Siddeshwar1625 commited on
Commit
b84dda3
·
verified ·
1 Parent(s): 67a1fe6

Intermediate checkpoint upload step=120 (answerer_train)

Browse files
.gitattributes CHANGED
@@ -60,3 +60,4 @@ self_play_hf_l40s_full/round_002/generator_train/final_model/tokenizer.json filt
60
  self_play_hf_l40s_full/round_002/answerer_train/checkpoint-30/tokenizer.json filter=lfs diff=lfs merge=lfs -text
61
  self_play_hf_l40s_full/round_002/answerer_train/checkpoint-60/tokenizer.json filter=lfs diff=lfs merge=lfs -text
62
  self_play_hf_l40s_full/round_002/answerer_train/checkpoint-90/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 
 
60
  self_play_hf_l40s_full/round_002/answerer_train/checkpoint-30/tokenizer.json filter=lfs diff=lfs merge=lfs -text
61
  self_play_hf_l40s_full/round_002/answerer_train/checkpoint-60/tokenizer.json filter=lfs diff=lfs merge=lfs -text
62
  self_play_hf_l40s_full/round_002/answerer_train/checkpoint-90/tokenizer.json filter=lfs diff=lfs merge=lfs -text
63
+ self_play_hf_l40s_full/round_002/answerer_train/checkpoint-120/tokenizer.json filter=lfs diff=lfs merge=lfs -text
self_play_hf_l40s_full/round_002/answerer_train/checkpoint-120/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 %}
self_play_hf_l40s_full/round_002/answerer_train/checkpoint-120/config.json ADDED
@@ -0,0 +1,57 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Qwen2ForCausalLM"
4
+ ],
5
+ "attention_dropout": 0.0,
6
+ "bos_token_id": null,
7
+ "dtype": "float32",
8
+ "eos_token_id": 151645,
9
+ "hidden_act": "silu",
10
+ "hidden_size": 896,
11
+ "initializer_range": 0.02,
12
+ "intermediate_size": 4864,
13
+ "layer_types": [
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
+ ],
39
+ "max_position_embeddings": 32768,
40
+ "max_window_layers": 21,
41
+ "model_type": "qwen2",
42
+ "num_attention_heads": 14,
43
+ "num_hidden_layers": 24,
44
+ "num_key_value_heads": 2,
45
+ "pad_token_id": 151643,
46
+ "rms_norm_eps": 1e-06,
47
+ "rope_parameters": {
48
+ "rope_theta": 1000000.0,
49
+ "rope_type": "default"
50
+ },
51
+ "sliding_window": null,
52
+ "tie_word_embeddings": true,
53
+ "transformers_version": "5.6.2",
54
+ "use_cache": false,
55
+ "use_sliding_window": false,
56
+ "vocab_size": 151936
57
+ }
self_play_hf_l40s_full/round_002/answerer_train/checkpoint-120/generation_config.json ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "do_sample": true,
3
+ "eos_token_id": [
4
+ 151645,
5
+ 151643
6
+ ],
7
+ "pad_token_id": 151643,
8
+ "repetition_penalty": 1.1,
9
+ "temperature": 0.7,
10
+ "top_k": 20,
11
+ "top_p": 0.8,
12
+ "transformers_version": "5.6.2"
13
+ }
self_play_hf_l40s_full/round_002/answerer_train/checkpoint-120/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:097a7b338a8299a40adb572018acf6eb77c485cd1c71368fb868e24a256728db
3
+ size 1976163472
self_play_hf_l40s_full/round_002/answerer_train/checkpoint-120/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ffd0822c8f0a10caf59bac4769cf62cc5496fd26dfe45477b78ade09e90072af
3
+ size 3952509771
self_play_hf_l40s_full/round_002/answerer_train/checkpoint-120/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:09d387fab291cf55ab5d51fbc665b5b617f39b612148cbfb2c8bec740f61e07b
3
+ size 14645
self_play_hf_l40s_full/round_002/answerer_train/checkpoint-120/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7278566d3e343c7c551e937aa2102c1aa1c1b0a72103eeaeaa73c012e20ea6df
3
+ size 1465
self_play_hf_l40s_full/round_002/answerer_train/checkpoint-120/tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3fd169731d2cbde95e10bf356d66d5997fd885dd8dbb6fb4684da3f23b2585d8
3
+ size 11421892
self_play_hf_l40s_full/round_002/answerer_train/checkpoint-120/tokenizer_config.json ADDED
@@ -0,0 +1,32 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_prefix_space": false,
3
+ "backend": "tokenizers",
4
+ "bos_token": null,
5
+ "clean_up_tokenization_spaces": false,
6
+ "eos_token": "<|im_end|>",
7
+ "errors": "replace",
8
+ "extra_special_tokens": [
9
+ "<|im_start|>",
10
+ "<|im_end|>",
11
+ "<|object_ref_start|>",
12
+ "<|object_ref_end|>",
13
+ "<|box_start|>",
14
+ "<|box_end|>",
15
+ "<|quad_start|>",
16
+ "<|quad_end|>",
17
+ "<|vision_start|>",
18
+ "<|vision_end|>",
19
+ "<|vision_pad|>",
20
+ "<|image_pad|>",
21
+ "<|video_pad|>"
22
+ ],
23
+ "is_local": true,
24
+ "local_files_only": false,
25
+ "model_max_length": 131072,
26
+ "pad_token": "<|endoftext|>",
27
+ "padding_side": "left",
28
+ "split_special_tokens": false,
29
+ "tokenizer_class": "Qwen2Tokenizer",
30
+ "truncation_side": "left",
31
+ "unk_token": null
32
+ }
self_play_hf_l40s_full/round_002/answerer_train/checkpoint-120/trainer_state.json ADDED
@@ -0,0 +1,2224 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 3.0,
6
+ "eval_steps": 500,
7
+ "global_step": 120,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "clip_ratio/high_max": 0.0,
14
+ "clip_ratio/high_mean": 0.0,
15
+ "clip_ratio/low_mean": 0.0,
16
+ "clip_ratio/low_min": 0.0,
17
+ "clip_ratio/region_mean": 0.0,
18
+ "completions/clipped_ratio": 1.0,
19
+ "completions/max_length": 256.0,
20
+ "completions/max_terminated_length": 0.0,
21
+ "completions/mean_length": 256.0,
22
+ "completions/mean_terminated_length": 0.0,
23
+ "completions/min_length": 256.0,
24
+ "completions/min_terminated_length": 0.0,
25
+ "entropy": 1.1812694072723389,
26
+ "epoch": 0.025,
27
+ "frac_reward_zero_std": 0.0,
28
+ "grad_norm": 3.56195330619812,
29
+ "kl": 0.0,
30
+ "learning_rate": 3e-06,
31
+ "loss": 0.12478145956993103,
32
+ "num_tokens": 9980.0,
33
+ "reward": 0.49252849817276,
34
+ "reward_std": 0.12672561407089233,
35
+ "rewards/AnswererRewardFunction/mean": 0.49252849817276,
36
+ "rewards/AnswererRewardFunction/std": 0.12672561407089233,
37
+ "step": 1,
38
+ "step_time": 8.06768666500011
39
+ },
40
+ {
41
+ "clip_ratio/high_max": 0.0244140625,
42
+ "clip_ratio/high_mean": 0.0244140625,
43
+ "clip_ratio/low_mean": 0.03125,
44
+ "clip_ratio/low_min": 0.03125,
45
+ "clip_ratio/region_mean": 0.0556640625,
46
+ "entropy": 1.0541718006134033,
47
+ "epoch": 0.05,
48
+ "grad_norm": 3.2018611431121826,
49
+ "kl": 0.07901028543710709,
50
+ "learning_rate": 2.9750000000000003e-06,
51
+ "loss": -0.11482742428779602,
52
+ "step": 2,
53
+ "step_time": 0.117666047000057
54
+ },
55
+ {
56
+ "clip_ratio/high_max": 0.0341796875,
57
+ "clip_ratio/high_mean": 0.0341796875,
58
+ "clip_ratio/low_mean": 0.056640625,
59
+ "clip_ratio/low_min": 0.056640625,
60
+ "clip_ratio/region_mean": 0.0908203125,
61
+ "entropy": 1.2944715023040771,
62
+ "epoch": 0.075,
63
+ "grad_norm": 3.514223575592041,
64
+ "kl": 0.32789888978004456,
65
+ "learning_rate": 2.9499999999999997e-06,
66
+ "loss": 0.06260380148887634,
67
+ "step": 3,
68
+ "step_time": 0.11041338499990161
69
+ },
70
+ {
71
+ "clip_ratio/high_max": 0.0400390625,
72
+ "clip_ratio/high_mean": 0.0400390625,
73
+ "clip_ratio/low_mean": 0.05859375,
74
+ "clip_ratio/low_min": 0.05859375,
75
+ "clip_ratio/region_mean": 0.0986328125,
76
+ "entropy": 1.2256050109863281,
77
+ "epoch": 0.1,
78
+ "grad_norm": 3.100332260131836,
79
+ "kl": 0.21089623868465424,
80
+ "learning_rate": 2.925e-06,
81
+ "loss": -0.05488990247249603,
82
+ "step": 4,
83
+ "step_time": 0.11791598300033002
84
+ },
85
+ {
86
+ "clip_ratio/high_max": 0.0,
87
+ "clip_ratio/high_mean": 0.0,
88
+ "clip_ratio/low_mean": 0.0,
89
+ "clip_ratio/low_min": 0.0,
90
+ "clip_ratio/region_mean": 0.0,
91
+ "completions/clipped_ratio": 1.0,
92
+ "completions/max_length": 256.0,
93
+ "completions/max_terminated_length": 0.0,
94
+ "completions/mean_length": 256.0,
95
+ "completions/mean_terminated_length": 0.0,
96
+ "completions/min_length": 256.0,
97
+ "completions/min_terminated_length": 0.0,
98
+ "entropy": 0.769734263420105,
99
+ "epoch": 0.125,
100
+ "frac_reward_zero_std": 0.5,
101
+ "grad_norm": 0.6963396668434143,
102
+ "kl": 0.026502244174480438,
103
+ "learning_rate": 2.9e-06,
104
+ "loss": 0.03130355477333069,
105
+ "num_tokens": 19780.0,
106
+ "reward": 0.4761159420013428,
107
+ "reward_std": 0.14743122458457947,
108
+ "rewards/AnswererRewardFunction/mean": 0.4761159420013428,
109
+ "rewards/AnswererRewardFunction/std": 0.14743123948574066,
110
+ "step": 5,
111
+ "step_time": 7.967737137000313
112
+ },
113
+ {
114
+ "clip_ratio/high_max": 0.0126953125,
115
+ "clip_ratio/high_mean": 0.0126953125,
116
+ "clip_ratio/low_mean": 0.0029296875,
117
+ "clip_ratio/low_min": 0.0029296875,
118
+ "clip_ratio/region_mean": 0.015625,
119
+ "entropy": 1.4138596057891846,
120
+ "epoch": 0.15,
121
+ "grad_norm": 3.238894462585449,
122
+ "kl": 0.047499917447566986,
123
+ "learning_rate": 2.875e-06,
124
+ "loss": -0.05275709182024002,
125
+ "step": 6,
126
+ "step_time": 0.1440943980001066
127
+ },
128
+ {
129
+ "clip_ratio/high_max": 0.0,
130
+ "clip_ratio/high_mean": 0.0,
131
+ "clip_ratio/low_mean": 0.005859375,
132
+ "clip_ratio/low_min": 0.005859375,
133
+ "clip_ratio/region_mean": 0.005859375,
134
+ "entropy": 1.1906815767288208,
135
+ "epoch": 0.175,
136
+ "grad_norm": 1.7127066850662231,
137
+ "kl": 0.06469351053237915,
138
+ "learning_rate": 2.85e-06,
139
+ "loss": 0.05530916154384613,
140
+ "step": 7,
141
+ "step_time": 0.11163456800022686
142
+ },
143
+ {
144
+ "clip_ratio/high_max": 0.0166015625,
145
+ "clip_ratio/high_mean": 0.0166015625,
146
+ "clip_ratio/low_mean": 0.0576171875,
147
+ "clip_ratio/low_min": 0.0576171875,
148
+ "clip_ratio/region_mean": 0.07421875,
149
+ "entropy": 1.191636323928833,
150
+ "epoch": 0.2,
151
+ "grad_norm": 2.787623643875122,
152
+ "kl": 0.048633381724357605,
153
+ "learning_rate": 2.825e-06,
154
+ "loss": -0.03133530914783478,
155
+ "step": 8,
156
+ "step_time": 0.11240155399991636
157
+ },
158
+ {
159
+ "clip_ratio/high_max": 0.0,
160
+ "clip_ratio/high_mean": 0.0,
161
+ "clip_ratio/low_mean": 0.0,
162
+ "clip_ratio/low_min": 0.0,
163
+ "clip_ratio/region_mean": 0.0,
164
+ "completions/clipped_ratio": 1.0,
165
+ "completions/max_length": 256.0,
166
+ "completions/max_terminated_length": 0.0,
167
+ "completions/mean_length": 256.0,
168
+ "completions/mean_terminated_length": 0.0,
169
+ "completions/min_length": 256.0,
170
+ "completions/min_terminated_length": 0.0,
171
+ "entropy": 1.061751127243042,
172
+ "epoch": 0.225,
173
+ "frac_reward_zero_std": 0.25,
174
+ "grad_norm": 3.077267646789551,
175
+ "kl": 0.06591981649398804,
176
+ "learning_rate": 2.8000000000000003e-06,
177
+ "loss": 0.025028277188539505,
178
+ "num_tokens": 29540.0,
179
+ "reward": 0.530171275138855,
180
+ "reward_std": 0.12436192482709885,
181
+ "rewards/AnswererRewardFunction/mean": 0.530171275138855,
182
+ "rewards/AnswererRewardFunction/std": 0.12436193227767944,
183
+ "step": 9,
184
+ "step_time": 8.047823582000092
185
+ },
186
+ {
187
+ "clip_ratio/high_max": 0.0029296875,
188
+ "clip_ratio/high_mean": 0.0029296875,
189
+ "clip_ratio/low_mean": 0.0029296875,
190
+ "clip_ratio/low_min": 0.0029296875,
191
+ "clip_ratio/region_mean": 0.005859375,
192
+ "entropy": 1.1717921495437622,
193
+ "epoch": 0.25,
194
+ "grad_norm": 1.6465528011322021,
195
+ "kl": 0.08923725038766861,
196
+ "learning_rate": 2.775e-06,
197
+ "loss": 0.016560696065425873,
198
+ "step": 10,
199
+ "step_time": 0.11130034100006014
200
+ },
201
+ {
202
+ "clip_ratio/high_max": 0.01171875,
203
+ "clip_ratio/high_mean": 0.01171875,
204
+ "clip_ratio/low_mean": 0.0166015625,
205
+ "clip_ratio/low_min": 0.0166015625,
206
+ "clip_ratio/region_mean": 0.0283203125,
207
+ "entropy": 1.2988202571868896,
208
+ "epoch": 0.275,
209
+ "grad_norm": 4.246798515319824,
210
+ "kl": 0.1189119964838028,
211
+ "learning_rate": 2.75e-06,
212
+ "loss": 0.0765266865491867,
213
+ "step": 11,
214
+ "step_time": 0.1099475139999413
215
+ },
216
+ {
217
+ "clip_ratio/high_max": 0.0244140625,
218
+ "clip_ratio/high_mean": 0.0244140625,
219
+ "clip_ratio/low_mean": 0.0,
220
+ "clip_ratio/low_min": 0.0,
221
+ "clip_ratio/region_mean": 0.0244140625,
222
+ "entropy": 1.2800407409667969,
223
+ "epoch": 0.3,
224
+ "grad_norm": 2.2184503078460693,
225
+ "kl": 0.12053709477186203,
226
+ "learning_rate": 2.725e-06,
227
+ "loss": -0.11435238271951675,
228
+ "step": 12,
229
+ "step_time": 0.11112866800021948
230
+ },
231
+ {
232
+ "clip_ratio/high_max": 0.0009765625,
233
+ "clip_ratio/high_mean": 0.0009765625,
234
+ "clip_ratio/low_mean": 0.0,
235
+ "clip_ratio/low_min": 0.0,
236
+ "clip_ratio/region_mean": 0.0009765625,
237
+ "completions/clipped_ratio": 1.0,
238
+ "completions/max_length": 256.0,
239
+ "completions/max_terminated_length": 0.0,
240
+ "completions/mean_length": 256.0,
241
+ "completions/mean_terminated_length": 0.0,
242
+ "completions/min_length": 256.0,
243
+ "completions/min_terminated_length": 0.0,
244
+ "entropy": 1.1048344373703003,
245
+ "epoch": 0.325,
246
+ "frac_reward_zero_std": 0.0,
247
+ "grad_norm": 2.5493807792663574,
248
+ "kl": 0.10315192490816116,
249
+ "learning_rate": 2.7e-06,
250
+ "loss": -0.14281664788722992,
251
+ "num_tokens": 39576.0,
252
+ "reward": 0.5976681113243103,
253
+ "reward_std": 0.13136300444602966,
254
+ "rewards/AnswererRewardFunction/mean": 0.5976681113243103,
255
+ "rewards/AnswererRewardFunction/std": 0.13136300444602966,
256
+ "step": 13,
257
+ "step_time": 8.073091011000088
258
+ },
259
+ {
260
+ "clip_ratio/high_max": 0.0078125,
261
+ "clip_ratio/high_mean": 0.0078125,
262
+ "clip_ratio/low_mean": 0.001953125,
263
+ "clip_ratio/low_min": 0.001953125,
264
+ "clip_ratio/region_mean": 0.009765625,
265
+ "entropy": 1.0296664237976074,
266
+ "epoch": 0.35,
267
+ "grad_norm": 3.767220973968506,
268
+ "kl": 0.07366035133600235,
269
+ "learning_rate": 2.6750000000000002e-06,
270
+ "loss": 0.05409611761569977,
271
+ "step": 14,
272
+ "step_time": 0.11164900099993247
273
+ },
274
+ {
275
+ "clip_ratio/high_max": 0.0400390625,
276
+ "clip_ratio/high_mean": 0.0400390625,
277
+ "clip_ratio/low_mean": 0.0048828125,
278
+ "clip_ratio/low_min": 0.0048828125,
279
+ "clip_ratio/region_mean": 0.044921875,
280
+ "entropy": 1.0058298110961914,
281
+ "epoch": 0.375,
282
+ "grad_norm": 2.3384571075439453,
283
+ "kl": 0.08770984411239624,
284
+ "learning_rate": 2.65e-06,
285
+ "loss": -0.029884662479162216,
286
+ "step": 15,
287
+ "step_time": 0.11125424400006523
288
+ },
289
+ {
290
+ "clip_ratio/high_max": 0.03125,
291
+ "clip_ratio/high_mean": 0.03125,
292
+ "clip_ratio/low_mean": 0.0087890625,
293
+ "clip_ratio/low_min": 0.0087890625,
294
+ "clip_ratio/region_mean": 0.0400390625,
295
+ "entropy": 1.0768989324569702,
296
+ "epoch": 0.4,
297
+ "grad_norm": 4.1971917152404785,
298
+ "kl": 0.10156311094760895,
299
+ "learning_rate": 2.6250000000000003e-06,
300
+ "loss": 0.12216067314147949,
301
+ "step": 16,
302
+ "step_time": 0.11126710500002446
303
+ },
304
+ {
305
+ "clip_ratio/high_max": 0.0,
306
+ "clip_ratio/high_mean": 0.0,
307
+ "clip_ratio/low_mean": 0.0,
308
+ "clip_ratio/low_min": 0.0,
309
+ "clip_ratio/region_mean": 0.0,
310
+ "completions/clipped_ratio": 1.0,
311
+ "completions/max_length": 256.0,
312
+ "completions/max_terminated_length": 0.0,
313
+ "completions/mean_length": 256.0,
314
+ "completions/mean_terminated_length": 0.0,
315
+ "completions/min_length": 256.0,
316
+ "completions/min_terminated_length": 0.0,
317
+ "entropy": 0.9591922760009766,
318
+ "epoch": 0.425,
319
+ "frac_reward_zero_std": 0.0,
320
+ "grad_norm": 3.6558687686920166,
321
+ "kl": 0.07310992479324341,
322
+ "learning_rate": 2.6e-06,
323
+ "loss": -0.07063402235507965,
324
+ "num_tokens": 49936.0,
325
+ "reward": 0.5328114032745361,
326
+ "reward_std": 0.13155435025691986,
327
+ "rewards/AnswererRewardFunction/mean": 0.5328114032745361,
328
+ "rewards/AnswererRewardFunction/std": 0.13155436515808105,
329
+ "step": 17,
330
+ "step_time": 8.044399631000033
331
+ },
332
+ {
333
+ "clip_ratio/high_max": 0.0,
334
+ "clip_ratio/high_mean": 0.0,
335
+ "clip_ratio/low_mean": 0.0107421875,
336
+ "clip_ratio/low_min": 0.0107421875,
337
+ "clip_ratio/region_mean": 0.0107421875,
338
+ "entropy": 0.7852188944816589,
339
+ "epoch": 0.45,
340
+ "grad_norm": 2.870166540145874,
341
+ "kl": 0.13418066501617432,
342
+ "learning_rate": 2.575e-06,
343
+ "loss": 0.1557493656873703,
344
+ "step": 18,
345
+ "step_time": 0.11281930699988152
346
+ },
347
+ {
348
+ "clip_ratio/high_max": 0.01953125,
349
+ "clip_ratio/high_mean": 0.01953125,
350
+ "clip_ratio/low_mean": 0.0126953125,
351
+ "clip_ratio/low_min": 0.0126953125,
352
+ "clip_ratio/region_mean": 0.0322265625,
353
+ "entropy": 1.009390115737915,
354
+ "epoch": 0.475,
355
+ "grad_norm": 2.2228500843048096,
356
+ "kl": 0.11547495424747467,
357
+ "learning_rate": 2.55e-06,
358
+ "loss": -0.05351074039936066,
359
+ "step": 19,
360
+ "step_time": 0.1127072539998153
361
+ },
362
+ {
363
+ "clip_ratio/high_max": 0.015625,
364
+ "clip_ratio/high_mean": 0.015625,
365
+ "clip_ratio/low_mean": 0.03125,
366
+ "clip_ratio/low_min": 0.03125,
367
+ "clip_ratio/region_mean": 0.046875,
368
+ "entropy": 1.210716962814331,
369
+ "epoch": 0.5,
370
+ "grad_norm": 3.261709213256836,
371
+ "kl": 0.17568016052246094,
372
+ "learning_rate": 2.525e-06,
373
+ "loss": -0.03016088344156742,
374
+ "step": 20,
375
+ "step_time": 0.11370289600017713
376
+ },
377
+ {
378
+ "clip_ratio/high_max": 0.0,
379
+ "clip_ratio/high_mean": 0.0,
380
+ "clip_ratio/low_mean": 0.0,
381
+ "clip_ratio/low_min": 0.0,
382
+ "clip_ratio/region_mean": 0.0,
383
+ "completions/clipped_ratio": 1.0,
384
+ "completions/max_length": 256.0,
385
+ "completions/max_terminated_length": 0.0,
386
+ "completions/mean_length": 256.0,
387
+ "completions/mean_terminated_length": 0.0,
388
+ "completions/min_length": 256.0,
389
+ "completions/min_terminated_length": 0.0,
390
+ "entropy": 1.0023118257522583,
391
+ "epoch": 0.525,
392
+ "frac_reward_zero_std": 0.0,
393
+ "grad_norm": 3.9749913215637207,
394
+ "kl": 0.2008446753025055,
395
+ "learning_rate": 2.5e-06,
396
+ "loss": -0.1282912790775299,
397
+ "num_tokens": 60108.0,
398
+ "reward": 0.6376597881317139,
399
+ "reward_std": 0.1541692167520523,
400
+ "rewards/AnswererRewardFunction/mean": 0.6376597881317139,
401
+ "rewards/AnswererRewardFunction/std": 0.1541692018508911,
402
+ "step": 21,
403
+ "step_time": 7.979511771000034
404
+ },
405
+ {
406
+ "clip_ratio/high_max": 0.0009765625,
407
+ "clip_ratio/high_mean": 0.0009765625,
408
+ "clip_ratio/low_mean": 0.0078125,
409
+ "clip_ratio/low_min": 0.0078125,
410
+ "clip_ratio/region_mean": 0.0087890625,
411
+ "entropy": 0.9689664840698242,
412
+ "epoch": 0.55,
413
+ "grad_norm": 1.8822933435440063,
414
+ "kl": 0.1251663863658905,
415
+ "learning_rate": 2.475e-06,
416
+ "loss": 0.06288938969373703,
417
+ "step": 22,
418
+ "step_time": 0.11127889299996241
419
+ },
420
+ {
421
+ "clip_ratio/high_max": 0.0068359375,
422
+ "clip_ratio/high_mean": 0.0068359375,
423
+ "clip_ratio/low_mean": 0.0205078125,
424
+ "clip_ratio/low_min": 0.0205078125,
425
+ "clip_ratio/region_mean": 0.02734375,
426
+ "entropy": 0.9676262140274048,
427
+ "epoch": 0.575,
428
+ "grad_norm": 3.6209120750427246,
429
+ "kl": 0.1173456609249115,
430
+ "learning_rate": 2.45e-06,
431
+ "loss": 0.07742585241794586,
432
+ "step": 23,
433
+ "step_time": 0.11072158999968451
434
+ },
435
+ {
436
+ "clip_ratio/high_max": 0.017578125,
437
+ "clip_ratio/high_mean": 0.017578125,
438
+ "clip_ratio/low_mean": 0.0283203125,
439
+ "clip_ratio/low_min": 0.0283203125,
440
+ "clip_ratio/region_mean": 0.0458984375,
441
+ "entropy": 0.9444276094436646,
442
+ "epoch": 0.6,
443
+ "grad_norm": 2.653747320175171,
444
+ "kl": 0.19196566939353943,
445
+ "learning_rate": 2.425e-06,
446
+ "loss": -0.010297899134457111,
447
+ "step": 24,
448
+ "step_time": 0.11332796499982578
449
+ },
450
+ {
451
+ "clip_ratio/high_max": 0.0,
452
+ "clip_ratio/high_mean": 0.0,
453
+ "clip_ratio/low_mean": 0.0,
454
+ "clip_ratio/low_min": 0.0,
455
+ "clip_ratio/region_mean": 0.0,
456
+ "completions/clipped_ratio": 1.0,
457
+ "completions/max_length": 256.0,
458
+ "completions/max_terminated_length": 0.0,
459
+ "completions/mean_length": 256.0,
460
+ "completions/mean_terminated_length": 0.0,
461
+ "completions/min_length": 256.0,
462
+ "completions/min_terminated_length": 0.0,
463
+ "entropy": 1.0263278484344482,
464
+ "epoch": 0.625,
465
+ "frac_reward_zero_std": 0.0,
466
+ "grad_norm": 3.7951908111572266,
467
+ "kl": 0.13174787163734436,
468
+ "learning_rate": 2.4000000000000003e-06,
469
+ "loss": -0.0016784188337624073,
470
+ "num_tokens": 70020.0,
471
+ "reward": 0.7374376058578491,
472
+ "reward_std": 0.1732868254184723,
473
+ "rewards/AnswererRewardFunction/mean": 0.7374376058578491,
474
+ "rewards/AnswererRewardFunction/std": 0.1732868254184723,
475
+ "step": 25,
476
+ "step_time": 7.9698032609999245
477
+ },
478
+ {
479
+ "clip_ratio/high_max": 0.0029296875,
480
+ "clip_ratio/high_mean": 0.0029296875,
481
+ "clip_ratio/low_mean": 0.0048828125,
482
+ "clip_ratio/low_min": 0.0048828125,
483
+ "clip_ratio/region_mean": 0.0078125,
484
+ "entropy": 1.0938024520874023,
485
+ "epoch": 0.65,
486
+ "grad_norm": 1.8259835243225098,
487
+ "kl": 0.15534526109695435,
488
+ "learning_rate": 2.375e-06,
489
+ "loss": 0.06504079699516296,
490
+ "step": 26,
491
+ "step_time": 0.11241862999986552
492
+ },
493
+ {
494
+ "clip_ratio/high_max": 0.0126953125,
495
+ "clip_ratio/high_mean": 0.0126953125,
496
+ "clip_ratio/low_mean": 0.0087890625,
497
+ "clip_ratio/low_min": 0.0087890625,
498
+ "clip_ratio/region_mean": 0.021484375,
499
+ "entropy": 0.9784930348396301,
500
+ "epoch": 0.675,
501
+ "grad_norm": 1.7331888675689697,
502
+ "kl": 0.1776759773492813,
503
+ "learning_rate": 2.35e-06,
504
+ "loss": -0.06545304507017136,
505
+ "step": 27,
506
+ "step_time": 0.11266273999990517
507
+ },
508
+ {
509
+ "clip_ratio/high_max": 0.0224609375,
510
+ "clip_ratio/high_mean": 0.0224609375,
511
+ "clip_ratio/low_mean": 0.0185546875,
512
+ "clip_ratio/low_min": 0.0185546875,
513
+ "clip_ratio/region_mean": 0.041015625,
514
+ "entropy": 0.8833986520767212,
515
+ "epoch": 0.7,
516
+ "grad_norm": 3.654691219329834,
517
+ "kl": 0.17693831026554108,
518
+ "learning_rate": 2.325e-06,
519
+ "loss": 0.005695457104593515,
520
+ "step": 28,
521
+ "step_time": 0.11081353399958971
522
+ },
523
+ {
524
+ "clip_ratio/high_max": 0.0,
525
+ "clip_ratio/high_mean": 0.0,
526
+ "clip_ratio/low_mean": 0.0009765625,
527
+ "clip_ratio/low_min": 0.0009765625,
528
+ "clip_ratio/region_mean": 0.0009765625,
529
+ "completions/clipped_ratio": 1.0,
530
+ "completions/max_length": 256.0,
531
+ "completions/max_terminated_length": 0.0,
532
+ "completions/mean_length": 256.0,
533
+ "completions/mean_terminated_length": 0.0,
534
+ "completions/min_length": 256.0,
535
+ "completions/min_terminated_length": 0.0,
536
+ "entropy": 0.9469633102416992,
537
+ "epoch": 0.725,
538
+ "frac_reward_zero_std": 0.0,
539
+ "grad_norm": 4.039062023162842,
540
+ "kl": 0.1610972285270691,
541
+ "learning_rate": 2.3000000000000004e-06,
542
+ "loss": 0.12820656597614288,
543
+ "num_tokens": 80300.0,
544
+ "reward": 0.77821946144104,
545
+ "reward_std": 0.19161619246006012,
546
+ "rewards/AnswererRewardFunction/mean": 0.77821946144104,
547
+ "rewards/AnswererRewardFunction/std": 0.19161619246006012,
548
+ "step": 29,
549
+ "step_time": 8.035392143000081
550
+ },
551
+ {
552
+ "clip_ratio/high_max": 0.0068359375,
553
+ "clip_ratio/high_mean": 0.0068359375,
554
+ "clip_ratio/low_mean": 0.0009765625,
555
+ "clip_ratio/low_min": 0.0009765625,
556
+ "clip_ratio/region_mean": 0.0078125,
557
+ "entropy": 1.054987907409668,
558
+ "epoch": 0.75,
559
+ "grad_norm": 3.0317585468292236,
560
+ "kl": 0.1419621706008911,
561
+ "learning_rate": 2.275e-06,
562
+ "loss": -0.03360540792346001,
563
+ "step": 30,
564
+ "step_time": 0.11271856700022909
565
+ },
566
+ {
567
+ "clip_ratio/high_max": 0.0126953125,
568
+ "clip_ratio/high_mean": 0.0126953125,
569
+ "clip_ratio/low_mean": 0.0068359375,
570
+ "clip_ratio/low_min": 0.0068359375,
571
+ "clip_ratio/region_mean": 0.01953125,
572
+ "entropy": 0.9814310073852539,
573
+ "epoch": 0.775,
574
+ "grad_norm": 2.4117178916931152,
575
+ "kl": 0.1381104737520218,
576
+ "learning_rate": 2.25e-06,
577
+ "loss": -0.002541737863793969,
578
+ "step": 31,
579
+ "step_time": 0.11368492900010096
580
+ },
581
+ {
582
+ "clip_ratio/high_max": 0.0146484375,
583
+ "clip_ratio/high_mean": 0.0146484375,
584
+ "clip_ratio/low_mean": 0.0078125,
585
+ "clip_ratio/low_min": 0.0078125,
586
+ "clip_ratio/region_mean": 0.0224609375,
587
+ "entropy": 0.907570481300354,
588
+ "epoch": 0.8,
589
+ "grad_norm": 1.7884786128997803,
590
+ "kl": 0.16731491684913635,
591
+ "learning_rate": 2.2250000000000003e-06,
592
+ "loss": -0.08919669687747955,
593
+ "step": 32,
594
+ "step_time": 0.11238900299986199
595
+ },
596
+ {
597
+ "clip_ratio/high_max": 0.0009765625,
598
+ "clip_ratio/high_mean": 0.0009765625,
599
+ "clip_ratio/low_mean": 0.0,
600
+ "clip_ratio/low_min": 0.0,
601
+ "clip_ratio/region_mean": 0.0009765625,
602
+ "completions/clipped_ratio": 1.0,
603
+ "completions/max_length": 256.0,
604
+ "completions/max_terminated_length": 0.0,
605
+ "completions/mean_length": 256.0,
606
+ "completions/mean_terminated_length": 0.0,
607
+ "completions/min_length": 256.0,
608
+ "completions/min_terminated_length": 0.0,
609
+ "entropy": 0.9146954417228699,
610
+ "epoch": 0.825,
611
+ "frac_reward_zero_std": 0.0,
612
+ "grad_norm": 3.307687520980835,
613
+ "kl": 0.19084078073501587,
614
+ "learning_rate": 2.1999999999999997e-06,
615
+ "loss": 0.0015747896395623684,
616
+ "num_tokens": 90656.0,
617
+ "reward": 0.7868335843086243,
618
+ "reward_std": 0.16629737615585327,
619
+ "rewards/AnswererRewardFunction/mean": 0.7868335843086243,
620
+ "rewards/AnswererRewardFunction/std": 0.16629737615585327,
621
+ "step": 33,
622
+ "step_time": 7.996074866000072
623
+ },
624
+ {
625
+ "clip_ratio/high_max": 0.001953125,
626
+ "clip_ratio/high_mean": 0.001953125,
627
+ "clip_ratio/low_mean": 0.0029296875,
628
+ "clip_ratio/low_min": 0.0029296875,
629
+ "clip_ratio/region_mean": 0.0048828125,
630
+ "entropy": 1.0972727537155151,
631
+ "epoch": 0.85,
632
+ "grad_norm": 4.319361686706543,
633
+ "kl": 0.18110132217407227,
634
+ "learning_rate": 2.175e-06,
635
+ "loss": 0.11823011934757233,
636
+ "step": 34,
637
+ "step_time": 0.11284558099987407
638
+ },
639
+ {
640
+ "clip_ratio/high_max": 0.0078125,
641
+ "clip_ratio/high_mean": 0.0078125,
642
+ "clip_ratio/low_mean": 0.0009765625,
643
+ "clip_ratio/low_min": 0.0009765625,
644
+ "clip_ratio/region_mean": 0.0087890625,
645
+ "entropy": 0.9452131986618042,
646
+ "epoch": 0.875,
647
+ "grad_norm": 2.554110288619995,
648
+ "kl": 0.14194214344024658,
649
+ "learning_rate": 2.15e-06,
650
+ "loss": 0.0009212729637511075,
651
+ "step": 35,
652
+ "step_time": 0.11485822300028303
653
+ },
654
+ {
655
+ "clip_ratio/high_max": 0.021484375,
656
+ "clip_ratio/high_mean": 0.021484375,
657
+ "clip_ratio/low_mean": 0.0068359375,
658
+ "clip_ratio/low_min": 0.0068359375,
659
+ "clip_ratio/region_mean": 0.0283203125,
660
+ "entropy": 1.001446008682251,
661
+ "epoch": 0.9,
662
+ "grad_norm": 2.39662504196167,
663
+ "kl": 0.17090469598770142,
664
+ "learning_rate": 2.125e-06,
665
+ "loss": -0.11863560974597931,
666
+ "step": 36,
667
+ "step_time": 0.1127581209998425
668
+ },
669
+ {
670
+ "clip_ratio/high_max": 0.0,
671
+ "clip_ratio/high_mean": 0.0,
672
+ "clip_ratio/low_mean": 0.0,
673
+ "clip_ratio/low_min": 0.0,
674
+ "clip_ratio/region_mean": 0.0,
675
+ "completions/clipped_ratio": 1.0,
676
+ "completions/max_length": 256.0,
677
+ "completions/max_terminated_length": 0.0,
678
+ "completions/mean_length": 256.0,
679
+ "completions/mean_terminated_length": 0.0,
680
+ "completions/min_length": 256.0,
681
+ "completions/min_terminated_length": 0.0,
682
+ "entropy": 0.8905212879180908,
683
+ "epoch": 0.925,
684
+ "frac_reward_zero_std": 0.5,
685
+ "grad_norm": 2.1671974658966064,
686
+ "kl": 0.1887168288230896,
687
+ "learning_rate": 2.1e-06,
688
+ "loss": -0.10587899386882782,
689
+ "num_tokens": 100516.0,
690
+ "reward": 0.7836564779281616,
691
+ "reward_std": 0.16003844141960144,
692
+ "rewards/AnswererRewardFunction/mean": 0.7836564779281616,
693
+ "rewards/AnswererRewardFunction/std": 0.16003844141960144,
694
+ "step": 37,
695
+ "step_time": 7.983559749000051
696
+ },
697
+ {
698
+ "clip_ratio/high_max": 0.005859375,
699
+ "clip_ratio/high_mean": 0.005859375,
700
+ "clip_ratio/low_mean": 0.0,
701
+ "clip_ratio/low_min": 0.0,
702
+ "clip_ratio/region_mean": 0.005859375,
703
+ "entropy": 0.9539603590965271,
704
+ "epoch": 0.95,
705
+ "grad_norm": 2.3658673763275146,
706
+ "kl": 0.21254020929336548,
707
+ "learning_rate": 2.075e-06,
708
+ "loss": -0.009321342222392559,
709
+ "step": 38,
710
+ "step_time": 0.11240537300000142
711
+ },
712
+ {
713
+ "clip_ratio/high_max": 0.017578125,
714
+ "clip_ratio/high_mean": 0.017578125,
715
+ "clip_ratio/low_mean": 0.0,
716
+ "clip_ratio/low_min": 0.0,
717
+ "clip_ratio/region_mean": 0.017578125,
718
+ "entropy": 0.8658829927444458,
719
+ "epoch": 0.975,
720
+ "grad_norm": 0.928003191947937,
721
+ "kl": 0.2214215099811554,
722
+ "learning_rate": 2.0500000000000003e-06,
723
+ "loss": -0.03125650808215141,
724
+ "step": 39,
725
+ "step_time": 0.11115311799994743
726
+ },
727
+ {
728
+ "clip_ratio/high_max": 0.0,
729
+ "clip_ratio/high_mean": 0.0,
730
+ "clip_ratio/low_mean": 0.009765625,
731
+ "clip_ratio/low_min": 0.009765625,
732
+ "clip_ratio/region_mean": 0.009765625,
733
+ "entropy": 0.8934024572372437,
734
+ "epoch": 1.0,
735
+ "grad_norm": 3.346177339553833,
736
+ "kl": 0.15183529257774353,
737
+ "learning_rate": 2.025e-06,
738
+ "loss": 0.1518702208995819,
739
+ "step": 40,
740
+ "step_time": 0.11065240799962339
741
+ },
742
+ {
743
+ "clip_ratio/high_max": 0.0,
744
+ "clip_ratio/high_mean": 0.0,
745
+ "clip_ratio/low_mean": 0.0,
746
+ "clip_ratio/low_min": 0.0,
747
+ "clip_ratio/region_mean": 0.0,
748
+ "completions/clipped_ratio": 1.0,
749
+ "completions/max_length": 256.0,
750
+ "completions/max_terminated_length": 0.0,
751
+ "completions/mean_length": 256.0,
752
+ "completions/mean_terminated_length": 0.0,
753
+ "completions/min_length": 256.0,
754
+ "completions/min_terminated_length": 0.0,
755
+ "entropy": 1.1226850748062134,
756
+ "epoch": 1.025,
757
+ "frac_reward_zero_std": 0.0,
758
+ "grad_norm": 3.663853883743286,
759
+ "kl": 0.2066955268383026,
760
+ "learning_rate": 2e-06,
761
+ "loss": 0.038323815912008286,
762
+ "num_tokens": 110068.0,
763
+ "reward": 0.6620498299598694,
764
+ "reward_std": 0.10584933310747147,
765
+ "rewards/AnswererRewardFunction/mean": 0.6620498299598694,
766
+ "rewards/AnswererRewardFunction/std": 0.10584934055805206,
767
+ "step": 41,
768
+ "step_time": 8.124446219999754
769
+ },
770
+ {
771
+ "clip_ratio/high_max": 0.0029296875,
772
+ "clip_ratio/high_mean": 0.0029296875,
773
+ "clip_ratio/low_mean": 0.005859375,
774
+ "clip_ratio/low_min": 0.005859375,
775
+ "clip_ratio/region_mean": 0.0087890625,
776
+ "entropy": 1.200155258178711,
777
+ "epoch": 1.05,
778
+ "grad_norm": 4.07336950302124,
779
+ "kl": 0.20725010335445404,
780
+ "learning_rate": 1.975e-06,
781
+ "loss": 0.17892315983772278,
782
+ "step": 42,
783
+ "step_time": 0.11061930799996844
784
+ },
785
+ {
786
+ "clip_ratio/high_max": 0.015625,
787
+ "clip_ratio/high_mean": 0.015625,
788
+ "clip_ratio/low_mean": 0.0,
789
+ "clip_ratio/low_min": 0.0,
790
+ "clip_ratio/region_mean": 0.015625,
791
+ "entropy": 1.1768546104431152,
792
+ "epoch": 1.075,
793
+ "grad_norm": 2.9877402782440186,
794
+ "kl": 0.21286916732788086,
795
+ "learning_rate": 1.95e-06,
796
+ "loss": -0.17067846655845642,
797
+ "step": 43,
798
+ "step_time": 0.10934286300016538
799
+ },
800
+ {
801
+ "clip_ratio/high_max": 0.015625,
802
+ "clip_ratio/high_mean": 0.015625,
803
+ "clip_ratio/low_mean": 0.0029296875,
804
+ "clip_ratio/low_min": 0.0029296875,
805
+ "clip_ratio/region_mean": 0.0185546875,
806
+ "entropy": 1.133467674255371,
807
+ "epoch": 1.1,
808
+ "grad_norm": 2.2016777992248535,
809
+ "kl": 0.1699901670217514,
810
+ "learning_rate": 1.925e-06,
811
+ "loss": -0.04436863958835602,
812
+ "step": 44,
813
+ "step_time": 0.10889889400004904
814
+ },
815
+ {
816
+ "clip_ratio/high_max": 0.0,
817
+ "clip_ratio/high_mean": 0.0,
818
+ "clip_ratio/low_mean": 0.0,
819
+ "clip_ratio/low_min": 0.0,
820
+ "clip_ratio/region_mean": 0.0,
821
+ "completions/clipped_ratio": 1.0,
822
+ "completions/max_length": 256.0,
823
+ "completions/max_terminated_length": 0.0,
824
+ "completions/mean_length": 256.0,
825
+ "completions/mean_terminated_length": 0.0,
826
+ "completions/min_length": 256.0,
827
+ "completions/min_terminated_length": 0.0,
828
+ "entropy": 1.4023334980010986,
829
+ "epoch": 1.125,
830
+ "frac_reward_zero_std": 0.5,
831
+ "grad_norm": 1.5270278453826904,
832
+ "kl": 0.17242750525474548,
833
+ "learning_rate": 1.9e-06,
834
+ "loss": -0.05998625606298447,
835
+ "num_tokens": 120076.0,
836
+ "reward": 0.7482740879058838,
837
+ "reward_std": 0.13831442594528198,
838
+ "rewards/AnswererRewardFunction/mean": 0.7482740879058838,
839
+ "rewards/AnswererRewardFunction/std": 0.13831445574760437,
840
+ "step": 45,
841
+ "step_time": 8.025359671999922
842
+ },
843
+ {
844
+ "clip_ratio/high_max": 0.0068359375,
845
+ "clip_ratio/high_mean": 0.0068359375,
846
+ "clip_ratio/low_mean": 0.0,
847
+ "clip_ratio/low_min": 0.0,
848
+ "clip_ratio/region_mean": 0.0068359375,
849
+ "entropy": 1.1902995109558105,
850
+ "epoch": 1.15,
851
+ "grad_norm": 2.9169154167175293,
852
+ "kl": 0.1808495670557022,
853
+ "learning_rate": 1.875e-06,
854
+ "loss": -0.12556925415992737,
855
+ "step": 46,
856
+ "step_time": 0.11420444699979271
857
+ },
858
+ {
859
+ "clip_ratio/high_max": 0.0,
860
+ "clip_ratio/high_mean": 0.0,
861
+ "clip_ratio/low_mean": 0.005859375,
862
+ "clip_ratio/low_min": 0.005859375,
863
+ "clip_ratio/region_mean": 0.005859375,
864
+ "entropy": 1.116856336593628,
865
+ "epoch": 1.175,
866
+ "grad_norm": 2.0427322387695312,
867
+ "kl": 0.22487860918045044,
868
+ "learning_rate": 1.85e-06,
869
+ "loss": 0.09498156607151031,
870
+ "step": 47,
871
+ "step_time": 0.11219538700015619
872
+ },
873
+ {
874
+ "clip_ratio/high_max": 0.0,
875
+ "clip_ratio/high_mean": 0.0,
876
+ "clip_ratio/low_mean": 0.001953125,
877
+ "clip_ratio/low_min": 0.001953125,
878
+ "clip_ratio/region_mean": 0.001953125,
879
+ "entropy": 1.2951176166534424,
880
+ "epoch": 1.2,
881
+ "grad_norm": 3.691594362258911,
882
+ "kl": 0.14247789978981018,
883
+ "learning_rate": 1.8249999999999999e-06,
884
+ "loss": 0.09663838893175125,
885
+ "step": 48,
886
+ "step_time": 0.1128243289999773
887
+ },
888
+ {
889
+ "clip_ratio/high_max": 0.0,
890
+ "clip_ratio/high_mean": 0.0,
891
+ "clip_ratio/low_mean": 0.0,
892
+ "clip_ratio/low_min": 0.0,
893
+ "clip_ratio/region_mean": 0.0,
894
+ "completions/clipped_ratio": 1.0,
895
+ "completions/max_length": 256.0,
896
+ "completions/max_terminated_length": 0.0,
897
+ "completions/mean_length": 256.0,
898
+ "completions/mean_terminated_length": 0.0,
899
+ "completions/min_length": 256.0,
900
+ "completions/min_terminated_length": 0.0,
901
+ "entropy": 1.3217719793319702,
902
+ "epoch": 1.225,
903
+ "frac_reward_zero_std": 0.25,
904
+ "grad_norm": 3.7451093196868896,
905
+ "kl": 0.18847626447677612,
906
+ "learning_rate": 1.8e-06,
907
+ "loss": 0.0004711912479251623,
908
+ "num_tokens": 130096.0,
909
+ "reward": 0.886970043182373,
910
+ "reward_std": 0.09752363711595535,
911
+ "rewards/AnswererRewardFunction/mean": 0.886970043182373,
912
+ "rewards/AnswererRewardFunction/std": 0.09752362966537476,
913
+ "step": 49,
914
+ "step_time": 8.057454567999685
915
+ },
916
+ {
917
+ "clip_ratio/high_max": 0.0009765625,
918
+ "clip_ratio/high_mean": 0.0009765625,
919
+ "clip_ratio/low_mean": 0.001953125,
920
+ "clip_ratio/low_min": 0.001953125,
921
+ "clip_ratio/region_mean": 0.0029296875,
922
+ "entropy": 1.2656468152999878,
923
+ "epoch": 1.25,
924
+ "grad_norm": 1.1844326257705688,
925
+ "kl": 0.18295341730117798,
926
+ "learning_rate": 1.7750000000000002e-06,
927
+ "loss": 0.0003048159123864025,
928
+ "step": 50,
929
+ "step_time": 0.11213290599971515
930
+ },
931
+ {
932
+ "clip_ratio/high_max": 0.0009765625,
933
+ "clip_ratio/high_mean": 0.0009765625,
934
+ "clip_ratio/low_mean": 0.0029296875,
935
+ "clip_ratio/low_min": 0.0029296875,
936
+ "clip_ratio/region_mean": 0.00390625,
937
+ "entropy": 1.2773690223693848,
938
+ "epoch": 1.275,
939
+ "grad_norm": 1.258259892463684,
940
+ "kl": 0.166593998670578,
941
+ "learning_rate": 1.7500000000000002e-06,
942
+ "loss": 0.00022288173204287887,
943
+ "step": 51,
944
+ "step_time": 0.11323672800017448
945
+ },
946
+ {
947
+ "clip_ratio/high_max": 0.0029296875,
948
+ "clip_ratio/high_mean": 0.0029296875,
949
+ "clip_ratio/low_mean": 0.009765625,
950
+ "clip_ratio/low_min": 0.009765625,
951
+ "clip_ratio/region_mean": 0.0126953125,
952
+ "entropy": 1.1527807712554932,
953
+ "epoch": 1.3,
954
+ "grad_norm": 2.9316976070404053,
955
+ "kl": 0.16980895400047302,
956
+ "learning_rate": 1.725e-06,
957
+ "loss": 0.0025419846642762423,
958
+ "step": 52,
959
+ "step_time": 0.11281484900018768
960
+ },
961
+ {
962
+ "clip_ratio/high_max": 0.0,
963
+ "clip_ratio/high_mean": 0.0,
964
+ "clip_ratio/low_mean": 0.0,
965
+ "clip_ratio/low_min": 0.0,
966
+ "clip_ratio/region_mean": 0.0,
967
+ "completions/clipped_ratio": 1.0,
968
+ "completions/max_length": 256.0,
969
+ "completions/max_terminated_length": 0.0,
970
+ "completions/mean_length": 256.0,
971
+ "completions/mean_terminated_length": 0.0,
972
+ "completions/min_length": 256.0,
973
+ "completions/min_terminated_length": 0.0,
974
+ "entropy": 1.3017133474349976,
975
+ "epoch": 1.325,
976
+ "frac_reward_zero_std": 0.75,
977
+ "grad_norm": 0.038915447890758514,
978
+ "kl": 0.15527546405792236,
979
+ "learning_rate": 1.7e-06,
980
+ "loss": 0.00038818863686174154,
981
+ "num_tokens": 140252.0,
982
+ "reward": 0.8427923917770386,
983
+ "reward_std": 0.11980737000703812,
984
+ "rewards/AnswererRewardFunction/mean": 0.8427923917770386,
985
+ "rewards/AnswererRewardFunction/std": 0.11980737000703812,
986
+ "step": 53,
987
+ "step_time": 8.022011856000063
988
+ },
989
+ {
990
+ "clip_ratio/high_max": 0.0,
991
+ "clip_ratio/high_mean": 0.0,
992
+ "clip_ratio/low_mean": 0.0,
993
+ "clip_ratio/low_min": 0.0,
994
+ "clip_ratio/region_mean": 0.0,
995
+ "entropy": 1.21080482006073,
996
+ "epoch": 1.35,
997
+ "grad_norm": 1.6995686292648315,
998
+ "kl": 0.14733567833900452,
999
+ "learning_rate": 1.675e-06,
1000
+ "loss": -0.05371169000864029,
1001
+ "step": 54,
1002
+ "step_time": 0.11191600300026039
1003
+ },
1004
+ {
1005
+ "clip_ratio/high_max": 0.0,
1006
+ "clip_ratio/high_mean": 0.0,
1007
+ "clip_ratio/low_mean": 0.00390625,
1008
+ "clip_ratio/low_min": 0.00390625,
1009
+ "clip_ratio/region_mean": 0.00390625,
1010
+ "entropy": 1.392147421836853,
1011
+ "epoch": 1.375,
1012
+ "grad_norm": 2.4972898960113525,
1013
+ "kl": 0.19401419162750244,
1014
+ "learning_rate": 1.65e-06,
1015
+ "loss": 0.10861188173294067,
1016
+ "step": 55,
1017
+ "step_time": 0.11128443200004767
1018
+ },
1019
+ {
1020
+ "clip_ratio/high_max": 0.001953125,
1021
+ "clip_ratio/high_mean": 0.001953125,
1022
+ "clip_ratio/low_mean": 0.0,
1023
+ "clip_ratio/low_min": 0.0,
1024
+ "clip_ratio/region_mean": 0.001953125,
1025
+ "entropy": 1.2331085205078125,
1026
+ "epoch": 1.4,
1027
+ "grad_norm": 1.5207850933074951,
1028
+ "kl": 0.14608821272850037,
1029
+ "learning_rate": 1.625e-06,
1030
+ "loss": -0.0538652203977108,
1031
+ "step": 56,
1032
+ "step_time": 0.11067563899996458
1033
+ },
1034
+ {
1035
+ "clip_ratio/high_max": 0.0,
1036
+ "clip_ratio/high_mean": 0.0,
1037
+ "clip_ratio/low_mean": 0.0,
1038
+ "clip_ratio/low_min": 0.0,
1039
+ "clip_ratio/region_mean": 0.0,
1040
+ "completions/clipped_ratio": 1.0,
1041
+ "completions/max_length": 256.0,
1042
+ "completions/max_terminated_length": 0.0,
1043
+ "completions/mean_length": 256.0,
1044
+ "completions/mean_terminated_length": 0.0,
1045
+ "completions/min_length": 256.0,
1046
+ "completions/min_terminated_length": 0.0,
1047
+ "entropy": 1.2587049007415771,
1048
+ "epoch": 1.425,
1049
+ "frac_reward_zero_std": 0.75,
1050
+ "grad_norm": 0.03795379400253296,
1051
+ "kl": 0.14604490995407104,
1052
+ "learning_rate": 1.6e-06,
1053
+ "loss": 0.0003651122679002583,
1054
+ "num_tokens": 150596.0,
1055
+ "reward": 0.8504944443702698,
1056
+ "reward_std": 0.12533700466156006,
1057
+ "rewards/AnswererRewardFunction/mean": 0.8504944443702698,
1058
+ "rewards/AnswererRewardFunction/std": 0.12533698976039886,
1059
+ "step": 57,
1060
+ "step_time": 7.993730275999951
1061
+ },
1062
+ {
1063
+ "clip_ratio/high_max": 0.0,
1064
+ "clip_ratio/high_mean": 0.0,
1065
+ "clip_ratio/low_mean": 0.0,
1066
+ "clip_ratio/low_min": 0.0,
1067
+ "clip_ratio/region_mean": 0.0,
1068
+ "entropy": 1.2113020420074463,
1069
+ "epoch": 1.45,
1070
+ "grad_norm": 3.0792369842529297,
1071
+ "kl": 0.17833790183067322,
1072
+ "learning_rate": 1.5750000000000002e-06,
1073
+ "loss": 0.09483250975608826,
1074
+ "step": 58,
1075
+ "step_time": 0.11180590300000404
1076
+ },
1077
+ {
1078
+ "clip_ratio/high_max": 0.0,
1079
+ "clip_ratio/high_mean": 0.0,
1080
+ "clip_ratio/low_mean": 0.0,
1081
+ "clip_ratio/low_min": 0.0,
1082
+ "clip_ratio/region_mean": 0.0,
1083
+ "entropy": 1.2311359643936157,
1084
+ "epoch": 1.475,
1085
+ "grad_norm": 0.9535032510757446,
1086
+ "kl": 0.2094941884279251,
1087
+ "learning_rate": 1.5500000000000002e-06,
1088
+ "loss": -0.03074605017900467,
1089
+ "step": 59,
1090
+ "step_time": 0.11118914000007862
1091
+ },
1092
+ {
1093
+ "clip_ratio/high_max": 0.0048828125,
1094
+ "clip_ratio/high_mean": 0.0048828125,
1095
+ "clip_ratio/low_mean": 0.0,
1096
+ "clip_ratio/low_min": 0.0,
1097
+ "clip_ratio/region_mean": 0.0048828125,
1098
+ "entropy": 1.2338982820510864,
1099
+ "epoch": 1.5,
1100
+ "grad_norm": 1.1395492553710938,
1101
+ "kl": 0.19933584332466125,
1102
+ "learning_rate": 1.525e-06,
1103
+ "loss": -0.061825230717659,
1104
+ "step": 60,
1105
+ "step_time": 0.1112795730000471
1106
+ },
1107
+ {
1108
+ "clip_ratio/high_max": 0.0,
1109
+ "clip_ratio/high_mean": 0.0,
1110
+ "clip_ratio/low_mean": 0.0,
1111
+ "clip_ratio/low_min": 0.0,
1112
+ "clip_ratio/region_mean": 0.0,
1113
+ "completions/clipped_ratio": 1.0,
1114
+ "completions/max_length": 256.0,
1115
+ "completions/max_terminated_length": 0.0,
1116
+ "completions/mean_length": 256.0,
1117
+ "completions/mean_terminated_length": 0.0,
1118
+ "completions/min_length": 256.0,
1119
+ "completions/min_terminated_length": 0.0,
1120
+ "entropy": 1.3512625694274902,
1121
+ "epoch": 1.525,
1122
+ "frac_reward_zero_std": 0.75,
1123
+ "grad_norm": 3.3047728538513184,
1124
+ "kl": 0.24816307425498962,
1125
+ "learning_rate": 1.5e-06,
1126
+ "loss": 0.06274406611919403,
1127
+ "num_tokens": 160792.0,
1128
+ "reward": 0.817285418510437,
1129
+ "reward_std": 0.12126864492893219,
1130
+ "rewards/AnswererRewardFunction/mean": 0.817285418510437,
1131
+ "rewards/AnswererRewardFunction/std": 0.12126864492893219,
1132
+ "step": 61,
1133
+ "step_time": 7.993776449000052
1134
+ },
1135
+ {
1136
+ "clip_ratio/high_max": 0.0,
1137
+ "clip_ratio/high_mean": 0.0,
1138
+ "clip_ratio/low_mean": 0.0,
1139
+ "clip_ratio/low_min": 0.0,
1140
+ "clip_ratio/region_mean": 0.0,
1141
+ "entropy": 1.407865047454834,
1142
+ "epoch": 1.55,
1143
+ "grad_norm": 0.8703442811965942,
1144
+ "kl": 0.15828697383403778,
1145
+ "learning_rate": 1.4749999999999999e-06,
1146
+ "loss": -0.030646126717329025,
1147
+ "step": 62,
1148
+ "step_time": 0.11262021999982608
1149
+ },
1150
+ {
1151
+ "clip_ratio/high_max": 0.0,
1152
+ "clip_ratio/high_mean": 0.0,
1153
+ "clip_ratio/low_mean": 0.0,
1154
+ "clip_ratio/low_min": 0.0,
1155
+ "clip_ratio/region_mean": 0.0,
1156
+ "entropy": 1.2677810192108154,
1157
+ "epoch": 1.575,
1158
+ "grad_norm": 0.0925767794251442,
1159
+ "kl": 0.17386655509471893,
1160
+ "learning_rate": 1.45e-06,
1161
+ "loss": 0.0004346664063632488,
1162
+ "step": 63,
1163
+ "step_time": 0.11218224999993254
1164
+ },
1165
+ {
1166
+ "clip_ratio/high_max": 0.0048828125,
1167
+ "clip_ratio/high_mean": 0.0048828125,
1168
+ "clip_ratio/low_mean": 0.0,
1169
+ "clip_ratio/low_min": 0.0,
1170
+ "clip_ratio/region_mean": 0.0048828125,
1171
+ "entropy": 1.3804912567138672,
1172
+ "epoch": 1.6,
1173
+ "grad_norm": 0.9855126142501831,
1174
+ "kl": 0.2777293026447296,
1175
+ "learning_rate": 1.425e-06,
1176
+ "loss": -0.03046189248561859,
1177
+ "step": 64,
1178
+ "step_time": 0.11151005800002167
1179
+ },
1180
+ {
1181
+ "clip_ratio/high_max": 0.0,
1182
+ "clip_ratio/high_mean": 0.0,
1183
+ "clip_ratio/low_mean": 0.0,
1184
+ "clip_ratio/low_min": 0.0,
1185
+ "clip_ratio/region_mean": 0.0,
1186
+ "completions/clipped_ratio": 1.0,
1187
+ "completions/max_length": 256.0,
1188
+ "completions/max_terminated_length": 0.0,
1189
+ "completions/mean_length": 256.0,
1190
+ "completions/mean_terminated_length": 0.0,
1191
+ "completions/min_length": 256.0,
1192
+ "completions/min_terminated_length": 0.0,
1193
+ "entropy": 1.6676212549209595,
1194
+ "epoch": 1.625,
1195
+ "frac_reward_zero_std": 0.75,
1196
+ "grad_norm": 0.847771942615509,
1197
+ "kl": 0.17978495359420776,
1198
+ "learning_rate": 1.4000000000000001e-06,
1199
+ "loss": 0.031675659120082855,
1200
+ "num_tokens": 170980.0,
1201
+ "reward": 0.8866766691207886,
1202
+ "reward_std": 0.10052888840436935,
1203
+ "rewards/AnswererRewardFunction/mean": 0.8866766691207886,
1204
+ "rewards/AnswererRewardFunction/std": 0.10052888840436935,
1205
+ "step": 65,
1206
+ "step_time": 8.045760322000206
1207
+ },
1208
+ {
1209
+ "clip_ratio/high_max": 0.0,
1210
+ "clip_ratio/high_mean": 0.0,
1211
+ "clip_ratio/low_mean": 0.0,
1212
+ "clip_ratio/low_min": 0.0,
1213
+ "clip_ratio/region_mean": 0.0,
1214
+ "entropy": 1.3201985359191895,
1215
+ "epoch": 1.65,
1216
+ "grad_norm": 2.5406768321990967,
1217
+ "kl": 0.1744946837425232,
1218
+ "learning_rate": 1.375e-06,
1219
+ "loss": -0.0932559221982956,
1220
+ "step": 66,
1221
+ "step_time": 0.11122614600026282
1222
+ },
1223
+ {
1224
+ "clip_ratio/high_max": 0.0,
1225
+ "clip_ratio/high_mean": 0.0,
1226
+ "clip_ratio/low_mean": 0.0,
1227
+ "clip_ratio/low_min": 0.0,
1228
+ "clip_ratio/region_mean": 0.0,
1229
+ "entropy": 1.192025899887085,
1230
+ "epoch": 1.675,
1231
+ "grad_norm": 0.04267319664359093,
1232
+ "kl": 0.1620822697877884,
1233
+ "learning_rate": 1.35e-06,
1234
+ "loss": 0.00040520564652979374,
1235
+ "step": 67,
1236
+ "step_time": 0.11081279699965307
1237
+ },
1238
+ {
1239
+ "clip_ratio/high_max": 0.0,
1240
+ "clip_ratio/high_mean": 0.0,
1241
+ "clip_ratio/low_mean": 0.0009765625,
1242
+ "clip_ratio/low_min": 0.0009765625,
1243
+ "clip_ratio/region_mean": 0.0009765625,
1244
+ "entropy": 1.0180904865264893,
1245
+ "epoch": 1.7,
1246
+ "grad_norm": 1.4091999530792236,
1247
+ "kl": 0.15127339959144592,
1248
+ "learning_rate": 1.325e-06,
1249
+ "loss": 0.06317397952079773,
1250
+ "step": 68,
1251
+ "step_time": 0.11124899700007518
1252
+ },
1253
+ {
1254
+ "clip_ratio/high_max": 0.0,
1255
+ "clip_ratio/high_mean": 0.0,
1256
+ "clip_ratio/low_mean": 0.0,
1257
+ "clip_ratio/low_min": 0.0,
1258
+ "clip_ratio/region_mean": 0.0,
1259
+ "completions/clipped_ratio": 1.0,
1260
+ "completions/max_length": 256.0,
1261
+ "completions/max_terminated_length": 0.0,
1262
+ "completions/mean_length": 256.0,
1263
+ "completions/mean_terminated_length": 0.0,
1264
+ "completions/min_length": 256.0,
1265
+ "completions/min_terminated_length": 0.0,
1266
+ "entropy": 1.2156814336776733,
1267
+ "epoch": 1.725,
1268
+ "frac_reward_zero_std": 0.75,
1269
+ "grad_norm": 0.04841400310397148,
1270
+ "kl": 0.1523037701845169,
1271
+ "learning_rate": 1.3e-06,
1272
+ "loss": 0.00038075941847637296,
1273
+ "num_tokens": 180980.0,
1274
+ "reward": 0.8108064532279968,
1275
+ "reward_std": 0.14502660930156708,
1276
+ "rewards/AnswererRewardFunction/mean": 0.8108064532279968,
1277
+ "rewards/AnswererRewardFunction/std": 0.14502662420272827,
1278
+ "step": 69,
1279
+ "step_time": 8.016459874999782
1280
+ },
1281
+ {
1282
+ "clip_ratio/high_max": 0.0,
1283
+ "clip_ratio/high_mean": 0.0,
1284
+ "clip_ratio/low_mean": 0.0,
1285
+ "clip_ratio/low_min": 0.0,
1286
+ "clip_ratio/region_mean": 0.0,
1287
+ "entropy": 1.2170944213867188,
1288
+ "epoch": 1.75,
1289
+ "grad_norm": 3.1129679679870605,
1290
+ "kl": 0.18750926852226257,
1291
+ "learning_rate": 1.275e-06,
1292
+ "loss": 0.06332284957170486,
1293
+ "step": 70,
1294
+ "step_time": 0.11227171000018643
1295
+ },
1296
+ {
1297
+ "clip_ratio/high_max": 0.005859375,
1298
+ "clip_ratio/high_mean": 0.005859375,
1299
+ "clip_ratio/low_mean": 0.0,
1300
+ "clip_ratio/low_min": 0.0,
1301
+ "clip_ratio/region_mean": 0.005859375,
1302
+ "entropy": 1.4517723321914673,
1303
+ "epoch": 1.775,
1304
+ "grad_norm": 1.4262726306915283,
1305
+ "kl": 0.22623388469219208,
1306
+ "learning_rate": 1.25e-06,
1307
+ "loss": -0.062118660658597946,
1308
+ "step": 71,
1309
+ "step_time": 0.11156967399983841
1310
+ },
1311
+ {
1312
+ "clip_ratio/high_max": 0.0,
1313
+ "clip_ratio/high_mean": 0.0,
1314
+ "clip_ratio/low_mean": 0.0,
1315
+ "clip_ratio/low_min": 0.0,
1316
+ "clip_ratio/region_mean": 0.0,
1317
+ "entropy": 1.1083641052246094,
1318
+ "epoch": 1.8,
1319
+ "grad_norm": 0.038435112684965134,
1320
+ "kl": 0.16960963606834412,
1321
+ "learning_rate": 1.225e-06,
1322
+ "loss": 0.00042402403778396547,
1323
+ "step": 72,
1324
+ "step_time": 0.11067348699998547
1325
+ },
1326
+ {
1327
+ "clip_ratio/high_max": 0.0,
1328
+ "clip_ratio/high_mean": 0.0,
1329
+ "clip_ratio/low_mean": 0.0,
1330
+ "clip_ratio/low_min": 0.0,
1331
+ "clip_ratio/region_mean": 0.0,
1332
+ "completions/clipped_ratio": 1.0,
1333
+ "completions/max_length": 256.0,
1334
+ "completions/max_terminated_length": 0.0,
1335
+ "completions/mean_length": 256.0,
1336
+ "completions/mean_terminated_length": 0.0,
1337
+ "completions/min_length": 256.0,
1338
+ "completions/min_terminated_length": 0.0,
1339
+ "entropy": 1.3423596620559692,
1340
+ "epoch": 1.825,
1341
+ "frac_reward_zero_std": 1.0,
1342
+ "grad_norm": 0.04306335747241974,
1343
+ "kl": 0.16442564129829407,
1344
+ "learning_rate": 1.2000000000000002e-06,
1345
+ "loss": 0.00041106410208158195,
1346
+ "num_tokens": 190860.0,
1347
+ "reward": 0.8227691054344177,
1348
+ "reward_std": 0.11510815471410751,
1349
+ "rewards/AnswererRewardFunction/mean": 0.8227691054344177,
1350
+ "rewards/AnswererRewardFunction/std": 0.11510813981294632,
1351
+ "step": 73,
1352
+ "step_time": 7.9909718600001725
1353
+ },
1354
+ {
1355
+ "clip_ratio/high_max": 0.0,
1356
+ "clip_ratio/high_mean": 0.0,
1357
+ "clip_ratio/low_mean": 0.0,
1358
+ "clip_ratio/low_min": 0.0,
1359
+ "clip_ratio/region_mean": 0.0,
1360
+ "entropy": 1.315840244293213,
1361
+ "epoch": 1.85,
1362
+ "grad_norm": 0.07536096125841141,
1363
+ "kl": 0.18558362126350403,
1364
+ "learning_rate": 1.175e-06,
1365
+ "loss": 0.0004639590042643249,
1366
+ "step": 74,
1367
+ "step_time": 0.11113172400018811
1368
+ },
1369
+ {
1370
+ "clip_ratio/high_max": 0.0,
1371
+ "clip_ratio/high_mean": 0.0,
1372
+ "clip_ratio/low_mean": 0.0,
1373
+ "clip_ratio/low_min": 0.0,
1374
+ "clip_ratio/region_mean": 0.0,
1375
+ "entropy": 1.2952380180358887,
1376
+ "epoch": 1.875,
1377
+ "grad_norm": 0.05677155405282974,
1378
+ "kl": 0.1575685441493988,
1379
+ "learning_rate": 1.1500000000000002e-06,
1380
+ "loss": 0.0003939213347621262,
1381
+ "step": 75,
1382
+ "step_time": 0.1102538669997557
1383
+ },
1384
+ {
1385
+ "clip_ratio/high_max": 0.0,
1386
+ "clip_ratio/high_mean": 0.0,
1387
+ "clip_ratio/low_mean": 0.0,
1388
+ "clip_ratio/low_min": 0.0,
1389
+ "clip_ratio/region_mean": 0.0,
1390
+ "entropy": 0.9605852365493774,
1391
+ "epoch": 1.9,
1392
+ "grad_norm": 0.053945671766996384,
1393
+ "kl": 0.15232446789741516,
1394
+ "learning_rate": 1.125e-06,
1395
+ "loss": 0.0003808111359830946,
1396
+ "step": 76,
1397
+ "step_time": 0.11566855000000942
1398
+ },
1399
+ {
1400
+ "clip_ratio/high_max": 0.0,
1401
+ "clip_ratio/high_mean": 0.0,
1402
+ "clip_ratio/low_mean": 0.0,
1403
+ "clip_ratio/low_min": 0.0,
1404
+ "clip_ratio/region_mean": 0.0,
1405
+ "completions/clipped_ratio": 1.0,
1406
+ "completions/max_length": 256.0,
1407
+ "completions/max_terminated_length": 0.0,
1408
+ "completions/mean_length": 256.0,
1409
+ "completions/mean_terminated_length": 0.0,
1410
+ "completions/min_length": 256.0,
1411
+ "completions/min_terminated_length": 0.0,
1412
+ "entropy": 0.9576239585876465,
1413
+ "epoch": 1.925,
1414
+ "frac_reward_zero_std": 1.0,
1415
+ "grad_norm": 0.04307965934276581,
1416
+ "kl": 0.15413150191307068,
1417
+ "learning_rate": 1.0999999999999998e-06,
1418
+ "loss": 0.0003853287489619106,
1419
+ "num_tokens": 201032.0,
1420
+ "reward": 0.9358674883842468,
1421
+ "reward_std": 0.0063578663393855095,
1422
+ "rewards/AnswererRewardFunction/mean": 0.9358674883842468,
1423
+ "rewards/AnswererRewardFunction/std": 0.006357857491821051,
1424
+ "step": 77,
1425
+ "step_time": 8.005991259999973
1426
+ },
1427
+ {
1428
+ "clip_ratio/high_max": 0.0,
1429
+ "clip_ratio/high_mean": 0.0,
1430
+ "clip_ratio/low_mean": 0.0,
1431
+ "clip_ratio/low_min": 0.0,
1432
+ "clip_ratio/region_mean": 0.0,
1433
+ "entropy": 1.2276498079299927,
1434
+ "epoch": 1.95,
1435
+ "grad_norm": 0.036765944212675095,
1436
+ "kl": 0.1328800469636917,
1437
+ "learning_rate": 1.075e-06,
1438
+ "loss": 0.0003322001139167696,
1439
+ "step": 78,
1440
+ "step_time": 0.115283160999752
1441
+ },
1442
+ {
1443
+ "clip_ratio/high_max": 0.0,
1444
+ "clip_ratio/high_mean": 0.0,
1445
+ "clip_ratio/low_mean": 0.0,
1446
+ "clip_ratio/low_min": 0.0,
1447
+ "clip_ratio/region_mean": 0.0,
1448
+ "entropy": 1.4125936031341553,
1449
+ "epoch": 1.975,
1450
+ "grad_norm": 0.038389381021261215,
1451
+ "kl": 0.2053934633731842,
1452
+ "learning_rate": 1.05e-06,
1453
+ "loss": 0.0005134836537763476,
1454
+ "step": 79,
1455
+ "step_time": 0.11063392400001248
1456
+ },
1457
+ {
1458
+ "clip_ratio/high_max": 0.0,
1459
+ "clip_ratio/high_mean": 0.0,
1460
+ "clip_ratio/low_mean": 0.0,
1461
+ "clip_ratio/low_min": 0.0,
1462
+ "clip_ratio/region_mean": 0.0,
1463
+ "entropy": 1.370209813117981,
1464
+ "epoch": 2.0,
1465
+ "grad_norm": 0.040238432586193085,
1466
+ "kl": 0.17255280911922455,
1467
+ "learning_rate": 1.0250000000000001e-06,
1468
+ "loss": 0.0004313820390962064,
1469
+ "step": 80,
1470
+ "step_time": 0.11083741000038572
1471
+ },
1472
+ {
1473
+ "clip_ratio/high_max": 0.0,
1474
+ "clip_ratio/high_mean": 0.0,
1475
+ "clip_ratio/low_mean": 0.0,
1476
+ "clip_ratio/low_min": 0.0,
1477
+ "clip_ratio/region_mean": 0.0,
1478
+ "completions/clipped_ratio": 1.0,
1479
+ "completions/max_length": 256.0,
1480
+ "completions/max_terminated_length": 0.0,
1481
+ "completions/mean_length": 256.0,
1482
+ "completions/mean_terminated_length": 0.0,
1483
+ "completions/min_length": 256.0,
1484
+ "completions/min_terminated_length": 0.0,
1485
+ "entropy": 1.2237610816955566,
1486
+ "epoch": 2.025,
1487
+ "frac_reward_zero_std": 1.0,
1488
+ "grad_norm": 0.0578920878469944,
1489
+ "kl": 0.19325289130210876,
1490
+ "learning_rate": 1e-06,
1491
+ "loss": 0.0004831322003155947,
1492
+ "num_tokens": 211156.0,
1493
+ "reward": 0.819347620010376,
1494
+ "reward_std": 0.11957820504903793,
1495
+ "rewards/AnswererRewardFunction/mean": 0.819347620010376,
1496
+ "rewards/AnswererRewardFunction/std": 0.11957821249961853,
1497
+ "step": 81,
1498
+ "step_time": 7.974696767999831
1499
+ },
1500
+ {
1501
+ "clip_ratio/high_max": 0.0,
1502
+ "clip_ratio/high_mean": 0.0,
1503
+ "clip_ratio/low_mean": 0.0,
1504
+ "clip_ratio/low_min": 0.0,
1505
+ "clip_ratio/region_mean": 0.0,
1506
+ "entropy": 1.3902840614318848,
1507
+ "epoch": 2.05,
1508
+ "grad_norm": 0.09637662768363953,
1509
+ "kl": 0.2233792543411255,
1510
+ "learning_rate": 9.75e-07,
1511
+ "loss": 0.0005584481405094266,
1512
+ "step": 82,
1513
+ "step_time": 0.11330902999998216
1514
+ },
1515
+ {
1516
+ "clip_ratio/high_max": 0.0,
1517
+ "clip_ratio/high_mean": 0.0,
1518
+ "clip_ratio/low_mean": 0.0,
1519
+ "clip_ratio/low_min": 0.0,
1520
+ "clip_ratio/region_mean": 0.0,
1521
+ "entropy": 1.3189706802368164,
1522
+ "epoch": 2.075,
1523
+ "grad_norm": 0.11303476989269257,
1524
+ "kl": 0.2002892792224884,
1525
+ "learning_rate": 9.5e-07,
1526
+ "loss": 0.0005007231957279146,
1527
+ "step": 83,
1528
+ "step_time": 0.11202955199996723
1529
+ },
1530
+ {
1531
+ "clip_ratio/high_max": 0.0,
1532
+ "clip_ratio/high_mean": 0.0,
1533
+ "clip_ratio/low_mean": 0.0,
1534
+ "clip_ratio/low_min": 0.0,
1535
+ "clip_ratio/region_mean": 0.0,
1536
+ "entropy": 1.545952320098877,
1537
+ "epoch": 2.1,
1538
+ "grad_norm": 0.11450417339801788,
1539
+ "kl": 0.2085442692041397,
1540
+ "learning_rate": 9.25e-07,
1541
+ "loss": 0.0005213606636971235,
1542
+ "step": 84,
1543
+ "step_time": 0.11084980799978439
1544
+ },
1545
+ {
1546
+ "clip_ratio/high_max": 0.0,
1547
+ "clip_ratio/high_mean": 0.0,
1548
+ "clip_ratio/low_mean": 0.0,
1549
+ "clip_ratio/low_min": 0.0,
1550
+ "clip_ratio/region_mean": 0.0,
1551
+ "completions/clipped_ratio": 1.0,
1552
+ "completions/max_length": 256.0,
1553
+ "completions/max_terminated_length": 0.0,
1554
+ "completions/mean_length": 256.0,
1555
+ "completions/mean_terminated_length": 0.0,
1556
+ "completions/min_length": 256.0,
1557
+ "completions/min_terminated_length": 0.0,
1558
+ "entropy": 1.2259491682052612,
1559
+ "epoch": 2.125,
1560
+ "frac_reward_zero_std": 0.75,
1561
+ "grad_norm": 1.3692517280578613,
1562
+ "kl": 0.22583863139152527,
1563
+ "learning_rate": 9e-07,
1564
+ "loss": 0.06301698833703995,
1565
+ "num_tokens": 220860.0,
1566
+ "reward": 0.7968939542770386,
1567
+ "reward_std": 0.10052310675382614,
1568
+ "rewards/AnswererRewardFunction/mean": 0.7968939542770386,
1569
+ "rewards/AnswererRewardFunction/std": 0.10052311420440674,
1570
+ "step": 85,
1571
+ "step_time": 8.069017658999655
1572
+ },
1573
+ {
1574
+ "clip_ratio/high_max": 0.0,
1575
+ "clip_ratio/high_mean": 0.0,
1576
+ "clip_ratio/low_mean": 0.0,
1577
+ "clip_ratio/low_min": 0.0,
1578
+ "clip_ratio/region_mean": 0.0,
1579
+ "entropy": 1.4456801414489746,
1580
+ "epoch": 2.15,
1581
+ "grad_norm": 0.04716620221734047,
1582
+ "kl": 0.18480879068374634,
1583
+ "learning_rate": 8.750000000000001e-07,
1584
+ "loss": 0.00046202196972444654,
1585
+ "step": 86,
1586
+ "step_time": 0.10887526600026831
1587
+ },
1588
+ {
1589
+ "clip_ratio/high_max": 0.0,
1590
+ "clip_ratio/high_mean": 0.0,
1591
+ "clip_ratio/low_mean": 0.0,
1592
+ "clip_ratio/low_min": 0.0,
1593
+ "clip_ratio/region_mean": 0.0,
1594
+ "entropy": 1.3787351846694946,
1595
+ "epoch": 2.175,
1596
+ "grad_norm": 0.13064244389533997,
1597
+ "kl": 0.2059539407491684,
1598
+ "learning_rate": 8.5e-07,
1599
+ "loss": 0.0005148848285898566,
1600
+ "step": 87,
1601
+ "step_time": 0.10871470299980501
1602
+ },
1603
+ {
1604
+ "clip_ratio/high_max": 0.0,
1605
+ "clip_ratio/high_mean": 0.0,
1606
+ "clip_ratio/low_mean": 0.0009765625,
1607
+ "clip_ratio/low_min": 0.0009765625,
1608
+ "clip_ratio/region_mean": 0.0009765625,
1609
+ "entropy": 1.4968674182891846,
1610
+ "epoch": 2.2,
1611
+ "grad_norm": 2.7954156398773193,
1612
+ "kl": 0.21777424216270447,
1613
+ "learning_rate": 8.25e-07,
1614
+ "loss": -0.06176532059907913,
1615
+ "step": 88,
1616
+ "step_time": 0.10914303899971856
1617
+ },
1618
+ {
1619
+ "clip_ratio/high_max": 0.0,
1620
+ "clip_ratio/high_mean": 0.0,
1621
+ "clip_ratio/low_mean": 0.0,
1622
+ "clip_ratio/low_min": 0.0,
1623
+ "clip_ratio/region_mean": 0.0,
1624
+ "completions/clipped_ratio": 1.0,
1625
+ "completions/max_length": 256.0,
1626
+ "completions/max_terminated_length": 0.0,
1627
+ "completions/mean_length": 256.0,
1628
+ "completions/mean_terminated_length": 0.0,
1629
+ "completions/min_length": 256.0,
1630
+ "completions/min_terminated_length": 0.0,
1631
+ "entropy": 1.4094501733779907,
1632
+ "epoch": 2.225,
1633
+ "frac_reward_zero_std": 0.75,
1634
+ "grad_norm": 1.311699628829956,
1635
+ "kl": 0.21720707416534424,
1636
+ "learning_rate": 8e-07,
1637
+ "loss": -0.06158065050840378,
1638
+ "num_tokens": 230668.0,
1639
+ "reward": 0.7586733102798462,
1640
+ "reward_std": 0.10710728168487549,
1641
+ "rewards/AnswererRewardFunction/mean": 0.7586733102798462,
1642
+ "rewards/AnswererRewardFunction/std": 0.10710727423429489,
1643
+ "step": 89,
1644
+ "step_time": 8.016205262000312
1645
+ },
1646
+ {
1647
+ "clip_ratio/high_max": 0.0009765625,
1648
+ "clip_ratio/high_mean": 0.0009765625,
1649
+ "clip_ratio/low_mean": 0.0,
1650
+ "clip_ratio/low_min": 0.0,
1651
+ "clip_ratio/region_mean": 0.0009765625,
1652
+ "entropy": 1.5809335708618164,
1653
+ "epoch": 2.25,
1654
+ "grad_norm": 1.1251577138900757,
1655
+ "kl": 0.18235114216804504,
1656
+ "learning_rate": 7.750000000000001e-07,
1657
+ "loss": -0.030835507437586784,
1658
+ "step": 90,
1659
+ "step_time": 0.1106916500002626
1660
+ },
1661
+ {
1662
+ "clip_ratio/high_max": 0.0,
1663
+ "clip_ratio/high_mean": 0.0,
1664
+ "clip_ratio/low_mean": 0.0009765625,
1665
+ "clip_ratio/low_min": 0.0009765625,
1666
+ "clip_ratio/region_mean": 0.0009765625,
1667
+ "entropy": 1.2390732765197754,
1668
+ "epoch": 2.275,
1669
+ "grad_norm": 2.447874069213867,
1670
+ "kl": 0.257110059261322,
1671
+ "learning_rate": 7.5e-07,
1672
+ "loss": 0.09387418627738953,
1673
+ "step": 91,
1674
+ "step_time": 0.11327933999973538
1675
+ },
1676
+ {
1677
+ "clip_ratio/high_max": 0.0,
1678
+ "clip_ratio/high_mean": 0.0,
1679
+ "clip_ratio/low_mean": 0.0,
1680
+ "clip_ratio/low_min": 0.0,
1681
+ "clip_ratio/region_mean": 0.0,
1682
+ "entropy": 1.2858381271362305,
1683
+ "epoch": 2.3,
1684
+ "grad_norm": 0.23859065771102905,
1685
+ "kl": 0.231545552611351,
1686
+ "learning_rate": 7.25e-07,
1687
+ "loss": 0.0005788638954982162,
1688
+ "step": 92,
1689
+ "step_time": 0.11277877800011993
1690
+ },
1691
+ {
1692
+ "clip_ratio/high_max": 0.0,
1693
+ "clip_ratio/high_mean": 0.0,
1694
+ "clip_ratio/low_mean": 0.0,
1695
+ "clip_ratio/low_min": 0.0,
1696
+ "clip_ratio/region_mean": 0.0,
1697
+ "completions/clipped_ratio": 1.0,
1698
+ "completions/max_length": 256.0,
1699
+ "completions/max_terminated_length": 0.0,
1700
+ "completions/mean_length": 256.0,
1701
+ "completions/mean_terminated_length": 0.0,
1702
+ "completions/min_length": 256.0,
1703
+ "completions/min_terminated_length": 0.0,
1704
+ "entropy": 1.2816568613052368,
1705
+ "epoch": 2.325,
1706
+ "frac_reward_zero_std": 1.0,
1707
+ "grad_norm": 0.04282540827989578,
1708
+ "kl": 0.2023722380399704,
1709
+ "learning_rate": 7.000000000000001e-07,
1710
+ "loss": 0.0005059306276962161,
1711
+ "num_tokens": 241160.0,
1712
+ "reward": 0.9323133230209351,
1713
+ "reward_std": 0.0,
1714
+ "rewards/AnswererRewardFunction/mean": 0.9323133230209351,
1715
+ "rewards/AnswererRewardFunction/std": 0.0,
1716
+ "step": 93,
1717
+ "step_time": 7.996730308000224
1718
+ },
1719
+ {
1720
+ "clip_ratio/high_max": 0.0,
1721
+ "clip_ratio/high_mean": 0.0,
1722
+ "clip_ratio/low_mean": 0.0,
1723
+ "clip_ratio/low_min": 0.0,
1724
+ "clip_ratio/region_mean": 0.0,
1725
+ "entropy": 1.417135238647461,
1726
+ "epoch": 2.35,
1727
+ "grad_norm": 0.054960086941719055,
1728
+ "kl": 0.17452408373355865,
1729
+ "learning_rate": 6.75e-07,
1730
+ "loss": 0.00043631019070744514,
1731
+ "step": 94,
1732
+ "step_time": 0.11194130800004132
1733
+ },
1734
+ {
1735
+ "clip_ratio/high_max": 0.0,
1736
+ "clip_ratio/high_mean": 0.0,
1737
+ "clip_ratio/low_mean": 0.0,
1738
+ "clip_ratio/low_min": 0.0,
1739
+ "clip_ratio/region_mean": 0.0,
1740
+ "entropy": 1.2406070232391357,
1741
+ "epoch": 2.375,
1742
+ "grad_norm": 0.05798948183655739,
1743
+ "kl": 0.19247736036777496,
1744
+ "learning_rate": 6.5e-07,
1745
+ "loss": 0.000481193361338228,
1746
+ "step": 95,
1747
+ "step_time": 0.11430945899974176
1748
+ },
1749
+ {
1750
+ "clip_ratio/high_max": 0.0,
1751
+ "clip_ratio/high_mean": 0.0,
1752
+ "clip_ratio/low_mean": 0.0,
1753
+ "clip_ratio/low_min": 0.0,
1754
+ "clip_ratio/region_mean": 0.0,
1755
+ "entropy": 1.332491159439087,
1756
+ "epoch": 2.4,
1757
+ "grad_norm": 0.04423392191529274,
1758
+ "kl": 0.15216711163520813,
1759
+ "learning_rate": 6.25e-07,
1760
+ "loss": 0.0003804177977144718,
1761
+ "step": 96,
1762
+ "step_time": 0.11453651299962075
1763
+ },
1764
+ {
1765
+ "clip_ratio/high_max": 0.0,
1766
+ "clip_ratio/high_mean": 0.0,
1767
+ "clip_ratio/low_mean": 0.0,
1768
+ "clip_ratio/low_min": 0.0,
1769
+ "clip_ratio/region_mean": 0.0,
1770
+ "completions/clipped_ratio": 1.0,
1771
+ "completions/max_length": 256.0,
1772
+ "completions/max_terminated_length": 0.0,
1773
+ "completions/mean_length": 256.0,
1774
+ "completions/mean_terminated_length": 0.0,
1775
+ "completions/min_length": 256.0,
1776
+ "completions/min_terminated_length": 0.0,
1777
+ "entropy": 1.3983848094940186,
1778
+ "epoch": 2.425,
1779
+ "frac_reward_zero_std": 1.0,
1780
+ "grad_norm": 0.048516470938920975,
1781
+ "kl": 0.14233297109603882,
1782
+ "learning_rate": 6.000000000000001e-07,
1783
+ "loss": 0.00035583245335146785,
1784
+ "num_tokens": 251204.0,
1785
+ "reward": 0.822901725769043,
1786
+ "reward_std": 0.1232716366648674,
1787
+ "rewards/AnswererRewardFunction/mean": 0.822901725769043,
1788
+ "rewards/AnswererRewardFunction/std": 0.123271644115448,
1789
+ "step": 97,
1790
+ "step_time": 7.996272254999894
1791
+ },
1792
+ {
1793
+ "clip_ratio/high_max": 0.0,
1794
+ "clip_ratio/high_mean": 0.0,
1795
+ "clip_ratio/low_mean": 0.0,
1796
+ "clip_ratio/low_min": 0.0,
1797
+ "clip_ratio/region_mean": 0.0,
1798
+ "entropy": 1.3374725580215454,
1799
+ "epoch": 2.45,
1800
+ "grad_norm": 0.04759443923830986,
1801
+ "kl": 0.18392232060432434,
1802
+ "learning_rate": 5.750000000000001e-07,
1803
+ "loss": 0.00045980580034665763,
1804
+ "step": 98,
1805
+ "step_time": 0.11136618299997281
1806
+ },
1807
+ {
1808
+ "clip_ratio/high_max": 0.0,
1809
+ "clip_ratio/high_mean": 0.0,
1810
+ "clip_ratio/low_mean": 0.0,
1811
+ "clip_ratio/low_min": 0.0,
1812
+ "clip_ratio/region_mean": 0.0,
1813
+ "entropy": 1.2736051082611084,
1814
+ "epoch": 2.475,
1815
+ "grad_norm": 0.11932854354381561,
1816
+ "kl": 0.21164000034332275,
1817
+ "learning_rate": 5.499999999999999e-07,
1818
+ "loss": 0.0005290999542921782,
1819
+ "step": 99,
1820
+ "step_time": 0.11165870800004996
1821
+ },
1822
+ {
1823
+ "clip_ratio/high_max": 0.0,
1824
+ "clip_ratio/high_mean": 0.0,
1825
+ "clip_ratio/low_mean": 0.0,
1826
+ "clip_ratio/low_min": 0.0,
1827
+ "clip_ratio/region_mean": 0.0,
1828
+ "entropy": 1.3093762397766113,
1829
+ "epoch": 2.5,
1830
+ "grad_norm": 0.1727670133113861,
1831
+ "kl": 0.23851320147514343,
1832
+ "learning_rate": 5.25e-07,
1833
+ "loss": 0.0005962830036878586,
1834
+ "step": 100,
1835
+ "step_time": 0.11271154099995329
1836
+ },
1837
+ {
1838
+ "clip_ratio/high_max": 0.0,
1839
+ "clip_ratio/high_mean": 0.0,
1840
+ "clip_ratio/low_mean": 0.0,
1841
+ "clip_ratio/low_min": 0.0,
1842
+ "clip_ratio/region_mean": 0.0,
1843
+ "completions/clipped_ratio": 1.0,
1844
+ "completions/max_length": 256.0,
1845
+ "completions/max_terminated_length": 0.0,
1846
+ "completions/mean_length": 256.0,
1847
+ "completions/mean_terminated_length": 0.0,
1848
+ "completions/min_length": 256.0,
1849
+ "completions/min_terminated_length": 0.0,
1850
+ "entropy": 1.2672348022460938,
1851
+ "epoch": 2.525,
1852
+ "frac_reward_zero_std": 0.5,
1853
+ "grad_norm": 1.4839428663253784,
1854
+ "kl": 0.19125787913799286,
1855
+ "learning_rate": 5e-07,
1856
+ "loss": -0.06198640912771225,
1857
+ "num_tokens": 261180.0,
1858
+ "reward": 0.8894580602645874,
1859
+ "reward_std": 0.14652438461780548,
1860
+ "rewards/AnswererRewardFunction/mean": 0.8894580602645874,
1861
+ "rewards/AnswererRewardFunction/std": 0.14652438461780548,
1862
+ "step": 101,
1863
+ "step_time": 8.041301798999939
1864
+ },
1865
+ {
1866
+ "clip_ratio/high_max": 0.001953125,
1867
+ "clip_ratio/high_mean": 0.001953125,
1868
+ "clip_ratio/low_mean": 0.0,
1869
+ "clip_ratio/low_min": 0.0,
1870
+ "clip_ratio/region_mean": 0.001953125,
1871
+ "entropy": 1.389164924621582,
1872
+ "epoch": 2.55,
1873
+ "grad_norm": 3.5372402667999268,
1874
+ "kl": 0.18080012500286102,
1875
+ "learning_rate": 4.75e-07,
1876
+ "loss": 0.031853705644607544,
1877
+ "step": 102,
1878
+ "step_time": 0.11171470999988742
1879
+ },
1880
+ {
1881
+ "clip_ratio/high_max": 0.0009765625,
1882
+ "clip_ratio/high_mean": 0.0009765625,
1883
+ "clip_ratio/low_mean": 0.0,
1884
+ "clip_ratio/low_min": 0.0,
1885
+ "clip_ratio/region_mean": 0.0009765625,
1886
+ "entropy": 1.111598253250122,
1887
+ "epoch": 2.575,
1888
+ "grad_norm": 2.8348701000213623,
1889
+ "kl": 0.2075410932302475,
1890
+ "learning_rate": 4.5e-07,
1891
+ "loss": 0.031495481729507446,
1892
+ "step": 103,
1893
+ "step_time": 0.11201832500000819
1894
+ },
1895
+ {
1896
+ "clip_ratio/high_max": 0.0,
1897
+ "clip_ratio/high_mean": 0.0,
1898
+ "clip_ratio/low_mean": 0.0,
1899
+ "clip_ratio/low_min": 0.0,
1900
+ "clip_ratio/region_mean": 0.0,
1901
+ "entropy": 1.4506624937057495,
1902
+ "epoch": 2.6,
1903
+ "grad_norm": 0.04200572893023491,
1904
+ "kl": 0.16218087077140808,
1905
+ "learning_rate": 4.25e-07,
1906
+ "loss": 0.0004054521559737623,
1907
+ "step": 104,
1908
+ "step_time": 0.11145493399999395
1909
+ },
1910
+ {
1911
+ "clip_ratio/high_max": 0.0,
1912
+ "clip_ratio/high_mean": 0.0,
1913
+ "clip_ratio/low_mean": 0.0,
1914
+ "clip_ratio/low_min": 0.0,
1915
+ "clip_ratio/region_mean": 0.0,
1916
+ "completions/clipped_ratio": 1.0,
1917
+ "completions/max_length": 256.0,
1918
+ "completions/max_terminated_length": 0.0,
1919
+ "completions/mean_length": 256.0,
1920
+ "completions/mean_terminated_length": 0.0,
1921
+ "completions/min_length": 256.0,
1922
+ "completions/min_terminated_length": 0.0,
1923
+ "entropy": 1.3652623891830444,
1924
+ "epoch": 2.625,
1925
+ "frac_reward_zero_std": 0.75,
1926
+ "grad_norm": 0.03890267759561539,
1927
+ "kl": 0.15317970514297485,
1928
+ "learning_rate": 4e-07,
1929
+ "loss": 0.00038294930709525943,
1930
+ "num_tokens": 271156.0,
1931
+ "reward": 0.8711539506912231,
1932
+ "reward_std": 0.10366164892911911,
1933
+ "rewards/AnswererRewardFunction/mean": 0.8711539506912231,
1934
+ "rewards/AnswererRewardFunction/std": 0.10366164147853851,
1935
+ "step": 105,
1936
+ "step_time": 8.082592108999961
1937
+ },
1938
+ {
1939
+ "clip_ratio/high_max": 0.0,
1940
+ "clip_ratio/high_mean": 0.0,
1941
+ "clip_ratio/low_mean": 0.0009765625,
1942
+ "clip_ratio/low_min": 0.0009765625,
1943
+ "clip_ratio/region_mean": 0.0009765625,
1944
+ "entropy": 1.4629085063934326,
1945
+ "epoch": 2.65,
1946
+ "grad_norm": 3.0278525352478027,
1947
+ "kl": 0.20959222316741943,
1948
+ "learning_rate": 3.75e-07,
1949
+ "loss": 0.06314528733491898,
1950
+ "step": 106,
1951
+ "step_time": 0.11130337599979612
1952
+ },
1953
+ {
1954
+ "clip_ratio/high_max": 0.0009765625,
1955
+ "clip_ratio/high_mean": 0.0009765625,
1956
+ "clip_ratio/low_mean": 0.0,
1957
+ "clip_ratio/low_min": 0.0,
1958
+ "clip_ratio/region_mean": 0.0009765625,
1959
+ "entropy": 1.1185526847839355,
1960
+ "epoch": 2.675,
1961
+ "grad_norm": 0.8490508794784546,
1962
+ "kl": 0.15610143542289734,
1963
+ "learning_rate": 3.5000000000000004e-07,
1964
+ "loss": -0.030861012637615204,
1965
+ "step": 107,
1966
+ "step_time": 0.11095707699996638
1967
+ },
1968
+ {
1969
+ "clip_ratio/high_max": 0.0,
1970
+ "clip_ratio/high_mean": 0.0,
1971
+ "clip_ratio/low_mean": 0.0,
1972
+ "clip_ratio/low_min": 0.0,
1973
+ "clip_ratio/region_mean": 0.0,
1974
+ "entropy": 1.3533684015274048,
1975
+ "epoch": 2.7,
1976
+ "grad_norm": 0.9826887249946594,
1977
+ "kl": 0.1878659874200821,
1978
+ "learning_rate": 3.25e-07,
1979
+ "loss": -0.030797190964221954,
1980
+ "step": 108,
1981
+ "step_time": 0.11087298700022075
1982
+ },
1983
+ {
1984
+ "clip_ratio/high_max": 0.0,
1985
+ "clip_ratio/high_mean": 0.0,
1986
+ "clip_ratio/low_mean": 0.0,
1987
+ "clip_ratio/low_min": 0.0,
1988
+ "clip_ratio/region_mean": 0.0,
1989
+ "completions/clipped_ratio": 1.0,
1990
+ "completions/max_length": 256.0,
1991
+ "completions/max_terminated_length": 0.0,
1992
+ "completions/mean_length": 256.0,
1993
+ "completions/mean_terminated_length": 0.0,
1994
+ "completions/min_length": 256.0,
1995
+ "completions/min_terminated_length": 0.0,
1996
+ "entropy": 1.4031682014465332,
1997
+ "epoch": 2.725,
1998
+ "frac_reward_zero_std": 1.0,
1999
+ "grad_norm": 0.06574197858572006,
2000
+ "kl": 0.18314988911151886,
2001
+ "learning_rate": 3.0000000000000004e-07,
2002
+ "loss": 0.000457874673884362,
2003
+ "num_tokens": 281352.0,
2004
+ "reward": 0.8702797293663025,
2005
+ "reward_std": 0.11096905916929245,
2006
+ "rewards/AnswererRewardFunction/mean": 0.8702797293663025,
2007
+ "rewards/AnswererRewardFunction/std": 0.11096906661987305,
2008
+ "step": 109,
2009
+ "step_time": 7.987654709000253
2010
+ },
2011
+ {
2012
+ "clip_ratio/high_max": 0.0,
2013
+ "clip_ratio/high_mean": 0.0,
2014
+ "clip_ratio/low_mean": 0.0,
2015
+ "clip_ratio/low_min": 0.0,
2016
+ "clip_ratio/region_mean": 0.0,
2017
+ "entropy": 1.5492128133773804,
2018
+ "epoch": 2.75,
2019
+ "grad_norm": 0.04146639257669449,
2020
+ "kl": 0.17638707160949707,
2021
+ "learning_rate": 2.7499999999999996e-07,
2022
+ "loss": 0.00044096767669543624,
2023
+ "step": 110,
2024
+ "step_time": 0.1102177429997937
2025
+ },
2026
+ {
2027
+ "clip_ratio/high_max": 0.0,
2028
+ "clip_ratio/high_mean": 0.0,
2029
+ "clip_ratio/low_mean": 0.0,
2030
+ "clip_ratio/low_min": 0.0,
2031
+ "clip_ratio/region_mean": 0.0,
2032
+ "entropy": 1.1322059631347656,
2033
+ "epoch": 2.775,
2034
+ "grad_norm": 0.045318856835365295,
2035
+ "kl": 0.17277760803699493,
2036
+ "learning_rate": 2.5e-07,
2037
+ "loss": 0.0004319440049584955,
2038
+ "step": 111,
2039
+ "step_time": 0.11153030900004524
2040
+ },
2041
+ {
2042
+ "clip_ratio/high_max": 0.0,
2043
+ "clip_ratio/high_mean": 0.0,
2044
+ "clip_ratio/low_mean": 0.0,
2045
+ "clip_ratio/low_min": 0.0,
2046
+ "clip_ratio/region_mean": 0.0,
2047
+ "entropy": 1.2663240432739258,
2048
+ "epoch": 2.8,
2049
+ "grad_norm": 0.036373987793922424,
2050
+ "kl": 0.12916100025177002,
2051
+ "learning_rate": 2.25e-07,
2052
+ "loss": 0.0003229025169275701,
2053
+ "step": 112,
2054
+ "step_time": 0.1101257909999731
2055
+ },
2056
+ {
2057
+ "clip_ratio/high_max": 0.0,
2058
+ "clip_ratio/high_mean": 0.0,
2059
+ "clip_ratio/low_mean": 0.0,
2060
+ "clip_ratio/low_min": 0.0,
2061
+ "clip_ratio/region_mean": 0.0,
2062
+ "completions/clipped_ratio": 1.0,
2063
+ "completions/max_length": 256.0,
2064
+ "completions/max_terminated_length": 0.0,
2065
+ "completions/mean_length": 256.0,
2066
+ "completions/mean_terminated_length": 0.0,
2067
+ "completions/min_length": 256.0,
2068
+ "completions/min_terminated_length": 0.0,
2069
+ "entropy": 1.3358656167984009,
2070
+ "epoch": 2.825,
2071
+ "frac_reward_zero_std": 1.0,
2072
+ "grad_norm": 0.03242846205830574,
2073
+ "kl": 0.137445330619812,
2074
+ "learning_rate": 2e-07,
2075
+ "loss": 0.00034361338475719094,
2076
+ "num_tokens": 291448.0,
2077
+ "reward": 0.8755199909210205,
2078
+ "reward_std": 0.10159505903720856,
2079
+ "rewards/AnswererRewardFunction/mean": 0.8755199909210205,
2080
+ "rewards/AnswererRewardFunction/std": 0.10159505903720856,
2081
+ "step": 113,
2082
+ "step_time": 7.975539141999889
2083
+ },
2084
+ {
2085
+ "clip_ratio/high_max": 0.0,
2086
+ "clip_ratio/high_mean": 0.0,
2087
+ "clip_ratio/low_mean": 0.0,
2088
+ "clip_ratio/low_min": 0.0,
2089
+ "clip_ratio/region_mean": 0.0,
2090
+ "entropy": 1.492983341217041,
2091
+ "epoch": 2.85,
2092
+ "grad_norm": 0.08132708072662354,
2093
+ "kl": 0.1894962638616562,
2094
+ "learning_rate": 1.7500000000000002e-07,
2095
+ "loss": 0.0004737406561616808,
2096
+ "step": 114,
2097
+ "step_time": 0.11158334499987177
2098
+ },
2099
+ {
2100
+ "clip_ratio/high_max": 0.0,
2101
+ "clip_ratio/high_mean": 0.0,
2102
+ "clip_ratio/low_mean": 0.0,
2103
+ "clip_ratio/low_min": 0.0,
2104
+ "clip_ratio/region_mean": 0.0,
2105
+ "entropy": 1.2457528114318848,
2106
+ "epoch": 2.875,
2107
+ "grad_norm": 0.08294132351875305,
2108
+ "kl": 0.17598748207092285,
2109
+ "learning_rate": 1.5000000000000002e-07,
2110
+ "loss": 0.0004399687168188393,
2111
+ "step": 115,
2112
+ "step_time": 0.1111911970001529
2113
+ },
2114
+ {
2115
+ "clip_ratio/high_max": 0.0,
2116
+ "clip_ratio/high_mean": 0.0,
2117
+ "clip_ratio/low_mean": 0.0,
2118
+ "clip_ratio/low_min": 0.0,
2119
+ "clip_ratio/region_mean": 0.0,
2120
+ "entropy": 1.3998165130615234,
2121
+ "epoch": 2.9,
2122
+ "grad_norm": 0.04320406913757324,
2123
+ "kl": 0.15890346467494965,
2124
+ "learning_rate": 1.25e-07,
2125
+ "loss": 0.00039725861279293895,
2126
+ "step": 116,
2127
+ "step_time": 0.11117327600004501
2128
+ },
2129
+ {
2130
+ "clip_ratio/high_max": 0.0,
2131
+ "clip_ratio/high_mean": 0.0,
2132
+ "clip_ratio/low_mean": 0.0,
2133
+ "clip_ratio/low_min": 0.0,
2134
+ "clip_ratio/region_mean": 0.0,
2135
+ "completions/clipped_ratio": 1.0,
2136
+ "completions/max_length": 256.0,
2137
+ "completions/max_terminated_length": 0.0,
2138
+ "completions/mean_length": 256.0,
2139
+ "completions/mean_terminated_length": 0.0,
2140
+ "completions/min_length": 256.0,
2141
+ "completions/min_terminated_length": 0.0,
2142
+ "entropy": 1.4396095275878906,
2143
+ "epoch": 2.925,
2144
+ "frac_reward_zero_std": 1.0,
2145
+ "grad_norm": 0.04113089293241501,
2146
+ "kl": 0.17974942922592163,
2147
+ "learning_rate": 1e-07,
2148
+ "loss": 0.0004493735614232719,
2149
+ "num_tokens": 301548.0,
2150
+ "reward": 0.8100648522377014,
2151
+ "reward_std": 0.12688994407653809,
2152
+ "rewards/AnswererRewardFunction/mean": 0.8100648522377014,
2153
+ "rewards/AnswererRewardFunction/std": 0.12688994407653809,
2154
+ "step": 117,
2155
+ "step_time": 7.944180141000288
2156
+ },
2157
+ {
2158
+ "clip_ratio/high_max": 0.0,
2159
+ "clip_ratio/high_mean": 0.0,
2160
+ "clip_ratio/low_mean": 0.0,
2161
+ "clip_ratio/low_min": 0.0,
2162
+ "clip_ratio/region_mean": 0.0,
2163
+ "entropy": 1.355454683303833,
2164
+ "epoch": 2.95,
2165
+ "grad_norm": 0.051230356097221375,
2166
+ "kl": 0.17781737446784973,
2167
+ "learning_rate": 7.500000000000001e-08,
2168
+ "loss": 0.00044454343151301146,
2169
+ "step": 118,
2170
+ "step_time": 0.11226749899969946
2171
+ },
2172
+ {
2173
+ "clip_ratio/high_max": 0.0,
2174
+ "clip_ratio/high_mean": 0.0,
2175
+ "clip_ratio/low_mean": 0.0,
2176
+ "clip_ratio/low_min": 0.0,
2177
+ "clip_ratio/region_mean": 0.0,
2178
+ "entropy": 1.5611437559127808,
2179
+ "epoch": 2.975,
2180
+ "grad_norm": 0.04255163297057152,
2181
+ "kl": 0.18407562375068665,
2182
+ "learning_rate": 5e-08,
2183
+ "loss": 0.00046018906868994236,
2184
+ "step": 119,
2185
+ "step_time": 0.11076878699986992
2186
+ },
2187
+ {
2188
+ "clip_ratio/high_max": 0.0,
2189
+ "clip_ratio/high_mean": 0.0,
2190
+ "clip_ratio/low_mean": 0.0,
2191
+ "clip_ratio/low_min": 0.0,
2192
+ "clip_ratio/region_mean": 0.0,
2193
+ "entropy": 1.446387767791748,
2194
+ "epoch": 3.0,
2195
+ "grad_norm": 0.041754741221666336,
2196
+ "kl": 0.1644238829612732,
2197
+ "learning_rate": 2.5e-08,
2198
+ "loss": 0.000411059707403183,
2199
+ "step": 120,
2200
+ "step_time": 0.11240428199971575
2201
+ }
2202
+ ],
2203
+ "logging_steps": 1,
2204
+ "max_steps": 120,
2205
+ "num_input_tokens_seen": 301548,
2206
+ "num_train_epochs": 3,
2207
+ "save_steps": 30,
2208
+ "stateful_callbacks": {
2209
+ "TrainerControl": {
2210
+ "args": {
2211
+ "should_epoch_stop": false,
2212
+ "should_evaluate": false,
2213
+ "should_log": false,
2214
+ "should_save": true,
2215
+ "should_training_stop": true
2216
+ },
2217
+ "attributes": {}
2218
+ }
2219
+ },
2220
+ "total_flos": 0.0,
2221
+ "train_batch_size": 4,
2222
+ "trial_name": null,
2223
+ "trial_params": null
2224
+ }
self_play_hf_l40s_full/round_002/answerer_train/checkpoint-120/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:07f752d454d10dde5c777669d2fdda366c9140734be3a0440014e6c662b1b868
3
+ size 7249