+ {
803
+ "epoch": 0.11534025374855825,
804
+ "grad_norm": 0.78125,
805
+ "kd_loss": 0.03054620325565338,
806
+ "learning_rate": 6.834598583458862e-06,
807
+ "loss": 0.0275,
808
+ "step": 400,
809
+ "student_loss": 0.0005328103434294462,
810
+ "teacher_loss": 0.00015425821766257286
811
+ },
812
+ {
813
+ "epoch": 0.11678200692041522,
814
+ "grad_norm": 0.640625,
815
+ "kd_loss": 0.02604236640036106,
816
+ "learning_rate": 6.7590398148958625e-06,
817
+ "loss": 0.0335,
818
+ "step": 405,
819
+ "student_loss": 0.011330639012157917,
820
+ "teacher_loss": 0.00021900788124185055
821
+ },
822
+ {
823
+ "epoch": 0.1182237600922722,
824
+ "grad_norm": 0.8046875,
825
+ "kd_loss": 0.03826223686337471,
826
+ "learning_rate": 6.6830197686277945e-06,
827
+ "loss": 0.0366,
828
+ "step": 410,
829
+ "student_loss": 0.014453927055001259,
830
+ "teacher_loss": 0.00844341516494751
831
+ },
832
+ {
833
+ "epoch": 0.11966551326412918,
834
+ "grad_norm": 1.203125,
835
+ "kd_loss": 0.021850954741239548,
836
+ "learning_rate": 6.6065583795942625e-06,
837
+ "loss": 0.0347,
838
+ "step": 415,
839
+ "student_loss": 0.004759110510349274,
840
+ "teacher_loss": 0.0034685542341321707
841
+ },
842
+ {
843
+ "epoch": 0.12110726643598616,
844
+ "grad_norm": 0.93359375,
845
+ "kd_loss": 0.045137468725442886,
846
+ "learning_rate": 6.52967569846937e-06,
847
+ "loss": 0.0331,
848
+ "step": 420,
849
+ "student_loss": 0.04586087912321091,
850
+ "teacher_loss": 0.021448055282235146
851
+ },
852
+ {
853
+ "epoch": 0.12254901960784313,
854
+ "grad_norm": 1.4453125,
855
+ "kd_loss": 0.014929288066923618,
856
+ "learning_rate": 6.452391886403767e-06,
857
+ "loss": 0.0299,
858
+ "step": 425,
859
+ "student_loss": 0.002178685739636421,
860
+ "teacher_loss": 0.0021052202209830284
861
+ },
862
+ {
863
+ "epoch": 0.12399077277970011,
864
+ "grad_norm": 1.3671875,
865
+ "kd_loss": 0.0547032505273819,
866
+ "learning_rate": 6.374727209737743e-06,
867
+ "loss": 0.0368,
868
+ "step": 430,
869
+ "student_loss": 0.05219801887869835,
870
+ "teacher_loss": 0.060560259968042374
871
+ },
872
+ {
873
+ "epoch": 0.1254325259515571,
874
+ "grad_norm": 1.453125,
875
+ "kd_loss": 0.01706705428659916,
876
+ "learning_rate": 6.296702034686726e-06,
877
+ "loss": 0.0301,
878
+ "step": 435,
879
+ "student_loss": 0.0008379022474400699,
880
+ "teacher_loss": 0.0006224109092727304
881
+ },
882
+ {
883
+ "epoch": 0.12687427912341406,
884
+ "grad_norm": 1.0,
885
+ "kd_loss": 0.03890637308359146,
886
+ "learning_rate": 6.218336822000598e-06,
887
+ "loss": 0.0314,
888
+ "step": 440,
889
+ "student_loss": 0.012142423540353775,
890
+ "teacher_loss": 0.007434291299432516
891
+ },
892
+ {
893
+ "epoch": 0.12831603229527105,
894
+ "grad_norm": 0.890625,
895
+ "kd_loss": 0.03179040178656578,
896
+ "learning_rate": 6.139652121598219e-06,
897
+ "loss": 0.0313,
898
+ "step": 445,
899
+ "student_loss": 0.03341586887836456,
900
+ "teacher_loss": 0.018860887736082077
901
+ },
902
+ {
903
+ "epoch": 0.12975778546712802,
904
+ "grad_norm": 0.69140625,
905
+ "kd_loss": 0.03629063814878464,
906
+ "learning_rate": 6.060668567178561e-06,
907
+ "loss": 0.0329,
908
+ "step": 450,
909
+ "student_loss": 0.0010012147249653935,
910
+ "teacher_loss": 0.0014839700888842344
911
+ },
912
+ {
913
+ "epoch": 0.131199538638985,
914
+ "grad_norm": 0.8828125,
915
+ "kd_loss": 0.02263510413467884,
916
+ "learning_rate": 5.981406870809889e-06,
917
+ "loss": 0.0326,
918
+ "step": 455,
919
+ "student_loss": 0.0007265008171088994,
920
+ "teacher_loss": 0.0003751025360543281
921
+ },
922
+ {
923
+ "epoch": 0.13264129181084197,
924
+ "grad_norm": 1.2109375,
925
+ "kd_loss": 0.0492476262152195,
926
+ "learning_rate": 5.9018878174983674e-06,
927
+ "loss": 0.0295,
928
+ "step": 460,
929
+ "student_loss": 0.005101657006889582,
930
+ "teacher_loss": 0.0002927044697571546
931
+ },
932
+ {
933
+ "epoch": 0.13408304498269896,
934
+ "grad_norm": 1.2578125,
935
+ "kd_loss": 0.031007954850792885,
936
+ "learning_rate": 5.822132259737565e-06,
937
+ "loss": 0.034,
938
+ "step": 465,
939
+ "student_loss": 0.00047535731573589146,
940
+ "teacher_loss": 0.00024468303308822215
941
+ },
942
+ {
943
+ "epoch": 0.13552479815455595,
944
+ "grad_norm": 1.7265625,
945
+ "kd_loss": 0.02650834433734417,
946
+ "learning_rate": 5.742161112040237e-06,
947
+ "loss": 0.0313,
948
+ "step": 470,
949
+ "student_loss": 0.000378905504476279,
950
+ "teacher_loss": 0.0002708406245801598
951
+ },
952
+ {
953
+ "epoch": 0.13696655132641292,
954
+ "grad_norm": 1.171875,
955
+ "kd_loss": 0.03383399918675423,
956
+ "learning_rate": 5.661995345453867e-06,
957
+ "loss": 0.0289,
958
+ "step": 475,
959
+ "student_loss": 0.0004789994563907385,
960
+ "teacher_loss": 0.0002899342798627913
961
+ },
962
+ {
963
+ "epoch": 0.1384083044982699,
964
+ "grad_norm": 1.1953125,
965
+ "kd_loss": 0.03584924340248108,
966
+ "learning_rate": 5.581655982061367e-06,
967
+ "loss": 0.0422,
968
+ "step": 480,
969
+ "student_loss": 0.03862505778670311,
970
+ "teacher_loss": 0.036061737686395645
971
+ },
972
+ {
973
+ "epoch": 0.13985005767012687,
974
+ "grad_norm": 1.203125,
975
+ "kd_loss": 0.020711667835712433,
976
+ "learning_rate": 5.501164089468406e-06,
977
+ "loss": 0.0313,
978
+ "step": 485,
979
+ "student_loss": 0.0004981390666216612,
980
+ "teacher_loss": 0.0005692155100405216
981
+ },
982
+ {
983
+ "epoch": 0.14129181084198386,
984
+ "grad_norm": 1.3984375,
985
+ "kd_loss": 0.027012551203370094,
986
+ "learning_rate": 5.4205407752787884e-06,
987
+ "loss": 0.0367,
988
+ "step": 490,
989
+ "student_loss": 0.0008920110412873328,
990
+ "teacher_loss": 0.0007303191814571619
991
+ },
992
+ {
993
+ "epoch": 0.14273356401384082,
994
+ "grad_norm": 1.3671875,
995
+ "kd_loss": 0.040799129754304886,
996
+ "learning_rate": 5.339807181559359e-06,
997
+ "loss": 0.0399,
998
+ "step": 495,
999
+ "student_loss": 0.0009312813053838909,
1000
+ "teacher_loss": 0.0005571586079895496
1001
+ },
1002
+ {
1003
+ "epoch": 0.14417531718569782,
1004
+ "grad_norm": 1.6875,
1005
+ "kd_loss": 0.03058801032602787,
1006
+ "learning_rate": 5.258984479295853e-06,
1007
+ "loss": 0.0316,
1008
+ "step": 500,
1009
+ "student_loss": 0.00041255459655076265,
1010
+ "teacher_loss": 0.00031951998244039714
1011
+ },
1012
+ {
1013
+ "epoch": 0.14561707035755478,
1014
+ "grad_norm": 1.734375,
1015
+ "kd_loss": 0.02832438424229622,
1016
+ "learning_rate": 5.1780938628411795e-06,
1017
+ "loss": 0.0402,
1018
+ "step": 505,
1019
+ "student_loss": 0.0005547971813939512,
1020
+ "teacher_loss": 0.0004580595705192536
1021
+ },
1022
+ {
1023
+ "epoch": 0.14705882352941177,
1024
+ "grad_norm": 2.28125,
1025
+ "kd_loss": 0.019691068679094315,
1026
+ "learning_rate": 5.097156544357567e-06,
1027
+ "loss": 0.032,
1028
+ "step": 510,
1029
+ "student_loss": 0.00035549805033952,
1030
+ "teacher_loss": 0.00029006152180954814
1031
+ },
1032
+ {
1033
+ "epoch": 0.14850057670126873,
1034
+ "grad_norm": 1.4609375,
1035
+ "kd_loss": 0.028027402237057686,
1036
+ "learning_rate": 5.016193748254045e-06,
1037
+ "loss": 0.0299,
1038
+ "step": 515,
1039
+ "student_loss": 0.016550345346331596,
1040
+ "teacher_loss": 0.00025012606056407094
1041
+ },
1042
+ {
1043
+ "epoch": 0.14994232987312572,
1044
+ "grad_norm": 1.7734375,
1045
+ "kd_loss": 0.06320768594741821,
1046
+ "learning_rate": 4.935226705620699e-06,
1047
+ "loss": 0.0344,
1048
+ "step": 520,
1049
+ "student_loss": 0.04100845754146576,
1050
+ "teacher_loss": 0.011567573063075542
1051
+ },
1052
+ {
1053
+ "epoch": 0.1513840830449827,
1054
+ "grad_norm": 0.859375,
1055
+ "kd_loss": 0.032060496509075165,
1056
+ "learning_rate": 4.8542766486612035e-06,
1057
+ "loss": 0.033,
1058
+ "step": 525,
1059
+ "student_loss": 0.0004020243068225682,
1060
+ "teacher_loss": 0.00020839735225308686
1061
+ },
1062
+ {
1063
+ "epoch": 0.15282583621683968,
1064
+ "grad_norm": 1.203125,
1065
+ "kd_loss": 0.024977991357445717,
1066
+ "learning_rate": 4.773364805125025e-06,
1067
+ "loss": 0.0309,
1068
+ "step": 530,
1069
+ "student_loss": 0.0005168431089259684,
1070
+ "teacher_loss": 0.00039579602889716625
1071
+ },
1072
+ {
1073
+ "epoch": 0.15426758938869667,
1074
+ "grad_norm": 2.046875,
1075
+ "kd_loss": 0.03925255313515663,
1076
+ "learning_rate": 4.6925123927408265e-06,
1077
+ "loss": 0.0308,
1078
+ "step": 535,
1079
+ "student_loss": 0.0007517871563322842,
1080
+ "teacher_loss": 0.00039392782491631806
1081
+ },
1082
+ {
1083
+ "epoch": 0.15570934256055363,
1084
+ "grad_norm": 1.8203125,
1085
+ "kd_loss": 0.05933701992034912,
1086
+ "learning_rate": 4.611740613652485e-06,
1087
+ "loss": 0.0286,
1088
+ "step": 540,
1089
+ "student_loss": 0.0131508968770504,
1090
+ "teacher_loss": 0.0005503469728864729
1091
+ },
1092
+ {
1093
+ "epoch": 0.15715109573241062,
1094
+ "grad_norm": 1.0390625,
1095
+ "kd_loss": 0.023665662854909897,
1096
+ "learning_rate": 4.531070648859186e-06,
1097
+ "loss": 0.0282,
1098
+ "step": 545,
1099
+ "student_loss": 0.0027271448634564877,
1100
+ "teacher_loss": 0.0027394567150622606
1101
+ },
1102
+ {
1103
+ "epoch": 0.15859284890426759,
1104
+ "grad_norm": 2.046875,
1105
+ "kd_loss": 0.05410351976752281,
1106
+ "learning_rate": 4.450523652661086e-06,
1107
+ "loss": 0.0276,
1108
+ "step": 550,
1109
+ "student_loss": 0.00023584096925333142,
1110
+ "teacher_loss": 0.0036266453098505735
1111
+ },
1112
+ {
1113
+ "epoch": 0.16003460207612458,
1114
+ "grad_norm": 1.1796875,
1115
+ "kd_loss": 0.08308840543031693,
1116
+ "learning_rate": 4.370120747111956e-06,
1117
+ "loss": 0.0327,
1118
+ "step": 555,
1119
+ "student_loss": 0.046959906816482544,
1120
+ "teacher_loss": 0.0017155115492641926
1121
+ },
1122
+ {
1123
+ "epoch": 0.16147635524798154,
1124
+ "grad_norm": 2.4375,
1125
+ "kd_loss": 0.0380968302488327,
1126
+ "learning_rate": 4.289883016480291e-06,
1127
+ "loss": 0.0349,
1128
+ "step": 560,
1129
+ "student_loss": 0.001130731194280088,
1130
+ "teacher_loss": 0.00036855213693343103
1131
+ },
1132
+ {
1133
+ "epoch": 0.16291810841983853,
1134
+ "grad_norm": 1.1328125,
1135
+ "kd_loss": 0.03884750232100487,
1136
+ "learning_rate": 4.209831501720328e-06,
1137
+ "loss": 0.0325,
1138
+ "step": 565,
1139
+ "student_loss": 0.02741910330951214,
1140
+ "teacher_loss": 0.019537584856152534
1141
+ },
1142
+ {
1143
+ "epoch": 0.1643598615916955,
1144
+ "grad_norm": 5.1875,
1145
+ "kd_loss": 0.0330473892390728,
1146
+ "learning_rate": 4.129987194954421e-06,
1147
+ "loss": 0.0356,
1148
+ "step": 570,
1149
+ "student_loss": 0.010371256619691849,
1150
+ "teacher_loss": 0.0070776245556771755
1151
+ },
1152
+ {
1153
+ "epoch": 0.16580161476355249,
1154
+ "grad_norm": 1.1015625,
1155
+ "kd_loss": 0.028591087087988853,
1156
+ "learning_rate": 4.050371033968216e-06,
1157
+ "loss": 0.0328,
1158
+ "step": 575,
1159
+ "student_loss": 0.000467249978100881,
1160
+ "teacher_loss": 0.0002494181098882109
1161
+ },
1162
+ {
1163
+ "epoch": 0.16724336793540945,
1164
+ "grad_norm": 0.80078125,
1165
+ "kd_loss": 0.02640584297478199,
1166
+ "learning_rate": 3.9710038967200825e-06,
1167
+ "loss": 0.0284,
1168
+ "step": 580,
1169
+ "student_loss": 0.0017712963744997978,
1170
+ "teacher_loss": 0.0019773358944803476
1171
+ },
1172
+ {
1173
+ "epoch": 0.16868512110726644,
1174
+ "grad_norm": 0.85546875,
1175
+ "kd_loss": 0.030399348586797714,
1176
+ "learning_rate": 3.89190659586623e-06,
1177
+ "loss": 0.0281,
1178
+ "step": 585,
1179
+ "student_loss": 0.0004607080190908164,
1180
+ "teacher_loss": 0.0002647584769874811
1181
+ },
1182
+ {
1183
+ "epoch": 0.1701268742791234,
1184
+ "grad_norm": 1.1328125,
1185
+ "kd_loss": 0.03226058557629585,
1186
+ "learning_rate": 3.8130998733029517e-06,
1187
+ "loss": 0.0302,
1188
+ "step": 590,
1189
+ "student_loss": 0.0005413145408965647,
1190
+ "teacher_loss": 0.0011012081522494555
1191
+ },
1192
+ {
1193
+ "epoch": 0.1715686274509804,
1194
+ "grad_norm": 1.8203125,
1195
+ "kd_loss": 0.023172084242105484,
1196
+ "learning_rate": 3.734604394727419e-06,
1197
+ "loss": 0.0325,
1198
+ "step": 595,
1199
+ "student_loss": 0.00059064372908324,
1200
+ "teacher_loss": 0.0005090903723612428
1201
+ },
1202
+ {
1203
+ "epoch": 0.17301038062283736,
1204
+ "grad_norm": 1.1640625,
1205
+ "kd_loss": 0.038484539836645126,
1206
+ "learning_rate": 3.656440744218464e-06,
1207
+ "loss": 0.0346,
1208
+ "step": 600,
1209
+ "student_loss": 0.041177455335855484,
1210
+ "teacher_loss": 0.004930450581014156
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_a0.5_b1.0_g0.5_ls2/checkpoint-600/vocab.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoints/codi3b_a0.5_b1.0_g0.5_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_a0.5_b1.0_g0.5_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_a0.5_b1.0_g0.5_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_a0.5_b1.0_g0.5_ls2/checkpoint-800/merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
checkpoints/codi3b_a0.5_b1.0_g0.5_ls2/checkpoint-800/pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9edcc00bfefc311d04e3e1a0d401e8fbed45afc64aae68aa648cda8c39bc16c0
3
+ size 6187858991