sirui6011 commited on
Commit
2da4a2f
·
verified ·
1 Parent(s): 0f328d4

add checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. .gitattributes +7 -0
  2. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-1000/added_tokens.json +35 -0
  3. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-1000/chat_template.jinja +54 -0
  4. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-1000/config.json +58 -0
  5. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-1000/merges.txt +0 -0
  6. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-1000/pytorch_model.bin +3 -0
  7. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-1000/special_tokens_map.json +31 -0
  8. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-1000/thought_projector.pt +3 -0
  9. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-1000/tokenizer.json +3 -0
  10. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-1000/tokenizer_config.json +295 -0
  11. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-1000/trainer_state.json +2046 -0
  12. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-1000/vocab.json +0 -0
  13. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-200/added_tokens.json +35 -0
  14. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-200/chat_template.jinja +54 -0
  15. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-200/config.json +58 -0
  16. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-200/merges.txt +0 -0
  17. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-200/pytorch_model.bin +3 -0
  18. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-200/special_tokens_map.json +31 -0
  19. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-200/thought_projector.pt +3 -0
  20. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-200/tokenizer.json +3 -0
  21. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-200/tokenizer_config.json +295 -0
  22. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-200/trainer_state.json +434 -0
  23. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-200/vocab.json +0 -0
  24. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-400/added_tokens.json +35 -0
  25. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-400/chat_template.jinja +54 -0
  26. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-400/config.json +58 -0
  27. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-400/merges.txt +0 -0
  28. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-400/pytorch_model.bin +3 -0
  29. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-400/special_tokens_map.json +31 -0
  30. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-400/thought_projector.pt +3 -0
  31. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-400/tokenizer.json +3 -0
  32. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-400/tokenizer_config.json +295 -0
  33. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-400/trainer_state.json +834 -0
  34. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-400/vocab.json +0 -0
  35. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-600/added_tokens.json +35 -0
  36. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-600/chat_template.jinja +54 -0
  37. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-600/config.json +58 -0
  38. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-600/merges.txt +0 -0
  39. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-600/pytorch_model.bin +3 -0
  40. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-600/special_tokens_map.json +31 -0
  41. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-600/thought_projector.pt +3 -0
  42. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-600/tokenizer.json +3 -0
  43. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-600/tokenizer_config.json +295 -0
  44. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-600/trainer_state.json +1234 -0
  45. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-600/vocab.json +0 -0
  46. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-800/added_tokens.json +35 -0
  47. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-800/chat_template.jinja +54 -0
  48. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-800/config.json +58 -0
  49. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-800/merges.txt +0 -0
  50. checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-800/pytorch_model.bin +3 -0
.gitattributes CHANGED
@@ -49,3 +49,10 @@ checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls1/checkpoint-1500/tokenizer.json filter=lf
49
  checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls1/checkpoint-2000/tokenizer.json filter=lfs diff=lfs merge=lfs -text
50
  checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls1/checkpoint-500/tokenizer.json filter=lfs diff=lfs merge=lfs -text
51
  checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls1/wandb/offline-run-20260619_215302-q1wgok6t/run-q1wgok6t.wandb filter=lfs diff=lfs merge=lfs -text
 
 
 
 
 
 
 
 
49
  checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls1/checkpoint-2000/tokenizer.json filter=lfs diff=lfs merge=lfs -text
50
  checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls1/checkpoint-500/tokenizer.json filter=lfs diff=lfs merge=lfs -text
51
  checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls1/wandb/offline-run-20260619_215302-q1wgok6t/run-q1wgok6t.wandb filter=lfs diff=lfs merge=lfs -text
52
+ checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-1000/tokenizer.json filter=lfs diff=lfs merge=lfs -text
53
+ checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-200/tokenizer.json filter=lfs diff=lfs merge=lfs -text
54
+ checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-400/tokenizer.json filter=lfs diff=lfs merge=lfs -text
55
+ checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-600/tokenizer.json filter=lfs diff=lfs merge=lfs -text
56
+ checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-800/tokenizer.json filter=lfs diff=lfs merge=lfs -text
57
+ checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/wandb/offline-run-20260620_130138-bqeecd43/run-bqeecd43.wandb filter=lfs diff=lfs merge=lfs -text
58
+ checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/wandb/offline-run-20260621_033730-ue0j104y/run-ue0j104y.wandb filter=lfs diff=lfs merge=lfs -text
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-1000/added_tokens.json ADDED
@@ -0,0 +1,35 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "</tool_call>": 151658,
3
+ "<tool_call>": 151657,
4
+ "<|action_sep|>": 151670,
5
+ "<|arg_sep|>": 151671,
6
+ "<|box_end|>": 151649,
7
+ "<|box_start|>": 151648,
8
+ "<|call_sep|>": 151666,
9
+ "<|end_of_text|>": 151673,
10
+ "<|endoftext|>": 151643,
11
+ "<|exception_sep|>": 151669,
12
+ "<|file_sep|>": 151664,
13
+ "<|fim_middle|>": 151660,
14
+ "<|fim_pad|>": 151662,
15
+ "<|fim_prefix|>": 151659,
16
+ "<|fim_suffix|>": 151661,
17
+ "<|frame_sep|>": 151672,
18
+ "<|im_end|>": 151645,
19
+ "<|im_start|>": 151644,
20
+ "<|image_pad|>": 151655,
21
+ "<|latent_end|>": 151675,
22
+ "<|latent_start|>": 151674,
23
+ "<|line_sep|>": 151667,
24
+ "<|object_ref_end|>": 151647,
25
+ "<|object_ref_start|>": 151646,
26
+ "<|quad_end|>": 151651,
27
+ "<|quad_start|>": 151650,
28
+ "<|repo_name|>": 151663,
29
+ "<|return_sep|>": 151668,
30
+ "<|trace_context_start|>": 151665,
31
+ "<|video_pad|>": 151656,
32
+ "<|vision_end|>": 151653,
33
+ "<|vision_pad|>": 151654,
34
+ "<|vision_start|>": 151652
35
+ }
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-1000/chat_template.jinja ADDED
@@ -0,0 +1,54 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {%- if tools %}
2
+ {{- '<|im_start|>system\n' }}
3
+ {%- if messages[0]['role'] == 'system' %}
4
+ {{- messages[0]['content'] }}
5
+ {%- else %}
6
+ {{- 'You are Qwen, created by Alibaba Cloud. You are a helpful assistant.' }}
7
+ {%- endif %}
8
+ {{- "\n\n# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
9
+ {%- for tool in tools %}
10
+ {{- "\n" }}
11
+ {{- tool | tojson }}
12
+ {%- endfor %}
13
+ {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
14
+ {%- else %}
15
+ {%- if messages[0]['role'] == 'system' %}
16
+ {{- '<|im_start|>system\n' + messages[0]['content'] + '<|im_end|>\n' }}
17
+ {%- else %}
18
+ {{- '<|im_start|>system\nYou are Qwen, created by Alibaba Cloud. You are a helpful assistant.<|im_end|>\n' }}
19
+ {%- endif %}
20
+ {%- endif %}
21
+ {%- for message in messages %}
22
+ {%- if (message.role == "user") or (message.role == "system" and not loop.first) or (message.role == "assistant" and not message.tool_calls) %}
23
+ {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
24
+ {%- elif message.role == "assistant" %}
25
+ {{- '<|im_start|>' + message.role }}
26
+ {%- if message.content %}
27
+ {{- '\n' + message.content }}
28
+ {%- endif %}
29
+ {%- for tool_call in message.tool_calls %}
30
+ {%- if tool_call.function is defined %}
31
+ {%- set tool_call = tool_call.function %}
32
+ {%- endif %}
33
+ {{- '\n<tool_call>\n{"name": "' }}
34
+ {{- tool_call.name }}
35
+ {{- '", "arguments": ' }}
36
+ {{- tool_call.arguments | tojson }}
37
+ {{- '}\n</tool_call>' }}
38
+ {%- endfor %}
39
+ {{- '<|im_end|>\n' }}
40
+ {%- elif message.role == "tool" %}
41
+ {%- if (loop.index0 == 0) or (messages[loop.index0 - 1].role != "tool") %}
42
+ {{- '<|im_start|>user' }}
43
+ {%- endif %}
44
+ {{- '\n<tool_response>\n' }}
45
+ {{- message.content }}
46
+ {{- '\n</tool_response>' }}
47
+ {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
48
+ {{- '<|im_end|>\n' }}
49
+ {%- endif %}
50
+ {%- endif %}
51
+ {%- endfor %}
52
+ {%- if add_generation_prompt %}
53
+ {{- '<|im_start|>assistant\n' }}
54
+ {%- endif %}
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-1000/config.json ADDED
@@ -0,0 +1,58 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Qwen2ForCausalLM"
4
+ ],
5
+ "attention_dropout": 0.0,
6
+ "dtype": "bfloat16",
7
+ "eos_token_id": 151643,
8
+ "hidden_act": "silu",
9
+ "hidden_size": 1536,
10
+ "initializer_range": 0.02,
11
+ "intermediate_size": 8960,
12
+ "layer_types": [
13
+ "full_attention",
14
+ "full_attention",
15
+ "full_attention",
16
+ "full_attention",
17
+ "full_attention",
18
+ "full_attention",
19
+ "full_attention",
20
+ "full_attention",
21
+ "full_attention",
22
+ "full_attention",
23
+ "full_attention",
24
+ "full_attention",
25
+ "full_attention",
26
+ "full_attention",
27
+ "full_attention",
28
+ "full_attention",
29
+ "full_attention",
30
+ "full_attention",
31
+ "full_attention",
32
+ "full_attention",
33
+ "full_attention",
34
+ "full_attention",
35
+ "full_attention",
36
+ "full_attention",
37
+ "full_attention",
38
+ "full_attention",
39
+ "full_attention",
40
+ "full_attention"
41
+ ],
42
+ "max_position_embeddings": 32768,
43
+ "max_window_layers": 28,
44
+ "model_type": "qwen2",
45
+ "num_attention_heads": 12,
46
+ "num_hidden_layers": 28,
47
+ "num_key_value_heads": 2,
48
+ "pad_token_id": 151643,
49
+ "rms_norm_eps": 1e-06,
50
+ "rope_scaling": null,
51
+ "rope_theta": 1000000.0,
52
+ "sliding_window": null,
53
+ "tie_word_embeddings": true,
54
+ "transformers_version": "4.57.6",
55
+ "use_cache": true,
56
+ "use_sliding_window": false,
57
+ "vocab_size": 151676
58
+ }
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-1000/merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-1000/pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6e0e3318e37fa1a259755f87da55e3348b8afa9883834de39a4cb2422e95c9be
3
+ size 3096275515
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-1000/special_tokens_map.json ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "additional_special_tokens": [
3
+ "<|im_start|>",
4
+ "<|im_end|>",
5
+ "<|object_ref_start|>",
6
+ "<|object_ref_end|>",
7
+ "<|box_start|>",
8
+ "<|box_end|>",
9
+ "<|quad_start|>",
10
+ "<|quad_end|>",
11
+ "<|vision_start|>",
12
+ "<|vision_end|>",
13
+ "<|vision_pad|>",
14
+ "<|image_pad|>",
15
+ "<|video_pad|>"
16
+ ],
17
+ "eos_token": {
18
+ "content": "<|endoftext|>",
19
+ "lstrip": false,
20
+ "normalized": false,
21
+ "rstrip": false,
22
+ "single_word": false
23
+ },
24
+ "pad_token": {
25
+ "content": "<|endoftext|>",
26
+ "lstrip": false,
27
+ "normalized": false,
28
+ "rstrip": false,
29
+ "single_word": false
30
+ }
31
+ }
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-1000/thought_projector.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9d29253b502e8e69a15fcfac33948fabd531aff3b6a6398fe133b3ae2c1fb7b8
3
+ size 9445953
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-1000/tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:83a790d654474f5dfe225f889afd0210313eb1083f942671f2c4b8e95a1c922b
3
+ size 11424004
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-1000/tokenizer_config.json ADDED
@@ -0,0 +1,295 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_bos_token": false,
3
+ "add_prefix_space": false,
4
+ "added_tokens_decoder": {
5
+ "151643": {
6
+ "content": "<|endoftext|>",
7
+ "lstrip": false,
8
+ "normalized": false,
9
+ "rstrip": false,
10
+ "single_word": false,
11
+ "special": true
12
+ },
13
+ "151644": {
14
+ "content": "<|im_start|>",
15
+ "lstrip": false,
16
+ "normalized": false,
17
+ "rstrip": false,
18
+ "single_word": false,
19
+ "special": true
20
+ },
21
+ "151645": {
22
+ "content": "<|im_end|>",
23
+ "lstrip": false,
24
+ "normalized": false,
25
+ "rstrip": false,
26
+ "single_word": false,
27
+ "special": true
28
+ },
29
+ "151646": {
30
+ "content": "<|object_ref_start|>",
31
+ "lstrip": false,
32
+ "normalized": false,
33
+ "rstrip": false,
34
+ "single_word": false,
35
+ "special": true
36
+ },
37
+ "151647": {
38
+ "content": "<|object_ref_end|>",
39
+ "lstrip": false,
40
+ "normalized": false,
41
+ "rstrip": false,
42
+ "single_word": false,
43
+ "special": true
44
+ },
45
+ "151648": {
46
+ "content": "<|box_start|>",
47
+ "lstrip": false,
48
+ "normalized": false,
49
+ "rstrip": false,
50
+ "single_word": false,
51
+ "special": true
52
+ },
53
+ "151649": {
54
+ "content": "<|box_end|>",
55
+ "lstrip": false,
56
+ "normalized": false,
57
+ "rstrip": false,
58
+ "single_word": false,
59
+ "special": true
60
+ },
61
+ "151650": {
62
+ "content": "<|quad_start|>",
63
+ "lstrip": false,
64
+ "normalized": false,
65
+ "rstrip": false,
66
+ "single_word": false,
67
+ "special": true
68
+ },
69
+ "151651": {
70
+ "content": "<|quad_end|>",
71
+ "lstrip": false,
72
+ "normalized": false,
73
+ "rstrip": false,
74
+ "single_word": false,
75
+ "special": true
76
+ },
77
+ "151652": {
78
+ "content": "<|vision_start|>",
79
+ "lstrip": false,
80
+ "normalized": false,
81
+ "rstrip": false,
82
+ "single_word": false,
83
+ "special": true
84
+ },
85
+ "151653": {
86
+ "content": "<|vision_end|>",
87
+ "lstrip": false,
88
+ "normalized": false,
89
+ "rstrip": false,
90
+ "single_word": false,
91
+ "special": true
92
+ },
93
+ "151654": {
94
+ "content": "<|vision_pad|>",
95
+ "lstrip": false,
96
+ "normalized": false,
97
+ "rstrip": false,
98
+ "single_word": false,
99
+ "special": true
100
+ },
101
+ "151655": {
102
+ "content": "<|image_pad|>",
103
+ "lstrip": false,
104
+ "normalized": false,
105
+ "rstrip": false,
106
+ "single_word": false,
107
+ "special": true
108
+ },
109
+ "151656": {
110
+ "content": "<|video_pad|>",
111
+ "lstrip": false,
112
+ "normalized": false,
113
+ "rstrip": false,
114
+ "single_word": false,
115
+ "special": true
116
+ },
117
+ "151657": {
118
+ "content": "<tool_call>",
119
+ "lstrip": false,
120
+ "normalized": false,
121
+ "rstrip": false,
122
+ "single_word": false,
123
+ "special": false
124
+ },
125
+ "151658": {
126
+ "content": "</tool_call>",
127
+ "lstrip": false,
128
+ "normalized": false,
129
+ "rstrip": false,
130
+ "single_word": false,
131
+ "special": false
132
+ },
133
+ "151659": {
134
+ "content": "<|fim_prefix|>",
135
+ "lstrip": false,
136
+ "normalized": false,
137
+ "rstrip": false,
138
+ "single_word": false,
139
+ "special": false
140
+ },
141
+ "151660": {
142
+ "content": "<|fim_middle|>",
143
+ "lstrip": false,
144
+ "normalized": false,
145
+ "rstrip": false,
146
+ "single_word": false,
147
+ "special": false
148
+ },
149
+ "151661": {
150
+ "content": "<|fim_suffix|>",
151
+ "lstrip": false,
152
+ "normalized": false,
153
+ "rstrip": false,
154
+ "single_word": false,
155
+ "special": false
156
+ },
157
+ "151662": {
158
+ "content": "<|fim_pad|>",
159
+ "lstrip": false,
160
+ "normalized": false,
161
+ "rstrip": false,
162
+ "single_word": false,
163
+ "special": false
164
+ },
165
+ "151663": {
166
+ "content": "<|repo_name|>",
167
+ "lstrip": false,
168
+ "normalized": false,
169
+ "rstrip": false,
170
+ "single_word": false,
171
+ "special": false
172
+ },
173
+ "151664": {
174
+ "content": "<|file_sep|>",
175
+ "lstrip": false,
176
+ "normalized": false,
177
+ "rstrip": false,
178
+ "single_word": false,
179
+ "special": false
180
+ },
181
+ "151665": {
182
+ "content": "<|trace_context_start|>",
183
+ "lstrip": false,
184
+ "normalized": false,
185
+ "rstrip": false,
186
+ "single_word": false,
187
+ "special": true
188
+ },
189
+ "151666": {
190
+ "content": "<|call_sep|>",
191
+ "lstrip": false,
192
+ "normalized": false,
193
+ "rstrip": false,
194
+ "single_word": false,
195
+ "special": true
196
+ },
197
+ "151667": {
198
+ "content": "<|line_sep|>",
199
+ "lstrip": false,
200
+ "normalized": false,
201
+ "rstrip": false,
202
+ "single_word": false,
203
+ "special": true
204
+ },
205
+ "151668": {
206
+ "content": "<|return_sep|>",
207
+ "lstrip": false,
208
+ "normalized": false,
209
+ "rstrip": false,
210
+ "single_word": false,
211
+ "special": true
212
+ },
213
+ "151669": {
214
+ "content": "<|exception_sep|>",
215
+ "lstrip": false,
216
+ "normalized": false,
217
+ "rstrip": false,
218
+ "single_word": false,
219
+ "special": true
220
+ },
221
+ "151670": {
222
+ "content": "<|action_sep|>",
223
+ "lstrip": false,
224
+ "normalized": false,
225
+ "rstrip": false,
226
+ "single_word": false,
227
+ "special": true
228
+ },
229
+ "151671": {
230
+ "content": "<|arg_sep|>",
231
+ "lstrip": false,
232
+ "normalized": false,
233
+ "rstrip": false,
234
+ "single_word": false,
235
+ "special": true
236
+ },
237
+ "151672": {
238
+ "content": "<|frame_sep|>",
239
+ "lstrip": false,
240
+ "normalized": false,
241
+ "rstrip": false,
242
+ "single_word": false,
243
+ "special": true
244
+ },
245
+ "151673": {
246
+ "content": "<|end_of_text|>",
247
+ "lstrip": false,
248
+ "normalized": false,
249
+ "rstrip": false,
250
+ "single_word": false,
251
+ "special": true
252
+ },
253
+ "151674": {
254
+ "content": "<|latent_start|>",
255
+ "lstrip": false,
256
+ "normalized": false,
257
+ "rstrip": false,
258
+ "single_word": false,
259
+ "special": true
260
+ },
261
+ "151675": {
262
+ "content": "<|latent_end|>",
263
+ "lstrip": false,
264
+ "normalized": false,
265
+ "rstrip": false,
266
+ "single_word": false,
267
+ "special": true
268
+ }
269
+ },
270
+ "additional_special_tokens": [
271
+ "<|im_start|>",
272
+ "<|im_end|>",
273
+ "<|object_ref_start|>",
274
+ "<|object_ref_end|>",
275
+ "<|box_start|>",
276
+ "<|box_end|>",
277
+ "<|quad_start|>",
278
+ "<|quad_end|>",
279
+ "<|vision_start|>",
280
+ "<|vision_end|>",
281
+ "<|vision_pad|>",
282
+ "<|image_pad|>",
283
+ "<|video_pad|>"
284
+ ],
285
+ "bos_token": null,
286
+ "clean_up_tokenization_spaces": false,
287
+ "eos_token": "<|endoftext|>",
288
+ "errors": "replace",
289
+ "extra_special_tokens": {},
290
+ "model_max_length": 32768,
291
+ "pad_token": "<|endoftext|>",
292
+ "split_special_tokens": false,
293
+ "tokenizer_class": "Qwen2Tokenizer",
294
+ "unk_token": null
295
+ }
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-1000/trainer_state.json ADDED
@@ -0,0 +1,2046 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 0.28835063437139563,
6
+ "eval_steps": 500,
7
+ "global_step": 1000,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.0014417531718569781,
14
+ "grad_norm": 52992.0,
15
+ "kd_loss": 0.3193017542362213,
16
+ "learning_rate": 1.3333333333333334e-06,
17
+ "loss": 0.473,
18
+ "step": 5,
19
+ "student_loss": 0.15727384388446808,
20
+ "teacher_loss": 0.0004137983196415007
21
+ },
22
+ {
23
+ "epoch": 0.0028835063437139563,
24
+ "grad_norm": 16121856.0,
25
+ "kd_loss": 0.3087984025478363,
26
+ "learning_rate": 3e-06,
27
+ "loss": 0.466,
28
+ "step": 10,
29
+ "student_loss": 0.2608874440193176,
30
+ "teacher_loss": 0.0007329508662223816
31
+ },
32
+ {
33
+ "epoch": 0.004325259515570935,
34
+ "grad_norm": 2240.0,
35
+ "kd_loss": 0.25935155153274536,
36
+ "learning_rate": 4.666666666666667e-06,
37
+ "loss": 0.4585,
38
+ "step": 15,
39
+ "student_loss": 0.1052168533205986,
40
+ "teacher_loss": 0.000609480426646769
41
+ },
42
+ {
43
+ "epoch": 0.0057670126874279125,
44
+ "grad_norm": 4194304.0,
45
+ "kd_loss": 0.25011566281318665,
46
+ "learning_rate": 6.333333333333333e-06,
47
+ "loss": 0.4707,
48
+ "step": 20,
49
+ "student_loss": 0.15271778404712677,
50
+ "teacher_loss": 0.0002972046786453575
51
+ },
52
+ {
53
+ "epoch": 0.00720876585928489,
54
+ "grad_norm": 41680896.0,
55
+ "kd_loss": 0.2633170485496521,
56
+ "learning_rate": 8.000000000000001e-06,
57
+ "loss": 0.4368,
58
+ "step": 25,
59
+ "student_loss": 0.09109362959861755,
60
+ "teacher_loss": 0.0009600125486031175
61
+ },
62
+ {
63
+ "epoch": 0.00865051903114187,
64
+ "grad_norm": 512.0,
65
+ "kd_loss": 0.2533612847328186,
66
+ "learning_rate": 9.666666666666667e-06,
67
+ "loss": 0.4187,
68
+ "step": 30,
69
+ "student_loss": 0.22675999999046326,
70
+ "teacher_loss": 0.00035076512722298503
71
+ },
72
+ {
73
+ "epoch": 0.010092272202998846,
74
+ "grad_norm": 48.75,
75
+ "kd_loss": 0.2179471105337143,
76
+ "learning_rate": 9.99958042442916e-06,
77
+ "loss": 0.3713,
78
+ "step": 35,
79
+ "student_loss": 0.09486500918865204,
80
+ "teacher_loss": 0.00026503618573769927
81
+ },
82
+ {
83
+ "epoch": 0.011534025374855825,
84
+ "grad_norm": 10.0,
85
+ "kd_loss": 0.17874062061309814,
86
+ "learning_rate": 9.997876019358083e-06,
87
+ "loss": 0.2647,
88
+ "step": 40,
89
+ "student_loss": 0.07879351079463959,
90
+ "teacher_loss": 0.05543303117156029
91
+ },
92
+ {
93
+ "epoch": 0.012975778546712802,
94
+ "grad_norm": 4.9375,
95
+ "kd_loss": 0.19365739822387695,
96
+ "learning_rate": 9.99486100792044e-06,
97
+ "loss": 0.1686,
98
+ "step": 45,
99
+ "student_loss": 0.07774724811315536,
100
+ "teacher_loss": 0.0008653572876937687
101
+ },
102
+ {
103
+ "epoch": 0.01441753171856978,
104
+ "grad_norm": 3.359375,
105
+ "kd_loss": 0.06108914315700531,
106
+ "learning_rate": 9.990536180750724e-06,
107
+ "loss": 0.1396,
108
+ "step": 50,
109
+ "student_loss": 0.008677379228174686,
110
+ "teacher_loss": 0.005272893700748682
111
+ },
112
+ {
113
+ "epoch": 0.015859284890426758,
114
+ "grad_norm": 3.71875,
115
+ "kd_loss": 0.04330182075500488,
116
+ "learning_rate": 9.984902671959911e-06,
117
+ "loss": 0.1065,
118
+ "step": 55,
119
+ "student_loss": 0.001041711075231433,
120
+ "teacher_loss": 0.0005072589847259223
121
+ },
122
+ {
123
+ "epoch": 0.01730103806228374,
124
+ "grad_norm": 4.125,
125
+ "kd_loss": 0.03823668882250786,
126
+ "learning_rate": 9.97796195883804e-06,
127
+ "loss": 0.0862,
128
+ "step": 60,
129
+ "student_loss": 0.0209807138890028,
130
+ "teacher_loss": 0.0038409351836889982
131
+ },
132
+ {
133
+ "epoch": 0.018742791234140715,
134
+ "grad_norm": 10.1875,
135
+ "kd_loss": 0.03922531381249428,
136
+ "learning_rate": 9.969715861466839e-06,
137
+ "loss": 0.0844,
138
+ "step": 65,
139
+ "student_loss": 0.04678678512573242,
140
+ "teacher_loss": 0.05432586744427681
141
+ },
142
+ {
143
+ "epoch": 0.020184544405997693,
144
+ "grad_norm": 2.03125,
145
+ "kd_loss": 0.0549621656537056,
146
+ "learning_rate": 9.96016654224243e-06,
147
+ "loss": 0.0797,
148
+ "step": 70,
149
+ "student_loss": 0.016930116340517998,
150
+ "teacher_loss": 0.0029678302817046642
151
+ },
152
+ {
153
+ "epoch": 0.02162629757785467,
154
+ "grad_norm": 2.5625,
155
+ "kd_loss": 0.023080041632056236,
156
+ "learning_rate": 9.94931650530827e-06,
157
+ "loss": 0.0756,
158
+ "step": 75,
159
+ "student_loss": 0.002210108330473304,
160
+ "teacher_loss": 0.0013834249693900347
161
+ },
162
+ {
163
+ "epoch": 0.02306805074971165,
164
+ "grad_norm": 2.296875,
165
+ "kd_loss": 0.025744518265128136,
166
+ "learning_rate": 9.93716859589851e-06,
167
+ "loss": 0.0644,
168
+ "step": 80,
169
+ "student_loss": 0.0015430074417963624,
170
+ "teacher_loss": 0.000761081581003964
171
+ },
172
+ {
173
+ "epoch": 0.024509803921568627,
174
+ "grad_norm": 3.125,
175
+ "kd_loss": 0.023152953013777733,
176
+ "learning_rate": 9.923725999591846e-06,
177
+ "loss": 0.0609,
178
+ "step": 85,
179
+ "student_loss": 0.0187687948346138,
180
+ "teacher_loss": 0.011518553830683231
181
+ },
182
+ {
183
+ "epoch": 0.025951557093425604,
184
+ "grad_norm": 1.8359375,
185
+ "kd_loss": 0.06841632723808289,
186
+ "learning_rate": 9.908992241476189e-06,
187
+ "loss": 0.0622,
188
+ "step": 90,
189
+ "student_loss": 0.05537872016429901,
190
+ "teacher_loss": 0.011890009045600891
191
+ },
192
+ {
193
+ "epoch": 0.027393310265282585,
194
+ "grad_norm": 4.21875,
195
+ "kd_loss": 0.041227906942367554,
196
+ "learning_rate": 9.892971185224244e-06,
197
+ "loss": 0.0626,
198
+ "step": 95,
199
+ "student_loss": 0.009465624578297138,
200
+ "teacher_loss": 0.005057212430983782
201
+ },
202
+ {
203
+ "epoch": 0.02883506343713956,
204
+ "grad_norm": 1.8359375,
205
+ "kd_loss": 0.021006152033805847,
206
+ "learning_rate": 9.875667032080354e-06,
207
+ "loss": 0.0538,
208
+ "step": 100,
209
+ "student_loss": 0.0007415671134367585,
210
+ "teacher_loss": 0.0005348982522264123
211
+ },
212
+ {
213
+ "epoch": 0.03027681660899654,
214
+ "grad_norm": 1.6328125,
215
+ "kd_loss": 0.035115621984004974,
216
+ "learning_rate": 9.857084319758772e-06,
217
+ "loss": 0.0541,
218
+ "step": 105,
219
+ "student_loss": 0.005187581293284893,
220
+ "teacher_loss": 0.00030648778192698956
221
+ },
222
+ {
223
+ "epoch": 0.031718569780853516,
224
+ "grad_norm": 2.953125,
225
+ "kd_loss": 0.03037514165043831,
226
+ "learning_rate": 9.837227921253747e-06,
227
+ "loss": 0.06,
228
+ "step": 110,
229
+ "student_loss": 0.031953420490026474,
230
+ "teacher_loss": 0.0175339188426733
231
+ },
232
+ {
233
+ "epoch": 0.03316032295271049,
234
+ "grad_norm": 1.484375,
235
+ "kd_loss": 0.023421702906489372,
236
+ "learning_rate": 9.816103043561648e-06,
237
+ "loss": 0.0563,
238
+ "step": 115,
239
+ "student_loss": 0.0006847245967946947,
240
+ "teacher_loss": 0.0003864571626763791
241
+ },
242
+ {
243
+ "epoch": 0.03460207612456748,
244
+ "grad_norm": 1.8203125,
245
+ "kd_loss": 0.018447211012244225,
246
+ "learning_rate": 9.79371522631553e-06,
247
+ "loss": 0.0478,
248
+ "step": 120,
249
+ "student_loss": 0.0015498347347602248,
250
+ "teacher_loss": 0.0010675351368263364
251
+ },
252
+ {
253
+ "epoch": 0.036043829296424454,
254
+ "grad_norm": 1.96875,
255
+ "kd_loss": 0.05630031228065491,
256
+ "learning_rate": 9.770070340332457e-06,
257
+ "loss": 0.0527,
258
+ "step": 125,
259
+ "student_loss": 0.04758762940764427,
260
+ "teacher_loss": 0.0004967509885318577
261
+ },
262
+ {
263
+ "epoch": 0.03748558246828143,
264
+ "grad_norm": 1.859375,
265
+ "kd_loss": 0.018152693286538124,
266
+ "learning_rate": 9.745174586073982e-06,
267
+ "loss": 0.0518,
268
+ "step": 130,
269
+ "student_loss": 0.0008988101035356522,
270
+ "teacher_loss": 0.0007278143893927336
271
+ },
272
+ {
273
+ "epoch": 0.03892733564013841,
274
+ "grad_norm": 1.6640625,
275
+ "kd_loss": 0.030904680490493774,
276
+ "learning_rate": 9.719034492020183e-06,
277
+ "loss": 0.0526,
278
+ "step": 135,
279
+ "student_loss": 0.0127235297113657,
280
+ "teacher_loss": 0.0011665972415357828
281
+ },
282
+ {
283
+ "epoch": 0.040369088811995385,
284
+ "grad_norm": 1.1640625,
285
+ "kd_loss": 0.01759810373187065,
286
+ "learning_rate": 9.691656912957686e-06,
287
+ "loss": 0.0494,
288
+ "step": 140,
289
+ "student_loss": 0.0008300879271700978,
290
+ "teacher_loss": 0.0005129938363097608
291
+ },
292
+ {
293
+ "epoch": 0.04181084198385236,
294
+ "grad_norm": 1.0234375,
295
+ "kd_loss": 0.061179179698228836,
296
+ "learning_rate": 9.663049028182112e-06,
297
+ "loss": 0.0457,
298
+ "step": 145,
299
+ "student_loss": 0.04404008761048317,
300
+ "teacher_loss": 0.009324059821665287
301
+ },
302
+ {
303
+ "epoch": 0.04325259515570934,
304
+ "grad_norm": 1.15625,
305
+ "kd_loss": 0.01291500125080347,
306
+ "learning_rate": 9.633218339615433e-06,
307
+ "loss": 0.0455,
308
+ "step": 150,
309
+ "student_loss": 0.0008404091349802911,
310
+ "teacher_loss": 0.0008450732566416264
311
+ },
312
+ {
313
+ "epoch": 0.04469434832756632,
314
+ "grad_norm": 1.4921875,
315
+ "kd_loss": 0.017479021102190018,
316
+ "learning_rate": 9.602172669838721e-06,
317
+ "loss": 0.0575,
318
+ "step": 155,
319
+ "student_loss": 0.0009151780977845192,
320
+ "teacher_loss": 0.0006974826683290303
321
+ },
322
+ {
323
+ "epoch": 0.0461361014994233,
324
+ "grad_norm": 1.671875,
325
+ "kd_loss": 0.022708255797624588,
326
+ "learning_rate": 9.569920160040815e-06,
327
+ "loss": 0.0503,
328
+ "step": 160,
329
+ "student_loss": 0.07660368084907532,
330
+ "teacher_loss": 0.04039459675550461
331
+ },
332
+ {
333
+ "epoch": 0.04757785467128028,
334
+ "grad_norm": 1.0703125,
335
+ "kd_loss": 0.027437202632427216,
336
+ "learning_rate": 9.536469267883432e-06,
337
+ "loss": 0.0451,
338
+ "step": 165,
339
+ "student_loss": 0.0005943832220509648,
340
+ "teacher_loss": 0.005336699075996876
341
+ },
342
+ {
343
+ "epoch": 0.049019607843137254,
344
+ "grad_norm": 1.71875,
345
+ "kd_loss": 0.014546042308211327,
346
+ "learning_rate": 9.501828765283295e-06,
347
+ "loss": 0.0459,
348
+ "step": 170,
349
+ "student_loss": 0.0006958618760108948,
350
+ "teacher_loss": 0.0005218263831920922
351
+ },
352
+ {
353
+ "epoch": 0.05046136101499423,
354
+ "grad_norm": 1.4453125,
355
+ "kd_loss": 0.02377346158027649,
356
+ "learning_rate": 9.466007736111846e-06,
357
+ "loss": 0.0465,
358
+ "step": 175,
359
+ "student_loss": 0.003214897122234106,
360
+ "teacher_loss": 0.0006713470793329179
361
+ },
362
+ {
363
+ "epoch": 0.05190311418685121,
364
+ "grad_norm": 1.296875,
365
+ "kd_loss": 0.02517828531563282,
366
+ "learning_rate": 9.429015573813163e-06,
367
+ "loss": 0.0454,
368
+ "step": 180,
369
+ "student_loss": 0.0007039233460091054,
370
+ "teacher_loss": 0.0003641469811554998
371
+ },
372
+ {
373
+ "epoch": 0.05334486735870819,
374
+ "grad_norm": 1.25,
375
+ "kd_loss": 0.029188504442572594,
376
+ "learning_rate": 9.390861978940687e-06,
377
+ "loss": 0.0473,
378
+ "step": 185,
379
+ "student_loss": 0.04238048940896988,
380
+ "teacher_loss": 0.00033479384728707373
381
+ },
382
+ {
383
+ "epoch": 0.05478662053056517,
384
+ "grad_norm": 1.484375,
385
+ "kd_loss": 0.03749082237482071,
386
+ "learning_rate": 9.351556956613423e-06,
387
+ "loss": 0.0436,
388
+ "step": 190,
389
+ "student_loss": 0.010270334780216217,
390
+ "teacher_loss": 0.0024021982681006193
391
+ },
392
+ {
393
+ "epoch": 0.056228373702422146,
394
+ "grad_norm": 2.0,
395
+ "kd_loss": 0.030336912721395493,
396
+ "learning_rate": 9.31111081389227e-06,
397
+ "loss": 0.0442,
398
+ "step": 195,
399
+ "student_loss": 0.010534510016441345,
400
+ "teacher_loss": 0.0003890562220476568
401
+ },
402
+ {
403
+ "epoch": 0.05767012687427912,
404
+ "grad_norm": 0.921875,
405
+ "kd_loss": 0.026554003357887268,
406
+ "learning_rate": 9.269534157077177e-06,
407
+ "loss": 0.0497,
408
+ "step": 200,
409
+ "student_loss": 0.0034760867711156607,
410
+ "teacher_loss": 0.00022423896007239819
411
+ },
412
+ {
413
+ "epoch": 0.0591118800461361,
414
+ "grad_norm": 1.9140625,
415
+ "kd_loss": 0.023410219699144363,
416
+ "learning_rate": 9.226837888925813e-06,
417
+ "loss": 0.0471,
418
+ "step": 205,
419
+ "student_loss": 0.03611442819237709,
420
+ "teacher_loss": 0.027781842276453972
421
+ },
422
+ {
423
+ "epoch": 0.06055363321799308,
424
+ "grad_norm": 1.7578125,
425
+ "kd_loss": 0.022867349907755852,
426
+ "learning_rate": 9.183033205794525e-06,
427
+ "loss": 0.0428,
428
+ "step": 210,
429
+ "student_loss": 0.001971556106582284,
430
+ "teacher_loss": 0.0009746765717864037
431
+ },
432
+ {
433
+ "epoch": 0.061995386389850055,
434
+ "grad_norm": 1.0234375,
435
+ "kd_loss": 0.028235338628292084,
436
+ "learning_rate": 9.13813159470227e-06,
437
+ "loss": 0.0449,
438
+ "step": 215,
439
+ "student_loss": 0.0009558866149745882,
440
+ "teacher_loss": 0.0004049334966111928
441
+ },
442
+ {
443
+ "epoch": 0.06343713956170703,
444
+ "grad_norm": 1.4296875,
445
+ "kd_loss": 0.03134317323565483,
446
+ "learning_rate": 9.092144830318357e-06,
447
+ "loss": 0.0438,
448
+ "step": 220,
449
+ "student_loss": 0.048195730894804,
450
+ "teacher_loss": 0.009161758236587048
451
+ },
452
+ {
453
+ "epoch": 0.06487889273356401,
454
+ "grad_norm": 1.1328125,
455
+ "kd_loss": 0.014049015007913113,
456
+ "learning_rate": 9.045084971874738e-06,
457
+ "loss": 0.0487,
458
+ "step": 225,
459
+ "student_loss": 0.05373133346438408,
460
+ "teacher_loss": 0.0009734037448652089
461
+ },
462
+ {
463
+ "epoch": 0.06632064590542099,
464
+ "grad_norm": 1.7265625,
465
+ "kd_loss": 0.021539395675063133,
466
+ "learning_rate": 8.99696436000368e-06,
467
+ "loss": 0.0384,
468
+ "step": 230,
469
+ "student_loss": 0.0008717047749087214,
470
+ "teacher_loss": 0.0028650674503296614
471
+ },
472
+ {
473
+ "epoch": 0.06776239907727798,
474
+ "grad_norm": 1.0390625,
475
+ "kd_loss": 0.045927971601486206,
476
+ "learning_rate": 8.947795613501658e-06,
477
+ "loss": 0.0397,
478
+ "step": 235,
479
+ "student_loss": 0.012143217958509922,
480
+ "teacher_loss": 0.0006723064580000937
481
+ },
482
+ {
483
+ "epoch": 0.06920415224913495,
484
+ "grad_norm": 0.78125,
485
+ "kd_loss": 0.01626853086054325,
486
+ "learning_rate": 8.897591626020284e-06,
487
+ "loss": 0.0425,
488
+ "step": 240,
489
+ "student_loss": 0.0004826921212952584,
490
+ "teacher_loss": 0.0003007515915669501
491
+ },
492
+ {
493
+ "epoch": 0.07064590542099193,
494
+ "grad_norm": 1.078125,
495
+ "kd_loss": 0.010876834392547607,
496
+ "learning_rate": 8.846365562685178e-06,
497
+ "loss": 0.0379,
498
+ "step": 245,
499
+ "student_loss": 0.00123109039850533,
500
+ "teacher_loss": 0.0017208561766892672
501
+ },
502
+ {
503
+ "epoch": 0.07208765859284891,
504
+ "grad_norm": 1.6484375,
505
+ "kd_loss": 0.012208702974021435,
506
+ "learning_rate": 8.794130856643635e-06,
507
+ "loss": 0.0402,
508
+ "step": 250,
509
+ "student_loss": 0.0006701876409351826,
510
+ "teacher_loss": 0.00039402375114150345
511
+ },
512
+ {
513
+ "epoch": 0.07352941176470588,
514
+ "grad_norm": 3.859375,
515
+ "kd_loss": 0.010964477434754372,
516
+ "learning_rate": 8.74090120554202e-06,
517
+ "loss": 0.0428,
518
+ "step": 255,
519
+ "student_loss": 0.0004243800649419427,
520
+ "teacher_loss": 0.0003773768839892
521
+ },
522
+ {
523
+ "epoch": 0.07497116493656286,
524
+ "grad_norm": 2.015625,
525
+ "kd_loss": 0.018098091706633568,
526
+ "learning_rate": 8.686690567933803e-06,
527
+ "loss": 0.0436,
528
+ "step": 260,
529
+ "student_loss": 0.040437813848257065,
530
+ "teacher_loss": 0.032756391912698746
531
+ },
532
+ {
533
+ "epoch": 0.07641291810841984,
534
+ "grad_norm": 1.125,
535
+ "kd_loss": 0.011841603554785252,
536
+ "learning_rate": 8.63151315961915e-06,
537
+ "loss": 0.0412,
538
+ "step": 265,
539
+ "student_loss": 0.0006605261587537825,
540
+ "teacher_loss": 0.0004997120704501867
541
+ },
542
+ {
543
+ "epoch": 0.07785467128027682,
544
+ "grad_norm": 1.046875,
545
+ "kd_loss": 0.01046097930520773,
546
+ "learning_rate": 8.575383449917103e-06,
547
+ "loss": 0.044,
548
+ "step": 270,
549
+ "student_loss": 0.0009966753423213959,
550
+ "teacher_loss": 0.0006801315466873348
551
+ },
552
+ {
553
+ "epoch": 0.07929642445213379,
554
+ "grad_norm": 1.515625,
555
+ "kd_loss": 0.01713527925312519,
556
+ "learning_rate": 8.518316157871232e-06,
557
+ "loss": 0.0403,
558
+ "step": 275,
559
+ "student_loss": 0.04822665825486183,
560
+ "teacher_loss": 0.03450855612754822
561
+ },
562
+ {
563
+ "epoch": 0.08073817762399077,
564
+ "grad_norm": 1.3828125,
565
+ "kd_loss": 0.021968362852931023,
566
+ "learning_rate": 8.460326248389825e-06,
567
+ "loss": 0.0401,
568
+ "step": 280,
569
+ "student_loss": 0.0006340977270156145,
570
+ "teacher_loss": 0.0001483594678575173
571
+ },
572
+ {
573
+ "epoch": 0.08217993079584775,
574
+ "grad_norm": 1.3203125,
575
+ "kd_loss": 0.01131309662014246,
576
+ "learning_rate": 8.401428928321607e-06,
577
+ "loss": 0.0435,
578
+ "step": 285,
579
+ "student_loss": 0.0005278023309074342,
580
+ "teacher_loss": 0.0004103762039449066
581
+ },
582
+ {
583
+ "epoch": 0.08362168396770472,
584
+ "grad_norm": 1.1015625,
585
+ "kd_loss": 0.00958277191966772,
586
+ "learning_rate": 8.341639642468002e-06,
587
+ "loss": 0.0469,
588
+ "step": 290,
589
+ "student_loss": 0.005627856124192476,
590
+ "teacher_loss": 0.0034064538776874542
591
+ },
592
+ {
593
+ "epoch": 0.0850634371395617,
594
+ "grad_norm": 1.140625,
595
+ "kd_loss": 0.009735080413520336,
596
+ "learning_rate": 8.280974069532999e-06,
597
+ "loss": 0.0409,
598
+ "step": 295,
599
+ "student_loss": 0.000509344507008791,
600
+ "teacher_loss": 0.0004550877492874861
601
+ },
602
+ {
603
+ "epoch": 0.08650519031141868,
604
+ "grad_norm": 1.09375,
605
+ "kd_loss": 0.031678151339292526,
606
+ "learning_rate": 8.219448118011687e-06,
607
+ "loss": 0.0379,
608
+ "step": 300,
609
+ "student_loss": 0.016466177999973297,
610
+ "teacher_loss": 0.002865313785150647
611
+ },
612
+ {
613
+ "epoch": 0.08794694348327567,
614
+ "grad_norm": 1.1015625,
615
+ "kd_loss": 0.013346894644200802,
616
+ "learning_rate": 8.157077922018537e-06,
617
+ "loss": 0.0353,
618
+ "step": 305,
619
+ "student_loss": 0.009980602189898491,
620
+ "teacher_loss": 0.0014113991055637598
621
+ },
622
+ {
623
+ "epoch": 0.08938869665513265,
624
+ "grad_norm": 1.90625,
625
+ "kd_loss": 0.010920979082584381,
626
+ "learning_rate": 8.093879837056486e-06,
627
+ "loss": 0.0384,
628
+ "step": 310,
629
+ "student_loss": 0.00036570793599821627,
630
+ "teacher_loss": 0.0002939177502412349
631
+ },
632
+ {
633
+ "epoch": 0.09083044982698962,
634
+ "grad_norm": 1.0078125,
635
+ "kd_loss": 0.021986963227391243,
636
+ "learning_rate": 8.029870435728018e-06,
637
+ "loss": 0.0385,
638
+ "step": 315,
639
+ "student_loss": 0.01466396264731884,
640
+ "teacher_loss": 0.0008777049370110035
641
+ },
642
+ {
643
+ "epoch": 0.0922722029988466,
644
+ "grad_norm": 1.640625,
645
+ "kd_loss": 0.014081098139286041,
646
+ "learning_rate": 7.965066503389264e-06,
647
+ "loss": 0.0417,
648
+ "step": 320,
649
+ "student_loss": 0.0004964058171026409,
650
+ "teacher_loss": 0.0003025036712642759
651
+ },
652
+ {
653
+ "epoch": 0.09371395617070358,
654
+ "grad_norm": 1.2421875,
655
+ "kd_loss": 0.03542112559080124,
656
+ "learning_rate": 7.89948503374835e-06,
657
+ "loss": 0.0378,
658
+ "step": 325,
659
+ "student_loss": 0.01556178368628025,
660
+ "teacher_loss": 0.018082890659570694
661
+ },
662
+ {
663
+ "epoch": 0.09515570934256055,
664
+ "grad_norm": 1.40625,
665
+ "kd_loss": 0.04085119441151619,
666
+ "learning_rate": 7.833143224409076e-06,
667
+ "loss": 0.0394,
668
+ "step": 330,
669
+ "student_loss": 0.024770811200141907,
670
+ "teacher_loss": 0.0021320905070751905
671
+ },
672
+ {
673
+ "epoch": 0.09659746251441753,
674
+ "grad_norm": 1.0234375,
675
+ "kd_loss": 0.008477268740534782,
676
+ "learning_rate": 7.766058472361154e-06,
677
+ "loss": 0.0338,
678
+ "step": 335,
679
+ "student_loss": 0.0009084849734790623,
680
+ "teacher_loss": 0.0007285018800757825
681
+ },
682
+ {
683
+ "epoch": 0.09803921568627451,
684
+ "grad_norm": 1.28125,
685
+ "kd_loss": 0.014947460032999516,
686
+ "learning_rate": 7.698248369418146e-06,
687
+ "loss": 0.0388,
688
+ "step": 340,
689
+ "student_loss": 0.009875037707388401,
690
+ "teacher_loss": 0.00507500022649765
691
+ },
692
+ {
693
+ "epoch": 0.09948096885813149,
694
+ "grad_norm": 1.2421875,
695
+ "kd_loss": 0.019748909398913383,
696
+ "learning_rate": 7.629730697604314e-06,
697
+ "loss": 0.0437,
698
+ "step": 345,
699
+ "student_loss": 0.008698839694261551,
700
+ "teacher_loss": 0.00021470236242748797
701
+ },
702
+ {
703
+ "epoch": 0.10092272202998846,
704
+ "grad_norm": 1.453125,
705
+ "kd_loss": 0.038638580590486526,
706
+ "learning_rate": 7.560523424491595e-06,
707
+ "loss": 0.0333,
708
+ "step": 350,
709
+ "student_loss": 0.007182905916124582,
710
+ "teacher_loss": 0.0004611952172126621
711
+ },
712
+ {
713
+ "epoch": 0.10236447520184544,
714
+ "grad_norm": 1.4375,
715
+ "kd_loss": 0.008461177349090576,
716
+ "learning_rate": 7.490644698487909e-06,
717
+ "loss": 0.0405,
718
+ "step": 355,
719
+ "student_loss": 0.0004199718532618135,
720
+ "teacher_loss": 0.00038222281727939844
721
+ },
722
+ {
723
+ "epoch": 0.10380622837370242,
724
+ "grad_norm": 1.1796875,
725
+ "kd_loss": 0.008598582819104195,
726
+ "learning_rate": 7.420112844078066e-06,
727
+ "loss": 0.0396,
728
+ "step": 360,
729
+ "student_loss": 0.00045404830598272383,
730
+ "teacher_loss": 0.0003997418680228293
731
+ },
732
+ {
733
+ "epoch": 0.1052479815455594,
734
+ "grad_norm": 1.78125,
735
+ "kd_loss": 0.018130140379071236,
736
+ "learning_rate": 7.348946357018479e-06,
737
+ "loss": 0.0425,
738
+ "step": 365,
739
+ "student_loss": 0.06869062036275864,
740
+ "teacher_loss": 0.027944643050432205
741
+ },
742
+ {
743
+ "epoch": 0.10668973471741638,
744
+ "grad_norm": 1.03125,
745
+ "kd_loss": 0.0350988432765007,
746
+ "learning_rate": 7.277163899486975e-06,
747
+ "loss": 0.0447,
748
+ "step": 370,
749
+ "student_loss": 0.015085455030202866,
750
+ "teacher_loss": 0.0004648211761377752
751
+ },
752
+ {
753
+ "epoch": 0.10813148788927336,
754
+ "grad_norm": 1.0703125,
755
+ "kd_loss": 0.01613399013876915,
756
+ "learning_rate": 7.204784295188959e-06,
757
+ "loss": 0.0365,
758
+ "step": 375,
759
+ "student_loss": 0.03554360568523407,
760
+ "teacher_loss": 0.021264519542455673
761
+ },
762
+ {
763
+ "epoch": 0.10957324106113034,
764
+ "grad_norm": 1.296875,
765
+ "kd_loss": 0.010999550111591816,
766
+ "learning_rate": 7.1318265244212305e-06,
767
+ "loss": 0.0381,
768
+ "step": 380,
769
+ "student_loss": 0.0004269553755875677,
770
+ "teacher_loss": 0.0003071910759899765
771
+ },
772
+ {
773
+ "epoch": 0.11101499423298732,
774
+ "grad_norm": 0.875,
775
+ "kd_loss": 0.015028773806989193,
776
+ "learning_rate": 7.05830971909472e-06,
777
+ "loss": 0.0349,
778
+ "step": 385,
779
+ "student_loss": 0.0005910908803343773,
780
+ "teacher_loss": 0.0003781717678066343
781
+ },
782
+ {
783
+ "epoch": 0.11245674740484429,
784
+ "grad_norm": 1.25,
785
+ "kd_loss": 0.010998338460922241,
786
+ "learning_rate": 6.9842531577174865e-06,
787
+ "loss": 0.033,
788
+ "step": 390,
789
+ "student_loss": 0.00025347189512103796,
790
+ "teacher_loss": 0.00023755666916258633
791
+ },
792
+ {
793
+ "epoch": 0.11389850057670127,
794
+ "grad_norm": 1.390625,
795
+ "kd_loss": 0.010839108377695084,
796
+ "learning_rate": 6.9096762603392595e-06,
797
+ "loss": 0.0393,
798
+ "step": 395,
799
+ "student_loss": 0.00046008970821276307,
800
+ "teacher_loss": 0.00034296512603759766
801
+ },
802
+ {
803
+ "epoch": 0.11534025374855825,
804
+ "grad_norm": 2.40625,
805
+ "kd_loss": 0.015689171850681305,
806
+ "learning_rate": 6.834598583458862e-06,
807
+ "loss": 0.0333,
808
+ "step": 400,
809
+ "student_loss": 0.0004951858427375555,
810
+ "teacher_loss": 0.0002840181696228683
811
+ },
812
+ {
813
+ "epoch": 0.11678200692041522,
814
+ "grad_norm": 1.328125,
815
+ "kd_loss": 0.014689328148961067,
816
+ "learning_rate": 6.7590398148958625e-06,
817
+ "loss": 0.0425,
818
+ "step": 405,
819
+ "student_loss": 0.005328211933374405,
820
+ "teacher_loss": 0.0002102136058965698
821
+ },
822
+ {
823
+ "epoch": 0.1182237600922722,
824
+ "grad_norm": 1.4609375,
825
+ "kd_loss": 0.026721687987446785,
826
+ "learning_rate": 6.6830197686277945e-06,
827
+ "loss": 0.0417,
828
+ "step": 410,
829
+ "student_loss": 0.015544440597295761,
830
+ "teacher_loss": 0.00868156086653471
831
+ },
832
+ {
833
+ "epoch": 0.11966551326412918,
834
+ "grad_norm": 1.1640625,
835
+ "kd_loss": 0.0165933258831501,
836
+ "learning_rate": 6.6065583795942625e-06,
837
+ "loss": 0.0409,
838
+ "step": 415,
839
+ "student_loss": 0.0018743074033409357,
840
+ "teacher_loss": 0.002361958846449852
841
+ },
842
+ {
843
+ "epoch": 0.12110726643598616,
844
+ "grad_norm": 1.34375,
845
+ "kd_loss": 0.02428903616964817,
846
+ "learning_rate": 6.52967569846937e-06,
847
+ "loss": 0.0376,
848
+ "step": 420,
849
+ "student_loss": 0.05027158558368683,
850
+ "teacher_loss": 0.022369280457496643
851
+ },
852
+ {
853
+ "epoch": 0.12254901960784313,
854
+ "grad_norm": 1.171875,
855
+ "kd_loss": 0.00989749375730753,
856
+ "learning_rate": 6.452391886403767e-06,
857
+ "loss": 0.0366,
858
+ "step": 425,
859
+ "student_loss": 0.001526529318653047,
860
+ "teacher_loss": 0.001954666106030345
861
+ },
862
+ {
863
+ "epoch": 0.12399077277970011,
864
+ "grad_norm": 1.40625,
865
+ "kd_loss": 0.025572113692760468,
866
+ "learning_rate": 6.374727209737743e-06,
867
+ "loss": 0.0457,
868
+ "step": 430,
869
+ "student_loss": 0.05225735530257225,
870
+ "teacher_loss": 0.06261400878429413
871
+ },
872
+ {
873
+ "epoch": 0.1254325259515571,
874
+ "grad_norm": 1.1953125,
875
+ "kd_loss": 0.009243296459317207,
876
+ "learning_rate": 6.296702034686726e-06,
877
+ "loss": 0.0351,
878
+ "step": 435,
879
+ "student_loss": 0.0006208799313753843,
880
+ "teacher_loss": 0.0005131844663992524
881
+ },
882
+ {
883
+ "epoch": 0.12687427912341406,
884
+ "grad_norm": 1.140625,
885
+ "kd_loss": 0.025882720947265625,
886
+ "learning_rate": 6.218336822000598e-06,
887
+ "loss": 0.0366,
888
+ "step": 440,
889
+ "student_loss": 0.022235309705138206,
890
+ "teacher_loss": 0.005665617063641548
891
+ },
892
+ {
893
+ "epoch": 0.12831603229527105,
894
+ "grad_norm": 1.5546875,
895
+ "kd_loss": 0.020934544503688812,
896
+ "learning_rate": 6.139652121598219e-06,
897
+ "loss": 0.0381,
898
+ "step": 445,
899
+ "student_loss": 0.02492375858128071,
900
+ "teacher_loss": 0.01704920269548893
901
+ },
902
+ {
903
+ "epoch": 0.12975778546712802,
904
+ "grad_norm": 1.078125,
905
+ "kd_loss": 0.021503543481230736,
906
+ "learning_rate": 6.060668567178561e-06,
907
+ "loss": 0.039,
908
+ "step": 450,
909
+ "student_loss": 0.0008401675149798393,
910
+ "teacher_loss": 0.0007966886623762548
911
+ },
912
+ {
913
+ "epoch": 0.131199538638985,
914
+ "grad_norm": 0.99609375,
915
+ "kd_loss": 0.011403748765587807,
916
+ "learning_rate": 5.981406870809889e-06,
917
+ "loss": 0.0383,
918
+ "step": 455,
919
+ "student_loss": 0.0014924939023330808,
920
+ "teacher_loss": 0.000647007254883647
921
+ },
922
+ {
923
+ "epoch": 0.13264129181084197,
924
+ "grad_norm": 1.515625,
925
+ "kd_loss": 0.013661489821970463,
926
+ "learning_rate": 5.9018878174983674e-06,
927
+ "loss": 0.0329,
928
+ "step": 460,
929
+ "student_loss": 0.0015512986574321985,
930
+ "teacher_loss": 0.0012551237596198916
931
+ },
932
+ {
933
+ "epoch": 0.13408304498269896,
934
+ "grad_norm": 1.265625,
935
+ "kd_loss": 0.020031817257404327,
936
+ "learning_rate": 5.822132259737565e-06,
937
+ "loss": 0.0399,
938
+ "step": 465,
939
+ "student_loss": 0.0005617028218694031,
940
+ "teacher_loss": 0.0006807463942095637
941
+ },
942
+ {
943
+ "epoch": 0.13552479815455595,
944
+ "grad_norm": 1.171875,
945
+ "kd_loss": 0.012775675393640995,
946
+ "learning_rate": 5.742161112040237e-06,
947
+ "loss": 0.0377,
948
+ "step": 470,
949
+ "student_loss": 0.00029168292530812323,
950
+ "teacher_loss": 0.00020419349311850965
951
+ },
952
+ {
953
+ "epoch": 0.13696655132641292,
954
+ "grad_norm": 0.87109375,
955
+ "kd_loss": 0.016445137560367584,
956
+ "learning_rate": 5.661995345453867e-06,
957
+ "loss": 0.0315,
958
+ "step": 475,
959
+ "student_loss": 0.0005847346619702876,
960
+ "teacher_loss": 0.0003634036984294653
961
+ },
962
+ {
963
+ "epoch": 0.1384083044982699,
964
+ "grad_norm": 0.95703125,
965
+ "kd_loss": 0.014534696005284786,
966
+ "learning_rate": 5.581655982061367e-06,
967
+ "loss": 0.0466,
968
+ "step": 480,
969
+ "student_loss": 0.0203537754714489,
970
+ "teacher_loss": 0.014939186163246632
971
+ },
972
+ {
973
+ "epoch": 0.13985005767012687,
974
+ "grad_norm": 1.1640625,
975
+ "kd_loss": 0.008392157033085823,
976
+ "learning_rate": 5.501164089468406e-06,
977
+ "loss": 0.0357,
978
+ "step": 485,
979
+ "student_loss": 0.0004224316217005253,
980
+ "teacher_loss": 0.00044211489148437977
981
+ },
982
+ {
983
+ "epoch": 0.14129181084198386,
984
+ "grad_norm": 1.8984375,
985
+ "kd_loss": 0.014247741550207138,
986
+ "learning_rate": 5.4205407752787884e-06,
987
+ "loss": 0.0428,
988
+ "step": 490,
989
+ "student_loss": 0.019939009100198746,
990
+ "teacher_loss": 0.009562309831380844
991
+ },
992
+ {
993
+ "epoch": 0.14273356401384082,
994
+ "grad_norm": 1.828125,
995
+ "kd_loss": 0.026821549981832504,
996
+ "learning_rate": 5.339807181559359e-06,
997
+ "loss": 0.0484,
998
+ "step": 495,
999
+ "student_loss": 0.01085620652884245,
1000
+ "teacher_loss": 0.001887582358904183
1001
+ },
1002
+ {
1003
+ "epoch": 0.14417531718569782,
1004
+ "grad_norm": 1.03125,
1005
+ "kd_loss": 0.011890986934304237,
1006
+ "learning_rate": 5.258984479295853e-06,
1007
+ "loss": 0.0366,
1008
+ "step": 500,
1009
+ "student_loss": 0.0007122078677639365,
1010
+ "teacher_loss": 0.0006097567966207862
1011
+ },
1012
+ {
1013
+ "epoch": 0.14561707035755478,
1014
+ "grad_norm": 1.0,
1015
+ "kd_loss": 0.01012763474136591,
1016
+ "learning_rate": 5.1780938628411795e-06,
1017
+ "loss": 0.0433,
1018
+ "step": 505,
1019
+ "student_loss": 0.0004120942030567676,
1020
+ "teacher_loss": 0.00030555034754797816
1021
+ },
1022
+ {
1023
+ "epoch": 0.14705882352941177,
1024
+ "grad_norm": 1.90625,
1025
+ "kd_loss": 0.010170744732022285,
1026
+ "learning_rate": 5.097156544357567e-06,
1027
+ "loss": 0.0382,
1028
+ "step": 510,
1029
+ "student_loss": 0.0002881335385609418,
1030
+ "teacher_loss": 0.000244438968366012
1031
+ },
1032
+ {
1033
+ "epoch": 0.14850057670126873,
1034
+ "grad_norm": 1.359375,
1035
+ "kd_loss": 0.0138414166867733,
1036
+ "learning_rate": 5.016193748254045e-06,
1037
+ "loss": 0.031,
1038
+ "step": 515,
1039
+ "student_loss": 0.005790602881461382,
1040
+ "teacher_loss": 0.0001912783191073686
1041
+ },
1042
+ {
1043
+ "epoch": 0.14994232987312572,
1044
+ "grad_norm": 1.9921875,
1045
+ "kd_loss": 0.03539569675922394,
1046
+ "learning_rate": 4.935226705620699e-06,
1047
+ "loss": 0.0397,
1048
+ "step": 520,
1049
+ "student_loss": 0.04242030531167984,
1050
+ "teacher_loss": 0.011500299908220768
1051
+ },
1052
+ {
1053
+ "epoch": 0.1513840830449827,
1054
+ "grad_norm": 0.67578125,
1055
+ "kd_loss": 0.018779262900352478,
1056
+ "learning_rate": 4.8542766486612035e-06,
1057
+ "loss": 0.0403,
1058
+ "step": 525,
1059
+ "student_loss": 0.00041962284012697637,
1060
+ "teacher_loss": 0.00021298698266036808
1061
+ },
1062
+ {
1063
+ "epoch": 0.15282583621683968,
1064
+ "grad_norm": 1.2421875,
1065
+ "kd_loss": 0.01180377509444952,
1066
+ "learning_rate": 4.773364805125025e-06,
1067
+ "loss": 0.0351,
1068
+ "step": 530,
1069
+ "student_loss": 0.00039459849358536303,
1070
+ "teacher_loss": 0.0003135094593744725
1071
+ },
1072
+ {
1073
+ "epoch": 0.15426758938869667,
1074
+ "grad_norm": 0.96484375,
1075
+ "kd_loss": 0.022030461579561234,
1076
+ "learning_rate": 4.6925123927408265e-06,
1077
+ "loss": 0.0343,
1078
+ "step": 535,
1079
+ "student_loss": 0.000637000659480691,
1080
+ "teacher_loss": 0.00034665301791392267
1081
+ },
1082
+ {
1083
+ "epoch": 0.15570934256055363,
1084
+ "grad_norm": 1.234375,
1085
+ "kd_loss": 0.01367215160280466,
1086
+ "learning_rate": 4.611740613652485e-06,
1087
+ "loss": 0.0333,
1088
+ "step": 540,
1089
+ "student_loss": 0.0006559583125635982,
1090
+ "teacher_loss": 0.0003204621316399425
1091
+ },
1092
+ {
1093
+ "epoch": 0.15715109573241062,
1094
+ "grad_norm": 2.234375,
1095
+ "kd_loss": 0.00888274796307087,
1096
+ "learning_rate": 4.531070648859186e-06,
1097
+ "loss": 0.0319,
1098
+ "step": 545,
1099
+ "student_loss": 0.01682436652481556,
1100
+ "teacher_loss": 0.013372495770454407
1101
+ },
1102
+ {
1103
+ "epoch": 0.15859284890426759,
1104
+ "grad_norm": 1.640625,
1105
+ "kd_loss": 0.026771951466798782,
1106
+ "learning_rate": 4.450523652661086e-06,
1107
+ "loss": 0.0306,
1108
+ "step": 550,
1109
+ "student_loss": 0.0001725157053442672,
1110
+ "teacher_loss": 0.012681440450251102
1111
+ },
1112
+ {
1113
+ "epoch": 0.16003460207612458,
1114
+ "grad_norm": 1.1015625,
1115
+ "kd_loss": 0.03767453879117966,
1116
+ "learning_rate": 4.370120747111956e-06,
1117
+ "loss": 0.0381,
1118
+ "step": 555,
1119
+ "student_loss": 0.009898228570818901,
1120
+ "teacher_loss": 0.0022960191126912832
1121
+ },
1122
+ {
1123
+ "epoch": 0.16147635524798154,
1124
+ "grad_norm": 1.734375,
1125
+ "kd_loss": 0.0351857915520668,
1126
+ "learning_rate": 4.289883016480291e-06,
1127
+ "loss": 0.0381,
1128
+ "step": 560,
1129
+ "student_loss": 0.027657803148031235,
1130
+ "teacher_loss": 0.00034597894409671426
1131
+ },
1132
+ {
1133
+ "epoch": 0.16291810841983853,
1134
+ "grad_norm": 1.4296875,
1135
+ "kd_loss": 0.019546959549188614,
1136
+ "learning_rate": 4.209831501720328e-06,
1137
+ "loss": 0.0346,
1138
+ "step": 565,
1139
+ "student_loss": 0.028599649667739868,
1140
+ "teacher_loss": 0.018787631765007973
1141
+ },
1142
+ {
1143
+ "epoch": 0.1643598615916955,
1144
+ "grad_norm": 1.5625,
1145
+ "kd_loss": 0.020644016563892365,
1146
+ "learning_rate": 4.129987194954421e-06,
1147
+ "loss": 0.0405,
1148
+ "step": 570,
1149
+ "student_loss": 0.010047786869108677,
1150
+ "teacher_loss": 0.006790643557906151
1151
+ },
1152
+ {
1153
+ "epoch": 0.16580161476355249,
1154
+ "grad_norm": 1.4921875,
1155
+ "kd_loss": 0.01332971639931202,
1156
+ "learning_rate": 4.050371033968216e-06,
1157
+ "loss": 0.0376,
1158
+ "step": 575,
1159
+ "student_loss": 0.0004119194345548749,
1160
+ "teacher_loss": 0.00025748086045496166
1161
+ },
1162
+ {
1163
+ "epoch": 0.16724336793540945,
1164
+ "grad_norm": 1.3515625,
1165
+ "kd_loss": 0.012037251144647598,
1166
+ "learning_rate": 3.9710038967200825e-06,
1167
+ "loss": 0.0342,
1168
+ "step": 580,
1169
+ "student_loss": 0.0030028573237359524,
1170
+ "teacher_loss": 0.0012405890738591552
1171
+ },
1172
+ {
1173
+ "epoch": 0.16868512110726644,
1174
+ "grad_norm": 0.88671875,
1175
+ "kd_loss": 0.01703016459941864,
1176
+ "learning_rate": 3.89190659586623e-06,
1177
+ "loss": 0.032,
1178
+ "step": 585,
1179
+ "student_loss": 0.0007975402986630797,
1180
+ "teacher_loss": 0.0002710323315113783
1181
+ },
1182
+ {
1183
+ "epoch": 0.1701268742791234,
1184
+ "grad_norm": 1.1171875,
1185
+ "kd_loss": 0.01272827759385109,
1186
+ "learning_rate": 3.8130998733029517e-06,
1187
+ "loss": 0.035,
1188
+ "step": 590,
1189
+ "student_loss": 0.000885407323949039,
1190
+ "teacher_loss": 0.00043826864566653967
1191
+ },
1192
+ {
1193
+ "epoch": 0.1715686274509804,
1194
+ "grad_norm": 1.3828125,
1195
+ "kd_loss": 0.009466774761676788,
1196
+ "learning_rate": 3.734604394727419e-06,
1197
+ "loss": 0.0345,
1198
+ "step": 595,
1199
+ "student_loss": 0.0005723866051994264,
1200
+ "teacher_loss": 0.0005249618552625179
1201
+ },
1202
+ {
1203
+ "epoch": 0.17301038062283736,
1204
+ "grad_norm": 1.4765625,
1205
+ "kd_loss": 0.020295076072216034,
1206
+ "learning_rate": 3.656440744218464e-06,
1207
+ "loss": 0.04,
1208
+ "step": 600,
1209
+ "student_loss": 0.04926498606801033,
1210
+ "teacher_loss": 0.00894973799586296
1211
+ },
1212
+ {
1213
+ "epoch": 0.17445213379469435,
1214
+ "grad_norm": 1.1484375,
1215
+ "kd_loss": 0.020013822242617607,
1216
+ "learning_rate": 3.578629418838757e-06,
1217
+ "loss": 0.0378,
1218
+ "step": 605,
1219
+ "student_loss": 0.06380540132522583,
1220
+ "teacher_loss": 0.03696807473897934
1221
+ },
1222
+ {
1223
+ "epoch": 0.17589388696655134,
1224
+ "grad_norm": 1.1796875,
1225
+ "kd_loss": 0.026783978566527367,
1226
+ "learning_rate": 3.5011908232598124e-06,
1227
+ "loss": 0.0342,
1228
+ "step": 610,
1229
+ "student_loss": 0.007228103466331959,
1230
+ "teacher_loss": 0.0016964355017989874
1231
+ },
1232
+ {
1233
+ "epoch": 0.1773356401384083,
1234
+ "grad_norm": 1.0625,
1235
+ "kd_loss": 0.013121667318046093,
1236
+ "learning_rate": 3.4241452644112085e-06,
1237
+ "loss": 0.0302,
1238
+ "step": 615,
1239
+ "student_loss": 0.000252434954745695,
1240
+ "teacher_loss": 0.0002282678324263543
1241
+ },
1242
+ {
1243
+ "epoch": 0.1787773933102653,
1244
+ "grad_norm": 1.0390625,
1245
+ "kd_loss": 0.01814274862408638,
1246
+ "learning_rate": 3.3475129461554567e-06,
1247
+ "loss": 0.0412,
1248
+ "step": 620,
1249
+ "student_loss": 0.010608767159283161,
1250
+ "teacher_loss": 0.006938976235687733
1251
+ },
1252
+ {
1253
+ "epoch": 0.18021914648212226,
1254
+ "grad_norm": 1.21875,
1255
+ "kd_loss": 0.016144433990120888,
1256
+ "learning_rate": 3.271313963989886e-06,
1257
+ "loss": 0.0318,
1258
+ "step": 625,
1259
+ "student_loss": 0.003376360982656479,
1260
+ "teacher_loss": 0.003628369187936187
1261
+ },
1262
+ {
1263
+ "epoch": 0.18166089965397925,
1264
+ "grad_norm": 1.6796875,
1265
+ "kd_loss": 0.01930667646229267,
1266
+ "learning_rate": 3.195568299776945e-06,
1267
+ "loss": 0.0392,
1268
+ "step": 630,
1269
+ "student_loss": 0.042429447174072266,
1270
+ "teacher_loss": 0.02139645628631115
1271
+ },
1272
+ {
1273
+ "epoch": 0.1831026528258362,
1274
+ "grad_norm": 1.1640625,
1275
+ "kd_loss": 0.015184500254690647,
1276
+ "learning_rate": 3.1202958165043053e-06,
1277
+ "loss": 0.0522,
1278
+ "step": 635,
1279
+ "student_loss": 0.0015469329664483666,
1280
+ "teacher_loss": 0.0002648464578669518
1281
+ },
1282
+ {
1283
+ "epoch": 0.1845444059976932,
1284
+ "grad_norm": 1.6875,
1285
+ "kd_loss": 0.01360522024333477,
1286
+ "learning_rate": 3.045516253076137e-06,
1287
+ "loss": 0.0321,
1288
+ "step": 640,
1289
+ "student_loss": 0.000489354133605957,
1290
+ "teacher_loss": 0.0003778399550355971
1291
+ },
1292
+ {
1293
+ "epoch": 0.18598615916955016,
1294
+ "grad_norm": 1.328125,
1295
+ "kd_loss": 0.016992032527923584,
1296
+ "learning_rate": 2.9712492191369245e-06,
1297
+ "loss": 0.0325,
1298
+ "step": 645,
1299
+ "student_loss": 0.0005057267262600362,
1300
+ "teacher_loss": 0.0007710132049396634
1301
+ },
1302
+ {
1303
+ "epoch": 0.18742791234140715,
1304
+ "grad_norm": 1.34375,
1305
+ "kd_loss": 0.009760286659002304,
1306
+ "learning_rate": 2.8975141899291777e-06,
1307
+ "loss": 0.0316,
1308
+ "step": 650,
1309
+ "student_loss": 0.0006846904871053994,
1310
+ "teacher_loss": 0.0006360806291922927
1311
+ },
1312
+ {
1313
+ "epoch": 0.18886966551326412,
1314
+ "grad_norm": 1.140625,
1315
+ "kd_loss": 0.01777932234108448,
1316
+ "learning_rate": 2.8243305011863843e-06,
1317
+ "loss": 0.035,
1318
+ "step": 655,
1319
+ "student_loss": 0.0004645780718419701,
1320
+ "teacher_loss": 0.0002376337070018053
1321
+ },
1322
+ {
1323
+ "epoch": 0.1903114186851211,
1324
+ "grad_norm": 1.328125,
1325
+ "kd_loss": 0.012804077006876469,
1326
+ "learning_rate": 2.751717344062552e-06,
1327
+ "loss": 0.0393,
1328
+ "step": 660,
1329
+ "student_loss": 0.0006086655776016414,
1330
+ "teacher_loss": 0.0005673821433447301
1331
+ },
1332
+ {
1333
+ "epoch": 0.19175317185697807,
1334
+ "grad_norm": 2.71875,
1335
+ "kd_loss": 0.04194798693060875,
1336
+ "learning_rate": 2.6796937600996587e-06,
1337
+ "loss": 0.0417,
1338
+ "step": 665,
1339
+ "student_loss": 0.004477351903915405,
1340
+ "teacher_loss": 0.0028319216798990965
1341
+ },
1342
+ {
1343
+ "epoch": 0.19319492502883506,
1344
+ "grad_norm": 1.4453125,
1345
+ "kd_loss": 0.020050913095474243,
1346
+ "learning_rate": 2.6082786362343377e-06,
1347
+ "loss": 0.0387,
1348
+ "step": 670,
1349
+ "student_loss": 0.008853067643940449,
1350
+ "teacher_loss": 0.005904260091483593
1351
+ },
1352
+ {
1353
+ "epoch": 0.19463667820069205,
1354
+ "grad_norm": 1.09375,
1355
+ "kd_loss": 0.008323847316205502,
1356
+ "learning_rate": 2.5374906998451094e-06,
1357
+ "loss": 0.0365,
1358
+ "step": 675,
1359
+ "student_loss": 0.0005440004752017558,
1360
+ "teacher_loss": 0.00048730146954767406
1361
+ },
1362
+ {
1363
+ "epoch": 0.19607843137254902,
1364
+ "grad_norm": 1.5234375,
1365
+ "kd_loss": 0.025357170030474663,
1366
+ "learning_rate": 2.467348513841447e-06,
1367
+ "loss": 0.033,
1368
+ "step": 680,
1369
+ "student_loss": 0.07211022824048996,
1370
+ "teacher_loss": 0.03345099464058876
1371
+ },
1372
+ {
1373
+ "epoch": 0.197520184544406,
1374
+ "grad_norm": 1.6171875,
1375
+ "kd_loss": 0.016125330701470375,
1376
+ "learning_rate": 2.3978704717959777e-06,
1377
+ "loss": 0.035,
1378
+ "step": 685,
1379
+ "student_loss": 0.035295210778713226,
1380
+ "teacher_loss": 0.02864968776702881
1381
+ },
1382
+ {
1383
+ "epoch": 0.19896193771626297,
1384
+ "grad_norm": 1.21875,
1385
+ "kd_loss": 0.042695559561252594,
1386
+ "learning_rate": 2.329074793121085e-06,
1387
+ "loss": 0.038,
1388
+ "step": 690,
1389
+ "student_loss": 0.015520169399678707,
1390
+ "teacher_loss": 0.006200406700372696
1391
+ },
1392
+ {
1393
+ "epoch": 0.20040369088811996,
1394
+ "grad_norm": 1.390625,
1395
+ "kd_loss": 0.016068480908870697,
1396
+ "learning_rate": 2.260979518291186e-06,
1397
+ "loss": 0.0406,
1398
+ "step": 695,
1399
+ "student_loss": 0.006739252246916294,
1400
+ "teacher_loss": 0.004444445949047804
1401
+ },
1402
+ {
1403
+ "epoch": 0.20184544405997693,
1404
+ "grad_norm": 1.5390625,
1405
+ "kd_loss": 0.010956254787743092,
1406
+ "learning_rate": 2.1936025041119268e-06,
1407
+ "loss": 0.0352,
1408
+ "step": 700,
1409
+ "student_loss": 0.0009333753259852529,
1410
+ "teacher_loss": 0.0003666435368359089
1411
+ },
1412
+ {
1413
+ "epoch": 0.20328719723183392,
1414
+ "grad_norm": 1.3828125,
1415
+ "kd_loss": 0.00898037664592266,
1416
+ "learning_rate": 2.1269614190375477e-06,
1417
+ "loss": 0.0344,
1418
+ "step": 705,
1419
+ "student_loss": 0.0006648478447459638,
1420
+ "teacher_loss": 0.0005693589337170124
1421
+ },
1422
+ {
1423
+ "epoch": 0.20472895040369088,
1424
+ "grad_norm": 1.5,
1425
+ "kd_loss": 0.041511427611112595,
1426
+ "learning_rate": 2.061073738537635e-06,
1427
+ "loss": 0.0396,
1428
+ "step": 710,
1429
+ "student_loss": 0.0163543913513422,
1430
+ "teacher_loss": 0.00382953486405313
1431
+ },
1432
+ {
1433
+ "epoch": 0.20617070357554787,
1434
+ "grad_norm": 2.0,
1435
+ "kd_loss": 0.014297292567789555,
1436
+ "learning_rate": 1.9959567405144825e-06,
1437
+ "loss": 0.0476,
1438
+ "step": 715,
1439
+ "student_loss": 0.004557440988719463,
1440
+ "teacher_loss": 0.009534560143947601
1441
+ },
1442
+ {
1443
+ "epoch": 0.20761245674740483,
1444
+ "grad_norm": 1.109375,
1445
+ "kd_loss": 0.013292558491230011,
1446
+ "learning_rate": 1.931627500772263e-06,
1447
+ "loss": 0.0378,
1448
+ "step": 720,
1449
+ "student_loss": 0.0005340321222320199,
1450
+ "teacher_loss": 0.000525432638823986
1451
+ },
1452
+ {
1453
+ "epoch": 0.20905420991926182,
1454
+ "grad_norm": 0.93359375,
1455
+ "kd_loss": 0.01108925137668848,
1456
+ "learning_rate": 1.8681028885391905e-06,
1457
+ "loss": 0.0297,
1458
+ "step": 725,
1459
+ "student_loss": 0.0002994259411934763,
1460
+ "teacher_loss": 0.0002954052179120481
1461
+ },
1462
+ {
1463
+ "epoch": 0.2104959630911188,
1464
+ "grad_norm": 1.2109375,
1465
+ "kd_loss": 0.01718929037451744,
1466
+ "learning_rate": 1.8053995620438625e-06,
1467
+ "loss": 0.0354,
1468
+ "step": 730,
1469
+ "student_loss": 0.00870412029325962,
1470
+ "teacher_loss": 0.0017060510581359267
1471
+ },
1472
+ {
1473
+ "epoch": 0.21193771626297578,
1474
+ "grad_norm": 1.515625,
1475
+ "kd_loss": 0.016554107889533043,
1476
+ "learning_rate": 1.743533964146924e-06,
1477
+ "loss": 0.041,
1478
+ "step": 735,
1479
+ "student_loss": 0.00043048406951129436,
1480
+ "teacher_loss": 0.0002133132511517033
1481
+ },
1482
+ {
1483
+ "epoch": 0.21337946943483277,
1484
+ "grad_norm": 0.9140625,
1485
+ "kd_loss": 0.023798726499080658,
1486
+ "learning_rate": 1.6825223180292138e-06,
1487
+ "loss": 0.0324,
1488
+ "step": 740,
1489
+ "student_loss": 0.0012331322068348527,
1490
+ "teacher_loss": 0.0018722045933827758
1491
+ },
1492
+ {
1493
+ "epoch": 0.21482122260668973,
1494
+ "grad_norm": 1.890625,
1495
+ "kd_loss": 0.014276617206633091,
1496
+ "learning_rate": 1.6223806229375182e-06,
1497
+ "loss": 0.0367,
1498
+ "step": 745,
1499
+ "student_loss": 0.040091373026371,
1500
+ "teacher_loss": 0.019951729103922844
1501
+ },
1502
+ {
1503
+ "epoch": 0.21626297577854672,
1504
+ "grad_norm": 1.2265625,
1505
+ "kd_loss": 0.03414074704051018,
1506
+ "learning_rate": 1.563124649989043e-06,
1507
+ "loss": 0.0366,
1508
+ "step": 750,
1509
+ "student_loss": 0.017988989129662514,
1510
+ "teacher_loss": 0.006315938197076321
1511
+ },
1512
+ {
1513
+ "epoch": 0.2177047289504037,
1514
+ "grad_norm": 1.9375,
1515
+ "kd_loss": 0.013627647422254086,
1516
+ "learning_rate": 1.5047699380357134e-06,
1517
+ "loss": 0.0418,
1518
+ "step": 755,
1519
+ "student_loss": 0.024281585589051247,
1520
+ "teacher_loss": 0.0027684655506163836
1521
+ },
1522
+ {
1523
+ "epoch": 0.21914648212226068,
1524
+ "grad_norm": 1.5,
1525
+ "kd_loss": 0.030101457610726357,
1526
+ "learning_rate": 1.4473317895893773e-06,
1527
+ "loss": 0.0317,
1528
+ "step": 760,
1529
+ "student_loss": 0.024278033524751663,
1530
+ "teacher_loss": 0.00042108772322535515
1531
+ },
1532
+ {
1533
+ "epoch": 0.22058823529411764,
1534
+ "grad_norm": 1.265625,
1535
+ "kd_loss": 0.03181413933634758,
1536
+ "learning_rate": 1.39082526680899e-06,
1537
+ "loss": 0.0404,
1538
+ "step": 765,
1539
+ "student_loss": 0.004552094731479883,
1540
+ "teacher_loss": 0.001384846051223576
1541
+ },
1542
+ {
1543
+ "epoch": 0.22202998846597463,
1544
+ "grad_norm": 1.0546875,
1545
+ "kd_loss": 0.03087952919304371,
1546
+ "learning_rate": 1.3352651875508204e-06,
1547
+ "loss": 0.0359,
1548
+ "step": 770,
1549
+ "student_loss": 0.00462262611836195,
1550
+ "teacher_loss": 0.0008631970267742872
1551
+ },
1552
+ {
1553
+ "epoch": 0.2234717416378316,
1554
+ "grad_norm": 1.0703125,
1555
+ "kd_loss": 0.01582036353647709,
1556
+ "learning_rate": 1.2806661214827286e-06,
1557
+ "loss": 0.0327,
1558
+ "step": 775,
1559
+ "student_loss": 0.00521172396838665,
1560
+ "teacher_loss": 0.002128555905073881
1561
+ },
1562
+ {
1563
+ "epoch": 0.22491349480968859,
1564
+ "grad_norm": 1.078125,
1565
+ "kd_loss": 0.017553336918354034,
1566
+ "learning_rate": 1.2270423862635188e-06,
1567
+ "loss": 0.0375,
1568
+ "step": 780,
1569
+ "student_loss": 0.0012290088925510645,
1570
+ "teacher_loss": 0.00023391979630105197
1571
+ },
1572
+ {
1573
+ "epoch": 0.22635524798154555,
1574
+ "grad_norm": 1.578125,
1575
+ "kd_loss": 0.009543483145534992,
1576
+ "learning_rate": 1.1744080437883859e-06,
1577
+ "loss": 0.0324,
1578
+ "step": 785,
1579
+ "student_loss": 0.00035309791564941406,
1580
+ "teacher_loss": 0.00032548559829592705
1581
+ },
1582
+ {
1583
+ "epoch": 0.22779700115340254,
1584
+ "grad_norm": 1.3203125,
1585
+ "kd_loss": 0.00901024416089058,
1586
+ "learning_rate": 1.1227768965014246e-06,
1587
+ "loss": 0.032,
1588
+ "step": 790,
1589
+ "student_loss": 0.000992319663055241,
1590
+ "teacher_loss": 0.0008018451044335961
1591
+ },
1592
+ {
1593
+ "epoch": 0.2292387543252595,
1594
+ "grad_norm": 1.25,
1595
+ "kd_loss": 0.013804268091917038,
1596
+ "learning_rate": 1.0721624837761768e-06,
1597
+ "loss": 0.0368,
1598
+ "step": 795,
1599
+ "student_loss": 0.00046814189408905804,
1600
+ "teacher_loss": 0.00019660216639749706
1601
+ },
1602
+ {
1603
+ "epoch": 0.2306805074971165,
1604
+ "grad_norm": 1.8046875,
1605
+ "kd_loss": 0.025934722274541855,
1606
+ "learning_rate": 1.0225780783651689e-06,
1607
+ "loss": 0.0436,
1608
+ "step": 800,
1609
+ "student_loss": 0.09193355590105057,
1610
+ "teacher_loss": 0.045340001583099365
1611
+ },
1612
+ {
1613
+ "epoch": 0.23212226066897348,
1614
+ "grad_norm": 1.40625,
1615
+ "kd_loss": 0.01032049860805273,
1616
+ "learning_rate": 9.740366829193587e-07,
1617
+ "loss": 0.0412,
1618
+ "step": 805,
1619
+ "student_loss": 0.0004188880557194352,
1620
+ "teacher_loss": 0.00034304597647860646
1621
+ },
1622
+ {
1623
+ "epoch": 0.23356401384083045,
1624
+ "grad_norm": 5.96875,
1625
+ "kd_loss": 0.01783071458339691,
1626
+ "learning_rate": 9.265510265784189e-07,
1627
+ "loss": 0.0363,
1628
+ "step": 810,
1629
+ "student_loss": 0.0005420530214905739,
1630
+ "teacher_loss": 0.00027590151876211166
1631
+ },
1632
+ {
1633
+ "epoch": 0.23500576701268744,
1634
+ "grad_norm": 1.9140625,
1635
+ "kd_loss": 0.011872808448970318,
1636
+ "learning_rate": 8.801335616327378e-07,
1637
+ "loss": 0.0351,
1638
+ "step": 815,
1639
+ "student_loss": 0.004859322682023048,
1640
+ "teacher_loss": 0.013330879621207714
1641
+ },
1642
+ {
1643
+ "epoch": 0.2364475201845444,
1644
+ "grad_norm": 1.2265625,
1645
+ "kd_loss": 0.01942305825650692,
1646
+ "learning_rate": 8.347964602580245e-07,
1647
+ "loss": 0.0359,
1648
+ "step": 820,
1649
+ "student_loss": 0.003229373600333929,
1650
+ "teacher_loss": 0.0003580668126232922
1651
+ },
1652
+ {
1653
+ "epoch": 0.2378892733564014,
1654
+ "grad_norm": 1.3828125,
1655
+ "kd_loss": 0.019788295030593872,
1656
+ "learning_rate": 7.905516113233652e-07,
1657
+ "loss": 0.0346,
1658
+ "step": 825,
1659
+ "student_loss": 0.0008779842173680663,
1660
+ "teacher_loss": 0.016724394634366035
1661
+ },
1662
+ {
1663
+ "epoch": 0.23933102652825836,
1664
+ "grad_norm": 1.421875,
1665
+ "kd_loss": 0.018033282831311226,
1666
+ "learning_rate": 7.474106172735746e-07,
1667
+ "loss": 0.0358,
1668
+ "step": 830,
1669
+ "student_loss": 0.008363536559045315,
1670
+ "teacher_loss": 0.005316237919032574
1671
+ },
1672
+ {
1673
+ "epoch": 0.24077277970011535,
1674
+ "grad_norm": 1.109375,
1675
+ "kd_loss": 0.016192706301808357,
1676
+ "learning_rate": 7.053847910866513e-07,
1677
+ "loss": 0.0315,
1678
+ "step": 835,
1679
+ "student_loss": 0.014144017361104488,
1680
+ "teacher_loss": 0.0013847900554537773
1681
+ },
1682
+ {
1683
+ "epoch": 0.2422145328719723,
1684
+ "grad_norm": 1.71875,
1685
+ "kd_loss": 0.019871558994054794,
1686
+ "learning_rate": 6.644851533071556e-07,
1687
+ "loss": 0.0359,
1688
+ "step": 840,
1689
+ "student_loss": 0.001529853092506528,
1690
+ "teacher_loss": 0.0001909300044644624
1691
+ },
1692
+ {
1693
+ "epoch": 0.2436562860438293,
1694
+ "grad_norm": 1.0234375,
1695
+ "kd_loss": 0.011251869611442089,
1696
+ "learning_rate": 6.24722429156251e-07,
1697
+ "loss": 0.0532,
1698
+ "step": 845,
1699
+ "student_loss": 0.0005619142320938408,
1700
+ "teacher_loss": 0.0004758336581289768
1701
+ },
1702
+ {
1703
+ "epoch": 0.24509803921568626,
1704
+ "grad_norm": 2.390625,
1705
+ "kd_loss": 0.02860349230468273,
1706
+ "learning_rate": 5.861070457192081e-07,
1707
+ "loss": 0.0371,
1708
+ "step": 850,
1709
+ "student_loss": 0.012859327718615532,
1710
+ "teacher_loss": 0.0003160137275699526
1711
+ },
1712
+ {
1713
+ "epoch": 0.24653979238754326,
1714
+ "grad_norm": 1.015625,
1715
+ "kd_loss": 0.016554828733205795,
1716
+ "learning_rate": 5.486491292110796e-07,
1717
+ "loss": 0.0368,
1718
+ "step": 855,
1719
+ "student_loss": 0.0005889005260542035,
1720
+ "teacher_loss": 0.0003898893191944808
1721
+ },
1722
+ {
1723
+ "epoch": 0.24798154555940022,
1724
+ "grad_norm": 0.78125,
1725
+ "kd_loss": 0.012213003821671009,
1726
+ "learning_rate": 5.123585023212785e-07,
1727
+ "loss": 0.0417,
1728
+ "step": 860,
1729
+ "student_loss": 0.0005682500777766109,
1730
+ "teacher_loss": 0.00047416481538675725
1731
+ },
1732
+ {
1733
+ "epoch": 0.2494232987312572,
1734
+ "grad_norm": 1.21875,
1735
+ "kd_loss": 0.01672137714922428,
1736
+ "learning_rate": 4.772446816377408e-07,
1737
+ "loss": 0.037,
1738
+ "step": 865,
1739
+ "student_loss": 0.0010720123536884785,
1740
+ "teacher_loss": 0.0008305139490403235
1741
+ },
1742
+ {
1743
+ "epoch": 0.2508650519031142,
1744
+ "grad_norm": 1.046875,
1745
+ "kd_loss": 0.018183542415499687,
1746
+ "learning_rate": 4.4331687515137614e-07,
1747
+ "loss": 0.0354,
1748
+ "step": 870,
1749
+ "student_loss": 0.001711298362351954,
1750
+ "teacher_loss": 0.0007006946834735572
1751
+ },
1752
+ {
1753
+ "epoch": 0.25230680507497116,
1754
+ "grad_norm": 1.53125,
1755
+ "kd_loss": 0.015019524842500687,
1756
+ "learning_rate": 4.1058397984142405e-07,
1757
+ "loss": 0.0375,
1758
+ "step": 875,
1759
+ "student_loss": 0.00019167628488503397,
1760
+ "teacher_loss": 0.0002023013075813651
1761
+ },
1762
+ {
1763
+ "epoch": 0.2537485582468281,
1764
+ "grad_norm": 1.0859375,
1765
+ "kd_loss": 0.016358323395252228,
1766
+ "learning_rate": 3.790545793423761e-07,
1767
+ "loss": 0.039,
1768
+ "step": 880,
1769
+ "student_loss": 0.00405404157936573,
1770
+ "teacher_loss": 0.0018468407215550542
1771
+ },
1772
+ {
1773
+ "epoch": 0.25519031141868515,
1774
+ "grad_norm": 1.28125,
1775
+ "kd_loss": 0.04253898561000824,
1776
+ "learning_rate": 3.4873694169306915e-07,
1777
+ "loss": 0.0316,
1778
+ "step": 885,
1779
+ "student_loss": 0.011344845406711102,
1780
+ "teacher_loss": 0.044258587062358856
1781
+ },
1782
+ {
1783
+ "epoch": 0.2566320645905421,
1784
+ "grad_norm": 1.3359375,
1785
+ "kd_loss": 0.009111128747463226,
1786
+ "learning_rate": 3.196390171685343e-07,
1787
+ "loss": 0.0369,
1788
+ "step": 890,
1789
+ "student_loss": 0.0003722853143699467,
1790
+ "teacher_loss": 0.00034851342206820846
1791
+ },
1792
+ {
1793
+ "epoch": 0.25807381776239907,
1794
+ "grad_norm": 1.4921875,
1795
+ "kd_loss": 0.02147591859102249,
1796
+ "learning_rate": 2.917684361951728e-07,
1797
+ "loss": 0.0352,
1798
+ "step": 895,
1799
+ "student_loss": 0.005593026988208294,
1800
+ "teacher_loss": 0.0024380958639085293
1801
+ },
1802
+ {
1803
+ "epoch": 0.25951557093425603,
1804
+ "grad_norm": 1.296875,
1805
+ "kd_loss": 0.017156513407826424,
1806
+ "learning_rate": 2.65132507349814e-07,
1807
+ "loss": 0.0382,
1808
+ "step": 900,
1809
+ "student_loss": 0.030825650319457054,
1810
+ "teacher_loss": 0.022559169679880142
1811
+ },
1812
+ {
1813
+ "epoch": 0.26095732410611305,
1814
+ "grad_norm": 1.296875,
1815
+ "kd_loss": 0.014595987275242805,
1816
+ "learning_rate": 2.397382154431621e-07,
1817
+ "loss": 0.0329,
1818
+ "step": 905,
1819
+ "student_loss": 0.0070388843305408955,
1820
+ "teacher_loss": 0.0033032468054443598
1821
+ },
1822
+ {
1823
+ "epoch": 0.26239907727797,
1824
+ "grad_norm": 1.1796875,
1825
+ "kd_loss": 0.00915343314409256,
1826
+ "learning_rate": 2.1559221968815547e-07,
1827
+ "loss": 0.0309,
1828
+ "step": 910,
1829
+ "student_loss": 0.0003465732734184712,
1830
+ "teacher_loss": 0.0003652082523331046
1831
+ },
1832
+ {
1833
+ "epoch": 0.263840830449827,
1834
+ "grad_norm": 1.4765625,
1835
+ "kd_loss": 0.047718290239572525,
1836
+ "learning_rate": 1.9270085195370048e-07,
1837
+ "loss": 0.0354,
1838
+ "step": 915,
1839
+ "student_loss": 0.07753528654575348,
1840
+ "teacher_loss": 0.00128907710313797
1841
+ },
1842
+ {
1843
+ "epoch": 0.26528258362168394,
1844
+ "grad_norm": 1.1171875,
1845
+ "kd_loss": 0.013440879993140697,
1846
+ "learning_rate": 1.7107011510424766e-07,
1847
+ "loss": 0.0341,
1848
+ "step": 920,
1849
+ "student_loss": 0.013975732028484344,
1850
+ "teacher_loss": 0.011724431999027729
1851
+ },
1852
+ {
1853
+ "epoch": 0.26672433679354096,
1854
+ "grad_norm": 0.96484375,
1855
+ "kd_loss": 0.010464920662343502,
1856
+ "learning_rate": 1.5070568142564912e-07,
1857
+ "loss": 0.0324,
1858
+ "step": 925,
1859
+ "student_loss": 0.0003131957200821489,
1860
+ "teacher_loss": 0.00029044810798950493
1861
+ },
1862
+ {
1863
+ "epoch": 0.2681660899653979,
1864
+ "grad_norm": 0.98828125,
1865
+ "kd_loss": 0.016116317361593246,
1866
+ "learning_rate": 1.3161289113769405e-07,
1867
+ "loss": 0.0326,
1868
+ "step": 930,
1869
+ "student_loss": 0.013056068681180477,
1870
+ "teacher_loss": 0.0016408703522756696
1871
+ },
1872
+ {
1873
+ "epoch": 0.2696078431372549,
1874
+ "grad_norm": 0.94921875,
1875
+ "kd_loss": 0.011886276304721832,
1876
+ "learning_rate": 1.1379675099373489e-07,
1877
+ "loss": 0.0334,
1878
+ "step": 935,
1879
+ "student_loss": 0.002202346222475171,
1880
+ "teacher_loss": 0.0023580535780638456
1881
+ },
1882
+ {
1883
+ "epoch": 0.2710495963091119,
1884
+ "grad_norm": 1.0390625,
1885
+ "kd_loss": 0.02240722067654133,
1886
+ "learning_rate": 9.726193296774767e-08,
1887
+ "loss": 0.0353,
1888
+ "step": 940,
1889
+ "student_loss": 0.008520473726093769,
1890
+ "teacher_loss": 0.000717982358764857
1891
+ },
1892
+ {
1893
+ "epoch": 0.27249134948096887,
1894
+ "grad_norm": 0.98046875,
1895
+ "kd_loss": 0.010714667849242687,
1896
+ "learning_rate": 8.201277302919086e-08,
1897
+ "loss": 0.0384,
1898
+ "step": 945,
1899
+ "student_loss": 0.003037899499759078,
1900
+ "teacher_loss": 0.001546806888654828
1901
+ },
1902
+ {
1903
+ "epoch": 0.27393310265282583,
1904
+ "grad_norm": 0.91796875,
1905
+ "kd_loss": 0.01626349613070488,
1906
+ "learning_rate": 6.805327000596995e-08,
1907
+ "loss": 0.0319,
1908
+ "step": 950,
1909
+ "student_loss": 0.00029273482505232096,
1910
+ "teacher_loss": 0.00015759514644742012
1911
+ },
1912
+ {
1913
+ "epoch": 0.2753748558246828,
1914
+ "grad_norm": 1.0625,
1915
+ "kd_loss": 0.016484661027789116,
1916
+ "learning_rate": 5.538708453581787e-08,
1917
+ "loss": 0.0353,
1918
+ "step": 955,
1919
+ "student_loss": 0.0004839501343667507,
1920
+ "teacher_loss": 0.00021457124967128038
1921
+ },
1922
+ {
1923
+ "epoch": 0.2768166089965398,
1924
+ "grad_norm": 1.609375,
1925
+ "kd_loss": 0.016130708158016205,
1926
+ "learning_rate": 4.40175381063529e-08,
1927
+ "loss": 0.039,
1928
+ "step": 960,
1929
+ "student_loss": 0.0031732963398098946,
1930
+ "teacher_loss": 0.0024704965762794018
1931
+ },
1932
+ {
1933
+ "epoch": 0.2782583621683968,
1934
+ "grad_norm": 1.1328125,
1935
+ "kd_loss": 0.022074589505791664,
1936
+ "learning_rate": 3.394761218407705e-08,
1937
+ "loss": 0.0341,
1938
+ "step": 965,
1939
+ "student_loss": 0.010913808830082417,
1940
+ "teacher_loss": 0.006757721304893494
1941
+ },
1942
+ {
1943
+ "epoch": 0.27970011534025374,
1944
+ "grad_norm": 1.34375,
1945
+ "kd_loss": 0.01725686714053154,
1946
+ "learning_rate": 2.5179947432540376e-08,
1947
+ "loss": 0.035,
1948
+ "step": 970,
1949
+ "student_loss": 0.00032446443219669163,
1950
+ "teacher_loss": 0.00020874734036624432
1951
+ },
1952
+ {
1953
+ "epoch": 0.2811418685121107,
1954
+ "grad_norm": 1.0,
1955
+ "kd_loss": 0.012324236333370209,
1956
+ "learning_rate": 1.7716843019867646e-08,
1957
+ "loss": 0.034,
1958
+ "step": 975,
1959
+ "student_loss": 0.02250129170715809,
1960
+ "teacher_loss": 0.0008003146504051983
1961
+ },
1962
+ {
1963
+ "epoch": 0.2825836216839677,
1964
+ "grad_norm": 0.97265625,
1965
+ "kd_loss": 0.01817537285387516,
1966
+ "learning_rate": 1.156025601584676e-08,
1967
+ "loss": 0.0372,
1968
+ "step": 980,
1969
+ "student_loss": 0.0004171262844465673,
1970
+ "teacher_loss": 0.00023037723440211266
1971
+ },
1972
+ {
1973
+ "epoch": 0.2840253748558247,
1974
+ "grad_norm": 1.203125,
1975
+ "kd_loss": 0.017179273068904877,
1976
+ "learning_rate": 6.711800878718144e-09,
1977
+ "loss": 0.0365,
1978
+ "step": 985,
1979
+ "student_loss": 0.000331983930664137,
1980
+ "teacher_loss": 0.000896628072950989
1981
+ },
1982
+ {
1983
+ "epoch": 0.28546712802768165,
1984
+ "grad_norm": 1.3671875,
1985
+ "kd_loss": 0.01191702764481306,
1986
+ "learning_rate": 3.1727490318111953e-09,
1987
+ "loss": 0.0315,
1988
+ "step": 990,
1989
+ "student_loss": 0.0003260337980464101,
1990
+ "teacher_loss": 0.00025129789719358087
1991
+ },
1992
+ {
1993
+ "epoch": 0.2869088811995386,
1994
+ "grad_norm": 1.4609375,
1995
+ "kd_loss": 0.027709918096661568,
1996
+ "learning_rate": 9.440285301370865e-10,
1997
+ "loss": 0.0405,
1998
+ "step": 995,
1999
+ "student_loss": 0.00899884570389986,
2000
+ "teacher_loss": 0.0006192626315169036
2001
+ },
2002
+ {
2003
+ "epoch": 0.28835063437139563,
2004
+ "grad_norm": 0.77734375,
2005
+ "kd_loss": 0.020871583372354507,
2006
+ "learning_rate": 2.622381702066523e-11,
2007
+ "loss": 0.0348,
2008
+ "step": 1000,
2009
+ "student_loss": 0.008410735987126827,
2010
+ "teacher_loss": 0.0003241391386836767
2011
+ },
2012
+ {
2013
+ "epoch": 0.28835063437139563,
2014
+ "kd_loss": 0.020871583372354507,
2015
+ "step": 1000,
2016
+ "student_loss": 0.008410735987126827,
2017
+ "teacher_loss": 0.0003241391386836767,
2018
+ "total_flos": 0.0,
2019
+ "train_loss": 0.05713042087852955,
2020
+ "train_runtime": 32582.8193,
2021
+ "train_samples_per_second": 0.491,
2022
+ "train_steps_per_second": 0.031
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/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-1000/vocab.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-200/added_tokens.json ADDED
@@ -0,0 +1,35 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "</tool_call>": 151658,
3
+ "<tool_call>": 151657,
4
+ "<|action_sep|>": 151670,
5
+ "<|arg_sep|>": 151671,
6
+ "<|box_end|>": 151649,
7
+ "<|box_start|>": 151648,
8
+ "<|call_sep|>": 151666,
9
+ "<|end_of_text|>": 151673,
10
+ "<|endoftext|>": 151643,
11
+ "<|exception_sep|>": 151669,
12
+ "<|file_sep|>": 151664,
13
+ "<|fim_middle|>": 151660,
14
+ "<|fim_pad|>": 151662,
15
+ "<|fim_prefix|>": 151659,
16
+ "<|fim_suffix|>": 151661,
17
+ "<|frame_sep|>": 151672,
18
+ "<|im_end|>": 151645,
19
+ "<|im_start|>": 151644,
20
+ "<|image_pad|>": 151655,
21
+ "<|latent_end|>": 151675,
22
+ "<|latent_start|>": 151674,
23
+ "<|line_sep|>": 151667,
24
+ "<|object_ref_end|>": 151647,
25
+ "<|object_ref_start|>": 151646,
26
+ "<|quad_end|>": 151651,
27
+ "<|quad_start|>": 151650,
28
+ "<|repo_name|>": 151663,
29
+ "<|return_sep|>": 151668,
30
+ "<|trace_context_start|>": 151665,
31
+ "<|video_pad|>": 151656,
32
+ "<|vision_end|>": 151653,
33
+ "<|vision_pad|>": 151654,
34
+ "<|vision_start|>": 151652
35
+ }
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-200/chat_template.jinja ADDED
@@ -0,0 +1,54 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {%- if tools %}
2
+ {{- '<|im_start|>system\n' }}
3
+ {%- if messages[0]['role'] == 'system' %}
4
+ {{- messages[0]['content'] }}
5
+ {%- else %}
6
+ {{- 'You are Qwen, created by Alibaba Cloud. You are a helpful assistant.' }}
7
+ {%- endif %}
8
+ {{- "\n\n# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
9
+ {%- for tool in tools %}
10
+ {{- "\n" }}
11
+ {{- tool | tojson }}
12
+ {%- endfor %}
13
+ {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
14
+ {%- else %}
15
+ {%- if messages[0]['role'] == 'system' %}
16
+ {{- '<|im_start|>system\n' + messages[0]['content'] + '<|im_end|>\n' }}
17
+ {%- else %}
18
+ {{- '<|im_start|>system\nYou are Qwen, created by Alibaba Cloud. You are a helpful assistant.<|im_end|>\n' }}
19
+ {%- endif %}
20
+ {%- endif %}
21
+ {%- for message in messages %}
22
+ {%- if (message.role == "user") or (message.role == "system" and not loop.first) or (message.role == "assistant" and not message.tool_calls) %}
23
+ {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
24
+ {%- elif message.role == "assistant" %}
25
+ {{- '<|im_start|>' + message.role }}
26
+ {%- if message.content %}
27
+ {{- '\n' + message.content }}
28
+ {%- endif %}
29
+ {%- for tool_call in message.tool_calls %}
30
+ {%- if tool_call.function is defined %}
31
+ {%- set tool_call = tool_call.function %}
32
+ {%- endif %}
33
+ {{- '\n<tool_call>\n{"name": "' }}
34
+ {{- tool_call.name }}
35
+ {{- '", "arguments": ' }}
36
+ {{- tool_call.arguments | tojson }}
37
+ {{- '}\n</tool_call>' }}
38
+ {%- endfor %}
39
+ {{- '<|im_end|>\n' }}
40
+ {%- elif message.role == "tool" %}
41
+ {%- if (loop.index0 == 0) or (messages[loop.index0 - 1].role != "tool") %}
42
+ {{- '<|im_start|>user' }}
43
+ {%- endif %}
44
+ {{- '\n<tool_response>\n' }}
45
+ {{- message.content }}
46
+ {{- '\n</tool_response>' }}
47
+ {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
48
+ {{- '<|im_end|>\n' }}
49
+ {%- endif %}
50
+ {%- endif %}
51
+ {%- endfor %}
52
+ {%- if add_generation_prompt %}
53
+ {{- '<|im_start|>assistant\n' }}
54
+ {%- endif %}
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-200/config.json ADDED
@@ -0,0 +1,58 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Qwen2ForCausalLM"
4
+ ],
5
+ "attention_dropout": 0.0,
6
+ "dtype": "bfloat16",
7
+ "eos_token_id": 151643,
8
+ "hidden_act": "silu",
9
+ "hidden_size": 1536,
10
+ "initializer_range": 0.02,
11
+ "intermediate_size": 8960,
12
+ "layer_types": [
13
+ "full_attention",
14
+ "full_attention",
15
+ "full_attention",
16
+ "full_attention",
17
+ "full_attention",
18
+ "full_attention",
19
+ "full_attention",
20
+ "full_attention",
21
+ "full_attention",
22
+ "full_attention",
23
+ "full_attention",
24
+ "full_attention",
25
+ "full_attention",
26
+ "full_attention",
27
+ "full_attention",
28
+ "full_attention",
29
+ "full_attention",
30
+ "full_attention",
31
+ "full_attention",
32
+ "full_attention",
33
+ "full_attention",
34
+ "full_attention",
35
+ "full_attention",
36
+ "full_attention",
37
+ "full_attention",
38
+ "full_attention",
39
+ "full_attention",
40
+ "full_attention"
41
+ ],
42
+ "max_position_embeddings": 32768,
43
+ "max_window_layers": 28,
44
+ "model_type": "qwen2",
45
+ "num_attention_heads": 12,
46
+ "num_hidden_layers": 28,
47
+ "num_key_value_heads": 2,
48
+ "pad_token_id": 151643,
49
+ "rms_norm_eps": 1e-06,
50
+ "rope_scaling": null,
51
+ "rope_theta": 1000000.0,
52
+ "sliding_window": null,
53
+ "tie_word_embeddings": true,
54
+ "transformers_version": "4.57.6",
55
+ "use_cache": true,
56
+ "use_sliding_window": false,
57
+ "vocab_size": 151676
58
+ }
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-200/merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-200/pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:12535971b61e18cd0094b867d83a111ff4684c13edee216d1be7223d0d041c5f
3
+ size 3096275515
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-200/special_tokens_map.json ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "additional_special_tokens": [
3
+ "<|im_start|>",
4
+ "<|im_end|>",
5
+ "<|object_ref_start|>",
6
+ "<|object_ref_end|>",
7
+ "<|box_start|>",
8
+ "<|box_end|>",
9
+ "<|quad_start|>",
10
+ "<|quad_end|>",
11
+ "<|vision_start|>",
12
+ "<|vision_end|>",
13
+ "<|vision_pad|>",
14
+ "<|image_pad|>",
15
+ "<|video_pad|>"
16
+ ],
17
+ "eos_token": {
18
+ "content": "<|endoftext|>",
19
+ "lstrip": false,
20
+ "normalized": false,
21
+ "rstrip": false,
22
+ "single_word": false
23
+ },
24
+ "pad_token": {
25
+ "content": "<|endoftext|>",
26
+ "lstrip": false,
27
+ "normalized": false,
28
+ "rstrip": false,
29
+ "single_word": false
30
+ }
31
+ }
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-200/thought_projector.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:454d5f5d4f97032f2c4ef6c292c2418e4799c6b9d0eedba3e52c675aa7ef61ad
3
+ size 9445953
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-200/tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:83a790d654474f5dfe225f889afd0210313eb1083f942671f2c4b8e95a1c922b
3
+ size 11424004
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-200/tokenizer_config.json ADDED
@@ -0,0 +1,295 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_bos_token": false,
3
+ "add_prefix_space": false,
4
+ "added_tokens_decoder": {
5
+ "151643": {
6
+ "content": "<|endoftext|>",
7
+ "lstrip": false,
8
+ "normalized": false,
9
+ "rstrip": false,
10
+ "single_word": false,
11
+ "special": true
12
+ },
13
+ "151644": {
14
+ "content": "<|im_start|>",
15
+ "lstrip": false,
16
+ "normalized": false,
17
+ "rstrip": false,
18
+ "single_word": false,
19
+ "special": true
20
+ },
21
+ "151645": {
22
+ "content": "<|im_end|>",
23
+ "lstrip": false,
24
+ "normalized": false,
25
+ "rstrip": false,
26
+ "single_word": false,
27
+ "special": true
28
+ },
29
+ "151646": {
30
+ "content": "<|object_ref_start|>",
31
+ "lstrip": false,
32
+ "normalized": false,
33
+ "rstrip": false,
34
+ "single_word": false,
35
+ "special": true
36
+ },
37
+ "151647": {
38
+ "content": "<|object_ref_end|>",
39
+ "lstrip": false,
40
+ "normalized": false,
41
+ "rstrip": false,
42
+ "single_word": false,
43
+ "special": true
44
+ },
45
+ "151648": {
46
+ "content": "<|box_start|>",
47
+ "lstrip": false,
48
+ "normalized": false,
49
+ "rstrip": false,
50
+ "single_word": false,
51
+ "special": true
52
+ },
53
+ "151649": {
54
+ "content": "<|box_end|>",
55
+ "lstrip": false,
56
+ "normalized": false,
57
+ "rstrip": false,
58
+ "single_word": false,
59
+ "special": true
60
+ },
61
+ "151650": {
62
+ "content": "<|quad_start|>",
63
+ "lstrip": false,
64
+ "normalized": false,
65
+ "rstrip": false,
66
+ "single_word": false,
67
+ "special": true
68
+ },
69
+ "151651": {
70
+ "content": "<|quad_end|>",
71
+ "lstrip": false,
72
+ "normalized": false,
73
+ "rstrip": false,
74
+ "single_word": false,
75
+ "special": true
76
+ },
77
+ "151652": {
78
+ "content": "<|vision_start|>",
79
+ "lstrip": false,
80
+ "normalized": false,
81
+ "rstrip": false,
82
+ "single_word": false,
83
+ "special": true
84
+ },
85
+ "151653": {
86
+ "content": "<|vision_end|>",
87
+ "lstrip": false,
88
+ "normalized": false,
89
+ "rstrip": false,
90
+ "single_word": false,
91
+ "special": true
92
+ },
93
+ "151654": {
94
+ "content": "<|vision_pad|>",
95
+ "lstrip": false,
96
+ "normalized": false,
97
+ "rstrip": false,
98
+ "single_word": false,
99
+ "special": true
100
+ },
101
+ "151655": {
102
+ "content": "<|image_pad|>",
103
+ "lstrip": false,
104
+ "normalized": false,
105
+ "rstrip": false,
106
+ "single_word": false,
107
+ "special": true
108
+ },
109
+ "151656": {
110
+ "content": "<|video_pad|>",
111
+ "lstrip": false,
112
+ "normalized": false,
113
+ "rstrip": false,
114
+ "single_word": false,
115
+ "special": true
116
+ },
117
+ "151657": {
118
+ "content": "<tool_call>",
119
+ "lstrip": false,
120
+ "normalized": false,
121
+ "rstrip": false,
122
+ "single_word": false,
123
+ "special": false
124
+ },
125
+ "151658": {
126
+ "content": "</tool_call>",
127
+ "lstrip": false,
128
+ "normalized": false,
129
+ "rstrip": false,
130
+ "single_word": false,
131
+ "special": false
132
+ },
133
+ "151659": {
134
+ "content": "<|fim_prefix|>",
135
+ "lstrip": false,
136
+ "normalized": false,
137
+ "rstrip": false,
138
+ "single_word": false,
139
+ "special": false
140
+ },
141
+ "151660": {
142
+ "content": "<|fim_middle|>",
143
+ "lstrip": false,
144
+ "normalized": false,
145
+ "rstrip": false,
146
+ "single_word": false,
147
+ "special": false
148
+ },
149
+ "151661": {
150
+ "content": "<|fim_suffix|>",
151
+ "lstrip": false,
152
+ "normalized": false,
153
+ "rstrip": false,
154
+ "single_word": false,
155
+ "special": false
156
+ },
157
+ "151662": {
158
+ "content": "<|fim_pad|>",
159
+ "lstrip": false,
160
+ "normalized": false,
161
+ "rstrip": false,
162
+ "single_word": false,
163
+ "special": false
164
+ },
165
+ "151663": {
166
+ "content": "<|repo_name|>",
167
+ "lstrip": false,
168
+ "normalized": false,
169
+ "rstrip": false,
170
+ "single_word": false,
171
+ "special": false
172
+ },
173
+ "151664": {
174
+ "content": "<|file_sep|>",
175
+ "lstrip": false,
176
+ "normalized": false,
177
+ "rstrip": false,
178
+ "single_word": false,
179
+ "special": false
180
+ },
181
+ "151665": {
182
+ "content": "<|trace_context_start|>",
183
+ "lstrip": false,
184
+ "normalized": false,
185
+ "rstrip": false,
186
+ "single_word": false,
187
+ "special": true
188
+ },
189
+ "151666": {
190
+ "content": "<|call_sep|>",
191
+ "lstrip": false,
192
+ "normalized": false,
193
+ "rstrip": false,
194
+ "single_word": false,
195
+ "special": true
196
+ },
197
+ "151667": {
198
+ "content": "<|line_sep|>",
199
+ "lstrip": false,
200
+ "normalized": false,
201
+ "rstrip": false,
202
+ "single_word": false,
203
+ "special": true
204
+ },
205
+ "151668": {
206
+ "content": "<|return_sep|>",
207
+ "lstrip": false,
208
+ "normalized": false,
209
+ "rstrip": false,
210
+ "single_word": false,
211
+ "special": true
212
+ },
213
+ "151669": {
214
+ "content": "<|exception_sep|>",
215
+ "lstrip": false,
216
+ "normalized": false,
217
+ "rstrip": false,
218
+ "single_word": false,
219
+ "special": true
220
+ },
221
+ "151670": {
222
+ "content": "<|action_sep|>",
223
+ "lstrip": false,
224
+ "normalized": false,
225
+ "rstrip": false,
226
+ "single_word": false,
227
+ "special": true
228
+ },
229
+ "151671": {
230
+ "content": "<|arg_sep|>",
231
+ "lstrip": false,
232
+ "normalized": false,
233
+ "rstrip": false,
234
+ "single_word": false,
235
+ "special": true
236
+ },
237
+ "151672": {
238
+ "content": "<|frame_sep|>",
239
+ "lstrip": false,
240
+ "normalized": false,
241
+ "rstrip": false,
242
+ "single_word": false,
243
+ "special": true
244
+ },
245
+ "151673": {
246
+ "content": "<|end_of_text|>",
247
+ "lstrip": false,
248
+ "normalized": false,
249
+ "rstrip": false,
250
+ "single_word": false,
251
+ "special": true
252
+ },
253
+ "151674": {
254
+ "content": "<|latent_start|>",
255
+ "lstrip": false,
256
+ "normalized": false,
257
+ "rstrip": false,
258
+ "single_word": false,
259
+ "special": true
260
+ },
261
+ "151675": {
262
+ "content": "<|latent_end|>",
263
+ "lstrip": false,
264
+ "normalized": false,
265
+ "rstrip": false,
266
+ "single_word": false,
267
+ "special": true
268
+ }
269
+ },
270
+ "additional_special_tokens": [
271
+ "<|im_start|>",
272
+ "<|im_end|>",
273
+ "<|object_ref_start|>",
274
+ "<|object_ref_end|>",
275
+ "<|box_start|>",
276
+ "<|box_end|>",
277
+ "<|quad_start|>",
278
+ "<|quad_end|>",
279
+ "<|vision_start|>",
280
+ "<|vision_end|>",
281
+ "<|vision_pad|>",
282
+ "<|image_pad|>",
283
+ "<|video_pad|>"
284
+ ],
285
+ "bos_token": null,
286
+ "clean_up_tokenization_spaces": false,
287
+ "eos_token": "<|endoftext|>",
288
+ "errors": "replace",
289
+ "extra_special_tokens": {},
290
+ "model_max_length": 32768,
291
+ "pad_token": "<|endoftext|>",
292
+ "split_special_tokens": false,
293
+ "tokenizer_class": "Qwen2Tokenizer",
294
+ "unk_token": null
295
+ }
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-200/trainer_state.json ADDED
@@ -0,0 +1,434 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 0.05767012687427912,
6
+ "eval_steps": 500,
7
+ "global_step": 200,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.0014417531718569781,
14
+ "grad_norm": 52992.0,
15
+ "kd_loss": 0.3193017542362213,
16
+ "learning_rate": 1.3333333333333334e-06,
17
+ "loss": 0.473,
18
+ "step": 5,
19
+ "student_loss": 0.15727384388446808,
20
+ "teacher_loss": 0.0004137983196415007
21
+ },
22
+ {
23
+ "epoch": 0.0028835063437139563,
24
+ "grad_norm": 16121856.0,
25
+ "kd_loss": 0.3087984025478363,
26
+ "learning_rate": 3e-06,
27
+ "loss": 0.466,
28
+ "step": 10,
29
+ "student_loss": 0.2608874440193176,
30
+ "teacher_loss": 0.0007329508662223816
31
+ },
32
+ {
33
+ "epoch": 0.004325259515570935,
34
+ "grad_norm": 2240.0,
35
+ "kd_loss": 0.25935155153274536,
36
+ "learning_rate": 4.666666666666667e-06,
37
+ "loss": 0.4585,
38
+ "step": 15,
39
+ "student_loss": 0.1052168533205986,
40
+ "teacher_loss": 0.000609480426646769
41
+ },
42
+ {
43
+ "epoch": 0.0057670126874279125,
44
+ "grad_norm": 4194304.0,
45
+ "kd_loss": 0.25011566281318665,
46
+ "learning_rate": 6.333333333333333e-06,
47
+ "loss": 0.4707,
48
+ "step": 20,
49
+ "student_loss": 0.15271778404712677,
50
+ "teacher_loss": 0.0002972046786453575
51
+ },
52
+ {
53
+ "epoch": 0.00720876585928489,
54
+ "grad_norm": 41680896.0,
55
+ "kd_loss": 0.2633170485496521,
56
+ "learning_rate": 8.000000000000001e-06,
57
+ "loss": 0.4368,
58
+ "step": 25,
59
+ "student_loss": 0.09109362959861755,
60
+ "teacher_loss": 0.0009600125486031175
61
+ },
62
+ {
63
+ "epoch": 0.00865051903114187,
64
+ "grad_norm": 512.0,
65
+ "kd_loss": 0.2533612847328186,
66
+ "learning_rate": 9.666666666666667e-06,
67
+ "loss": 0.4187,
68
+ "step": 30,
69
+ "student_loss": 0.22675999999046326,
70
+ "teacher_loss": 0.00035076512722298503
71
+ },
72
+ {
73
+ "epoch": 0.010092272202998846,
74
+ "grad_norm": 48.75,
75
+ "kd_loss": 0.2179471105337143,
76
+ "learning_rate": 9.99958042442916e-06,
77
+ "loss": 0.3713,
78
+ "step": 35,
79
+ "student_loss": 0.09486500918865204,
80
+ "teacher_loss": 0.00026503618573769927
81
+ },
82
+ {
83
+ "epoch": 0.011534025374855825,
84
+ "grad_norm": 10.0,
85
+ "kd_loss": 0.17874062061309814,
86
+ "learning_rate": 9.997876019358083e-06,
87
+ "loss": 0.2647,
88
+ "step": 40,
89
+ "student_loss": 0.07879351079463959,
90
+ "teacher_loss": 0.05543303117156029
91
+ },
92
+ {
93
+ "epoch": 0.012975778546712802,
94
+ "grad_norm": 4.9375,
95
+ "kd_loss": 0.19365739822387695,
96
+ "learning_rate": 9.99486100792044e-06,
97
+ "loss": 0.1686,
98
+ "step": 45,
99
+ "student_loss": 0.07774724811315536,
100
+ "teacher_loss": 0.0008653572876937687
101
+ },
102
+ {
103
+ "epoch": 0.01441753171856978,
104
+ "grad_norm": 3.359375,
105
+ "kd_loss": 0.06108914315700531,
106
+ "learning_rate": 9.990536180750724e-06,
107
+ "loss": 0.1396,
108
+ "step": 50,
109
+ "student_loss": 0.008677379228174686,
110
+ "teacher_loss": 0.005272893700748682
111
+ },
112
+ {
113
+ "epoch": 0.015859284890426758,
114
+ "grad_norm": 3.71875,
115
+ "kd_loss": 0.04330182075500488,
116
+ "learning_rate": 9.984902671959911e-06,
117
+ "loss": 0.1065,
118
+ "step": 55,
119
+ "student_loss": 0.001041711075231433,
120
+ "teacher_loss": 0.0005072589847259223
121
+ },
122
+ {
123
+ "epoch": 0.01730103806228374,
124
+ "grad_norm": 4.125,
125
+ "kd_loss": 0.03823668882250786,
126
+ "learning_rate": 9.97796195883804e-06,
127
+ "loss": 0.0862,
128
+ "step": 60,
129
+ "student_loss": 0.0209807138890028,
130
+ "teacher_loss": 0.0038409351836889982
131
+ },
132
+ {
133
+ "epoch": 0.018742791234140715,
134
+ "grad_norm": 10.1875,
135
+ "kd_loss": 0.03922531381249428,
136
+ "learning_rate": 9.969715861466839e-06,
137
+ "loss": 0.0844,
138
+ "step": 65,
139
+ "student_loss": 0.04678678512573242,
140
+ "teacher_loss": 0.05432586744427681
141
+ },
142
+ {
143
+ "epoch": 0.020184544405997693,
144
+ "grad_norm": 2.03125,
145
+ "kd_loss": 0.0549621656537056,
146
+ "learning_rate": 9.96016654224243e-06,
147
+ "loss": 0.0797,
148
+ "step": 70,
149
+ "student_loss": 0.016930116340517998,
150
+ "teacher_loss": 0.0029678302817046642
151
+ },
152
+ {
153
+ "epoch": 0.02162629757785467,
154
+ "grad_norm": 2.5625,
155
+ "kd_loss": 0.023080041632056236,
156
+ "learning_rate": 9.94931650530827e-06,
157
+ "loss": 0.0756,
158
+ "step": 75,
159
+ "student_loss": 0.002210108330473304,
160
+ "teacher_loss": 0.0013834249693900347
161
+ },
162
+ {
163
+ "epoch": 0.02306805074971165,
164
+ "grad_norm": 2.296875,
165
+ "kd_loss": 0.025744518265128136,
166
+ "learning_rate": 9.93716859589851e-06,
167
+ "loss": 0.0644,
168
+ "step": 80,
169
+ "student_loss": 0.0015430074417963624,
170
+ "teacher_loss": 0.000761081581003964
171
+ },
172
+ {
173
+ "epoch": 0.024509803921568627,
174
+ "grad_norm": 3.125,
175
+ "kd_loss": 0.023152953013777733,
176
+ "learning_rate": 9.923725999591846e-06,
177
+ "loss": 0.0609,
178
+ "step": 85,
179
+ "student_loss": 0.0187687948346138,
180
+ "teacher_loss": 0.011518553830683231
181
+ },
182
+ {
183
+ "epoch": 0.025951557093425604,
184
+ "grad_norm": 1.8359375,
185
+ "kd_loss": 0.06841632723808289,
186
+ "learning_rate": 9.908992241476189e-06,
187
+ "loss": 0.0622,
188
+ "step": 90,
189
+ "student_loss": 0.05537872016429901,
190
+ "teacher_loss": 0.011890009045600891
191
+ },
192
+ {
193
+ "epoch": 0.027393310265282585,
194
+ "grad_norm": 4.21875,
195
+ "kd_loss": 0.041227906942367554,
196
+ "learning_rate": 9.892971185224244e-06,
197
+ "loss": 0.0626,
198
+ "step": 95,
199
+ "student_loss": 0.009465624578297138,
200
+ "teacher_loss": 0.005057212430983782
201
+ },
202
+ {
203
+ "epoch": 0.02883506343713956,
204
+ "grad_norm": 1.8359375,
205
+ "kd_loss": 0.021006152033805847,
206
+ "learning_rate": 9.875667032080354e-06,
207
+ "loss": 0.0538,
208
+ "step": 100,
209
+ "student_loss": 0.0007415671134367585,
210
+ "teacher_loss": 0.0005348982522264123
211
+ },
212
+ {
213
+ "epoch": 0.03027681660899654,
214
+ "grad_norm": 1.6328125,
215
+ "kd_loss": 0.035115621984004974,
216
+ "learning_rate": 9.857084319758772e-06,
217
+ "loss": 0.0541,
218
+ "step": 105,
219
+ "student_loss": 0.005187581293284893,
220
+ "teacher_loss": 0.00030648778192698956
221
+ },
222
+ {
223
+ "epoch": 0.031718569780853516,
224
+ "grad_norm": 2.953125,
225
+ "kd_loss": 0.03037514165043831,
226
+ "learning_rate": 9.837227921253747e-06,
227
+ "loss": 0.06,
228
+ "step": 110,
229
+ "student_loss": 0.031953420490026474,
230
+ "teacher_loss": 0.0175339188426733
231
+ },
232
+ {
233
+ "epoch": 0.03316032295271049,
234
+ "grad_norm": 1.484375,
235
+ "kd_loss": 0.023421702906489372,
236
+ "learning_rate": 9.816103043561648e-06,
237
+ "loss": 0.0563,
238
+ "step": 115,
239
+ "student_loss": 0.0006847245967946947,
240
+ "teacher_loss": 0.0003864571626763791
241
+ },
242
+ {
243
+ "epoch": 0.03460207612456748,
244
+ "grad_norm": 1.8203125,
245
+ "kd_loss": 0.018447211012244225,
246
+ "learning_rate": 9.79371522631553e-06,
247
+ "loss": 0.0478,
248
+ "step": 120,
249
+ "student_loss": 0.0015498347347602248,
250
+ "teacher_loss": 0.0010675351368263364
251
+ },
252
+ {
253
+ "epoch": 0.036043829296424454,
254
+ "grad_norm": 1.96875,
255
+ "kd_loss": 0.05630031228065491,
256
+ "learning_rate": 9.770070340332457e-06,
257
+ "loss": 0.0527,
258
+ "step": 125,
259
+ "student_loss": 0.04758762940764427,
260
+ "teacher_loss": 0.0004967509885318577
261
+ },
262
+ {
263
+ "epoch": 0.03748558246828143,
264
+ "grad_norm": 1.859375,
265
+ "kd_loss": 0.018152693286538124,
266
+ "learning_rate": 9.745174586073982e-06,
267
+ "loss": 0.0518,
268
+ "step": 130,
269
+ "student_loss": 0.0008988101035356522,
270
+ "teacher_loss": 0.0007278143893927336
271
+ },
272
+ {
273
+ "epoch": 0.03892733564013841,
274
+ "grad_norm": 1.6640625,
275
+ "kd_loss": 0.030904680490493774,
276
+ "learning_rate": 9.719034492020183e-06,
277
+ "loss": 0.0526,
278
+ "step": 135,
279
+ "student_loss": 0.0127235297113657,
280
+ "teacher_loss": 0.0011665972415357828
281
+ },
282
+ {
283
+ "epoch": 0.040369088811995385,
284
+ "grad_norm": 1.1640625,
285
+ "kd_loss": 0.01759810373187065,
286
+ "learning_rate": 9.691656912957686e-06,
287
+ "loss": 0.0494,
288
+ "step": 140,
289
+ "student_loss": 0.0008300879271700978,
290
+ "teacher_loss": 0.0005129938363097608
291
+ },
292
+ {
293
+ "epoch": 0.04181084198385236,
294
+ "grad_norm": 1.0234375,
295
+ "kd_loss": 0.061179179698228836,
296
+ "learning_rate": 9.663049028182112e-06,
297
+ "loss": 0.0457,
298
+ "step": 145,
299
+ "student_loss": 0.04404008761048317,
300
+ "teacher_loss": 0.009324059821665287
301
+ },
302
+ {
303
+ "epoch": 0.04325259515570934,
304
+ "grad_norm": 1.15625,
305
+ "kd_loss": 0.01291500125080347,
306
+ "learning_rate": 9.633218339615433e-06,
307
+ "loss": 0.0455,
308
+ "step": 150,
309
+ "student_loss": 0.0008404091349802911,
310
+ "teacher_loss": 0.0008450732566416264
311
+ },
312
+ {
313
+ "epoch": 0.04469434832756632,
314
+ "grad_norm": 1.4921875,
315
+ "kd_loss": 0.017479021102190018,
316
+ "learning_rate": 9.602172669838721e-06,
317
+ "loss": 0.0575,
318
+ "step": 155,
319
+ "student_loss": 0.0009151780977845192,
320
+ "teacher_loss": 0.0006974826683290303
321
+ },
322
+ {
323
+ "epoch": 0.0461361014994233,
324
+ "grad_norm": 1.671875,
325
+ "kd_loss": 0.022708255797624588,
326
+ "learning_rate": 9.569920160040815e-06,
327
+ "loss": 0.0503,
328
+ "step": 160,
329
+ "student_loss": 0.07660368084907532,
330
+ "teacher_loss": 0.04039459675550461
331
+ },
332
+ {
333
+ "epoch": 0.04757785467128028,
334
+ "grad_norm": 1.0703125,
335
+ "kd_loss": 0.027437202632427216,
336
+ "learning_rate": 9.536469267883432e-06,
337
+ "loss": 0.0451,
338
+ "step": 165,
339
+ "student_loss": 0.0005943832220509648,
340
+ "teacher_loss": 0.005336699075996876
341
+ },
342
+ {
343
+ "epoch": 0.049019607843137254,
344
+ "grad_norm": 1.71875,
345
+ "kd_loss": 0.014546042308211327,
346
+ "learning_rate": 9.501828765283295e-06,
347
+ "loss": 0.0459,
348
+ "step": 170,
349
+ "student_loss": 0.0006958618760108948,
350
+ "teacher_loss": 0.0005218263831920922
351
+ },
352
+ {
353
+ "epoch": 0.05046136101499423,
354
+ "grad_norm": 1.4453125,
355
+ "kd_loss": 0.02377346158027649,
356
+ "learning_rate": 9.466007736111846e-06,
357
+ "loss": 0.0465,
358
+ "step": 175,
359
+ "student_loss": 0.003214897122234106,
360
+ "teacher_loss": 0.0006713470793329179
361
+ },
362
+ {
363
+ "epoch": 0.05190311418685121,
364
+ "grad_norm": 1.296875,
365
+ "kd_loss": 0.02517828531563282,
366
+ "learning_rate": 9.429015573813163e-06,
367
+ "loss": 0.0454,
368
+ "step": 180,
369
+ "student_loss": 0.0007039233460091054,
370
+ "teacher_loss": 0.0003641469811554998
371
+ },
372
+ {
373
+ "epoch": 0.05334486735870819,
374
+ "grad_norm": 1.25,
375
+ "kd_loss": 0.029188504442572594,
376
+ "learning_rate": 9.390861978940687e-06,
377
+ "loss": 0.0473,
378
+ "step": 185,
379
+ "student_loss": 0.04238048940896988,
380
+ "teacher_loss": 0.00033479384728707373
381
+ },
382
+ {
383
+ "epoch": 0.05478662053056517,
384
+ "grad_norm": 1.484375,
385
+ "kd_loss": 0.03749082237482071,
386
+ "learning_rate": 9.351556956613423e-06,
387
+ "loss": 0.0436,
388
+ "step": 190,
389
+ "student_loss": 0.010270334780216217,
390
+ "teacher_loss": 0.0024021982681006193
391
+ },
392
+ {
393
+ "epoch": 0.056228373702422146,
394
+ "grad_norm": 2.0,
395
+ "kd_loss": 0.030336912721395493,
396
+ "learning_rate": 9.31111081389227e-06,
397
+ "loss": 0.0442,
398
+ "step": 195,
399
+ "student_loss": 0.010534510016441345,
400
+ "teacher_loss": 0.0003890562220476568
401
+ },
402
+ {
403
+ "epoch": 0.05767012687427912,
404
+ "grad_norm": 0.921875,
405
+ "kd_loss": 0.026554003357887268,
406
+ "learning_rate": 9.269534157077177e-06,
407
+ "loss": 0.0497,
408
+ "step": 200,
409
+ "student_loss": 0.0034760867711156607,
410
+ "teacher_loss": 0.00022423896007239819
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/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-200/vocab.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-400/added_tokens.json ADDED
@@ -0,0 +1,35 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "</tool_call>": 151658,
3
+ "<tool_call>": 151657,
4
+ "<|action_sep|>": 151670,
5
+ "<|arg_sep|>": 151671,
6
+ "<|box_end|>": 151649,
7
+ "<|box_start|>": 151648,
8
+ "<|call_sep|>": 151666,
9
+ "<|end_of_text|>": 151673,
10
+ "<|endoftext|>": 151643,
11
+ "<|exception_sep|>": 151669,
12
+ "<|file_sep|>": 151664,
13
+ "<|fim_middle|>": 151660,
14
+ "<|fim_pad|>": 151662,
15
+ "<|fim_prefix|>": 151659,
16
+ "<|fim_suffix|>": 151661,
17
+ "<|frame_sep|>": 151672,
18
+ "<|im_end|>": 151645,
19
+ "<|im_start|>": 151644,
20
+ "<|image_pad|>": 151655,
21
+ "<|latent_end|>": 151675,
22
+ "<|latent_start|>": 151674,
23
+ "<|line_sep|>": 151667,
24
+ "<|object_ref_end|>": 151647,
25
+ "<|object_ref_start|>": 151646,
26
+ "<|quad_end|>": 151651,
27
+ "<|quad_start|>": 151650,
28
+ "<|repo_name|>": 151663,
29
+ "<|return_sep|>": 151668,
30
+ "<|trace_context_start|>": 151665,
31
+ "<|video_pad|>": 151656,
32
+ "<|vision_end|>": 151653,
33
+ "<|vision_pad|>": 151654,
34
+ "<|vision_start|>": 151652
35
+ }
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-400/chat_template.jinja ADDED
@@ -0,0 +1,54 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {%- if tools %}
2
+ {{- '<|im_start|>system\n' }}
3
+ {%- if messages[0]['role'] == 'system' %}
4
+ {{- messages[0]['content'] }}
5
+ {%- else %}
6
+ {{- 'You are Qwen, created by Alibaba Cloud. You are a helpful assistant.' }}
7
+ {%- endif %}
8
+ {{- "\n\n# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
9
+ {%- for tool in tools %}
10
+ {{- "\n" }}
11
+ {{- tool | tojson }}
12
+ {%- endfor %}
13
+ {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
14
+ {%- else %}
15
+ {%- if messages[0]['role'] == 'system' %}
16
+ {{- '<|im_start|>system\n' + messages[0]['content'] + '<|im_end|>\n' }}
17
+ {%- else %}
18
+ {{- '<|im_start|>system\nYou are Qwen, created by Alibaba Cloud. You are a helpful assistant.<|im_end|>\n' }}
19
+ {%- endif %}
20
+ {%- endif %}
21
+ {%- for message in messages %}
22
+ {%- if (message.role == "user") or (message.role == "system" and not loop.first) or (message.role == "assistant" and not message.tool_calls) %}
23
+ {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
24
+ {%- elif message.role == "assistant" %}
25
+ {{- '<|im_start|>' + message.role }}
26
+ {%- if message.content %}
27
+ {{- '\n' + message.content }}
28
+ {%- endif %}
29
+ {%- for tool_call in message.tool_calls %}
30
+ {%- if tool_call.function is defined %}
31
+ {%- set tool_call = tool_call.function %}
32
+ {%- endif %}
33
+ {{- '\n<tool_call>\n{"name": "' }}
34
+ {{- tool_call.name }}
35
+ {{- '", "arguments": ' }}
36
+ {{- tool_call.arguments | tojson }}
37
+ {{- '}\n</tool_call>' }}
38
+ {%- endfor %}
39
+ {{- '<|im_end|>\n' }}
40
+ {%- elif message.role == "tool" %}
41
+ {%- if (loop.index0 == 0) or (messages[loop.index0 - 1].role != "tool") %}
42
+ {{- '<|im_start|>user' }}
43
+ {%- endif %}
44
+ {{- '\n<tool_response>\n' }}
45
+ {{- message.content }}
46
+ {{- '\n</tool_response>' }}
47
+ {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
48
+ {{- '<|im_end|>\n' }}
49
+ {%- endif %}
50
+ {%- endif %}
51
+ {%- endfor %}
52
+ {%- if add_generation_prompt %}
53
+ {{- '<|im_start|>assistant\n' }}
54
+ {%- endif %}
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-400/config.json ADDED
@@ -0,0 +1,58 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Qwen2ForCausalLM"
4
+ ],
5
+ "attention_dropout": 0.0,
6
+ "dtype": "bfloat16",
7
+ "eos_token_id": 151643,
8
+ "hidden_act": "silu",
9
+ "hidden_size": 1536,
10
+ "initializer_range": 0.02,
11
+ "intermediate_size": 8960,
12
+ "layer_types": [
13
+ "full_attention",
14
+ "full_attention",
15
+ "full_attention",
16
+ "full_attention",
17
+ "full_attention",
18
+ "full_attention",
19
+ "full_attention",
20
+ "full_attention",
21
+ "full_attention",
22
+ "full_attention",
23
+ "full_attention",
24
+ "full_attention",
25
+ "full_attention",
26
+ "full_attention",
27
+ "full_attention",
28
+ "full_attention",
29
+ "full_attention",
30
+ "full_attention",
31
+ "full_attention",
32
+ "full_attention",
33
+ "full_attention",
34
+ "full_attention",
35
+ "full_attention",
36
+ "full_attention",
37
+ "full_attention",
38
+ "full_attention",
39
+ "full_attention",
40
+ "full_attention"
41
+ ],
42
+ "max_position_embeddings": 32768,
43
+ "max_window_layers": 28,
44
+ "model_type": "qwen2",
45
+ "num_attention_heads": 12,
46
+ "num_hidden_layers": 28,
47
+ "num_key_value_heads": 2,
48
+ "pad_token_id": 151643,
49
+ "rms_norm_eps": 1e-06,
50
+ "rope_scaling": null,
51
+ "rope_theta": 1000000.0,
52
+ "sliding_window": null,
53
+ "tie_word_embeddings": true,
54
+ "transformers_version": "4.57.6",
55
+ "use_cache": true,
56
+ "use_sliding_window": false,
57
+ "vocab_size": 151676
58
+ }
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-400/merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-400/pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:99416fef5a180a223d39281c8652c677e98b3f665130cbcf397aa051ef9253a2
3
+ size 3096275515
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-400/special_tokens_map.json ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "additional_special_tokens": [
3
+ "<|im_start|>",
4
+ "<|im_end|>",
5
+ "<|object_ref_start|>",
6
+ "<|object_ref_end|>",
7
+ "<|box_start|>",
8
+ "<|box_end|>",
9
+ "<|quad_start|>",
10
+ "<|quad_end|>",
11
+ "<|vision_start|>",
12
+ "<|vision_end|>",
13
+ "<|vision_pad|>",
14
+ "<|image_pad|>",
15
+ "<|video_pad|>"
16
+ ],
17
+ "eos_token": {
18
+ "content": "<|endoftext|>",
19
+ "lstrip": false,
20
+ "normalized": false,
21
+ "rstrip": false,
22
+ "single_word": false
23
+ },
24
+ "pad_token": {
25
+ "content": "<|endoftext|>",
26
+ "lstrip": false,
27
+ "normalized": false,
28
+ "rstrip": false,
29
+ "single_word": false
30
+ }
31
+ }
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-400/thought_projector.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4d96e186b0d1d947b9da71c4baff93cdf3c8f2746fae25062cf9ad1925b99b38
3
+ size 9445953
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-400/tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:83a790d654474f5dfe225f889afd0210313eb1083f942671f2c4b8e95a1c922b
3
+ size 11424004
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-400/tokenizer_config.json ADDED
@@ -0,0 +1,295 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_bos_token": false,
3
+ "add_prefix_space": false,
4
+ "added_tokens_decoder": {
5
+ "151643": {
6
+ "content": "<|endoftext|>",
7
+ "lstrip": false,
8
+ "normalized": false,
9
+ "rstrip": false,
10
+ "single_word": false,
11
+ "special": true
12
+ },
13
+ "151644": {
14
+ "content": "<|im_start|>",
15
+ "lstrip": false,
16
+ "normalized": false,
17
+ "rstrip": false,
18
+ "single_word": false,
19
+ "special": true
20
+ },
21
+ "151645": {
22
+ "content": "<|im_end|>",
23
+ "lstrip": false,
24
+ "normalized": false,
25
+ "rstrip": false,
26
+ "single_word": false,
27
+ "special": true
28
+ },
29
+ "151646": {
30
+ "content": "<|object_ref_start|>",
31
+ "lstrip": false,
32
+ "normalized": false,
33
+ "rstrip": false,
34
+ "single_word": false,
35
+ "special": true
36
+ },
37
+ "151647": {
38
+ "content": "<|object_ref_end|>",
39
+ "lstrip": false,
40
+ "normalized": false,
41
+ "rstrip": false,
42
+ "single_word": false,
43
+ "special": true
44
+ },
45
+ "151648": {
46
+ "content": "<|box_start|>",
47
+ "lstrip": false,
48
+ "normalized": false,
49
+ "rstrip": false,
50
+ "single_word": false,
51
+ "special": true
52
+ },
53
+ "151649": {
54
+ "content": "<|box_end|>",
55
+ "lstrip": false,
56
+ "normalized": false,
57
+ "rstrip": false,
58
+ "single_word": false,
59
+ "special": true
60
+ },
61
+ "151650": {
62
+ "content": "<|quad_start|>",
63
+ "lstrip": false,
64
+ "normalized": false,
65
+ "rstrip": false,
66
+ "single_word": false,
67
+ "special": true
68
+ },
69
+ "151651": {
70
+ "content": "<|quad_end|>",
71
+ "lstrip": false,
72
+ "normalized": false,
73
+ "rstrip": false,
74
+ "single_word": false,
75
+ "special": true
76
+ },
77
+ "151652": {
78
+ "content": "<|vision_start|>",
79
+ "lstrip": false,
80
+ "normalized": false,
81
+ "rstrip": false,
82
+ "single_word": false,
83
+ "special": true
84
+ },
85
+ "151653": {
86
+ "content": "<|vision_end|>",
87
+ "lstrip": false,
88
+ "normalized": false,
89
+ "rstrip": false,
90
+ "single_word": false,
91
+ "special": true
92
+ },
93
+ "151654": {
94
+ "content": "<|vision_pad|>",
95
+ "lstrip": false,
96
+ "normalized": false,
97
+ "rstrip": false,
98
+ "single_word": false,
99
+ "special": true
100
+ },
101
+ "151655": {
102
+ "content": "<|image_pad|>",
103
+ "lstrip": false,
104
+ "normalized": false,
105
+ "rstrip": false,
106
+ "single_word": false,
107
+ "special": true
108
+ },
109
+ "151656": {
110
+ "content": "<|video_pad|>",
111
+ "lstrip": false,
112
+ "normalized": false,
113
+ "rstrip": false,
114
+ "single_word": false,
115
+ "special": true
116
+ },
117
+ "151657": {
118
+ "content": "<tool_call>",
119
+ "lstrip": false,
120
+ "normalized": false,
121
+ "rstrip": false,
122
+ "single_word": false,
123
+ "special": false
124
+ },
125
+ "151658": {
126
+ "content": "</tool_call>",
127
+ "lstrip": false,
128
+ "normalized": false,
129
+ "rstrip": false,
130
+ "single_word": false,
131
+ "special": false
132
+ },
133
+ "151659": {
134
+ "content": "<|fim_prefix|>",
135
+ "lstrip": false,
136
+ "normalized": false,
137
+ "rstrip": false,
138
+ "single_word": false,
139
+ "special": false
140
+ },
141
+ "151660": {
142
+ "content": "<|fim_middle|>",
143
+ "lstrip": false,
144
+ "normalized": false,
145
+ "rstrip": false,
146
+ "single_word": false,
147
+ "special": false
148
+ },
149
+ "151661": {
150
+ "content": "<|fim_suffix|>",
151
+ "lstrip": false,
152
+ "normalized": false,
153
+ "rstrip": false,
154
+ "single_word": false,
155
+ "special": false
156
+ },
157
+ "151662": {
158
+ "content": "<|fim_pad|>",
159
+ "lstrip": false,
160
+ "normalized": false,
161
+ "rstrip": false,
162
+ "single_word": false,
163
+ "special": false
164
+ },
165
+ "151663": {
166
+ "content": "<|repo_name|>",
167
+ "lstrip": false,
168
+ "normalized": false,
169
+ "rstrip": false,
170
+ "single_word": false,
171
+ "special": false
172
+ },
173
+ "151664": {
174
+ "content": "<|file_sep|>",
175
+ "lstrip": false,
176
+ "normalized": false,
177
+ "rstrip": false,
178
+ "single_word": false,
179
+ "special": false
180
+ },
181
+ "151665": {
182
+ "content": "<|trace_context_start|>",
183
+ "lstrip": false,
184
+ "normalized": false,
185
+ "rstrip": false,
186
+ "single_word": false,
187
+ "special": true
188
+ },
189
+ "151666": {
190
+ "content": "<|call_sep|>",
191
+ "lstrip": false,
192
+ "normalized": false,
193
+ "rstrip": false,
194
+ "single_word": false,
195
+ "special": true
196
+ },
197
+ "151667": {
198
+ "content": "<|line_sep|>",
199
+ "lstrip": false,
200
+ "normalized": false,
201
+ "rstrip": false,
202
+ "single_word": false,
203
+ "special": true
204
+ },
205
+ "151668": {
206
+ "content": "<|return_sep|>",
207
+ "lstrip": false,
208
+ "normalized": false,
209
+ "rstrip": false,
210
+ "single_word": false,
211
+ "special": true
212
+ },
213
+ "151669": {
214
+ "content": "<|exception_sep|>",
215
+ "lstrip": false,
216
+ "normalized": false,
217
+ "rstrip": false,
218
+ "single_word": false,
219
+ "special": true
220
+ },
221
+ "151670": {
222
+ "content": "<|action_sep|>",
223
+ "lstrip": false,
224
+ "normalized": false,
225
+ "rstrip": false,
226
+ "single_word": false,
227
+ "special": true
228
+ },
229
+ "151671": {
230
+ "content": "<|arg_sep|>",
231
+ "lstrip": false,
232
+ "normalized": false,
233
+ "rstrip": false,
234
+ "single_word": false,
235
+ "special": true
236
+ },
237
+ "151672": {
238
+ "content": "<|frame_sep|>",
239
+ "lstrip": false,
240
+ "normalized": false,
241
+ "rstrip": false,
242
+ "single_word": false,
243
+ "special": true
244
+ },
245
+ "151673": {
246
+ "content": "<|end_of_text|>",
247
+ "lstrip": false,
248
+ "normalized": false,
249
+ "rstrip": false,
250
+ "single_word": false,
251
+ "special": true
252
+ },
253
+ "151674": {
254
+ "content": "<|latent_start|>",
255
+ "lstrip": false,
256
+ "normalized": false,
257
+ "rstrip": false,
258
+ "single_word": false,
259
+ "special": true
260
+ },
261
+ "151675": {
262
+ "content": "<|latent_end|>",
263
+ "lstrip": false,
264
+ "normalized": false,
265
+ "rstrip": false,
266
+ "single_word": false,
267
+ "special": true
268
+ }
269
+ },
270
+ "additional_special_tokens": [
271
+ "<|im_start|>",
272
+ "<|im_end|>",
273
+ "<|object_ref_start|>",
274
+ "<|object_ref_end|>",
275
+ "<|box_start|>",
276
+ "<|box_end|>",
277
+ "<|quad_start|>",
278
+ "<|quad_end|>",
279
+ "<|vision_start|>",
280
+ "<|vision_end|>",
281
+ "<|vision_pad|>",
282
+ "<|image_pad|>",
283
+ "<|video_pad|>"
284
+ ],
285
+ "bos_token": null,
286
+ "clean_up_tokenization_spaces": false,
287
+ "eos_token": "<|endoftext|>",
288
+ "errors": "replace",
289
+ "extra_special_tokens": {},
290
+ "model_max_length": 32768,
291
+ "pad_token": "<|endoftext|>",
292
+ "split_special_tokens": false,
293
+ "tokenizer_class": "Qwen2Tokenizer",
294
+ "unk_token": null
295
+ }
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-400/trainer_state.json ADDED
@@ -0,0 +1,834 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 0.11534025374855825,
6
+ "eval_steps": 500,
7
+ "global_step": 400,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.0014417531718569781,
14
+ "grad_norm": 52992.0,
15
+ "kd_loss": 0.3193017542362213,
16
+ "learning_rate": 1.3333333333333334e-06,
17
+ "loss": 0.473,
18
+ "step": 5,
19
+ "student_loss": 0.15727384388446808,
20
+ "teacher_loss": 0.0004137983196415007
21
+ },
22
+ {
23
+ "epoch": 0.0028835063437139563,
24
+ "grad_norm": 16121856.0,
25
+ "kd_loss": 0.3087984025478363,
26
+ "learning_rate": 3e-06,
27
+ "loss": 0.466,
28
+ "step": 10,
29
+ "student_loss": 0.2608874440193176,
30
+ "teacher_loss": 0.0007329508662223816
31
+ },
32
+ {
33
+ "epoch": 0.004325259515570935,
34
+ "grad_norm": 2240.0,
35
+ "kd_loss": 0.25935155153274536,
36
+ "learning_rate": 4.666666666666667e-06,
37
+ "loss": 0.4585,
38
+ "step": 15,
39
+ "student_loss": 0.1052168533205986,
40
+ "teacher_loss": 0.000609480426646769
41
+ },
42
+ {
43
+ "epoch": 0.0057670126874279125,
44
+ "grad_norm": 4194304.0,
45
+ "kd_loss": 0.25011566281318665,
46
+ "learning_rate": 6.333333333333333e-06,
47
+ "loss": 0.4707,
48
+ "step": 20,
49
+ "student_loss": 0.15271778404712677,
50
+ "teacher_loss": 0.0002972046786453575
51
+ },
52
+ {
53
+ "epoch": 0.00720876585928489,
54
+ "grad_norm": 41680896.0,
55
+ "kd_loss": 0.2633170485496521,
56
+ "learning_rate": 8.000000000000001e-06,
57
+ "loss": 0.4368,
58
+ "step": 25,
59
+ "student_loss": 0.09109362959861755,
60
+ "teacher_loss": 0.0009600125486031175
61
+ },
62
+ {
63
+ "epoch": 0.00865051903114187,
64
+ "grad_norm": 512.0,
65
+ "kd_loss": 0.2533612847328186,
66
+ "learning_rate": 9.666666666666667e-06,
67
+ "loss": 0.4187,
68
+ "step": 30,
69
+ "student_loss": 0.22675999999046326,
70
+ "teacher_loss": 0.00035076512722298503
71
+ },
72
+ {
73
+ "epoch": 0.010092272202998846,
74
+ "grad_norm": 48.75,
75
+ "kd_loss": 0.2179471105337143,
76
+ "learning_rate": 9.99958042442916e-06,
77
+ "loss": 0.3713,
78
+ "step": 35,
79
+ "student_loss": 0.09486500918865204,
80
+ "teacher_loss": 0.00026503618573769927
81
+ },
82
+ {
83
+ "epoch": 0.011534025374855825,
84
+ "grad_norm": 10.0,
85
+ "kd_loss": 0.17874062061309814,
86
+ "learning_rate": 9.997876019358083e-06,
87
+ "loss": 0.2647,
88
+ "step": 40,
89
+ "student_loss": 0.07879351079463959,
90
+ "teacher_loss": 0.05543303117156029
91
+ },
92
+ {
93
+ "epoch": 0.012975778546712802,
94
+ "grad_norm": 4.9375,
95
+ "kd_loss": 0.19365739822387695,
96
+ "learning_rate": 9.99486100792044e-06,
97
+ "loss": 0.1686,
98
+ "step": 45,
99
+ "student_loss": 0.07774724811315536,
100
+ "teacher_loss": 0.0008653572876937687
101
+ },
102
+ {
103
+ "epoch": 0.01441753171856978,
104
+ "grad_norm": 3.359375,
105
+ "kd_loss": 0.06108914315700531,
106
+ "learning_rate": 9.990536180750724e-06,
107
+ "loss": 0.1396,
108
+ "step": 50,
109
+ "student_loss": 0.008677379228174686,
110
+ "teacher_loss": 0.005272893700748682
111
+ },
112
+ {
113
+ "epoch": 0.015859284890426758,
114
+ "grad_norm": 3.71875,
115
+ "kd_loss": 0.04330182075500488,
116
+ "learning_rate": 9.984902671959911e-06,
117
+ "loss": 0.1065,
118
+ "step": 55,
119
+ "student_loss": 0.001041711075231433,
120
+ "teacher_loss": 0.0005072589847259223
121
+ },
122
+ {
123
+ "epoch": 0.01730103806228374,
124
+ "grad_norm": 4.125,
125
+ "kd_loss": 0.03823668882250786,
126
+ "learning_rate": 9.97796195883804e-06,
127
+ "loss": 0.0862,
128
+ "step": 60,
129
+ "student_loss": 0.0209807138890028,
130
+ "teacher_loss": 0.0038409351836889982
131
+ },
132
+ {
133
+ "epoch": 0.018742791234140715,
134
+ "grad_norm": 10.1875,
135
+ "kd_loss": 0.03922531381249428,
136
+ "learning_rate": 9.969715861466839e-06,
137
+ "loss": 0.0844,
138
+ "step": 65,
139
+ "student_loss": 0.04678678512573242,
140
+ "teacher_loss": 0.05432586744427681
141
+ },
142
+ {
143
+ "epoch": 0.020184544405997693,
144
+ "grad_norm": 2.03125,
145
+ "kd_loss": 0.0549621656537056,
146
+ "learning_rate": 9.96016654224243e-06,
147
+ "loss": 0.0797,
148
+ "step": 70,
149
+ "student_loss": 0.016930116340517998,
150
+ "teacher_loss": 0.0029678302817046642
151
+ },
152
+ {
153
+ "epoch": 0.02162629757785467,
154
+ "grad_norm": 2.5625,
155
+ "kd_loss": 0.023080041632056236,
156
+ "learning_rate": 9.94931650530827e-06,
157
+ "loss": 0.0756,
158
+ "step": 75,
159
+ "student_loss": 0.002210108330473304,
160
+ "teacher_loss": 0.0013834249693900347
161
+ },
162
+ {
163
+ "epoch": 0.02306805074971165,
164
+ "grad_norm": 2.296875,
165
+ "kd_loss": 0.025744518265128136,
166
+ "learning_rate": 9.93716859589851e-06,
167
+ "loss": 0.0644,
168
+ "step": 80,
169
+ "student_loss": 0.0015430074417963624,
170
+ "teacher_loss": 0.000761081581003964
171
+ },
172
+ {
173
+ "epoch": 0.024509803921568627,
174
+ "grad_norm": 3.125,
175
+ "kd_loss": 0.023152953013777733,
176
+ "learning_rate": 9.923725999591846e-06,
177
+ "loss": 0.0609,
178
+ "step": 85,
179
+ "student_loss": 0.0187687948346138,
180
+ "teacher_loss": 0.011518553830683231
181
+ },
182
+ {
183
+ "epoch": 0.025951557093425604,
184
+ "grad_norm": 1.8359375,
185
+ "kd_loss": 0.06841632723808289,
186
+ "learning_rate": 9.908992241476189e-06,
187
+ "loss": 0.0622,
188
+ "step": 90,
189
+ "student_loss": 0.05537872016429901,
190
+ "teacher_loss": 0.011890009045600891
191
+ },
192
+ {
193
+ "epoch": 0.027393310265282585,
194
+ "grad_norm": 4.21875,
195
+ "kd_loss": 0.041227906942367554,
196
+ "learning_rate": 9.892971185224244e-06,
197
+ "loss": 0.0626,
198
+ "step": 95,
199
+ "student_loss": 0.009465624578297138,
200
+ "teacher_loss": 0.005057212430983782
201
+ },
202
+ {
203
+ "epoch": 0.02883506343713956,
204
+ "grad_norm": 1.8359375,
205
+ "kd_loss": 0.021006152033805847,
206
+ "learning_rate": 9.875667032080354e-06,
207
+ "loss": 0.0538,
208
+ "step": 100,
209
+ "student_loss": 0.0007415671134367585,
210
+ "teacher_loss": 0.0005348982522264123
211
+ },
212
+ {
213
+ "epoch": 0.03027681660899654,
214
+ "grad_norm": 1.6328125,
215
+ "kd_loss": 0.035115621984004974,
216
+ "learning_rate": 9.857084319758772e-06,
217
+ "loss": 0.0541,
218
+ "step": 105,
219
+ "student_loss": 0.005187581293284893,
220
+ "teacher_loss": 0.00030648778192698956
221
+ },
222
+ {
223
+ "epoch": 0.031718569780853516,
224
+ "grad_norm": 2.953125,
225
+ "kd_loss": 0.03037514165043831,
226
+ "learning_rate": 9.837227921253747e-06,
227
+ "loss": 0.06,
228
+ "step": 110,
229
+ "student_loss": 0.031953420490026474,
230
+ "teacher_loss": 0.0175339188426733
231
+ },
232
+ {
233
+ "epoch": 0.03316032295271049,
234
+ "grad_norm": 1.484375,
235
+ "kd_loss": 0.023421702906489372,
236
+ "learning_rate": 9.816103043561648e-06,
237
+ "loss": 0.0563,
238
+ "step": 115,
239
+ "student_loss": 0.0006847245967946947,
240
+ "teacher_loss": 0.0003864571626763791
241
+ },
242
+ {
243
+ "epoch": 0.03460207612456748,
244
+ "grad_norm": 1.8203125,
245
+ "kd_loss": 0.018447211012244225,
246
+ "learning_rate": 9.79371522631553e-06,
247
+ "loss": 0.0478,
248
+ "step": 120,
249
+ "student_loss": 0.0015498347347602248,
250
+ "teacher_loss": 0.0010675351368263364
251
+ },
252
+ {
253
+ "epoch": 0.036043829296424454,
254
+ "grad_norm": 1.96875,
255
+ "kd_loss": 0.05630031228065491,
256
+ "learning_rate": 9.770070340332457e-06,
257
+ "loss": 0.0527,
258
+ "step": 125,
259
+ "student_loss": 0.04758762940764427,
260
+ "teacher_loss": 0.0004967509885318577
261
+ },
262
+ {
263
+ "epoch": 0.03748558246828143,
264
+ "grad_norm": 1.859375,
265
+ "kd_loss": 0.018152693286538124,
266
+ "learning_rate": 9.745174586073982e-06,
267
+ "loss": 0.0518,
268
+ "step": 130,
269
+ "student_loss": 0.0008988101035356522,
270
+ "teacher_loss": 0.0007278143893927336
271
+ },
272
+ {
273
+ "epoch": 0.03892733564013841,
274
+ "grad_norm": 1.6640625,
275
+ "kd_loss": 0.030904680490493774,
276
+ "learning_rate": 9.719034492020183e-06,
277
+ "loss": 0.0526,
278
+ "step": 135,
279
+ "student_loss": 0.0127235297113657,
280
+ "teacher_loss": 0.0011665972415357828
281
+ },
282
+ {
283
+ "epoch": 0.040369088811995385,
284
+ "grad_norm": 1.1640625,
285
+ "kd_loss": 0.01759810373187065,
286
+ "learning_rate": 9.691656912957686e-06,
287
+ "loss": 0.0494,
288
+ "step": 140,
289
+ "student_loss": 0.0008300879271700978,
290
+ "teacher_loss": 0.0005129938363097608
291
+ },
292
+ {
293
+ "epoch": 0.04181084198385236,
294
+ "grad_norm": 1.0234375,
295
+ "kd_loss": 0.061179179698228836,
296
+ "learning_rate": 9.663049028182112e-06,
297
+ "loss": 0.0457,
298
+ "step": 145,
299
+ "student_loss": 0.04404008761048317,
300
+ "teacher_loss": 0.009324059821665287
301
+ },
302
+ {
303
+ "epoch": 0.04325259515570934,
304
+ "grad_norm": 1.15625,
305
+ "kd_loss": 0.01291500125080347,
306
+ "learning_rate": 9.633218339615433e-06,
307
+ "loss": 0.0455,
308
+ "step": 150,
309
+ "student_loss": 0.0008404091349802911,
310
+ "teacher_loss": 0.0008450732566416264
311
+ },
312
+ {
313
+ "epoch": 0.04469434832756632,
314
+ "grad_norm": 1.4921875,
315
+ "kd_loss": 0.017479021102190018,
316
+ "learning_rate": 9.602172669838721e-06,
317
+ "loss": 0.0575,
318
+ "step": 155,
319
+ "student_loss": 0.0009151780977845192,
320
+ "teacher_loss": 0.0006974826683290303
321
+ },
322
+ {
323
+ "epoch": 0.0461361014994233,
324
+ "grad_norm": 1.671875,
325
+ "kd_loss": 0.022708255797624588,
326
+ "learning_rate": 9.569920160040815e-06,
327
+ "loss": 0.0503,
328
+ "step": 160,
329
+ "student_loss": 0.07660368084907532,
330
+ "teacher_loss": 0.04039459675550461
331
+ },
332
+ {
333
+ "epoch": 0.04757785467128028,
334
+ "grad_norm": 1.0703125,
335
+ "kd_loss": 0.027437202632427216,
336
+ "learning_rate": 9.536469267883432e-06,
337
+ "loss": 0.0451,
338
+ "step": 165,
339
+ "student_loss": 0.0005943832220509648,
340
+ "teacher_loss": 0.005336699075996876
341
+ },
342
+ {
343
+ "epoch": 0.049019607843137254,
344
+ "grad_norm": 1.71875,
345
+ "kd_loss": 0.014546042308211327,
346
+ "learning_rate": 9.501828765283295e-06,
347
+ "loss": 0.0459,
348
+ "step": 170,
349
+ "student_loss": 0.0006958618760108948,
350
+ "teacher_loss": 0.0005218263831920922
351
+ },
352
+ {
353
+ "epoch": 0.05046136101499423,
354
+ "grad_norm": 1.4453125,
355
+ "kd_loss": 0.02377346158027649,
356
+ "learning_rate": 9.466007736111846e-06,
357
+ "loss": 0.0465,
358
+ "step": 175,
359
+ "student_loss": 0.003214897122234106,
360
+ "teacher_loss": 0.0006713470793329179
361
+ },
362
+ {
363
+ "epoch": 0.05190311418685121,
364
+ "grad_norm": 1.296875,
365
+ "kd_loss": 0.02517828531563282,
366
+ "learning_rate": 9.429015573813163e-06,
367
+ "loss": 0.0454,
368
+ "step": 180,
369
+ "student_loss": 0.0007039233460091054,
370
+ "teacher_loss": 0.0003641469811554998
371
+ },
372
+ {
373
+ "epoch": 0.05334486735870819,
374
+ "grad_norm": 1.25,
375
+ "kd_loss": 0.029188504442572594,
376
+ "learning_rate": 9.390861978940687e-06,
377
+ "loss": 0.0473,
378
+ "step": 185,
379
+ "student_loss": 0.04238048940896988,
380
+ "teacher_loss": 0.00033479384728707373
381
+ },
382
+ {
383
+ "epoch": 0.05478662053056517,
384
+ "grad_norm": 1.484375,
385
+ "kd_loss": 0.03749082237482071,
386
+ "learning_rate": 9.351556956613423e-06,
387
+ "loss": 0.0436,
388
+ "step": 190,
389
+ "student_loss": 0.010270334780216217,
390
+ "teacher_loss": 0.0024021982681006193
391
+ },
392
+ {
393
+ "epoch": 0.056228373702422146,
394
+ "grad_norm": 2.0,
395
+ "kd_loss": 0.030336912721395493,
396
+ "learning_rate": 9.31111081389227e-06,
397
+ "loss": 0.0442,
398
+ "step": 195,
399
+ "student_loss": 0.010534510016441345,
400
+ "teacher_loss": 0.0003890562220476568
401
+ },
402
+ {
403
+ "epoch": 0.05767012687427912,
404
+ "grad_norm": 0.921875,
405
+ "kd_loss": 0.026554003357887268,
406
+ "learning_rate": 9.269534157077177e-06,
407
+ "loss": 0.0497,
408
+ "step": 200,
409
+ "student_loss": 0.0034760867711156607,
410
+ "teacher_loss": 0.00022423896007239819
411
+ },
412
+ {
413
+ "epoch": 0.0591118800461361,
414
+ "grad_norm": 1.9140625,
415
+ "kd_loss": 0.023410219699144363,
416
+ "learning_rate": 9.226837888925813e-06,
417
+ "loss": 0.0471,
418
+ "step": 205,
419
+ "student_loss": 0.03611442819237709,
420
+ "teacher_loss": 0.027781842276453972
421
+ },
422
+ {
423
+ "epoch": 0.06055363321799308,
424
+ "grad_norm": 1.7578125,
425
+ "kd_loss": 0.022867349907755852,
426
+ "learning_rate": 9.183033205794525e-06,
427
+ "loss": 0.0428,
428
+ "step": 210,
429
+ "student_loss": 0.001971556106582284,
430
+ "teacher_loss": 0.0009746765717864037
431
+ },
432
+ {
433
+ "epoch": 0.061995386389850055,
434
+ "grad_norm": 1.0234375,
435
+ "kd_loss": 0.028235338628292084,
436
+ "learning_rate": 9.13813159470227e-06,
437
+ "loss": 0.0449,
438
+ "step": 215,
439
+ "student_loss": 0.0009558866149745882,
440
+ "teacher_loss": 0.0004049334966111928
441
+ },
442
+ {
443
+ "epoch": 0.06343713956170703,
444
+ "grad_norm": 1.4296875,
445
+ "kd_loss": 0.03134317323565483,
446
+ "learning_rate": 9.092144830318357e-06,
447
+ "loss": 0.0438,
448
+ "step": 220,
449
+ "student_loss": 0.048195730894804,
450
+ "teacher_loss": 0.009161758236587048
451
+ },
452
+ {
453
+ "epoch": 0.06487889273356401,
454
+ "grad_norm": 1.1328125,
455
+ "kd_loss": 0.014049015007913113,
456
+ "learning_rate": 9.045084971874738e-06,
457
+ "loss": 0.0487,
458
+ "step": 225,
459
+ "student_loss": 0.05373133346438408,
460
+ "teacher_loss": 0.0009734037448652089
461
+ },
462
+ {
463
+ "epoch": 0.06632064590542099,
464
+ "grad_norm": 1.7265625,
465
+ "kd_loss": 0.021539395675063133,
466
+ "learning_rate": 8.99696436000368e-06,
467
+ "loss": 0.0384,
468
+ "step": 230,
469
+ "student_loss": 0.0008717047749087214,
470
+ "teacher_loss": 0.0028650674503296614
471
+ },
472
+ {
473
+ "epoch": 0.06776239907727798,
474
+ "grad_norm": 1.0390625,
475
+ "kd_loss": 0.045927971601486206,
476
+ "learning_rate": 8.947795613501658e-06,
477
+ "loss": 0.0397,
478
+ "step": 235,
479
+ "student_loss": 0.012143217958509922,
480
+ "teacher_loss": 0.0006723064580000937
481
+ },
482
+ {
483
+ "epoch": 0.06920415224913495,
484
+ "grad_norm": 0.78125,
485
+ "kd_loss": 0.01626853086054325,
486
+ "learning_rate": 8.897591626020284e-06,
487
+ "loss": 0.0425,
488
+ "step": 240,
489
+ "student_loss": 0.0004826921212952584,
490
+ "teacher_loss": 0.0003007515915669501
491
+ },
492
+ {
493
+ "epoch": 0.07064590542099193,
494
+ "grad_norm": 1.078125,
495
+ "kd_loss": 0.010876834392547607,
496
+ "learning_rate": 8.846365562685178e-06,
497
+ "loss": 0.0379,
498
+ "step": 245,
499
+ "student_loss": 0.00123109039850533,
500
+ "teacher_loss": 0.0017208561766892672
501
+ },
502
+ {
503
+ "epoch": 0.07208765859284891,
504
+ "grad_norm": 1.6484375,
505
+ "kd_loss": 0.012208702974021435,
506
+ "learning_rate": 8.794130856643635e-06,
507
+ "loss": 0.0402,
508
+ "step": 250,
509
+ "student_loss": 0.0006701876409351826,
510
+ "teacher_loss": 0.00039402375114150345
511
+ },
512
+ {
513
+ "epoch": 0.07352941176470588,
514
+ "grad_norm": 3.859375,
515
+ "kd_loss": 0.010964477434754372,
516
+ "learning_rate": 8.74090120554202e-06,
517
+ "loss": 0.0428,
518
+ "step": 255,
519
+ "student_loss": 0.0004243800649419427,
520
+ "teacher_loss": 0.0003773768839892
521
+ },
522
+ {
523
+ "epoch": 0.07497116493656286,
524
+ "grad_norm": 2.015625,
525
+ "kd_loss": 0.018098091706633568,
526
+ "learning_rate": 8.686690567933803e-06,
527
+ "loss": 0.0436,
528
+ "step": 260,
529
+ "student_loss": 0.040437813848257065,
530
+ "teacher_loss": 0.032756391912698746
531
+ },
532
+ {
533
+ "epoch": 0.07641291810841984,
534
+ "grad_norm": 1.125,
535
+ "kd_loss": 0.011841603554785252,
536
+ "learning_rate": 8.63151315961915e-06,
537
+ "loss": 0.0412,
538
+ "step": 265,
539
+ "student_loss": 0.0006605261587537825,
540
+ "teacher_loss": 0.0004997120704501867
541
+ },
542
+ {
543
+ "epoch": 0.07785467128027682,
544
+ "grad_norm": 1.046875,
545
+ "kd_loss": 0.01046097930520773,
546
+ "learning_rate": 8.575383449917103e-06,
547
+ "loss": 0.044,
548
+ "step": 270,
549
+ "student_loss": 0.0009966753423213959,
550
+ "teacher_loss": 0.0006801315466873348
551
+ },
552
+ {
553
+ "epoch": 0.07929642445213379,
554
+ "grad_norm": 1.515625,
555
+ "kd_loss": 0.01713527925312519,
556
+ "learning_rate": 8.518316157871232e-06,
557
+ "loss": 0.0403,
558
+ "step": 275,
559
+ "student_loss": 0.04822665825486183,
560
+ "teacher_loss": 0.03450855612754822
561
+ },
562
+ {
563
+ "epoch": 0.08073817762399077,
564
+ "grad_norm": 1.3828125,
565
+ "kd_loss": 0.021968362852931023,
566
+ "learning_rate": 8.460326248389825e-06,
567
+ "loss": 0.0401,
568
+ "step": 280,
569
+ "student_loss": 0.0006340977270156145,
570
+ "teacher_loss": 0.0001483594678575173
571
+ },
572
+ {
573
+ "epoch": 0.08217993079584775,
574
+ "grad_norm": 1.3203125,
575
+ "kd_loss": 0.01131309662014246,
576
+ "learning_rate": 8.401428928321607e-06,
577
+ "loss": 0.0435,
578
+ "step": 285,
579
+ "student_loss": 0.0005278023309074342,
580
+ "teacher_loss": 0.0004103762039449066
581
+ },
582
+ {
583
+ "epoch": 0.08362168396770472,
584
+ "grad_norm": 1.1015625,
585
+ "kd_loss": 0.00958277191966772,
586
+ "learning_rate": 8.341639642468002e-06,
587
+ "loss": 0.0469,
588
+ "step": 290,
589
+ "student_loss": 0.005627856124192476,
590
+ "teacher_loss": 0.0034064538776874542
591
+ },
592
+ {
593
+ "epoch": 0.0850634371395617,
594
+ "grad_norm": 1.140625,
595
+ "kd_loss": 0.009735080413520336,
596
+ "learning_rate": 8.280974069532999e-06,
597
+ "loss": 0.0409,
598
+ "step": 295,
599
+ "student_loss": 0.000509344507008791,
600
+ "teacher_loss": 0.0004550877492874861
601
+ },
602
+ {
603
+ "epoch": 0.08650519031141868,
604
+ "grad_norm": 1.09375,
605
+ "kd_loss": 0.031678151339292526,
606
+ "learning_rate": 8.219448118011687e-06,
607
+ "loss": 0.0379,
608
+ "step": 300,
609
+ "student_loss": 0.016466177999973297,
610
+ "teacher_loss": 0.002865313785150647
611
+ },
612
+ {
613
+ "epoch": 0.08794694348327567,
614
+ "grad_norm": 1.1015625,
615
+ "kd_loss": 0.013346894644200802,
616
+ "learning_rate": 8.157077922018537e-06,
617
+ "loss": 0.0353,
618
+ "step": 305,
619
+ "student_loss": 0.009980602189898491,
620
+ "teacher_loss": 0.0014113991055637598
621
+ },
622
+ {
623
+ "epoch": 0.08938869665513265,
624
+ "grad_norm": 1.90625,
625
+ "kd_loss": 0.010920979082584381,
626
+ "learning_rate": 8.093879837056486e-06,
627
+ "loss": 0.0384,
628
+ "step": 310,
629
+ "student_loss": 0.00036570793599821627,
630
+ "teacher_loss": 0.0002939177502412349
631
+ },
632
+ {
633
+ "epoch": 0.09083044982698962,
634
+ "grad_norm": 1.0078125,
635
+ "kd_loss": 0.021986963227391243,
636
+ "learning_rate": 8.029870435728018e-06,
637
+ "loss": 0.0385,
638
+ "step": 315,
639
+ "student_loss": 0.01466396264731884,
640
+ "teacher_loss": 0.0008777049370110035
641
+ },
642
+ {
643
+ "epoch": 0.0922722029988466,
644
+ "grad_norm": 1.640625,
645
+ "kd_loss": 0.014081098139286041,
646
+ "learning_rate": 7.965066503389264e-06,
647
+ "loss": 0.0417,
648
+ "step": 320,
649
+ "student_loss": 0.0004964058171026409,
650
+ "teacher_loss": 0.0003025036712642759
651
+ },
652
+ {
653
+ "epoch": 0.09371395617070358,
654
+ "grad_norm": 1.2421875,
655
+ "kd_loss": 0.03542112559080124,
656
+ "learning_rate": 7.89948503374835e-06,
657
+ "loss": 0.0378,
658
+ "step": 325,
659
+ "student_loss": 0.01556178368628025,
660
+ "teacher_loss": 0.018082890659570694
661
+ },
662
+ {
663
+ "epoch": 0.09515570934256055,
664
+ "grad_norm": 1.40625,
665
+ "kd_loss": 0.04085119441151619,
666
+ "learning_rate": 7.833143224409076e-06,
667
+ "loss": 0.0394,
668
+ "step": 330,
669
+ "student_loss": 0.024770811200141907,
670
+ "teacher_loss": 0.0021320905070751905
671
+ },
672
+ {
673
+ "epoch": 0.09659746251441753,
674
+ "grad_norm": 1.0234375,
675
+ "kd_loss": 0.008477268740534782,
676
+ "learning_rate": 7.766058472361154e-06,
677
+ "loss": 0.0338,
678
+ "step": 335,
679
+ "student_loss": 0.0009084849734790623,
680
+ "teacher_loss": 0.0007285018800757825
681
+ },
682
+ {
683
+ "epoch": 0.09803921568627451,
684
+ "grad_norm": 1.28125,
685
+ "kd_loss": 0.014947460032999516,
686
+ "learning_rate": 7.698248369418146e-06,
687
+ "loss": 0.0388,
688
+ "step": 340,
689
+ "student_loss": 0.009875037707388401,
690
+ "teacher_loss": 0.00507500022649765
691
+ },
692
+ {
693
+ "epoch": 0.09948096885813149,
694
+ "grad_norm": 1.2421875,
695
+ "kd_loss": 0.019748909398913383,
696
+ "learning_rate": 7.629730697604314e-06,
697
+ "loss": 0.0437,
698
+ "step": 345,
699
+ "student_loss": 0.008698839694261551,
700
+ "teacher_loss": 0.00021470236242748797
701
+ },
702
+ {
703
+ "epoch": 0.10092272202998846,
704
+ "grad_norm": 1.453125,
705
+ "kd_loss": 0.038638580590486526,
706
+ "learning_rate": 7.560523424491595e-06,
707
+ "loss": 0.0333,
708
+ "step": 350,
709
+ "student_loss": 0.007182905916124582,
710
+ "teacher_loss": 0.0004611952172126621
711
+ },
712
+ {
713
+ "epoch": 0.10236447520184544,
714
+ "grad_norm": 1.4375,
715
+ "kd_loss": 0.008461177349090576,
716
+ "learning_rate": 7.490644698487909e-06,
717
+ "loss": 0.0405,
718
+ "step": 355,
719
+ "student_loss": 0.0004199718532618135,
720
+ "teacher_loss": 0.00038222281727939844
721
+ },
722
+ {
723
+ "epoch": 0.10380622837370242,
724
+ "grad_norm": 1.1796875,
725
+ "kd_loss": 0.008598582819104195,
726
+ "learning_rate": 7.420112844078066e-06,
727
+ "loss": 0.0396,
728
+ "step": 360,
729
+ "student_loss": 0.00045404830598272383,
730
+ "teacher_loss": 0.0003997418680228293
731
+ },
732
+ {
733
+ "epoch": 0.1052479815455594,
734
+ "grad_norm": 1.78125,
735
+ "kd_loss": 0.018130140379071236,
736
+ "learning_rate": 7.348946357018479e-06,
737
+ "loss": 0.0425,
738
+ "step": 365,
739
+ "student_loss": 0.06869062036275864,
740
+ "teacher_loss": 0.027944643050432205
741
+ },
742
+ {
743
+ "epoch": 0.10668973471741638,
744
+ "grad_norm": 1.03125,
745
+ "kd_loss": 0.0350988432765007,
746
+ "learning_rate": 7.277163899486975e-06,
747
+ "loss": 0.0447,
748
+ "step": 370,
749
+ "student_loss": 0.015085455030202866,
750
+ "teacher_loss": 0.0004648211761377752
751
+ },
752
+ {
753
+ "epoch": 0.10813148788927336,
754
+ "grad_norm": 1.0703125,
755
+ "kd_loss": 0.01613399013876915,
756
+ "learning_rate": 7.204784295188959e-06,
757
+ "loss": 0.0365,
758
+ "step": 375,
759
+ "student_loss": 0.03554360568523407,
760
+ "teacher_loss": 0.021264519542455673
761
+ },
762
+ {
763
+ "epoch": 0.10957324106113034,
764
+ "grad_norm": 1.296875,
765
+ "kd_loss": 0.010999550111591816,
766
+ "learning_rate": 7.1318265244212305e-06,
767
+ "loss": 0.0381,
768
+ "step": 380,
769
+ "student_loss": 0.0004269553755875677,
770
+ "teacher_loss": 0.0003071910759899765
771
+ },
772
+ {
773
+ "epoch": 0.11101499423298732,
774
+ "grad_norm": 0.875,
775
+ "kd_loss": 0.015028773806989193,
776
+ "learning_rate": 7.05830971909472e-06,
777
+ "loss": 0.0349,
778
+ "step": 385,
779
+ "student_loss": 0.0005910908803343773,
780
+ "teacher_loss": 0.0003781717678066343
781
+ },
782
+ {
783
+ "epoch": 0.11245674740484429,
784
+ "grad_norm": 1.25,
785
+ "kd_loss": 0.010998338460922241,
786
+ "learning_rate": 6.9842531577174865e-06,
787
+ "loss": 0.033,
788
+ "step": 390,
789
+ "student_loss": 0.00025347189512103796,
790
+ "teacher_loss": 0.00023755666916258633
791
+ },
792
+ {
793
+ "epoch": 0.11389850057670127,
794
+ "grad_norm": 1.390625,
795
+ "kd_loss": 0.010839108377695084,
796
+ "learning_rate": 6.9096762603392595e-06,
797
+ "loss": 0.0393,
798
+ "step": 395,
799
+ "student_loss": 0.00046008970821276307,
800
+ "teacher_loss": 0.00034296512603759766
801
+ },
802
+ {
803
+ "epoch": 0.11534025374855825,
804
+ "grad_norm": 2.40625,
805
+ "kd_loss": 0.015689171850681305,
806
+ "learning_rate": 6.834598583458862e-06,
807
+ "loss": 0.0333,
808
+ "step": 400,
809
+ "student_loss": 0.0004951858427375555,
810
+ "teacher_loss": 0.0002840181696228683
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/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-400/vocab.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-600/added_tokens.json ADDED
@@ -0,0 +1,35 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "</tool_call>": 151658,
3
+ "<tool_call>": 151657,
4
+ "<|action_sep|>": 151670,
5
+ "<|arg_sep|>": 151671,
6
+ "<|box_end|>": 151649,
7
+ "<|box_start|>": 151648,
8
+ "<|call_sep|>": 151666,
9
+ "<|end_of_text|>": 151673,
10
+ "<|endoftext|>": 151643,
11
+ "<|exception_sep|>": 151669,
12
+ "<|file_sep|>": 151664,
13
+ "<|fim_middle|>": 151660,
14
+ "<|fim_pad|>": 151662,
15
+ "<|fim_prefix|>": 151659,
16
+ "<|fim_suffix|>": 151661,
17
+ "<|frame_sep|>": 151672,
18
+ "<|im_end|>": 151645,
19
+ "<|im_start|>": 151644,
20
+ "<|image_pad|>": 151655,
21
+ "<|latent_end|>": 151675,
22
+ "<|latent_start|>": 151674,
23
+ "<|line_sep|>": 151667,
24
+ "<|object_ref_end|>": 151647,
25
+ "<|object_ref_start|>": 151646,
26
+ "<|quad_end|>": 151651,
27
+ "<|quad_start|>": 151650,
28
+ "<|repo_name|>": 151663,
29
+ "<|return_sep|>": 151668,
30
+ "<|trace_context_start|>": 151665,
31
+ "<|video_pad|>": 151656,
32
+ "<|vision_end|>": 151653,
33
+ "<|vision_pad|>": 151654,
34
+ "<|vision_start|>": 151652
35
+ }
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-600/chat_template.jinja ADDED
@@ -0,0 +1,54 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {%- if tools %}
2
+ {{- '<|im_start|>system\n' }}
3
+ {%- if messages[0]['role'] == 'system' %}
4
+ {{- messages[0]['content'] }}
5
+ {%- else %}
6
+ {{- 'You are Qwen, created by Alibaba Cloud. You are a helpful assistant.' }}
7
+ {%- endif %}
8
+ {{- "\n\n# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
9
+ {%- for tool in tools %}
10
+ {{- "\n" }}
11
+ {{- tool | tojson }}
12
+ {%- endfor %}
13
+ {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
14
+ {%- else %}
15
+ {%- if messages[0]['role'] == 'system' %}
16
+ {{- '<|im_start|>system\n' + messages[0]['content'] + '<|im_end|>\n' }}
17
+ {%- else %}
18
+ {{- '<|im_start|>system\nYou are Qwen, created by Alibaba Cloud. You are a helpful assistant.<|im_end|>\n' }}
19
+ {%- endif %}
20
+ {%- endif %}
21
+ {%- for message in messages %}
22
+ {%- if (message.role == "user") or (message.role == "system" and not loop.first) or (message.role == "assistant" and not message.tool_calls) %}
23
+ {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
24
+ {%- elif message.role == "assistant" %}
25
+ {{- '<|im_start|>' + message.role }}
26
+ {%- if message.content %}
27
+ {{- '\n' + message.content }}
28
+ {%- endif %}
29
+ {%- for tool_call in message.tool_calls %}
30
+ {%- if tool_call.function is defined %}
31
+ {%- set tool_call = tool_call.function %}
32
+ {%- endif %}
33
+ {{- '\n<tool_call>\n{"name": "' }}
34
+ {{- tool_call.name }}
35
+ {{- '", "arguments": ' }}
36
+ {{- tool_call.arguments | tojson }}
37
+ {{- '}\n</tool_call>' }}
38
+ {%- endfor %}
39
+ {{- '<|im_end|>\n' }}
40
+ {%- elif message.role == "tool" %}
41
+ {%- if (loop.index0 == 0) or (messages[loop.index0 - 1].role != "tool") %}
42
+ {{- '<|im_start|>user' }}
43
+ {%- endif %}
44
+ {{- '\n<tool_response>\n' }}
45
+ {{- message.content }}
46
+ {{- '\n</tool_response>' }}
47
+ {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
48
+ {{- '<|im_end|>\n' }}
49
+ {%- endif %}
50
+ {%- endif %}
51
+ {%- endfor %}
52
+ {%- if add_generation_prompt %}
53
+ {{- '<|im_start|>assistant\n' }}
54
+ {%- endif %}
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-600/config.json ADDED
@@ -0,0 +1,58 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Qwen2ForCausalLM"
4
+ ],
5
+ "attention_dropout": 0.0,
6
+ "dtype": "bfloat16",
7
+ "eos_token_id": 151643,
8
+ "hidden_act": "silu",
9
+ "hidden_size": 1536,
10
+ "initializer_range": 0.02,
11
+ "intermediate_size": 8960,
12
+ "layer_types": [
13
+ "full_attention",
14
+ "full_attention",
15
+ "full_attention",
16
+ "full_attention",
17
+ "full_attention",
18
+ "full_attention",
19
+ "full_attention",
20
+ "full_attention",
21
+ "full_attention",
22
+ "full_attention",
23
+ "full_attention",
24
+ "full_attention",
25
+ "full_attention",
26
+ "full_attention",
27
+ "full_attention",
28
+ "full_attention",
29
+ "full_attention",
30
+ "full_attention",
31
+ "full_attention",
32
+ "full_attention",
33
+ "full_attention",
34
+ "full_attention",
35
+ "full_attention",
36
+ "full_attention",
37
+ "full_attention",
38
+ "full_attention",
39
+ "full_attention",
40
+ "full_attention"
41
+ ],
42
+ "max_position_embeddings": 32768,
43
+ "max_window_layers": 28,
44
+ "model_type": "qwen2",
45
+ "num_attention_heads": 12,
46
+ "num_hidden_layers": 28,
47
+ "num_key_value_heads": 2,
48
+ "pad_token_id": 151643,
49
+ "rms_norm_eps": 1e-06,
50
+ "rope_scaling": null,
51
+ "rope_theta": 1000000.0,
52
+ "sliding_window": null,
53
+ "tie_word_embeddings": true,
54
+ "transformers_version": "4.57.6",
55
+ "use_cache": true,
56
+ "use_sliding_window": false,
57
+ "vocab_size": 151676
58
+ }
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-600/merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-600/pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7448e9232596f2baf886f818457931ae5d797620a7cab584453a45dfe3ae64a5
3
+ size 3096275515
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-600/special_tokens_map.json ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "additional_special_tokens": [
3
+ "<|im_start|>",
4
+ "<|im_end|>",
5
+ "<|object_ref_start|>",
6
+ "<|object_ref_end|>",
7
+ "<|box_start|>",
8
+ "<|box_end|>",
9
+ "<|quad_start|>",
10
+ "<|quad_end|>",
11
+ "<|vision_start|>",
12
+ "<|vision_end|>",
13
+ "<|vision_pad|>",
14
+ "<|image_pad|>",
15
+ "<|video_pad|>"
16
+ ],
17
+ "eos_token": {
18
+ "content": "<|endoftext|>",
19
+ "lstrip": false,
20
+ "normalized": false,
21
+ "rstrip": false,
22
+ "single_word": false
23
+ },
24
+ "pad_token": {
25
+ "content": "<|endoftext|>",
26
+ "lstrip": false,
27
+ "normalized": false,
28
+ "rstrip": false,
29
+ "single_word": false
30
+ }
31
+ }
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-600/thought_projector.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1d6ca7d2abfac84407f2a77a3ef77cdfbd0a7e0c484f220f0e4616bbfd040b71
3
+ size 9445953
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-600/tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:83a790d654474f5dfe225f889afd0210313eb1083f942671f2c4b8e95a1c922b
3
+ size 11424004
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-600/tokenizer_config.json ADDED
@@ -0,0 +1,295 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_bos_token": false,
3
+ "add_prefix_space": false,
4
+ "added_tokens_decoder": {
5
+ "151643": {
6
+ "content": "<|endoftext|>",
7
+ "lstrip": false,
8
+ "normalized": false,
9
+ "rstrip": false,
10
+ "single_word": false,
11
+ "special": true
12
+ },
13
+ "151644": {
14
+ "content": "<|im_start|>",
15
+ "lstrip": false,
16
+ "normalized": false,
17
+ "rstrip": false,
18
+ "single_word": false,
19
+ "special": true
20
+ },
21
+ "151645": {
22
+ "content": "<|im_end|>",
23
+ "lstrip": false,
24
+ "normalized": false,
25
+ "rstrip": false,
26
+ "single_word": false,
27
+ "special": true
28
+ },
29
+ "151646": {
30
+ "content": "<|object_ref_start|>",
31
+ "lstrip": false,
32
+ "normalized": false,
33
+ "rstrip": false,
34
+ "single_word": false,
35
+ "special": true
36
+ },
37
+ "151647": {
38
+ "content": "<|object_ref_end|>",
39
+ "lstrip": false,
40
+ "normalized": false,
41
+ "rstrip": false,
42
+ "single_word": false,
43
+ "special": true
44
+ },
45
+ "151648": {
46
+ "content": "<|box_start|>",
47
+ "lstrip": false,
48
+ "normalized": false,
49
+ "rstrip": false,
50
+ "single_word": false,
51
+ "special": true
52
+ },
53
+ "151649": {
54
+ "content": "<|box_end|>",
55
+ "lstrip": false,
56
+ "normalized": false,
57
+ "rstrip": false,
58
+ "single_word": false,
59
+ "special": true
60
+ },
61
+ "151650": {
62
+ "content": "<|quad_start|>",
63
+ "lstrip": false,
64
+ "normalized": false,
65
+ "rstrip": false,
66
+ "single_word": false,
67
+ "special": true
68
+ },
69
+ "151651": {
70
+ "content": "<|quad_end|>",
71
+ "lstrip": false,
72
+ "normalized": false,
73
+ "rstrip": false,
74
+ "single_word": false,
75
+ "special": true
76
+ },
77
+ "151652": {
78
+ "content": "<|vision_start|>",
79
+ "lstrip": false,
80
+ "normalized": false,
81
+ "rstrip": false,
82
+ "single_word": false,
83
+ "special": true
84
+ },
85
+ "151653": {
86
+ "content": "<|vision_end|>",
87
+ "lstrip": false,
88
+ "normalized": false,
89
+ "rstrip": false,
90
+ "single_word": false,
91
+ "special": true
92
+ },
93
+ "151654": {
94
+ "content": "<|vision_pad|>",
95
+ "lstrip": false,
96
+ "normalized": false,
97
+ "rstrip": false,
98
+ "single_word": false,
99
+ "special": true
100
+ },
101
+ "151655": {
102
+ "content": "<|image_pad|>",
103
+ "lstrip": false,
104
+ "normalized": false,
105
+ "rstrip": false,
106
+ "single_word": false,
107
+ "special": true
108
+ },
109
+ "151656": {
110
+ "content": "<|video_pad|>",
111
+ "lstrip": false,
112
+ "normalized": false,
113
+ "rstrip": false,
114
+ "single_word": false,
115
+ "special": true
116
+ },
117
+ "151657": {
118
+ "content": "<tool_call>",
119
+ "lstrip": false,
120
+ "normalized": false,
121
+ "rstrip": false,
122
+ "single_word": false,
123
+ "special": false
124
+ },
125
+ "151658": {
126
+ "content": "</tool_call>",
127
+ "lstrip": false,
128
+ "normalized": false,
129
+ "rstrip": false,
130
+ "single_word": false,
131
+ "special": false
132
+ },
133
+ "151659": {
134
+ "content": "<|fim_prefix|>",
135
+ "lstrip": false,
136
+ "normalized": false,
137
+ "rstrip": false,
138
+ "single_word": false,
139
+ "special": false
140
+ },
141
+ "151660": {
142
+ "content": "<|fim_middle|>",
143
+ "lstrip": false,
144
+ "normalized": false,
145
+ "rstrip": false,
146
+ "single_word": false,
147
+ "special": false
148
+ },
149
+ "151661": {
150
+ "content": "<|fim_suffix|>",
151
+ "lstrip": false,
152
+ "normalized": false,
153
+ "rstrip": false,
154
+ "single_word": false,
155
+ "special": false
156
+ },
157
+ "151662": {
158
+ "content": "<|fim_pad|>",
159
+ "lstrip": false,
160
+ "normalized": false,
161
+ "rstrip": false,
162
+ "single_word": false,
163
+ "special": false
164
+ },
165
+ "151663": {
166
+ "content": "<|repo_name|>",
167
+ "lstrip": false,
168
+ "normalized": false,
169
+ "rstrip": false,
170
+ "single_word": false,
171
+ "special": false
172
+ },
173
+ "151664": {
174
+ "content": "<|file_sep|>",
175
+ "lstrip": false,
176
+ "normalized": false,
177
+ "rstrip": false,
178
+ "single_word": false,
179
+ "special": false
180
+ },
181
+ "151665": {
182
+ "content": "<|trace_context_start|>",
183
+ "lstrip": false,
184
+ "normalized": false,
185
+ "rstrip": false,
186
+ "single_word": false,
187
+ "special": true
188
+ },
189
+ "151666": {
190
+ "content": "<|call_sep|>",
191
+ "lstrip": false,
192
+ "normalized": false,
193
+ "rstrip": false,
194
+ "single_word": false,
195
+ "special": true
196
+ },
197
+ "151667": {
198
+ "content": "<|line_sep|>",
199
+ "lstrip": false,
200
+ "normalized": false,
201
+ "rstrip": false,
202
+ "single_word": false,
203
+ "special": true
204
+ },
205
+ "151668": {
206
+ "content": "<|return_sep|>",
207
+ "lstrip": false,
208
+ "normalized": false,
209
+ "rstrip": false,
210
+ "single_word": false,
211
+ "special": true
212
+ },
213
+ "151669": {
214
+ "content": "<|exception_sep|>",
215
+ "lstrip": false,
216
+ "normalized": false,
217
+ "rstrip": false,
218
+ "single_word": false,
219
+ "special": true
220
+ },
221
+ "151670": {
222
+ "content": "<|action_sep|>",
223
+ "lstrip": false,
224
+ "normalized": false,
225
+ "rstrip": false,
226
+ "single_word": false,
227
+ "special": true
228
+ },
229
+ "151671": {
230
+ "content": "<|arg_sep|>",
231
+ "lstrip": false,
232
+ "normalized": false,
233
+ "rstrip": false,
234
+ "single_word": false,
235
+ "special": true
236
+ },
237
+ "151672": {
238
+ "content": "<|frame_sep|>",
239
+ "lstrip": false,
240
+ "normalized": false,
241
+ "rstrip": false,
242
+ "single_word": false,
243
+ "special": true
244
+ },
245
+ "151673": {
246
+ "content": "<|end_of_text|>",
247
+ "lstrip": false,
248
+ "normalized": false,
249
+ "rstrip": false,
250
+ "single_word": false,
251
+ "special": true
252
+ },
253
+ "151674": {
254
+ "content": "<|latent_start|>",
255
+ "lstrip": false,
256
+ "normalized": false,
257
+ "rstrip": false,
258
+ "single_word": false,
259
+ "special": true
260
+ },
261
+ "151675": {
262
+ "content": "<|latent_end|>",
263
+ "lstrip": false,
264
+ "normalized": false,
265
+ "rstrip": false,
266
+ "single_word": false,
267
+ "special": true
268
+ }
269
+ },
270
+ "additional_special_tokens": [
271
+ "<|im_start|>",
272
+ "<|im_end|>",
273
+ "<|object_ref_start|>",
274
+ "<|object_ref_end|>",
275
+ "<|box_start|>",
276
+ "<|box_end|>",
277
+ "<|quad_start|>",
278
+ "<|quad_end|>",
279
+ "<|vision_start|>",
280
+ "<|vision_end|>",
281
+ "<|vision_pad|>",
282
+ "<|image_pad|>",
283
+ "<|video_pad|>"
284
+ ],
285
+ "bos_token": null,
286
+ "clean_up_tokenization_spaces": false,
287
+ "eos_token": "<|endoftext|>",
288
+ "errors": "replace",
289
+ "extra_special_tokens": {},
290
+ "model_max_length": 32768,
291
+ "pad_token": "<|endoftext|>",
292
+ "split_special_tokens": false,
293
+ "tokenizer_class": "Qwen2Tokenizer",
294
+ "unk_token": null
295
+ }
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-600/trainer_state.json ADDED
@@ -0,0 +1,1234 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 0.17301038062283736,
6
+ "eval_steps": 500,
7
+ "global_step": 600,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.0014417531718569781,
14
+ "grad_norm": 52992.0,
15
+ "kd_loss": 0.3193017542362213,
16
+ "learning_rate": 1.3333333333333334e-06,
17
+ "loss": 0.473,
18
+ "step": 5,
19
+ "student_loss": 0.15727384388446808,
20
+ "teacher_loss": 0.0004137983196415007
21
+ },
22
+ {
23
+ "epoch": 0.0028835063437139563,
24
+ "grad_norm": 16121856.0,
25
+ "kd_loss": 0.3087984025478363,
26
+ "learning_rate": 3e-06,
27
+ "loss": 0.466,
28
+ "step": 10,
29
+ "student_loss": 0.2608874440193176,
30
+ "teacher_loss": 0.0007329508662223816
31
+ },
32
+ {
33
+ "epoch": 0.004325259515570935,
34
+ "grad_norm": 2240.0,
35
+ "kd_loss": 0.25935155153274536,
36
+ "learning_rate": 4.666666666666667e-06,
37
+ "loss": 0.4585,
38
+ "step": 15,
39
+ "student_loss": 0.1052168533205986,
40
+ "teacher_loss": 0.000609480426646769
41
+ },
42
+ {
43
+ "epoch": 0.0057670126874279125,
44
+ "grad_norm": 4194304.0,
45
+ "kd_loss": 0.25011566281318665,
46
+ "learning_rate": 6.333333333333333e-06,
47
+ "loss": 0.4707,
48
+ "step": 20,
49
+ "student_loss": 0.15271778404712677,
50
+ "teacher_loss": 0.0002972046786453575
51
+ },
52
+ {
53
+ "epoch": 0.00720876585928489,
54
+ "grad_norm": 41680896.0,
55
+ "kd_loss": 0.2633170485496521,
56
+ "learning_rate": 8.000000000000001e-06,
57
+ "loss": 0.4368,
58
+ "step": 25,
59
+ "student_loss": 0.09109362959861755,
60
+ "teacher_loss": 0.0009600125486031175
61
+ },
62
+ {
63
+ "epoch": 0.00865051903114187,
64
+ "grad_norm": 512.0,
65
+ "kd_loss": 0.2533612847328186,
66
+ "learning_rate": 9.666666666666667e-06,
67
+ "loss": 0.4187,
68
+ "step": 30,
69
+ "student_loss": 0.22675999999046326,
70
+ "teacher_loss": 0.00035076512722298503
71
+ },
72
+ {
73
+ "epoch": 0.010092272202998846,
74
+ "grad_norm": 48.75,
75
+ "kd_loss": 0.2179471105337143,
76
+ "learning_rate": 9.99958042442916e-06,
77
+ "loss": 0.3713,
78
+ "step": 35,
79
+ "student_loss": 0.09486500918865204,
80
+ "teacher_loss": 0.00026503618573769927
81
+ },
82
+ {
83
+ "epoch": 0.011534025374855825,
84
+ "grad_norm": 10.0,
85
+ "kd_loss": 0.17874062061309814,
86
+ "learning_rate": 9.997876019358083e-06,
87
+ "loss": 0.2647,
88
+ "step": 40,
89
+ "student_loss": 0.07879351079463959,
90
+ "teacher_loss": 0.05543303117156029
91
+ },
92
+ {
93
+ "epoch": 0.012975778546712802,
94
+ "grad_norm": 4.9375,
95
+ "kd_loss": 0.19365739822387695,
96
+ "learning_rate": 9.99486100792044e-06,
97
+ "loss": 0.1686,
98
+ "step": 45,
99
+ "student_loss": 0.07774724811315536,
100
+ "teacher_loss": 0.0008653572876937687
101
+ },
102
+ {
103
+ "epoch": 0.01441753171856978,
104
+ "grad_norm": 3.359375,
105
+ "kd_loss": 0.06108914315700531,
106
+ "learning_rate": 9.990536180750724e-06,
107
+ "loss": 0.1396,
108
+ "step": 50,
109
+ "student_loss": 0.008677379228174686,
110
+ "teacher_loss": 0.005272893700748682
111
+ },
112
+ {
113
+ "epoch": 0.015859284890426758,
114
+ "grad_norm": 3.71875,
115
+ "kd_loss": 0.04330182075500488,
116
+ "learning_rate": 9.984902671959911e-06,
117
+ "loss": 0.1065,
118
+ "step": 55,
119
+ "student_loss": 0.001041711075231433,
120
+ "teacher_loss": 0.0005072589847259223
121
+ },
122
+ {
123
+ "epoch": 0.01730103806228374,
124
+ "grad_norm": 4.125,
125
+ "kd_loss": 0.03823668882250786,
126
+ "learning_rate": 9.97796195883804e-06,
127
+ "loss": 0.0862,
128
+ "step": 60,
129
+ "student_loss": 0.0209807138890028,
130
+ "teacher_loss": 0.0038409351836889982
131
+ },
132
+ {
133
+ "epoch": 0.018742791234140715,
134
+ "grad_norm": 10.1875,
135
+ "kd_loss": 0.03922531381249428,
136
+ "learning_rate": 9.969715861466839e-06,
137
+ "loss": 0.0844,
138
+ "step": 65,
139
+ "student_loss": 0.04678678512573242,
140
+ "teacher_loss": 0.05432586744427681
141
+ },
142
+ {
143
+ "epoch": 0.020184544405997693,
144
+ "grad_norm": 2.03125,
145
+ "kd_loss": 0.0549621656537056,
146
+ "learning_rate": 9.96016654224243e-06,
147
+ "loss": 0.0797,
148
+ "step": 70,
149
+ "student_loss": 0.016930116340517998,
150
+ "teacher_loss": 0.0029678302817046642
151
+ },
152
+ {
153
+ "epoch": 0.02162629757785467,
154
+ "grad_norm": 2.5625,
155
+ "kd_loss": 0.023080041632056236,
156
+ "learning_rate": 9.94931650530827e-06,
157
+ "loss": 0.0756,
158
+ "step": 75,
159
+ "student_loss": 0.002210108330473304,
160
+ "teacher_loss": 0.0013834249693900347
161
+ },
162
+ {
163
+ "epoch": 0.02306805074971165,
164
+ "grad_norm": 2.296875,
165
+ "kd_loss": 0.025744518265128136,
166
+ "learning_rate": 9.93716859589851e-06,
167
+ "loss": 0.0644,
168
+ "step": 80,
169
+ "student_loss": 0.0015430074417963624,
170
+ "teacher_loss": 0.000761081581003964
171
+ },
172
+ {
173
+ "epoch": 0.024509803921568627,
174
+ "grad_norm": 3.125,
175
+ "kd_loss": 0.023152953013777733,
176
+ "learning_rate": 9.923725999591846e-06,
177
+ "loss": 0.0609,
178
+ "step": 85,
179
+ "student_loss": 0.0187687948346138,
180
+ "teacher_loss": 0.011518553830683231
181
+ },
182
+ {
183
+ "epoch": 0.025951557093425604,
184
+ "grad_norm": 1.8359375,
185
+ "kd_loss": 0.06841632723808289,
186
+ "learning_rate": 9.908992241476189e-06,
187
+ "loss": 0.0622,
188
+ "step": 90,
189
+ "student_loss": 0.05537872016429901,
190
+ "teacher_loss": 0.011890009045600891
191
+ },
192
+ {
193
+ "epoch": 0.027393310265282585,
194
+ "grad_norm": 4.21875,
195
+ "kd_loss": 0.041227906942367554,
196
+ "learning_rate": 9.892971185224244e-06,
197
+ "loss": 0.0626,
198
+ "step": 95,
199
+ "student_loss": 0.009465624578297138,
200
+ "teacher_loss": 0.005057212430983782
201
+ },
202
+ {
203
+ "epoch": 0.02883506343713956,
204
+ "grad_norm": 1.8359375,
205
+ "kd_loss": 0.021006152033805847,
206
+ "learning_rate": 9.875667032080354e-06,
207
+ "loss": 0.0538,
208
+ "step": 100,
209
+ "student_loss": 0.0007415671134367585,
210
+ "teacher_loss": 0.0005348982522264123
211
+ },
212
+ {
213
+ "epoch": 0.03027681660899654,
214
+ "grad_norm": 1.6328125,
215
+ "kd_loss": 0.035115621984004974,
216
+ "learning_rate": 9.857084319758772e-06,
217
+ "loss": 0.0541,
218
+ "step": 105,
219
+ "student_loss": 0.005187581293284893,
220
+ "teacher_loss": 0.00030648778192698956
221
+ },
222
+ {
223
+ "epoch": 0.031718569780853516,
224
+ "grad_norm": 2.953125,
225
+ "kd_loss": 0.03037514165043831,
226
+ "learning_rate": 9.837227921253747e-06,
227
+ "loss": 0.06,
228
+ "step": 110,
229
+ "student_loss": 0.031953420490026474,
230
+ "teacher_loss": 0.0175339188426733
231
+ },
232
+ {
233
+ "epoch": 0.03316032295271049,
234
+ "grad_norm": 1.484375,
235
+ "kd_loss": 0.023421702906489372,
236
+ "learning_rate": 9.816103043561648e-06,
237
+ "loss": 0.0563,
238
+ "step": 115,
239
+ "student_loss": 0.0006847245967946947,
240
+ "teacher_loss": 0.0003864571626763791
241
+ },
242
+ {
243
+ "epoch": 0.03460207612456748,
244
+ "grad_norm": 1.8203125,
245
+ "kd_loss": 0.018447211012244225,
246
+ "learning_rate": 9.79371522631553e-06,
247
+ "loss": 0.0478,
248
+ "step": 120,
249
+ "student_loss": 0.0015498347347602248,
250
+ "teacher_loss": 0.0010675351368263364
251
+ },
252
+ {
253
+ "epoch": 0.036043829296424454,
254
+ "grad_norm": 1.96875,
255
+ "kd_loss": 0.05630031228065491,
256
+ "learning_rate": 9.770070340332457e-06,
257
+ "loss": 0.0527,
258
+ "step": 125,
259
+ "student_loss": 0.04758762940764427,
260
+ "teacher_loss": 0.0004967509885318577
261
+ },
262
+ {
263
+ "epoch": 0.03748558246828143,
264
+ "grad_norm": 1.859375,
265
+ "kd_loss": 0.018152693286538124,
266
+ "learning_rate": 9.745174586073982e-06,
267
+ "loss": 0.0518,
268
+ "step": 130,
269
+ "student_loss": 0.0008988101035356522,
270
+ "teacher_loss": 0.0007278143893927336
271
+ },
272
+ {
273
+ "epoch": 0.03892733564013841,
274
+ "grad_norm": 1.6640625,
275
+ "kd_loss": 0.030904680490493774,
276
+ "learning_rate": 9.719034492020183e-06,
277
+ "loss": 0.0526,
278
+ "step": 135,
279
+ "student_loss": 0.0127235297113657,
280
+ "teacher_loss": 0.0011665972415357828
281
+ },
282
+ {
283
+ "epoch": 0.040369088811995385,
284
+ "grad_norm": 1.1640625,
285
+ "kd_loss": 0.01759810373187065,
286
+ "learning_rate": 9.691656912957686e-06,
287
+ "loss": 0.0494,
288
+ "step": 140,
289
+ "student_loss": 0.0008300879271700978,
290
+ "teacher_loss": 0.0005129938363097608
291
+ },
292
+ {
293
+ "epoch": 0.04181084198385236,
294
+ "grad_norm": 1.0234375,
295
+ "kd_loss": 0.061179179698228836,
296
+ "learning_rate": 9.663049028182112e-06,
297
+ "loss": 0.0457,
298
+ "step": 145,
299
+ "student_loss": 0.04404008761048317,
300
+ "teacher_loss": 0.009324059821665287
301
+ },
302
+ {
303
+ "epoch": 0.04325259515570934,
304
+ "grad_norm": 1.15625,
305
+ "kd_loss": 0.01291500125080347,
306
+ "learning_rate": 9.633218339615433e-06,
307
+ "loss": 0.0455,
308
+ "step": 150,
309
+ "student_loss": 0.0008404091349802911,
310
+ "teacher_loss": 0.0008450732566416264
311
+ },
312
+ {
313
+ "epoch": 0.04469434832756632,
314
+ "grad_norm": 1.4921875,
315
+ "kd_loss": 0.017479021102190018,
316
+ "learning_rate": 9.602172669838721e-06,
317
+ "loss": 0.0575,
318
+ "step": 155,
319
+ "student_loss": 0.0009151780977845192,
320
+ "teacher_loss": 0.0006974826683290303
321
+ },
322
+ {
323
+ "epoch": 0.0461361014994233,
324
+ "grad_norm": 1.671875,
325
+ "kd_loss": 0.022708255797624588,
326
+ "learning_rate": 9.569920160040815e-06,
327
+ "loss": 0.0503,
328
+ "step": 160,
329
+ "student_loss": 0.07660368084907532,
330
+ "teacher_loss": 0.04039459675550461
331
+ },
332
+ {
333
+ "epoch": 0.04757785467128028,
334
+ "grad_norm": 1.0703125,
335
+ "kd_loss": 0.027437202632427216,
336
+ "learning_rate": 9.536469267883432e-06,
337
+ "loss": 0.0451,
338
+ "step": 165,
339
+ "student_loss": 0.0005943832220509648,
340
+ "teacher_loss": 0.005336699075996876
341
+ },
342
+ {
343
+ "epoch": 0.049019607843137254,
344
+ "grad_norm": 1.71875,
345
+ "kd_loss": 0.014546042308211327,
346
+ "learning_rate": 9.501828765283295e-06,
347
+ "loss": 0.0459,
348
+ "step": 170,
349
+ "student_loss": 0.0006958618760108948,
350
+ "teacher_loss": 0.0005218263831920922
351
+ },
352
+ {
353
+ "epoch": 0.05046136101499423,
354
+ "grad_norm": 1.4453125,
355
+ "kd_loss": 0.02377346158027649,
356
+ "learning_rate": 9.466007736111846e-06,
357
+ "loss": 0.0465,
358
+ "step": 175,
359
+ "student_loss": 0.003214897122234106,
360
+ "teacher_loss": 0.0006713470793329179
361
+ },
362
+ {
363
+ "epoch": 0.05190311418685121,
364
+ "grad_norm": 1.296875,
365
+ "kd_loss": 0.02517828531563282,
366
+ "learning_rate": 9.429015573813163e-06,
367
+ "loss": 0.0454,
368
+ "step": 180,
369
+ "student_loss": 0.0007039233460091054,
370
+ "teacher_loss": 0.0003641469811554998
371
+ },
372
+ {
373
+ "epoch": 0.05334486735870819,
374
+ "grad_norm": 1.25,
375
+ "kd_loss": 0.029188504442572594,
376
+ "learning_rate": 9.390861978940687e-06,
377
+ "loss": 0.0473,
378
+ "step": 185,
379
+ "student_loss": 0.04238048940896988,
380
+ "teacher_loss": 0.00033479384728707373
381
+ },
382
+ {
383
+ "epoch": 0.05478662053056517,
384
+ "grad_norm": 1.484375,
385
+ "kd_loss": 0.03749082237482071,
386
+ "learning_rate": 9.351556956613423e-06,
387
+ "loss": 0.0436,
388
+ "step": 190,
389
+ "student_loss": 0.010270334780216217,
390
+ "teacher_loss": 0.0024021982681006193
391
+ },
392
+ {
393
+ "epoch": 0.056228373702422146,
394
+ "grad_norm": 2.0,
395
+ "kd_loss": 0.030336912721395493,
396
+ "learning_rate": 9.31111081389227e-06,
397
+ "loss": 0.0442,
398
+ "step": 195,
399
+ "student_loss": 0.010534510016441345,
400
+ "teacher_loss": 0.0003890562220476568
401
+ },
402
+ {
403
+ "epoch": 0.05767012687427912,
404
+ "grad_norm": 0.921875,
405
+ "kd_loss": 0.026554003357887268,
406
+ "learning_rate": 9.269534157077177e-06,
407
+ "loss": 0.0497,
408
+ "step": 200,
409
+ "student_loss": 0.0034760867711156607,
410
+ "teacher_loss": 0.00022423896007239819
411
+ },
412
+ {
413
+ "epoch": 0.0591118800461361,
414
+ "grad_norm": 1.9140625,
415
+ "kd_loss": 0.023410219699144363,
416
+ "learning_rate": 9.226837888925813e-06,
417
+ "loss": 0.0471,
418
+ "step": 205,
419
+ "student_loss": 0.03611442819237709,
420
+ "teacher_loss": 0.027781842276453972
421
+ },
422
+ {
423
+ "epoch": 0.06055363321799308,
424
+ "grad_norm": 1.7578125,
425
+ "kd_loss": 0.022867349907755852,
426
+ "learning_rate": 9.183033205794525e-06,
427
+ "loss": 0.0428,
428
+ "step": 210,
429
+ "student_loss": 0.001971556106582284,
430
+ "teacher_loss": 0.0009746765717864037
431
+ },
432
+ {
433
+ "epoch": 0.061995386389850055,
434
+ "grad_norm": 1.0234375,
435
+ "kd_loss": 0.028235338628292084,
436
+ "learning_rate": 9.13813159470227e-06,
437
+ "loss": 0.0449,
438
+ "step": 215,
439
+ "student_loss": 0.0009558866149745882,
440
+ "teacher_loss": 0.0004049334966111928
441
+ },
442
+ {
443
+ "epoch": 0.06343713956170703,
444
+ "grad_norm": 1.4296875,
445
+ "kd_loss": 0.03134317323565483,
446
+ "learning_rate": 9.092144830318357e-06,
447
+ "loss": 0.0438,
448
+ "step": 220,
449
+ "student_loss": 0.048195730894804,
450
+ "teacher_loss": 0.009161758236587048
451
+ },
452
+ {
453
+ "epoch": 0.06487889273356401,
454
+ "grad_norm": 1.1328125,
455
+ "kd_loss": 0.014049015007913113,
456
+ "learning_rate": 9.045084971874738e-06,
457
+ "loss": 0.0487,
458
+ "step": 225,
459
+ "student_loss": 0.05373133346438408,
460
+ "teacher_loss": 0.0009734037448652089
461
+ },
462
+ {
463
+ "epoch": 0.06632064590542099,
464
+ "grad_norm": 1.7265625,
465
+ "kd_loss": 0.021539395675063133,
466
+ "learning_rate": 8.99696436000368e-06,
467
+ "loss": 0.0384,
468
+ "step": 230,
469
+ "student_loss": 0.0008717047749087214,
470
+ "teacher_loss": 0.0028650674503296614
471
+ },
472
+ {
473
+ "epoch": 0.06776239907727798,
474
+ "grad_norm": 1.0390625,
475
+ "kd_loss": 0.045927971601486206,
476
+ "learning_rate": 8.947795613501658e-06,
477
+ "loss": 0.0397,
478
+ "step": 235,
479
+ "student_loss": 0.012143217958509922,
480
+ "teacher_loss": 0.0006723064580000937
481
+ },
482
+ {
483
+ "epoch": 0.06920415224913495,
484
+ "grad_norm": 0.78125,
485
+ "kd_loss": 0.01626853086054325,
486
+ "learning_rate": 8.897591626020284e-06,
487
+ "loss": 0.0425,
488
+ "step": 240,
489
+ "student_loss": 0.0004826921212952584,
490
+ "teacher_loss": 0.0003007515915669501
491
+ },
492
+ {
493
+ "epoch": 0.07064590542099193,
494
+ "grad_norm": 1.078125,
495
+ "kd_loss": 0.010876834392547607,
496
+ "learning_rate": 8.846365562685178e-06,
497
+ "loss": 0.0379,
498
+ "step": 245,
499
+ "student_loss": 0.00123109039850533,
500
+ "teacher_loss": 0.0017208561766892672
501
+ },
502
+ {
503
+ "epoch": 0.07208765859284891,
504
+ "grad_norm": 1.6484375,
505
+ "kd_loss": 0.012208702974021435,
506
+ "learning_rate": 8.794130856643635e-06,
507
+ "loss": 0.0402,
508
+ "step": 250,
509
+ "student_loss": 0.0006701876409351826,
510
+ "teacher_loss": 0.00039402375114150345
511
+ },
512
+ {
513
+ "epoch": 0.07352941176470588,
514
+ "grad_norm": 3.859375,
515
+ "kd_loss": 0.010964477434754372,
516
+ "learning_rate": 8.74090120554202e-06,
517
+ "loss": 0.0428,
518
+ "step": 255,
519
+ "student_loss": 0.0004243800649419427,
520
+ "teacher_loss": 0.0003773768839892
521
+ },
522
+ {
523
+ "epoch": 0.07497116493656286,
524
+ "grad_norm": 2.015625,
525
+ "kd_loss": 0.018098091706633568,
526
+ "learning_rate": 8.686690567933803e-06,
527
+ "loss": 0.0436,
528
+ "step": 260,
529
+ "student_loss": 0.040437813848257065,
530
+ "teacher_loss": 0.032756391912698746
531
+ },
532
+ {
533
+ "epoch": 0.07641291810841984,
534
+ "grad_norm": 1.125,
535
+ "kd_loss": 0.011841603554785252,
536
+ "learning_rate": 8.63151315961915e-06,
537
+ "loss": 0.0412,
538
+ "step": 265,
539
+ "student_loss": 0.0006605261587537825,
540
+ "teacher_loss": 0.0004997120704501867
541
+ },
542
+ {
543
+ "epoch": 0.07785467128027682,
544
+ "grad_norm": 1.046875,
545
+ "kd_loss": 0.01046097930520773,
546
+ "learning_rate": 8.575383449917103e-06,
547
+ "loss": 0.044,
548
+ "step": 270,
549
+ "student_loss": 0.0009966753423213959,
550
+ "teacher_loss": 0.0006801315466873348
551
+ },
552
+ {
553
+ "epoch": 0.07929642445213379,
554
+ "grad_norm": 1.515625,
555
+ "kd_loss": 0.01713527925312519,
556
+ "learning_rate": 8.518316157871232e-06,
557
+ "loss": 0.0403,
558
+ "step": 275,
559
+ "student_loss": 0.04822665825486183,
560
+ "teacher_loss": 0.03450855612754822
561
+ },
562
+ {
563
+ "epoch": 0.08073817762399077,
564
+ "grad_norm": 1.3828125,
565
+ "kd_loss": 0.021968362852931023,
566
+ "learning_rate": 8.460326248389825e-06,
567
+ "loss": 0.0401,
568
+ "step": 280,
569
+ "student_loss": 0.0006340977270156145,
570
+ "teacher_loss": 0.0001483594678575173
571
+ },
572
+ {
573
+ "epoch": 0.08217993079584775,
574
+ "grad_norm": 1.3203125,
575
+ "kd_loss": 0.01131309662014246,
576
+ "learning_rate": 8.401428928321607e-06,
577
+ "loss": 0.0435,
578
+ "step": 285,
579
+ "student_loss": 0.0005278023309074342,
580
+ "teacher_loss": 0.0004103762039449066
581
+ },
582
+ {
583
+ "epoch": 0.08362168396770472,
584
+ "grad_norm": 1.1015625,
585
+ "kd_loss": 0.00958277191966772,
586
+ "learning_rate": 8.341639642468002e-06,
587
+ "loss": 0.0469,
588
+ "step": 290,
589
+ "student_loss": 0.005627856124192476,
590
+ "teacher_loss": 0.0034064538776874542
591
+ },
592
+ {
593
+ "epoch": 0.0850634371395617,
594
+ "grad_norm": 1.140625,
595
+ "kd_loss": 0.009735080413520336,
596
+ "learning_rate": 8.280974069532999e-06,
597
+ "loss": 0.0409,
598
+ "step": 295,
599
+ "student_loss": 0.000509344507008791,
600
+ "teacher_loss": 0.0004550877492874861
601
+ },
602
+ {
603
+ "epoch": 0.08650519031141868,
604
+ "grad_norm": 1.09375,
605
+ "kd_loss": 0.031678151339292526,
606
+ "learning_rate": 8.219448118011687e-06,
607
+ "loss": 0.0379,
608
+ "step": 300,
609
+ "student_loss": 0.016466177999973297,
610
+ "teacher_loss": 0.002865313785150647
611
+ },
612
+ {
613
+ "epoch": 0.08794694348327567,
614
+ "grad_norm": 1.1015625,
615
+ "kd_loss": 0.013346894644200802,
616
+ "learning_rate": 8.157077922018537e-06,
617
+ "loss": 0.0353,
618
+ "step": 305,
619
+ "student_loss": 0.009980602189898491,
620
+ "teacher_loss": 0.0014113991055637598
621
+ },
622
+ {
623
+ "epoch": 0.08938869665513265,
624
+ "grad_norm": 1.90625,
625
+ "kd_loss": 0.010920979082584381,
626
+ "learning_rate": 8.093879837056486e-06,
627
+ "loss": 0.0384,
628
+ "step": 310,
629
+ "student_loss": 0.00036570793599821627,
630
+ "teacher_loss": 0.0002939177502412349
631
+ },
632
+ {
633
+ "epoch": 0.09083044982698962,
634
+ "grad_norm": 1.0078125,
635
+ "kd_loss": 0.021986963227391243,
636
+ "learning_rate": 8.029870435728018e-06,
637
+ "loss": 0.0385,
638
+ "step": 315,
639
+ "student_loss": 0.01466396264731884,
640
+ "teacher_loss": 0.0008777049370110035
641
+ },
642
+ {
643
+ "epoch": 0.0922722029988466,
644
+ "grad_norm": 1.640625,
645
+ "kd_loss": 0.014081098139286041,
646
+ "learning_rate": 7.965066503389264e-06,
647
+ "loss": 0.0417,
648
+ "step": 320,
649
+ "student_loss": 0.0004964058171026409,
650
+ "teacher_loss": 0.0003025036712642759
651
+ },
652
+ {
653
+ "epoch": 0.09371395617070358,
654
+ "grad_norm": 1.2421875,
655
+ "kd_loss": 0.03542112559080124,
656
+ "learning_rate": 7.89948503374835e-06,
657
+ "loss": 0.0378,
658
+ "step": 325,
659
+ "student_loss": 0.01556178368628025,
660
+ "teacher_loss": 0.018082890659570694
661
+ },
662
+ {
663
+ "epoch": 0.09515570934256055,
664
+ "grad_norm": 1.40625,
665
+ "kd_loss": 0.04085119441151619,
666
+ "learning_rate": 7.833143224409076e-06,
667
+ "loss": 0.0394,
668
+ "step": 330,
669
+ "student_loss": 0.024770811200141907,
670
+ "teacher_loss": 0.0021320905070751905
671
+ },
672
+ {
673
+ "epoch": 0.09659746251441753,
674
+ "grad_norm": 1.0234375,
675
+ "kd_loss": 0.008477268740534782,
676
+ "learning_rate": 7.766058472361154e-06,
677
+ "loss": 0.0338,
678
+ "step": 335,
679
+ "student_loss": 0.0009084849734790623,
680
+ "teacher_loss": 0.0007285018800757825
681
+ },
682
+ {
683
+ "epoch": 0.09803921568627451,
684
+ "grad_norm": 1.28125,
685
+ "kd_loss": 0.014947460032999516,
686
+ "learning_rate": 7.698248369418146e-06,
687
+ "loss": 0.0388,
688
+ "step": 340,
689
+ "student_loss": 0.009875037707388401,
690
+ "teacher_loss": 0.00507500022649765
691
+ },
692
+ {
693
+ "epoch": 0.09948096885813149,
694
+ "grad_norm": 1.2421875,
695
+ "kd_loss": 0.019748909398913383,
696
+ "learning_rate": 7.629730697604314e-06,
697
+ "loss": 0.0437,
698
+ "step": 345,
699
+ "student_loss": 0.008698839694261551,
700
+ "teacher_loss": 0.00021470236242748797
701
+ },
702
+ {
703
+ "epoch": 0.10092272202998846,
704
+ "grad_norm": 1.453125,
705
+ "kd_loss": 0.038638580590486526,
706
+ "learning_rate": 7.560523424491595e-06,
707
+ "loss": 0.0333,
708
+ "step": 350,
709
+ "student_loss": 0.007182905916124582,
710
+ "teacher_loss": 0.0004611952172126621
711
+ },
712
+ {
713
+ "epoch": 0.10236447520184544,
714
+ "grad_norm": 1.4375,
715
+ "kd_loss": 0.008461177349090576,
716
+ "learning_rate": 7.490644698487909e-06,
717
+ "loss": 0.0405,
718
+ "step": 355,
719
+ "student_loss": 0.0004199718532618135,
720
+ "teacher_loss": 0.00038222281727939844
721
+ },
722
+ {
723
+ "epoch": 0.10380622837370242,
724
+ "grad_norm": 1.1796875,
725
+ "kd_loss": 0.008598582819104195,
726
+ "learning_rate": 7.420112844078066e-06,
727
+ "loss": 0.0396,
728
+ "step": 360,
729
+ "student_loss": 0.00045404830598272383,
730
+ "teacher_loss": 0.0003997418680228293
731
+ },
732
+ {
733
+ "epoch": 0.1052479815455594,
734
+ "grad_norm": 1.78125,
735
+ "kd_loss": 0.018130140379071236,
736
+ "learning_rate": 7.348946357018479e-06,
737
+ "loss": 0.0425,
738
+ "step": 365,
739
+ "student_loss": 0.06869062036275864,
740
+ "teacher_loss": 0.027944643050432205
741
+ },
742
+ {
743
+ "epoch": 0.10668973471741638,
744
+ "grad_norm": 1.03125,
745
+ "kd_loss": 0.0350988432765007,
746
+ "learning_rate": 7.277163899486975e-06,
747
+ "loss": 0.0447,
748
+ "step": 370,
749
+ "student_loss": 0.015085455030202866,
750
+ "teacher_loss": 0.0004648211761377752
751
+ },
752
+ {
753
+ "epoch": 0.10813148788927336,
754
+ "grad_norm": 1.0703125,
755
+ "kd_loss": 0.01613399013876915,
756
+ "learning_rate": 7.204784295188959e-06,
757
+ "loss": 0.0365,
758
+ "step": 375,
759
+ "student_loss": 0.03554360568523407,
760
+ "teacher_loss": 0.021264519542455673
761
+ },
762
+ {
763
+ "epoch": 0.10957324106113034,
764
+ "grad_norm": 1.296875,
765
+ "kd_loss": 0.010999550111591816,
766
+ "learning_rate": 7.1318265244212305e-06,
767
+ "loss": 0.0381,
768
+ "step": 380,
769
+ "student_loss": 0.0004269553755875677,
770
+ "teacher_loss": 0.0003071910759899765
771
+ },
772
+ {
773
+ "epoch": 0.11101499423298732,
774
+ "grad_norm": 0.875,
775
+ "kd_loss": 0.015028773806989193,
776
+ "learning_rate": 7.05830971909472e-06,
777
+ "loss": 0.0349,
778
+ "step": 385,
779
+ "student_loss": 0.0005910908803343773,
780
+ "teacher_loss": 0.0003781717678066343
781
+ },
782
+ {
783
+ "epoch": 0.11245674740484429,
784
+ "grad_norm": 1.25,
785
+ "kd_loss": 0.010998338460922241,
786
+ "learning_rate": 6.9842531577174865e-06,
787
+ "loss": 0.033,
788
+ "step": 390,
789
+ "student_loss": 0.00025347189512103796,
790
+ "teacher_loss": 0.00023755666916258633
791
+ },
792
+ {
793
+ "epoch": 0.11389850057670127,
794
+ "grad_norm": 1.390625,
795
+ "kd_loss": 0.010839108377695084,
796
+ "learning_rate": 6.9096762603392595e-06,
797
+ "loss": 0.0393,
798
+ "step": 395,
799
+ "student_loss": 0.00046008970821276307,
800
+ "teacher_loss": 0.00034296512603759766
801
+ },
802
+ {
803
+ "epoch": 0.11534025374855825,
804
+ "grad_norm": 2.40625,
805
+ "kd_loss": 0.015689171850681305,
806
+ "learning_rate": 6.834598583458862e-06,
807
+ "loss": 0.0333,
808
+ "step": 400,
809
+ "student_loss": 0.0004951858427375555,
810
+ "teacher_loss": 0.0002840181696228683
811
+ },
812
+ {
813
+ "epoch": 0.11678200692041522,
814
+ "grad_norm": 1.328125,
815
+ "kd_loss": 0.014689328148961067,
816
+ "learning_rate": 6.7590398148958625e-06,
817
+ "loss": 0.0425,
818
+ "step": 405,
819
+ "student_loss": 0.005328211933374405,
820
+ "teacher_loss": 0.0002102136058965698
821
+ },
822
+ {
823
+ "epoch": 0.1182237600922722,
824
+ "grad_norm": 1.4609375,
825
+ "kd_loss": 0.026721687987446785,
826
+ "learning_rate": 6.6830197686277945e-06,
827
+ "loss": 0.0417,
828
+ "step": 410,
829
+ "student_loss": 0.015544440597295761,
830
+ "teacher_loss": 0.00868156086653471
831
+ },
832
+ {
833
+ "epoch": 0.11966551326412918,
834
+ "grad_norm": 1.1640625,
835
+ "kd_loss": 0.0165933258831501,
836
+ "learning_rate": 6.6065583795942625e-06,
837
+ "loss": 0.0409,
838
+ "step": 415,
839
+ "student_loss": 0.0018743074033409357,
840
+ "teacher_loss": 0.002361958846449852
841
+ },
842
+ {
843
+ "epoch": 0.12110726643598616,
844
+ "grad_norm": 1.34375,
845
+ "kd_loss": 0.02428903616964817,
846
+ "learning_rate": 6.52967569846937e-06,
847
+ "loss": 0.0376,
848
+ "step": 420,
849
+ "student_loss": 0.05027158558368683,
850
+ "teacher_loss": 0.022369280457496643
851
+ },
852
+ {
853
+ "epoch": 0.12254901960784313,
854
+ "grad_norm": 1.171875,
855
+ "kd_loss": 0.00989749375730753,
856
+ "learning_rate": 6.452391886403767e-06,
857
+ "loss": 0.0366,
858
+ "step": 425,
859
+ "student_loss": 0.001526529318653047,
860
+ "teacher_loss": 0.001954666106030345
861
+ },
862
+ {
863
+ "epoch": 0.12399077277970011,
864
+ "grad_norm": 1.40625,
865
+ "kd_loss": 0.025572113692760468,
866
+ "learning_rate": 6.374727209737743e-06,
867
+ "loss": 0.0457,
868
+ "step": 430,
869
+ "student_loss": 0.05225735530257225,
870
+ "teacher_loss": 0.06261400878429413
871
+ },
872
+ {
873
+ "epoch": 0.1254325259515571,
874
+ "grad_norm": 1.1953125,
875
+ "kd_loss": 0.009243296459317207,
876
+ "learning_rate": 6.296702034686726e-06,
877
+ "loss": 0.0351,
878
+ "step": 435,
879
+ "student_loss": 0.0006208799313753843,
880
+ "teacher_loss": 0.0005131844663992524
881
+ },
882
+ {
883
+ "epoch": 0.12687427912341406,
884
+ "grad_norm": 1.140625,
885
+ "kd_loss": 0.025882720947265625,
886
+ "learning_rate": 6.218336822000598e-06,
887
+ "loss": 0.0366,
888
+ "step": 440,
889
+ "student_loss": 0.022235309705138206,
890
+ "teacher_loss": 0.005665617063641548
891
+ },
892
+ {
893
+ "epoch": 0.12831603229527105,
894
+ "grad_norm": 1.5546875,
895
+ "kd_loss": 0.020934544503688812,
896
+ "learning_rate": 6.139652121598219e-06,
897
+ "loss": 0.0381,
898
+ "step": 445,
899
+ "student_loss": 0.02492375858128071,
900
+ "teacher_loss": 0.01704920269548893
901
+ },
902
+ {
903
+ "epoch": 0.12975778546712802,
904
+ "grad_norm": 1.078125,
905
+ "kd_loss": 0.021503543481230736,
906
+ "learning_rate": 6.060668567178561e-06,
907
+ "loss": 0.039,
908
+ "step": 450,
909
+ "student_loss": 0.0008401675149798393,
910
+ "teacher_loss": 0.0007966886623762548
911
+ },
912
+ {
913
+ "epoch": 0.131199538638985,
914
+ "grad_norm": 0.99609375,
915
+ "kd_loss": 0.011403748765587807,
916
+ "learning_rate": 5.981406870809889e-06,
917
+ "loss": 0.0383,
918
+ "step": 455,
919
+ "student_loss": 0.0014924939023330808,
920
+ "teacher_loss": 0.000647007254883647
921
+ },
922
+ {
923
+ "epoch": 0.13264129181084197,
924
+ "grad_norm": 1.515625,
925
+ "kd_loss": 0.013661489821970463,
926
+ "learning_rate": 5.9018878174983674e-06,
927
+ "loss": 0.0329,
928
+ "step": 460,
929
+ "student_loss": 0.0015512986574321985,
930
+ "teacher_loss": 0.0012551237596198916
931
+ },
932
+ {
933
+ "epoch": 0.13408304498269896,
934
+ "grad_norm": 1.265625,
935
+ "kd_loss": 0.020031817257404327,
936
+ "learning_rate": 5.822132259737565e-06,
937
+ "loss": 0.0399,
938
+ "step": 465,
939
+ "student_loss": 0.0005617028218694031,
940
+ "teacher_loss": 0.0006807463942095637
941
+ },
942
+ {
943
+ "epoch": 0.13552479815455595,
944
+ "grad_norm": 1.171875,
945
+ "kd_loss": 0.012775675393640995,
946
+ "learning_rate": 5.742161112040237e-06,
947
+ "loss": 0.0377,
948
+ "step": 470,
949
+ "student_loss": 0.00029168292530812323,
950
+ "teacher_loss": 0.00020419349311850965
951
+ },
952
+ {
953
+ "epoch": 0.13696655132641292,
954
+ "grad_norm": 0.87109375,
955
+ "kd_loss": 0.016445137560367584,
956
+ "learning_rate": 5.661995345453867e-06,
957
+ "loss": 0.0315,
958
+ "step": 475,
959
+ "student_loss": 0.0005847346619702876,
960
+ "teacher_loss": 0.0003634036984294653
961
+ },
962
+ {
963
+ "epoch": 0.1384083044982699,
964
+ "grad_norm": 0.95703125,
965
+ "kd_loss": 0.014534696005284786,
966
+ "learning_rate": 5.581655982061367e-06,
967
+ "loss": 0.0466,
968
+ "step": 480,
969
+ "student_loss": 0.0203537754714489,
970
+ "teacher_loss": 0.014939186163246632
971
+ },
972
+ {
973
+ "epoch": 0.13985005767012687,
974
+ "grad_norm": 1.1640625,
975
+ "kd_loss": 0.008392157033085823,
976
+ "learning_rate": 5.501164089468406e-06,
977
+ "loss": 0.0357,
978
+ "step": 485,
979
+ "student_loss": 0.0004224316217005253,
980
+ "teacher_loss": 0.00044211489148437977
981
+ },
982
+ {
983
+ "epoch": 0.14129181084198386,
984
+ "grad_norm": 1.8984375,
985
+ "kd_loss": 0.014247741550207138,
986
+ "learning_rate": 5.4205407752787884e-06,
987
+ "loss": 0.0428,
988
+ "step": 490,
989
+ "student_loss": 0.019939009100198746,
990
+ "teacher_loss": 0.009562309831380844
991
+ },
992
+ {
993
+ "epoch": 0.14273356401384082,
994
+ "grad_norm": 1.828125,
995
+ "kd_loss": 0.026821549981832504,
996
+ "learning_rate": 5.339807181559359e-06,
997
+ "loss": 0.0484,
998
+ "step": 495,
999
+ "student_loss": 0.01085620652884245,
1000
+ "teacher_loss": 0.001887582358904183
1001
+ },
1002
+ {
1003
+ "epoch": 0.14417531718569782,
1004
+ "grad_norm": 1.03125,
1005
+ "kd_loss": 0.011890986934304237,
1006
+ "learning_rate": 5.258984479295853e-06,
1007
+ "loss": 0.0366,
1008
+ "step": 500,
1009
+ "student_loss": 0.0007122078677639365,
1010
+ "teacher_loss": 0.0006097567966207862
1011
+ },
1012
+ {
1013
+ "epoch": 0.14561707035755478,
1014
+ "grad_norm": 1.0,
1015
+ "kd_loss": 0.01012763474136591,
1016
+ "learning_rate": 5.1780938628411795e-06,
1017
+ "loss": 0.0433,
1018
+ "step": 505,
1019
+ "student_loss": 0.0004120942030567676,
1020
+ "teacher_loss": 0.00030555034754797816
1021
+ },
1022
+ {
1023
+ "epoch": 0.14705882352941177,
1024
+ "grad_norm": 1.90625,
1025
+ "kd_loss": 0.010170744732022285,
1026
+ "learning_rate": 5.097156544357567e-06,
1027
+ "loss": 0.0382,
1028
+ "step": 510,
1029
+ "student_loss": 0.0002881335385609418,
1030
+ "teacher_loss": 0.000244438968366012
1031
+ },
1032
+ {
1033
+ "epoch": 0.14850057670126873,
1034
+ "grad_norm": 1.359375,
1035
+ "kd_loss": 0.0138414166867733,
1036
+ "learning_rate": 5.016193748254045e-06,
1037
+ "loss": 0.031,
1038
+ "step": 515,
1039
+ "student_loss": 0.005790602881461382,
1040
+ "teacher_loss": 0.0001912783191073686
1041
+ },
1042
+ {
1043
+ "epoch": 0.14994232987312572,
1044
+ "grad_norm": 1.9921875,
1045
+ "kd_loss": 0.03539569675922394,
1046
+ "learning_rate": 4.935226705620699e-06,
1047
+ "loss": 0.0397,
1048
+ "step": 520,
1049
+ "student_loss": 0.04242030531167984,
1050
+ "teacher_loss": 0.011500299908220768
1051
+ },
1052
+ {
1053
+ "epoch": 0.1513840830449827,
1054
+ "grad_norm": 0.67578125,
1055
+ "kd_loss": 0.018779262900352478,
1056
+ "learning_rate": 4.8542766486612035e-06,
1057
+ "loss": 0.0403,
1058
+ "step": 525,
1059
+ "student_loss": 0.00041962284012697637,
1060
+ "teacher_loss": 0.00021298698266036808
1061
+ },
1062
+ {
1063
+ "epoch": 0.15282583621683968,
1064
+ "grad_norm": 1.2421875,
1065
+ "kd_loss": 0.01180377509444952,
1066
+ "learning_rate": 4.773364805125025e-06,
1067
+ "loss": 0.0351,
1068
+ "step": 530,
1069
+ "student_loss": 0.00039459849358536303,
1070
+ "teacher_loss": 0.0003135094593744725
1071
+ },
1072
+ {
1073
+ "epoch": 0.15426758938869667,
1074
+ "grad_norm": 0.96484375,
1075
+ "kd_loss": 0.022030461579561234,
1076
+ "learning_rate": 4.6925123927408265e-06,
1077
+ "loss": 0.0343,
1078
+ "step": 535,
1079
+ "student_loss": 0.000637000659480691,
1080
+ "teacher_loss": 0.00034665301791392267
1081
+ },
1082
+ {
1083
+ "epoch": 0.15570934256055363,
1084
+ "grad_norm": 1.234375,
1085
+ "kd_loss": 0.01367215160280466,
1086
+ "learning_rate": 4.611740613652485e-06,
1087
+ "loss": 0.0333,
1088
+ "step": 540,
1089
+ "student_loss": 0.0006559583125635982,
1090
+ "teacher_loss": 0.0003204621316399425
1091
+ },
1092
+ {
1093
+ "epoch": 0.15715109573241062,
1094
+ "grad_norm": 2.234375,
1095
+ "kd_loss": 0.00888274796307087,
1096
+ "learning_rate": 4.531070648859186e-06,
1097
+ "loss": 0.0319,
1098
+ "step": 545,
1099
+ "student_loss": 0.01682436652481556,
1100
+ "teacher_loss": 0.013372495770454407
1101
+ },
1102
+ {
1103
+ "epoch": 0.15859284890426759,
1104
+ "grad_norm": 1.640625,
1105
+ "kd_loss": 0.026771951466798782,
1106
+ "learning_rate": 4.450523652661086e-06,
1107
+ "loss": 0.0306,
1108
+ "step": 550,
1109
+ "student_loss": 0.0001725157053442672,
1110
+ "teacher_loss": 0.012681440450251102
1111
+ },
1112
+ {
1113
+ "epoch": 0.16003460207612458,
1114
+ "grad_norm": 1.1015625,
1115
+ "kd_loss": 0.03767453879117966,
1116
+ "learning_rate": 4.370120747111956e-06,
1117
+ "loss": 0.0381,
1118
+ "step": 555,
1119
+ "student_loss": 0.009898228570818901,
1120
+ "teacher_loss": 0.0022960191126912832
1121
+ },
1122
+ {
1123
+ "epoch": 0.16147635524798154,
1124
+ "grad_norm": 1.734375,
1125
+ "kd_loss": 0.0351857915520668,
1126
+ "learning_rate": 4.289883016480291e-06,
1127
+ "loss": 0.0381,
1128
+ "step": 560,
1129
+ "student_loss": 0.027657803148031235,
1130
+ "teacher_loss": 0.00034597894409671426
1131
+ },
1132
+ {
1133
+ "epoch": 0.16291810841983853,
1134
+ "grad_norm": 1.4296875,
1135
+ "kd_loss": 0.019546959549188614,
1136
+ "learning_rate": 4.209831501720328e-06,
1137
+ "loss": 0.0346,
1138
+ "step": 565,
1139
+ "student_loss": 0.028599649667739868,
1140
+ "teacher_loss": 0.018787631765007973
1141
+ },
1142
+ {
1143
+ "epoch": 0.1643598615916955,
1144
+ "grad_norm": 1.5625,
1145
+ "kd_loss": 0.020644016563892365,
1146
+ "learning_rate": 4.129987194954421e-06,
1147
+ "loss": 0.0405,
1148
+ "step": 570,
1149
+ "student_loss": 0.010047786869108677,
1150
+ "teacher_loss": 0.006790643557906151
1151
+ },
1152
+ {
1153
+ "epoch": 0.16580161476355249,
1154
+ "grad_norm": 1.4921875,
1155
+ "kd_loss": 0.01332971639931202,
1156
+ "learning_rate": 4.050371033968216e-06,
1157
+ "loss": 0.0376,
1158
+ "step": 575,
1159
+ "student_loss": 0.0004119194345548749,
1160
+ "teacher_loss": 0.00025748086045496166
1161
+ },
1162
+ {
1163
+ "epoch": 0.16724336793540945,
1164
+ "grad_norm": 1.3515625,
1165
+ "kd_loss": 0.012037251144647598,
1166
+ "learning_rate": 3.9710038967200825e-06,
1167
+ "loss": 0.0342,
1168
+ "step": 580,
1169
+ "student_loss": 0.0030028573237359524,
1170
+ "teacher_loss": 0.0012405890738591552
1171
+ },
1172
+ {
1173
+ "epoch": 0.16868512110726644,
1174
+ "grad_norm": 0.88671875,
1175
+ "kd_loss": 0.01703016459941864,
1176
+ "learning_rate": 3.89190659586623e-06,
1177
+ "loss": 0.032,
1178
+ "step": 585,
1179
+ "student_loss": 0.0007975402986630797,
1180
+ "teacher_loss": 0.0002710323315113783
1181
+ },
1182
+ {
1183
+ "epoch": 0.1701268742791234,
1184
+ "grad_norm": 1.1171875,
1185
+ "kd_loss": 0.01272827759385109,
1186
+ "learning_rate": 3.8130998733029517e-06,
1187
+ "loss": 0.035,
1188
+ "step": 590,
1189
+ "student_loss": 0.000885407323949039,
1190
+ "teacher_loss": 0.00043826864566653967
1191
+ },
1192
+ {
1193
+ "epoch": 0.1715686274509804,
1194
+ "grad_norm": 1.3828125,
1195
+ "kd_loss": 0.009466774761676788,
1196
+ "learning_rate": 3.734604394727419e-06,
1197
+ "loss": 0.0345,
1198
+ "step": 595,
1199
+ "student_loss": 0.0005723866051994264,
1200
+ "teacher_loss": 0.0005249618552625179
1201
+ },
1202
+ {
1203
+ "epoch": 0.17301038062283736,
1204
+ "grad_norm": 1.4765625,
1205
+ "kd_loss": 0.020295076072216034,
1206
+ "learning_rate": 3.656440744218464e-06,
1207
+ "loss": 0.04,
1208
+ "step": 600,
1209
+ "student_loss": 0.04926498606801033,
1210
+ "teacher_loss": 0.00894973799586296
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/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-600/vocab.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-800/added_tokens.json ADDED
@@ -0,0 +1,35 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "</tool_call>": 151658,
3
+ "<tool_call>": 151657,
4
+ "<|action_sep|>": 151670,
5
+ "<|arg_sep|>": 151671,
6
+ "<|box_end|>": 151649,
7
+ "<|box_start|>": 151648,
8
+ "<|call_sep|>": 151666,
9
+ "<|end_of_text|>": 151673,
10
+ "<|endoftext|>": 151643,
11
+ "<|exception_sep|>": 151669,
12
+ "<|file_sep|>": 151664,
13
+ "<|fim_middle|>": 151660,
14
+ "<|fim_pad|>": 151662,
15
+ "<|fim_prefix|>": 151659,
16
+ "<|fim_suffix|>": 151661,
17
+ "<|frame_sep|>": 151672,
18
+ "<|im_end|>": 151645,
19
+ "<|im_start|>": 151644,
20
+ "<|image_pad|>": 151655,
21
+ "<|latent_end|>": 151675,
22
+ "<|latent_start|>": 151674,
23
+ "<|line_sep|>": 151667,
24
+ "<|object_ref_end|>": 151647,
25
+ "<|object_ref_start|>": 151646,
26
+ "<|quad_end|>": 151651,
27
+ "<|quad_start|>": 151650,
28
+ "<|repo_name|>": 151663,
29
+ "<|return_sep|>": 151668,
30
+ "<|trace_context_start|>": 151665,
31
+ "<|video_pad|>": 151656,
32
+ "<|vision_end|>": 151653,
33
+ "<|vision_pad|>": 151654,
34
+ "<|vision_start|>": 151652
35
+ }
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-800/chat_template.jinja ADDED
@@ -0,0 +1,54 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {%- if tools %}
2
+ {{- '<|im_start|>system\n' }}
3
+ {%- if messages[0]['role'] == 'system' %}
4
+ {{- messages[0]['content'] }}
5
+ {%- else %}
6
+ {{- 'You are Qwen, created by Alibaba Cloud. You are a helpful assistant.' }}
7
+ {%- endif %}
8
+ {{- "\n\n# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
9
+ {%- for tool in tools %}
10
+ {{- "\n" }}
11
+ {{- tool | tojson }}
12
+ {%- endfor %}
13
+ {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
14
+ {%- else %}
15
+ {%- if messages[0]['role'] == 'system' %}
16
+ {{- '<|im_start|>system\n' + messages[0]['content'] + '<|im_end|>\n' }}
17
+ {%- else %}
18
+ {{- '<|im_start|>system\nYou are Qwen, created by Alibaba Cloud. You are a helpful assistant.<|im_end|>\n' }}
19
+ {%- endif %}
20
+ {%- endif %}
21
+ {%- for message in messages %}
22
+ {%- if (message.role == "user") or (message.role == "system" and not loop.first) or (message.role == "assistant" and not message.tool_calls) %}
23
+ {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
24
+ {%- elif message.role == "assistant" %}
25
+ {{- '<|im_start|>' + message.role }}
26
+ {%- if message.content %}
27
+ {{- '\n' + message.content }}
28
+ {%- endif %}
29
+ {%- for tool_call in message.tool_calls %}
30
+ {%- if tool_call.function is defined %}
31
+ {%- set tool_call = tool_call.function %}
32
+ {%- endif %}
33
+ {{- '\n<tool_call>\n{"name": "' }}
34
+ {{- tool_call.name }}
35
+ {{- '", "arguments": ' }}
36
+ {{- tool_call.arguments | tojson }}
37
+ {{- '}\n</tool_call>' }}
38
+ {%- endfor %}
39
+ {{- '<|im_end|>\n' }}
40
+ {%- elif message.role == "tool" %}
41
+ {%- if (loop.index0 == 0) or (messages[loop.index0 - 1].role != "tool") %}
42
+ {{- '<|im_start|>user' }}
43
+ {%- endif %}
44
+ {{- '\n<tool_response>\n' }}
45
+ {{- message.content }}
46
+ {{- '\n</tool_response>' }}
47
+ {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
48
+ {{- '<|im_end|>\n' }}
49
+ {%- endif %}
50
+ {%- endif %}
51
+ {%- endfor %}
52
+ {%- if add_generation_prompt %}
53
+ {{- '<|im_start|>assistant\n' }}
54
+ {%- endif %}
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-800/config.json ADDED
@@ -0,0 +1,58 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Qwen2ForCausalLM"
4
+ ],
5
+ "attention_dropout": 0.0,
6
+ "dtype": "bfloat16",
7
+ "eos_token_id": 151643,
8
+ "hidden_act": "silu",
9
+ "hidden_size": 1536,
10
+ "initializer_range": 0.02,
11
+ "intermediate_size": 8960,
12
+ "layer_types": [
13
+ "full_attention",
14
+ "full_attention",
15
+ "full_attention",
16
+ "full_attention",
17
+ "full_attention",
18
+ "full_attention",
19
+ "full_attention",
20
+ "full_attention",
21
+ "full_attention",
22
+ "full_attention",
23
+ "full_attention",
24
+ "full_attention",
25
+ "full_attention",
26
+ "full_attention",
27
+ "full_attention",
28
+ "full_attention",
29
+ "full_attention",
30
+ "full_attention",
31
+ "full_attention",
32
+ "full_attention",
33
+ "full_attention",
34
+ "full_attention",
35
+ "full_attention",
36
+ "full_attention",
37
+ "full_attention",
38
+ "full_attention",
39
+ "full_attention",
40
+ "full_attention"
41
+ ],
42
+ "max_position_embeddings": 32768,
43
+ "max_window_layers": 28,
44
+ "model_type": "qwen2",
45
+ "num_attention_heads": 12,
46
+ "num_hidden_layers": 28,
47
+ "num_key_value_heads": 2,
48
+ "pad_token_id": 151643,
49
+ "rms_norm_eps": 1e-06,
50
+ "rope_scaling": null,
51
+ "rope_theta": 1000000.0,
52
+ "sliding_window": null,
53
+ "tie_word_embeddings": true,
54
+ "transformers_version": "4.57.6",
55
+ "use_cache": true,
56
+ "use_sliding_window": false,
57
+ "vocab_size": 151676
58
+ }
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-800/merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
checkpoints/codi1.5b_a1.0_b1.0_g1.0_ls2/checkpoint-800/pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:660adf38ba4f7e0e04fe43e4f496edca9831d68d262d9fb06c3fa4aff27cad64
3
+ size 3096275515