Siddeshwar1625 commited on
Commit
7dc0e22
·
verified ·
1 Parent(s): 41cb25c

Intermediate checkpoint upload step=120 (answerer_train)

Browse files
.gitattributes CHANGED
@@ -80,3 +80,4 @@ self_play_hf_l40s_full/round_004/generator_train/final_model/tokenizer.json filt
80
  self_play_hf_l40s_full/round_004/answerer_train/checkpoint-30/tokenizer.json filter=lfs diff=lfs merge=lfs -text
81
  self_play_hf_l40s_full/round_004/answerer_train/checkpoint-60/tokenizer.json filter=lfs diff=lfs merge=lfs -text
82
  self_play_hf_l40s_full/round_004/answerer_train/checkpoint-90/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 
 
80
  self_play_hf_l40s_full/round_004/answerer_train/checkpoint-30/tokenizer.json filter=lfs diff=lfs merge=lfs -text
81
  self_play_hf_l40s_full/round_004/answerer_train/checkpoint-60/tokenizer.json filter=lfs diff=lfs merge=lfs -text
82
  self_play_hf_l40s_full/round_004/answerer_train/checkpoint-90/tokenizer.json filter=lfs diff=lfs merge=lfs -text
83
+ self_play_hf_l40s_full/round_004/answerer_train/checkpoint-120/tokenizer.json filter=lfs diff=lfs merge=lfs -text
self_play_hf_l40s_full/round_004/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_004/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_004/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_004/answerer_train/checkpoint-120/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:48e200a9bce551e2816c8ae9aa3a9439d68e1400fec13b52a770efb923f55060
3
+ size 1976163472
self_play_hf_l40s_full/round_004/answerer_train/checkpoint-120/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:08fb5a16e7147ef26f7d4ccda154b4b022bf29fe0273ec63345061e26db03ec7
3
+ size 3952509771
self_play_hf_l40s_full/round_004/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_004/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_004/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_004/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_004/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": 1.808485507965088,
26
+ "epoch": 0.025,
27
+ "frac_reward_zero_std": 0.75,
28
+ "grad_norm": 1.511763095855713,
29
+ "kl": 0.0,
30
+ "learning_rate": 3e-06,
31
+ "loss": -0.062454625964164734,
32
+ "num_tokens": 10016.0,
33
+ "reward": 0.796283483505249,
34
+ "reward_std": 0.15437306463718414,
35
+ "rewards/AnswererRewardFunction/mean": 0.796283483505249,
36
+ "rewards/AnswererRewardFunction/std": 0.15437307953834534,
37
+ "step": 1,
38
+ "step_time": 7.744606889000352
39
+ },
40
+ {
41
+ "clip_ratio/high_max": 0.0,
42
+ "clip_ratio/high_mean": 0.0,
43
+ "clip_ratio/low_mean": 0.0,
44
+ "clip_ratio/low_min": 0.0,
45
+ "clip_ratio/region_mean": 0.0,
46
+ "entropy": 2.061605215072632,
47
+ "epoch": 0.05,
48
+ "grad_norm": 0.030779404565691948,
49
+ "kl": 0.023394376039505005,
50
+ "learning_rate": 2.9750000000000003e-06,
51
+ "loss": 5.848593355040066e-05,
52
+ "step": 2,
53
+ "step_time": 0.11778782400051568
54
+ },
55
+ {
56
+ "clip_ratio/high_max": 0.0,
57
+ "clip_ratio/high_mean": 0.0,
58
+ "clip_ratio/low_mean": 0.029296875,
59
+ "clip_ratio/low_min": 0.029296875,
60
+ "clip_ratio/region_mean": 0.029296875,
61
+ "entropy": 2.177046537399292,
62
+ "epoch": 0.075,
63
+ "grad_norm": 3.1303863525390625,
64
+ "kl": 0.08087262511253357,
65
+ "learning_rate": 2.9499999999999997e-06,
66
+ "loss": 0.09753571450710297,
67
+ "step": 3,
68
+ "step_time": 0.1114002790000086
69
+ },
70
+ {
71
+ "clip_ratio/high_max": 0.0361328125,
72
+ "clip_ratio/high_mean": 0.0361328125,
73
+ "clip_ratio/low_mean": 0.0,
74
+ "clip_ratio/low_min": 0.0,
75
+ "clip_ratio/region_mean": 0.0361328125,
76
+ "entropy": 2.0383293628692627,
77
+ "epoch": 0.1,
78
+ "grad_norm": 0.9711113572120667,
79
+ "kl": 0.13760070502758026,
80
+ "learning_rate": 2.925e-06,
81
+ "loss": -0.030087757855653763,
82
+ "step": 4,
83
+ "step_time": 0.11656204800056003
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": 2.2891623973846436,
99
+ "epoch": 0.125,
100
+ "frac_reward_zero_std": 0.25,
101
+ "grad_norm": 3.584872007369995,
102
+ "kl": 0.0656374841928482,
103
+ "learning_rate": 2.9e-06,
104
+ "loss": -0.00024243921507149935,
105
+ "num_tokens": 19780.0,
106
+ "reward": 0.7218090295791626,
107
+ "reward_std": 0.14923875033855438,
108
+ "rewards/AnswererRewardFunction/mean": 0.7218090295791626,
109
+ "rewards/AnswererRewardFunction/std": 0.14923876523971558,
110
+ "step": 5,
111
+ "step_time": 7.795673902999624
112
+ },
113
+ {
114
+ "clip_ratio/high_max": 0.0078125,
115
+ "clip_ratio/high_mean": 0.0078125,
116
+ "clip_ratio/low_mean": 0.0,
117
+ "clip_ratio/low_min": 0.0,
118
+ "clip_ratio/region_mean": 0.0078125,
119
+ "entropy": 2.2073974609375,
120
+ "epoch": 0.15,
121
+ "grad_norm": 0.9116904139518738,
122
+ "kl": 0.08468662202358246,
123
+ "learning_rate": 2.875e-06,
124
+ "loss": -0.031071916222572327,
125
+ "step": 6,
126
+ "step_time": 0.1147218090000024
127
+ },
128
+ {
129
+ "clip_ratio/high_max": 0.0458984375,
130
+ "clip_ratio/high_mean": 0.0458984375,
131
+ "clip_ratio/low_mean": 0.0048828125,
132
+ "clip_ratio/low_min": 0.0048828125,
133
+ "clip_ratio/region_mean": 0.05078125,
134
+ "entropy": 1.828927755355835,
135
+ "epoch": 0.175,
136
+ "grad_norm": 2.9821267127990723,
137
+ "kl": 0.10931183397769928,
138
+ "learning_rate": 2.85e-06,
139
+ "loss": -0.052874594926834106,
140
+ "step": 7,
141
+ "step_time": 0.11118124299991905
142
+ },
143
+ {
144
+ "clip_ratio/high_max": 0.060546875,
145
+ "clip_ratio/high_mean": 0.060546875,
146
+ "clip_ratio/low_mean": 0.0693359375,
147
+ "clip_ratio/low_min": 0.0693359375,
148
+ "clip_ratio/region_mean": 0.1298828125,
149
+ "entropy": 2.239431619644165,
150
+ "epoch": 0.2,
151
+ "grad_norm": 3.761061668395996,
152
+ "kl": 0.11145439743995667,
153
+ "learning_rate": 2.825e-06,
154
+ "loss": 0.08703643828630447,
155
+ "step": 8,
156
+ "step_time": 0.11017163199994684
157
+ },
158
+ {
159
+ "clip_ratio/high_max": 0.001953125,
160
+ "clip_ratio/high_mean": 0.001953125,
161
+ "clip_ratio/low_mean": 0.0,
162
+ "clip_ratio/low_min": 0.0,
163
+ "clip_ratio/region_mean": 0.001953125,
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": 2.2865660190582275,
172
+ "epoch": 0.225,
173
+ "frac_reward_zero_std": 0.25,
174
+ "grad_norm": 1.6357184648513794,
175
+ "kl": 0.10884629935026169,
176
+ "learning_rate": 2.8000000000000003e-06,
177
+ "loss": 0.03159352019429207,
178
+ "num_tokens": 29592.0,
179
+ "reward": 0.6900650858879089,
180
+ "reward_std": 0.1505999118089676,
181
+ "rewards/AnswererRewardFunction/mean": 0.6900650858879089,
182
+ "rewards/AnswererRewardFunction/std": 0.1505999118089676,
183
+ "step": 9,
184
+ "step_time": 7.738351545999649
185
+ },
186
+ {
187
+ "clip_ratio/high_max": 0.001953125,
188
+ "clip_ratio/high_mean": 0.001953125,
189
+ "clip_ratio/low_mean": 0.0078125,
190
+ "clip_ratio/low_min": 0.0078125,
191
+ "clip_ratio/region_mean": 0.009765625,
192
+ "entropy": 2.2734012603759766,
193
+ "epoch": 0.25,
194
+ "grad_norm": 4.225009918212891,
195
+ "kl": 0.14073708653450012,
196
+ "learning_rate": 2.775e-06,
197
+ "loss": 0.1410004049539566,
198
+ "step": 10,
199
+ "step_time": 0.1489122450002469
200
+ },
201
+ {
202
+ "clip_ratio/high_max": 0.0263671875,
203
+ "clip_ratio/high_mean": 0.0263671875,
204
+ "clip_ratio/low_mean": 0.015625,
205
+ "clip_ratio/low_min": 0.015625,
206
+ "clip_ratio/region_mean": 0.0419921875,
207
+ "entropy": 2.2513952255249023,
208
+ "epoch": 0.275,
209
+ "grad_norm": 3.980247735977173,
210
+ "kl": 0.12798824906349182,
211
+ "learning_rate": 2.75e-06,
212
+ "loss": -0.13872209191322327,
213
+ "step": 11,
214
+ "step_time": 0.11282693799967092
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": 2.080702781677246,
223
+ "epoch": 0.3,
224
+ "grad_norm": 0.9219491481781006,
225
+ "kl": 0.12932462990283966,
226
+ "learning_rate": 2.725e-06,
227
+ "loss": -0.0311975609511137,
228
+ "step": 12,
229
+ "step_time": 0.11239939000006416
230
+ },
231
+ {
232
+ "clip_ratio/high_max": 0.0,
233
+ "clip_ratio/high_mean": 0.0,
234
+ "clip_ratio/low_mean": 0.0,
235
+ "clip_ratio/low_min": 0.0,
236
+ "clip_ratio/region_mean": 0.0,
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": 1.9584801197052002,
245
+ "epoch": 0.325,
246
+ "frac_reward_zero_std": 0.75,
247
+ "grad_norm": 0.03335069492459297,
248
+ "kl": 0.1676618754863739,
249
+ "learning_rate": 2.7e-06,
250
+ "loss": 0.0004191546468064189,
251
+ "num_tokens": 39564.0,
252
+ "reward": 0.8116456270217896,
253
+ "reward_std": 0.11149715632200241,
254
+ "rewards/AnswererRewardFunction/mean": 0.8116456270217896,
255
+ "rewards/AnswererRewardFunction/std": 0.11149716377258301,
256
+ "step": 13,
257
+ "step_time": 7.745207232000212
258
+ },
259
+ {
260
+ "clip_ratio/high_max": 0.0,
261
+ "clip_ratio/high_mean": 0.0,
262
+ "clip_ratio/low_mean": 0.0,
263
+ "clip_ratio/low_min": 0.0,
264
+ "clip_ratio/region_mean": 0.0,
265
+ "entropy": 2.2673592567443848,
266
+ "epoch": 0.35,
267
+ "grad_norm": 0.04296092316508293,
268
+ "kl": 0.1360304206609726,
269
+ "learning_rate": 2.6750000000000002e-06,
270
+ "loss": 0.0003400760469958186,
271
+ "step": 14,
272
+ "step_time": 0.11098732899972674
273
+ },
274
+ {
275
+ "clip_ratio/high_max": 0.0029296875,
276
+ "clip_ratio/high_mean": 0.0029296875,
277
+ "clip_ratio/low_mean": 0.0,
278
+ "clip_ratio/low_min": 0.0,
279
+ "clip_ratio/region_mean": 0.0029296875,
280
+ "entropy": 2.088756799697876,
281
+ "epoch": 0.375,
282
+ "grad_norm": 0.9206492900848389,
283
+ "kl": 0.1489439606666565,
284
+ "learning_rate": 2.65e-06,
285
+ "loss": -0.031076664105057716,
286
+ "step": 15,
287
+ "step_time": 0.11077636499976506
288
+ },
289
+ {
290
+ "clip_ratio/high_max": 0.0263671875,
291
+ "clip_ratio/high_mean": 0.0263671875,
292
+ "clip_ratio/low_mean": 0.0078125,
293
+ "clip_ratio/low_min": 0.0078125,
294
+ "clip_ratio/region_mean": 0.0341796875,
295
+ "entropy": 2.237659454345703,
296
+ "epoch": 0.4,
297
+ "grad_norm": 3.1504709720611572,
298
+ "kl": 0.16802266240119934,
299
+ "learning_rate": 2.6250000000000003e-06,
300
+ "loss": 0.03298211842775345,
301
+ "step": 16,
302
+ "step_time": 0.11010160100067878
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": 2.0111589431762695,
318
+ "epoch": 0.425,
319
+ "frac_reward_zero_std": 1.0,
320
+ "grad_norm": 0.06525206565856934,
321
+ "kl": 0.15062233805656433,
322
+ "learning_rate": 2.6e-06,
323
+ "loss": 0.0003765558358281851,
324
+ "num_tokens": 49924.0,
325
+ "reward": 0.9323133230209351,
326
+ "reward_std": 0.0,
327
+ "rewards/AnswererRewardFunction/mean": 0.9323133230209351,
328
+ "rewards/AnswererRewardFunction/std": 0.0,
329
+ "step": 17,
330
+ "step_time": 7.76931892399989
331
+ },
332
+ {
333
+ "clip_ratio/high_max": 0.0,
334
+ "clip_ratio/high_mean": 0.0,
335
+ "clip_ratio/low_mean": 0.0,
336
+ "clip_ratio/low_min": 0.0,
337
+ "clip_ratio/region_mean": 0.0,
338
+ "entropy": 1.9240071773529053,
339
+ "epoch": 0.45,
340
+ "grad_norm": 0.03619404137134552,
341
+ "kl": 0.1399194598197937,
342
+ "learning_rate": 2.575e-06,
343
+ "loss": 0.0003497986472211778,
344
+ "step": 18,
345
+ "step_time": 0.11182963699957327
346
+ },
347
+ {
348
+ "clip_ratio/high_max": 0.0,
349
+ "clip_ratio/high_mean": 0.0,
350
+ "clip_ratio/low_mean": 0.0,
351
+ "clip_ratio/low_min": 0.0,
352
+ "clip_ratio/region_mean": 0.0,
353
+ "entropy": 1.7755590677261353,
354
+ "epoch": 0.475,
355
+ "grad_norm": 0.046630166471004486,
356
+ "kl": 0.1555626392364502,
357
+ "learning_rate": 2.55e-06,
358
+ "loss": 0.00038890657015144825,
359
+ "step": 19,
360
+ "step_time": 0.11152786100046796
361
+ },
362
+ {
363
+ "clip_ratio/high_max": 0.0,
364
+ "clip_ratio/high_mean": 0.0,
365
+ "clip_ratio/low_mean": 0.0,
366
+ "clip_ratio/low_min": 0.0,
367
+ "clip_ratio/region_mean": 0.0,
368
+ "entropy": 2.1256823539733887,
369
+ "epoch": 0.5,
370
+ "grad_norm": 0.04450196027755737,
371
+ "kl": 0.1300331950187683,
372
+ "learning_rate": 2.525e-06,
373
+ "loss": 0.0003250829759053886,
374
+ "step": 20,
375
+ "step_time": 0.11261919400021725
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": 2.0368661880493164,
391
+ "epoch": 0.525,
392
+ "frac_reward_zero_std": 1.0,
393
+ "grad_norm": 0.03705684095621109,
394
+ "kl": 0.16164252161979675,
395
+ "learning_rate": 2.5e-06,
396
+ "loss": 0.00040410630754195154,
397
+ "num_tokens": 60044.0,
398
+ "reward": 0.819347620010376,
399
+ "reward_std": 0.11957820504903793,
400
+ "rewards/AnswererRewardFunction/mean": 0.819347620010376,
401
+ "rewards/AnswererRewardFunction/std": 0.11957821249961853,
402
+ "step": 21,
403
+ "step_time": 7.867706271000316
404
+ },
405
+ {
406
+ "clip_ratio/high_max": 0.0,
407
+ "clip_ratio/high_mean": 0.0,
408
+ "clip_ratio/low_mean": 0.0,
409
+ "clip_ratio/low_min": 0.0,
410
+ "clip_ratio/region_mean": 0.0,
411
+ "entropy": 1.8853070735931396,
412
+ "epoch": 0.55,
413
+ "grad_norm": 0.055804409086704254,
414
+ "kl": 0.16001088917255402,
415
+ "learning_rate": 2.475e-06,
416
+ "loss": 0.00040002723108045757,
417
+ "step": 22,
418
+ "step_time": 0.11128929000005883
419
+ },
420
+ {
421
+ "clip_ratio/high_max": 0.0,
422
+ "clip_ratio/high_mean": 0.0,
423
+ "clip_ratio/low_mean": 0.0,
424
+ "clip_ratio/low_min": 0.0,
425
+ "clip_ratio/region_mean": 0.0,
426
+ "entropy": 1.9584823846817017,
427
+ "epoch": 0.575,
428
+ "grad_norm": 0.04979805648326874,
429
+ "kl": 0.16607147455215454,
430
+ "learning_rate": 2.45e-06,
431
+ "loss": 0.00041517868521623313,
432
+ "step": 23,
433
+ "step_time": 0.11279742299939244
434
+ },
435
+ {
436
+ "clip_ratio/high_max": 0.0,
437
+ "clip_ratio/high_mean": 0.0,
438
+ "clip_ratio/low_mean": 0.0,
439
+ "clip_ratio/low_min": 0.0,
440
+ "clip_ratio/region_mean": 0.0,
441
+ "entropy": 1.9190866947174072,
442
+ "epoch": 0.6,
443
+ "grad_norm": 0.03331170231103897,
444
+ "kl": 0.1263861209154129,
445
+ "learning_rate": 2.425e-06,
446
+ "loss": 0.00031596526969224215,
447
+ "step": 24,
448
+ "step_time": 0.11192457400011335
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": 1.9539425373077393,
464
+ "epoch": 0.625,
465
+ "frac_reward_zero_std": 0.75,
466
+ "grad_norm": 1.0814502239227295,
467
+ "kl": 0.17089709639549255,
468
+ "learning_rate": 2.4000000000000003e-06,
469
+ "loss": 0.03165344148874283,
470
+ "num_tokens": 70048.0,
471
+ "reward": 0.8298832774162292,
472
+ "reward_std": 0.12205864489078522,
473
+ "rewards/AnswererRewardFunction/mean": 0.8298832774162292,
474
+ "rewards/AnswererRewardFunction/std": 0.12205864489078522,
475
+ "step": 25,
476
+ "step_time": 7.7907283719996485
477
+ },
478
+ {
479
+ "clip_ratio/high_max": 0.0,
480
+ "clip_ratio/high_mean": 0.0,
481
+ "clip_ratio/low_mean": 0.001953125,
482
+ "clip_ratio/low_min": 0.001953125,
483
+ "clip_ratio/region_mean": 0.001953125,
484
+ "entropy": 1.988601565361023,
485
+ "epoch": 0.65,
486
+ "grad_norm": 1.3814729452133179,
487
+ "kl": 0.18201372027397156,
488
+ "learning_rate": 2.375e-06,
489
+ "loss": 0.06283903121948242,
490
+ "step": 26,
491
+ "step_time": 0.11003638100009994
492
+ },
493
+ {
494
+ "clip_ratio/high_max": 0.001953125,
495
+ "clip_ratio/high_mean": 0.001953125,
496
+ "clip_ratio/low_mean": 0.0,
497
+ "clip_ratio/low_min": 0.0,
498
+ "clip_ratio/region_mean": 0.001953125,
499
+ "entropy": 2.075544834136963,
500
+ "epoch": 0.675,
501
+ "grad_norm": 3.0482325553894043,
502
+ "kl": 0.18647681176662445,
503
+ "learning_rate": 2.35e-06,
504
+ "loss": -0.09349231421947479,
505
+ "step": 27,
506
+ "step_time": 0.11002873100005672
507
+ },
508
+ {
509
+ "clip_ratio/high_max": 0.0,
510
+ "clip_ratio/high_mean": 0.0,
511
+ "clip_ratio/low_mean": 0.0,
512
+ "clip_ratio/low_min": 0.0,
513
+ "clip_ratio/region_mean": 0.0,
514
+ "entropy": 2.125493049621582,
515
+ "epoch": 0.7,
516
+ "grad_norm": 0.04627307131886482,
517
+ "kl": 0.19537895917892456,
518
+ "learning_rate": 2.325e-06,
519
+ "loss": 0.000488447374664247,
520
+ "step": 28,
521
+ "step_time": 0.1143881190000684
522
+ },
523
+ {
524
+ "clip_ratio/high_max": 0.0,
525
+ "clip_ratio/high_mean": 0.0,
526
+ "clip_ratio/low_mean": 0.0,
527
+ "clip_ratio/low_min": 0.0,
528
+ "clip_ratio/region_mean": 0.0,
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": 2.03523588180542,
537
+ "epoch": 0.725,
538
+ "frac_reward_zero_std": 0.75,
539
+ "grad_norm": 1.3853849172592163,
540
+ "kl": 0.1463131308555603,
541
+ "learning_rate": 2.3000000000000004e-06,
542
+ "loss": -0.06208648160099983,
543
+ "num_tokens": 80268.0,
544
+ "reward": 0.8684389591217041,
545
+ "reward_std": 0.09924228489398956,
546
+ "rewards/AnswererRewardFunction/mean": 0.8684389591217041,
547
+ "rewards/AnswererRewardFunction/std": 0.09924228489398956,
548
+ "step": 29,
549
+ "step_time": 7.750608561000263
550
+ },
551
+ {
552
+ "clip_ratio/high_max": 0.0,
553
+ "clip_ratio/high_mean": 0.0,
554
+ "clip_ratio/low_mean": 0.0,
555
+ "clip_ratio/low_min": 0.0,
556
+ "clip_ratio/region_mean": 0.0,
557
+ "entropy": 2.0841426849365234,
558
+ "epoch": 0.75,
559
+ "grad_norm": 2.837942361831665,
560
+ "kl": 0.1490478813648224,
561
+ "learning_rate": 2.275e-06,
562
+ "loss": 0.09484589099884033,
563
+ "step": 30,
564
+ "step_time": 0.11101365800004714
565
+ },
566
+ {
567
+ "clip_ratio/high_max": 0.0,
568
+ "clip_ratio/high_mean": 0.0,
569
+ "clip_ratio/low_mean": 0.0,
570
+ "clip_ratio/low_min": 0.0,
571
+ "clip_ratio/region_mean": 0.0,
572
+ "entropy": 2.026797294616699,
573
+ "epoch": 0.775,
574
+ "grad_norm": 0.045485783368349075,
575
+ "kl": 0.15276005864143372,
576
+ "learning_rate": 2.25e-06,
577
+ "loss": 0.0003819001140072942,
578
+ "step": 31,
579
+ "step_time": 0.11208413100030157
580
+ },
581
+ {
582
+ "clip_ratio/high_max": 0.0048828125,
583
+ "clip_ratio/high_mean": 0.0048828125,
584
+ "clip_ratio/low_mean": 0.0,
585
+ "clip_ratio/low_min": 0.0,
586
+ "clip_ratio/region_mean": 0.0048828125,
587
+ "entropy": 2.1123099327087402,
588
+ "epoch": 0.8,
589
+ "grad_norm": 0.9081559777259827,
590
+ "kl": 0.13755789399147034,
591
+ "learning_rate": 2.2250000000000003e-06,
592
+ "loss": -0.030523225665092468,
593
+ "step": 32,
594
+ "step_time": 0.11418771500029834
595
+ },
596
+ {
597
+ "clip_ratio/high_max": 0.0,
598
+ "clip_ratio/high_mean": 0.0,
599
+ "clip_ratio/low_mean": 0.0,
600
+ "clip_ratio/low_min": 0.0,
601
+ "clip_ratio/region_mean": 0.0,
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": 1.6823270320892334,
610
+ "epoch": 0.825,
611
+ "frac_reward_zero_std": 0.75,
612
+ "grad_norm": 2.9273855686187744,
613
+ "kl": 0.14870840311050415,
614
+ "learning_rate": 2.1999999999999997e-06,
615
+ "loss": -0.06208054721355438,
616
+ "num_tokens": 90568.0,
617
+ "reward": 0.8832219839096069,
618
+ "reward_std": 0.10554318875074387,
619
+ "rewards/AnswererRewardFunction/mean": 0.8832219839096069,
620
+ "rewards/AnswererRewardFunction/std": 0.10554319620132446,
621
+ "step": 33,
622
+ "step_time": 7.824478043999989
623
+ },
624
+ {
625
+ "clip_ratio/high_max": 0.0,
626
+ "clip_ratio/high_mean": 0.0,
627
+ "clip_ratio/low_mean": 0.0,
628
+ "clip_ratio/low_min": 0.0,
629
+ "clip_ratio/region_mean": 0.0,
630
+ "entropy": 2.322800397872925,
631
+ "epoch": 0.85,
632
+ "grad_norm": 1.030830979347229,
633
+ "kl": 0.1581311970949173,
634
+ "learning_rate": 2.175e-06,
635
+ "loss": 0.031573764979839325,
636
+ "step": 34,
637
+ "step_time": 0.1109909240003617
638
+ },
639
+ {
640
+ "clip_ratio/high_max": 0.0,
641
+ "clip_ratio/high_mean": 0.0,
642
+ "clip_ratio/low_mean": 0.001953125,
643
+ "clip_ratio/low_min": 0.001953125,
644
+ "clip_ratio/region_mean": 0.001953125,
645
+ "entropy": 2.079770088195801,
646
+ "epoch": 0.875,
647
+ "grad_norm": 1.0320844650268555,
648
+ "kl": 0.17237909138202667,
649
+ "learning_rate": 2.15e-06,
650
+ "loss": 0.03202669695019722,
651
+ "step": 35,
652
+ "step_time": 0.1104884630003653
653
+ },
654
+ {
655
+ "clip_ratio/high_max": 0.0,
656
+ "clip_ratio/high_mean": 0.0,
657
+ "clip_ratio/low_mean": 0.0,
658
+ "clip_ratio/low_min": 0.0,
659
+ "clip_ratio/region_mean": 0.0,
660
+ "entropy": 1.9677340984344482,
661
+ "epoch": 0.9,
662
+ "grad_norm": 0.05142974480986595,
663
+ "kl": 0.18628746271133423,
664
+ "learning_rate": 2.125e-06,
665
+ "loss": 0.000465718621853739,
666
+ "step": 36,
667
+ "step_time": 0.11126610900009837
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": 1.691863775253296,
683
+ "epoch": 0.925,
684
+ "frac_reward_zero_std": 0.75,
685
+ "grad_norm": 0.8771184086799622,
686
+ "kl": 0.15660348534584045,
687
+ "learning_rate": 2.1e-06,
688
+ "loss": 0.03148450702428818,
689
+ "num_tokens": 100596.0,
690
+ "reward": 0.8247425556182861,
691
+ "reward_std": 0.13356582820415497,
692
+ "rewards/AnswererRewardFunction/mean": 0.8247425556182861,
693
+ "rewards/AnswererRewardFunction/std": 0.13356582820415497,
694
+ "step": 37,
695
+ "step_time": 7.7617874139996275
696
+ },
697
+ {
698
+ "clip_ratio/high_max": 0.0,
699
+ "clip_ratio/high_mean": 0.0,
700
+ "clip_ratio/low_mean": 0.001953125,
701
+ "clip_ratio/low_min": 0.001953125,
702
+ "clip_ratio/region_mean": 0.001953125,
703
+ "entropy": 2.1614623069763184,
704
+ "epoch": 0.95,
705
+ "grad_norm": 0.9001947641372681,
706
+ "kl": 0.23196256160736084,
707
+ "learning_rate": 2.075e-06,
708
+ "loss": 0.03180602565407753,
709
+ "step": 38,
710
+ "step_time": 0.1102340079996793
711
+ },
712
+ {
713
+ "clip_ratio/high_max": 0.0,
714
+ "clip_ratio/high_mean": 0.0,
715
+ "clip_ratio/low_mean": 0.0009765625,
716
+ "clip_ratio/low_min": 0.0009765625,
717
+ "clip_ratio/region_mean": 0.0009765625,
718
+ "entropy": 1.8785492181777954,
719
+ "epoch": 0.975,
720
+ "grad_norm": 1.0380373001098633,
721
+ "kl": 0.1671696901321411,
722
+ "learning_rate": 2.0500000000000003e-06,
723
+ "loss": 0.03167015686631203,
724
+ "step": 39,
725
+ "step_time": 0.11017617600009544
726
+ },
727
+ {
728
+ "clip_ratio/high_max": 0.009765625,
729
+ "clip_ratio/high_mean": 0.009765625,
730
+ "clip_ratio/low_mean": 0.0,
731
+ "clip_ratio/low_min": 0.0,
732
+ "clip_ratio/region_mean": 0.009765625,
733
+ "entropy": 1.7612617015838623,
734
+ "epoch": 1.0,
735
+ "grad_norm": 2.548452138900757,
736
+ "kl": 0.20340929925441742,
737
+ "learning_rate": 2.025e-06,
738
+ "loss": -0.0916656106710434,
739
+ "step": 40,
740
+ "step_time": 0.11005827400003909
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": 1.935651183128357,
756
+ "epoch": 1.025,
757
+ "frac_reward_zero_std": 0.5,
758
+ "grad_norm": 0.9923874139785767,
759
+ "kl": 0.22261759638786316,
760
+ "learning_rate": 2e-06,
761
+ "loss": -0.0307498499751091,
762
+ "num_tokens": 110180.0,
763
+ "reward": 0.7413970232009888,
764
+ "reward_std": 0.10247324407100677,
765
+ "rewards/AnswererRewardFunction/mean": 0.7413970232009888,
766
+ "rewards/AnswererRewardFunction/std": 0.10247325897216797,
767
+ "step": 41,
768
+ "step_time": 7.812748901999839
769
+ },
770
+ {
771
+ "clip_ratio/high_max": 0.0,
772
+ "clip_ratio/high_mean": 0.0,
773
+ "clip_ratio/low_mean": 0.0029296875,
774
+ "clip_ratio/low_min": 0.0029296875,
775
+ "clip_ratio/region_mean": 0.0029296875,
776
+ "entropy": 1.7066082954406738,
777
+ "epoch": 1.05,
778
+ "grad_norm": 2.7330000400543213,
779
+ "kl": 0.1961289346218109,
780
+ "learning_rate": 1.975e-06,
781
+ "loss": 0.06310413777828217,
782
+ "step": 42,
783
+ "step_time": 0.10875788399971498
784
+ },
785
+ {
786
+ "clip_ratio/high_max": 0.001953125,
787
+ "clip_ratio/high_mean": 0.001953125,
788
+ "clip_ratio/low_mean": 0.0,
789
+ "clip_ratio/low_min": 0.0,
790
+ "clip_ratio/region_mean": 0.001953125,
791
+ "entropy": 1.829827070236206,
792
+ "epoch": 1.075,
793
+ "grad_norm": 1.2887893915176392,
794
+ "kl": 0.1938922256231308,
795
+ "learning_rate": 1.95e-06,
796
+ "loss": -0.061667174100875854,
797
+ "step": 43,
798
+ "step_time": 0.10980056400057947
799
+ },
800
+ {
801
+ "clip_ratio/high_max": 0.0107421875,
802
+ "clip_ratio/high_mean": 0.0107421875,
803
+ "clip_ratio/low_mean": 0.01171875,
804
+ "clip_ratio/low_min": 0.01171875,
805
+ "clip_ratio/region_mean": 0.0224609375,
806
+ "entropy": 2.006434440612793,
807
+ "epoch": 1.1,
808
+ "grad_norm": 2.9883673191070557,
809
+ "kl": 0.20987190306186676,
810
+ "learning_rate": 1.925e-06,
811
+ "loss": 0.03190411627292633,
812
+ "step": 44,
813
+ "step_time": 0.10804392199952417
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": 1.8556946516036987,
829
+ "epoch": 1.125,
830
+ "frac_reward_zero_std": 1.0,
831
+ "grad_norm": 0.0585976205766201,
832
+ "kl": 0.17927253246307373,
833
+ "learning_rate": 1.9e-06,
834
+ "loss": 0.00044818135211244226,
835
+ "num_tokens": 120220.0,
836
+ "reward": 0.8883568644523621,
837
+ "reward_std": 0.08731486648321152,
838
+ "rewards/AnswererRewardFunction/mean": 0.8883568644523621,
839
+ "rewards/AnswererRewardFunction/std": 0.08731484413146973,
840
+ "step": 45,
841
+ "step_time": 7.789618280000468
842
+ },
843
+ {
844
+ "clip_ratio/high_max": 0.0,
845
+ "clip_ratio/high_mean": 0.0,
846
+ "clip_ratio/low_mean": 0.0,
847
+ "clip_ratio/low_min": 0.0,
848
+ "clip_ratio/region_mean": 0.0,
849
+ "entropy": 1.771440863609314,
850
+ "epoch": 1.15,
851
+ "grad_norm": 0.05603954941034317,
852
+ "kl": 0.1786380112171173,
853
+ "learning_rate": 1.875e-06,
854
+ "loss": 0.00044659501872956753,
855
+ "step": 46,
856
+ "step_time": 0.11135138499957975
857
+ },
858
+ {
859
+ "clip_ratio/high_max": 0.0,
860
+ "clip_ratio/high_mean": 0.0,
861
+ "clip_ratio/low_mean": 0.0,
862
+ "clip_ratio/low_min": 0.0,
863
+ "clip_ratio/region_mean": 0.0,
864
+ "entropy": 1.9898463487625122,
865
+ "epoch": 1.175,
866
+ "grad_norm": 0.08282279223203659,
867
+ "kl": 0.21690496802330017,
868
+ "learning_rate": 1.85e-06,
869
+ "loss": 0.0005422623944468796,
870
+ "step": 47,
871
+ "step_time": 0.11093376600001648
872
+ },
873
+ {
874
+ "clip_ratio/high_max": 0.0,
875
+ "clip_ratio/high_mean": 0.0,
876
+ "clip_ratio/low_mean": 0.0,
877
+ "clip_ratio/low_min": 0.0,
878
+ "clip_ratio/region_mean": 0.0,
879
+ "entropy": 1.7627424001693726,
880
+ "epoch": 1.2,
881
+ "grad_norm": 0.06751802563667297,
882
+ "kl": 0.16698461771011353,
883
+ "learning_rate": 1.8249999999999999e-06,
884
+ "loss": 0.00041746150236576796,
885
+ "step": 48,
886
+ "step_time": 0.11264505899998767
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": 1.8894991874694824,
902
+ "epoch": 1.225,
903
+ "frac_reward_zero_std": 0.75,
904
+ "grad_norm": 0.03931179642677307,
905
+ "kl": 0.16673244535923004,
906
+ "learning_rate": 1.8e-06,
907
+ "loss": 0.0004168311133980751,
908
+ "num_tokens": 130268.0,
909
+ "reward": 0.88603675365448,
910
+ "reward_std": 0.10359743237495422,
911
+ "rewards/AnswererRewardFunction/mean": 0.88603675365448,
912
+ "rewards/AnswererRewardFunction/std": 0.10359743237495422,
913
+ "step": 49,
914
+ "step_time": 7.824710242999572
915
+ },
916
+ {
917
+ "clip_ratio/high_max": 0.0,
918
+ "clip_ratio/high_mean": 0.0,
919
+ "clip_ratio/low_mean": 0.0,
920
+ "clip_ratio/low_min": 0.0,
921
+ "clip_ratio/region_mean": 0.0,
922
+ "entropy": 1.8960050344467163,
923
+ "epoch": 1.25,
924
+ "grad_norm": 2.838279962539673,
925
+ "kl": 0.18674209713935852,
926
+ "learning_rate": 1.7750000000000002e-06,
927
+ "loss": 0.0941702127456665,
928
+ "step": 50,
929
+ "step_time": 0.11219863099995564
930
+ },
931
+ {
932
+ "clip_ratio/high_max": 0.0,
933
+ "clip_ratio/high_mean": 0.0,
934
+ "clip_ratio/low_mean": 0.0,
935
+ "clip_ratio/low_min": 0.0,
936
+ "clip_ratio/region_mean": 0.0,
937
+ "entropy": 1.9281809329986572,
938
+ "epoch": 1.275,
939
+ "grad_norm": 0.046354807913303375,
940
+ "kl": 0.171075239777565,
941
+ "learning_rate": 1.7500000000000002e-06,
942
+ "loss": 0.0004276880936231464,
943
+ "step": 51,
944
+ "step_time": 0.1119742769997174
945
+ },
946
+ {
947
+ "clip_ratio/high_max": 0.01171875,
948
+ "clip_ratio/high_mean": 0.01171875,
949
+ "clip_ratio/low_mean": 0.0,
950
+ "clip_ratio/low_min": 0.0,
951
+ "clip_ratio/region_mean": 0.01171875,
952
+ "entropy": 1.631166696548462,
953
+ "epoch": 1.3,
954
+ "grad_norm": 1.6717921495437622,
955
+ "kl": 0.16682168841362,
956
+ "learning_rate": 1.725e-06,
957
+ "loss": -0.09329331666231155,
958
+ "step": 52,
959
+ "step_time": 0.11258080000061454
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": 1.8790663480758667,
975
+ "epoch": 1.325,
976
+ "frac_reward_zero_std": 1.0,
977
+ "grad_norm": 0.06409338861703873,
978
+ "kl": 0.24286241829395294,
979
+ "learning_rate": 1.7e-06,
980
+ "loss": 0.0006071560783311725,
981
+ "num_tokens": 140416.0,
982
+ "reward": 0.8848026990890503,
983
+ "reward_std": 0.08498957008123398,
984
+ "rewards/AnswererRewardFunction/mean": 0.8848026990890503,
985
+ "rewards/AnswererRewardFunction/std": 0.08498956263065338,
986
+ "step": 53,
987
+ "step_time": 7.8205216610003845
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": 1.8914015293121338,
996
+ "epoch": 1.35,
997
+ "grad_norm": 0.07145795971155167,
998
+ "kl": 0.21129751205444336,
999
+ "learning_rate": 1.675e-06,
1000
+ "loss": 0.000528243777807802,
1001
+ "step": 54,
1002
+ "step_time": 0.1118343839998488
1003
+ },
1004
+ {
1005
+ "clip_ratio/high_max": 0.0,
1006
+ "clip_ratio/high_mean": 0.0,
1007
+ "clip_ratio/low_mean": 0.0,
1008
+ "clip_ratio/low_min": 0.0,
1009
+ "clip_ratio/region_mean": 0.0,
1010
+ "entropy": 1.757217288017273,
1011
+ "epoch": 1.375,
1012
+ "grad_norm": 0.038478150963783264,
1013
+ "kl": 0.14117839932441711,
1014
+ "learning_rate": 1.65e-06,
1015
+ "loss": 0.0003529459936544299,
1016
+ "step": 55,
1017
+ "step_time": 0.11172699199960334
1018
+ },
1019
+ {
1020
+ "clip_ratio/high_max": 0.0,
1021
+ "clip_ratio/high_mean": 0.0,
1022
+ "clip_ratio/low_mean": 0.0,
1023
+ "clip_ratio/low_min": 0.0,
1024
+ "clip_ratio/region_mean": 0.0,
1025
+ "entropy": 1.9119102954864502,
1026
+ "epoch": 1.4,
1027
+ "grad_norm": 0.0431695431470871,
1028
+ "kl": 0.17614459991455078,
1029
+ "learning_rate": 1.625e-06,
1030
+ "loss": 0.00044036147301085293,
1031
+ "step": 56,
1032
+ "step_time": 0.11101481700006843
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": 1.9821627140045166,
1048
+ "epoch": 1.425,
1049
+ "frac_reward_zero_std": 0.75,
1050
+ "grad_norm": 3.133037805557251,
1051
+ "kl": 0.17368242144584656,
1052
+ "learning_rate": 1.6e-06,
1053
+ "loss": 0.09411265701055527,
1054
+ "num_tokens": 150768.0,
1055
+ "reward": 0.9159495830535889,
1056
+ "reward_std": 0.06545510143041611,
1057
+ "rewards/AnswererRewardFunction/mean": 0.9159495830535889,
1058
+ "rewards/AnswererRewardFunction/std": 0.0654551088809967,
1059
+ "step": 57,
1060
+ "step_time": 7.76614765100021
1061
+ },
1062
+ {
1063
+ "clip_ratio/high_max": 0.001953125,
1064
+ "clip_ratio/high_mean": 0.001953125,
1065
+ "clip_ratio/low_mean": 0.0,
1066
+ "clip_ratio/low_min": 0.0,
1067
+ "clip_ratio/region_mean": 0.001953125,
1068
+ "entropy": 1.7750331163406372,
1069
+ "epoch": 1.45,
1070
+ "grad_norm": 0.953976571559906,
1071
+ "kl": 0.19966614246368408,
1072
+ "learning_rate": 1.5750000000000002e-06,
1073
+ "loss": -0.03101617470383644,
1074
+ "step": 58,
1075
+ "step_time": 0.1098387490001187
1076
+ },
1077
+ {
1078
+ "clip_ratio/high_max": 0.0009765625,
1079
+ "clip_ratio/high_mean": 0.0009765625,
1080
+ "clip_ratio/low_mean": 0.0,
1081
+ "clip_ratio/low_min": 0.0,
1082
+ "clip_ratio/region_mean": 0.0009765625,
1083
+ "entropy": 1.8459761142730713,
1084
+ "epoch": 1.475,
1085
+ "grad_norm": 0.8789482116699219,
1086
+ "kl": 0.1772395670413971,
1087
+ "learning_rate": 1.5500000000000002e-06,
1088
+ "loss": -0.030849514529109,
1089
+ "step": 59,
1090
+ "step_time": 0.10980367899992416
1091
+ },
1092
+ {
1093
+ "clip_ratio/high_max": 0.0029296875,
1094
+ "clip_ratio/high_mean": 0.0029296875,
1095
+ "clip_ratio/low_mean": 0.0,
1096
+ "clip_ratio/low_min": 0.0,
1097
+ "clip_ratio/region_mean": 0.0029296875,
1098
+ "entropy": 1.8890777826309204,
1099
+ "epoch": 1.5,
1100
+ "grad_norm": 0.938976526260376,
1101
+ "kl": 0.17906132340431213,
1102
+ "learning_rate": 1.525e-06,
1103
+ "loss": -0.030837498605251312,
1104
+ "step": 60,
1105
+ "step_time": 0.11071035700024368
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": 1.8429818153381348,
1121
+ "epoch": 1.525,
1122
+ "frac_reward_zero_std": 1.0,
1123
+ "grad_norm": 0.04960624501109123,
1124
+ "kl": 0.2106022834777832,
1125
+ "learning_rate": 1.5e-06,
1126
+ "loss": 0.0005265056970529258,
1127
+ "num_tokens": 160864.0,
1128
+ "reward": 0.8755199909210205,
1129
+ "reward_std": 0.10159505903720856,
1130
+ "rewards/AnswererRewardFunction/mean": 0.8755199909210205,
1131
+ "rewards/AnswererRewardFunction/std": 0.10159505903720856,
1132
+ "step": 61,
1133
+ "step_time": 7.777418299000601
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": 1.6882359981536865,
1142
+ "epoch": 1.55,
1143
+ "grad_norm": 0.06488100439310074,
1144
+ "kl": 0.18747568130493164,
1145
+ "learning_rate": 1.4749999999999999e-06,
1146
+ "loss": 0.00046868916251696646,
1147
+ "step": 62,
1148
+ "step_time": 0.1100167260001399
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": 1.9549317359924316,
1157
+ "epoch": 1.575,
1158
+ "grad_norm": 0.06376447528600693,
1159
+ "kl": 0.19842228293418884,
1160
+ "learning_rate": 1.45e-06,
1161
+ "loss": 0.0004960556980222464,
1162
+ "step": 63,
1163
+ "step_time": 0.10958483499962313
1164
+ },
1165
+ {
1166
+ "clip_ratio/high_max": 0.0,
1167
+ "clip_ratio/high_mean": 0.0,
1168
+ "clip_ratio/low_mean": 0.0,
1169
+ "clip_ratio/low_min": 0.0,
1170
+ "clip_ratio/region_mean": 0.0,
1171
+ "entropy": 1.7855771780014038,
1172
+ "epoch": 1.6,
1173
+ "grad_norm": 0.047496143728494644,
1174
+ "kl": 0.19172421097755432,
1175
+ "learning_rate": 1.425e-06,
1176
+ "loss": 0.00047931051813066006,
1177
+ "step": 64,
1178
+ "step_time": 0.11118459000044822
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": 1.8505133390426636,
1194
+ "epoch": 1.625,
1195
+ "frac_reward_zero_std": 1.0,
1196
+ "grad_norm": 0.050711821764707565,
1197
+ "kl": 0.17194393277168274,
1198
+ "learning_rate": 1.4000000000000001e-06,
1199
+ "loss": 0.0004298597923479974,
1200
+ "num_tokens": 171008.0,
1201
+ "reward": 0.819347620010376,
1202
+ "reward_std": 0.11957820504903793,
1203
+ "rewards/AnswererRewardFunction/mean": 0.819347620010376,
1204
+ "rewards/AnswererRewardFunction/std": 0.11957821249961853,
1205
+ "step": 65,
1206
+ "step_time": 7.7349191860002975
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": 1.8674921989440918,
1215
+ "epoch": 1.65,
1216
+ "grad_norm": 0.06205311790108681,
1217
+ "kl": 0.22628569602966309,
1218
+ "learning_rate": 1.375e-06,
1219
+ "loss": 0.0005657142028212547,
1220
+ "step": 66,
1221
+ "step_time": 0.11190354400059732
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": 1.7050056457519531,
1230
+ "epoch": 1.675,
1231
+ "grad_norm": 0.040369097143411636,
1232
+ "kl": 0.1742599755525589,
1233
+ "learning_rate": 1.35e-06,
1234
+ "loss": 0.0004356499412097037,
1235
+ "step": 67,
1236
+ "step_time": 0.1103675430003932
1237
+ },
1238
+ {
1239
+ "clip_ratio/high_max": 0.0,
1240
+ "clip_ratio/high_mean": 0.0,
1241
+ "clip_ratio/low_mean": 0.0,
1242
+ "clip_ratio/low_min": 0.0,
1243
+ "clip_ratio/region_mean": 0.0,
1244
+ "entropy": 1.6987855434417725,
1245
+ "epoch": 1.7,
1246
+ "grad_norm": 0.0360160693526268,
1247
+ "kl": 0.1615583896636963,
1248
+ "learning_rate": 1.325e-06,
1249
+ "loss": 0.00040389594505541027,
1250
+ "step": 68,
1251
+ "step_time": 0.10992557299960026
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": 1.701324701309204,
1267
+ "epoch": 1.725,
1268
+ "frac_reward_zero_std": 1.0,
1269
+ "grad_norm": 0.04009808227419853,
1270
+ "kl": 0.18351677060127258,
1271
+ "learning_rate": 1.3e-06,
1272
+ "loss": 0.0004587919102050364,
1273
+ "num_tokens": 181088.0,
1274
+ "reward": 0.8702797293663025,
1275
+ "reward_std": 0.11096905916929245,
1276
+ "rewards/AnswererRewardFunction/mean": 0.8702797293663025,
1277
+ "rewards/AnswererRewardFunction/std": 0.11096906661987305,
1278
+ "step": 69,
1279
+ "step_time": 7.762835034000091
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": 1.7760753631591797,
1288
+ "epoch": 1.75,
1289
+ "grad_norm": 0.0532107912003994,
1290
+ "kl": 0.19955360889434814,
1291
+ "learning_rate": 1.275e-06,
1292
+ "loss": 0.0004988840082660317,
1293
+ "step": 70,
1294
+ "step_time": 0.11100815299960232
1295
+ },
1296
+ {
1297
+ "clip_ratio/high_max": 0.0,
1298
+ "clip_ratio/high_mean": 0.0,
1299
+ "clip_ratio/low_mean": 0.0,
1300
+ "clip_ratio/low_min": 0.0,
1301
+ "clip_ratio/region_mean": 0.0,
1302
+ "entropy": 1.8080956935882568,
1303
+ "epoch": 1.775,
1304
+ "grad_norm": 0.04277520254254341,
1305
+ "kl": 0.18510785698890686,
1306
+ "learning_rate": 1.25e-06,
1307
+ "loss": 0.00046276964712888,
1308
+ "step": 71,
1309
+ "step_time": 0.1123151739993773
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": 1.7996044158935547,
1318
+ "epoch": 1.8,
1319
+ "grad_norm": 0.05546512082219124,
1320
+ "kl": 0.18127581477165222,
1321
+ "learning_rate": 1.225e-06,
1322
+ "loss": 0.00045318951015360653,
1323
+ "step": 72,
1324
+ "step_time": 0.11235005500020634
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": 1.673140048980713,
1340
+ "epoch": 1.825,
1341
+ "frac_reward_zero_std": 1.0,
1342
+ "grad_norm": 0.05601564049720764,
1343
+ "kl": 0.17777645587921143,
1344
+ "learning_rate": 1.2000000000000002e-06,
1345
+ "loss": 0.0004444411606527865,
1346
+ "num_tokens": 191124.0,
1347
+ "reward": 0.9533497095108032,
1348
+ "reward_std": 0.02976372465491295,
1349
+ "rewards/AnswererRewardFunction/mean": 0.9533497095108032,
1350
+ "rewards/AnswererRewardFunction/std": 0.029763715341687202,
1351
+ "step": 73,
1352
+ "step_time": 7.778424793999875
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": 1.648258924484253,
1361
+ "epoch": 1.85,
1362
+ "grad_norm": 0.04714483022689819,
1363
+ "kl": 0.18713414669036865,
1364
+ "learning_rate": 1.175e-06,
1365
+ "loss": 0.0004678354016505182,
1366
+ "step": 74,
1367
+ "step_time": 0.11056567400009953
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": 1.5942027568817139,
1376
+ "epoch": 1.875,
1377
+ "grad_norm": 0.07229740172624588,
1378
+ "kl": 0.17846925556659698,
1379
+ "learning_rate": 1.1500000000000002e-06,
1380
+ "loss": 0.00044617310049943626,
1381
+ "step": 75,
1382
+ "step_time": 0.10988077099955262
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": 1.8294134140014648,
1391
+ "epoch": 1.9,
1392
+ "grad_norm": 0.07372913509607315,
1393
+ "kl": 0.19231390953063965,
1394
+ "learning_rate": 1.125e-06,
1395
+ "loss": 0.0004807847726624459,
1396
+ "step": 76,
1397
+ "step_time": 0.11050818300009269
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": 1.8716906309127808,
1413
+ "epoch": 1.925,
1414
+ "frac_reward_zero_std": 1.0,
1415
+ "grad_norm": 0.04397920146584511,
1416
+ "kl": 0.20210802555084229,
1417
+ "learning_rate": 1.0999999999999998e-06,
1418
+ "loss": 0.0005052700289525092,
1419
+ "num_tokens": 201192.0,
1420
+ "reward": 0.8848026990890503,
1421
+ "reward_std": 0.08498957008123398,
1422
+ "rewards/AnswererRewardFunction/mean": 0.8848026990890503,
1423
+ "rewards/AnswererRewardFunction/std": 0.08498956263065338,
1424
+ "step": 77,
1425
+ "step_time": 7.766447917999358
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": 1.8421611785888672,
1434
+ "epoch": 1.95,
1435
+ "grad_norm": 0.04258396103978157,
1436
+ "kl": 0.16801506280899048,
1437
+ "learning_rate": 1.075e-06,
1438
+ "loss": 0.0004200376570224762,
1439
+ "step": 78,
1440
+ "step_time": 0.11128133199963486
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": 2.0212643146514893,
1449
+ "epoch": 1.975,
1450
+ "grad_norm": 0.037206318229436874,
1451
+ "kl": 0.17133170366287231,
1452
+ "learning_rate": 1.05e-06,
1453
+ "loss": 0.00042832925100810826,
1454
+ "step": 79,
1455
+ "step_time": 0.11063859800015052
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": 1.9713196754455566,
1464
+ "epoch": 2.0,
1465
+ "grad_norm": 0.06273569911718369,
1466
+ "kl": 0.20357024669647217,
1467
+ "learning_rate": 1.0250000000000001e-06,
1468
+ "loss": 0.0005089255864731967,
1469
+ "step": 80,
1470
+ "step_time": 0.11058984600003896
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": 1.7217357158660889,
1486
+ "epoch": 2.025,
1487
+ "frac_reward_zero_std": 0.75,
1488
+ "grad_norm": 1.6169368028640747,
1489
+ "kl": 0.16916689276695251,
1490
+ "learning_rate": 1e-06,
1491
+ "loss": 0.054513730108737946,
1492
+ "num_tokens": 211296.0,
1493
+ "reward": 0.8427923917770386,
1494
+ "reward_std": 0.11980737000703812,
1495
+ "rewards/AnswererRewardFunction/mean": 0.8427923917770386,
1496
+ "rewards/AnswererRewardFunction/std": 0.11980737000703812,
1497
+ "step": 81,
1498
+ "step_time": 7.74370976299997
1499
+ },
1500
+ {
1501
+ "clip_ratio/high_max": 0.0,
1502
+ "clip_ratio/high_mean": 0.0,
1503
+ "clip_ratio/low_mean": 0.0009765625,
1504
+ "clip_ratio/low_min": 0.0009765625,
1505
+ "clip_ratio/region_mean": 0.0009765625,
1506
+ "entropy": 1.7494382858276367,
1507
+ "epoch": 2.05,
1508
+ "grad_norm": 1.5589256286621094,
1509
+ "kl": 0.18270084261894226,
1510
+ "learning_rate": 9.75e-07,
1511
+ "loss": 0.05478590354323387,
1512
+ "step": 82,
1513
+ "step_time": 0.11115284799961955
1514
+ },
1515
+ {
1516
+ "clip_ratio/high_max": 0.001953125,
1517
+ "clip_ratio/high_mean": 0.001953125,
1518
+ "clip_ratio/low_mean": 0.0,
1519
+ "clip_ratio/low_min": 0.0,
1520
+ "clip_ratio/region_mean": 0.001953125,
1521
+ "entropy": 1.879669427871704,
1522
+ "epoch": 2.075,
1523
+ "grad_norm": 1.6640474796295166,
1524
+ "kl": 0.17780490219593048,
1525
+ "learning_rate": 9.5e-07,
1526
+ "loss": -0.053716279566287994,
1527
+ "step": 83,
1528
+ "step_time": 0.1114861039995958
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": 2.063556432723999,
1537
+ "epoch": 2.1,
1538
+ "grad_norm": 1.6819794178009033,
1539
+ "kl": 0.2023072987794876,
1540
+ "learning_rate": 9.25e-07,
1541
+ "loss": -0.05334728956222534,
1542
+ "step": 84,
1543
+ "step_time": 0.11098489600044559
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": 1.6534061431884766,
1559
+ "epoch": 2.125,
1560
+ "frac_reward_zero_std": 1.0,
1561
+ "grad_norm": 0.04503380134701729,
1562
+ "kl": 0.15787094831466675,
1563
+ "learning_rate": 9e-07,
1564
+ "loss": 0.00039467739406973124,
1565
+ "num_tokens": 221124.0,
1566
+ "reward": 0.8965563178062439,
1567
+ "reward_std": 0.11728676408529282,
1568
+ "rewards/AnswererRewardFunction/mean": 0.8965563178062439,
1569
+ "rewards/AnswererRewardFunction/std": 0.11728675663471222,
1570
+ "step": 85,
1571
+ "step_time": 7.803010329999779
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": 1.734236240386963,
1580
+ "epoch": 2.15,
1581
+ "grad_norm": 0.053943030536174774,
1582
+ "kl": 0.1871510148048401,
1583
+ "learning_rate": 8.750000000000001e-07,
1584
+ "loss": 0.00046787751489318907,
1585
+ "step": 86,
1586
+ "step_time": 0.11089296099999046
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": 1.8206863403320312,
1595
+ "epoch": 2.175,
1596
+ "grad_norm": 0.0429137758910656,
1597
+ "kl": 0.18653716146945953,
1598
+ "learning_rate": 8.5e-07,
1599
+ "loss": 0.0004663429281208664,
1600
+ "step": 87,
1601
+ "step_time": 0.109580606000236
1602
+ },
1603
+ {
1604
+ "clip_ratio/high_max": 0.0,
1605
+ "clip_ratio/high_mean": 0.0,
1606
+ "clip_ratio/low_mean": 0.0,
1607
+ "clip_ratio/low_min": 0.0,
1608
+ "clip_ratio/region_mean": 0.0,
1609
+ "entropy": 1.8033851385116577,
1610
+ "epoch": 2.2,
1611
+ "grad_norm": 0.032541193068027496,
1612
+ "kl": 0.1465151309967041,
1613
+ "learning_rate": 8.25e-07,
1614
+ "loss": 0.0003662878298200667,
1615
+ "step": 88,
1616
+ "step_time": 0.10985735099984595
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": 1.7311502695083618,
1632
+ "epoch": 2.225,
1633
+ "frac_reward_zero_std": 1.0,
1634
+ "grad_norm": 0.04454684257507324,
1635
+ "kl": 0.18079979717731476,
1636
+ "learning_rate": 8e-07,
1637
+ "loss": 0.00045199948363006115,
1638
+ "num_tokens": 230912.0,
1639
+ "reward": 0.8309686183929443,
1640
+ "reward_std": 0.14328104257583618,
1641
+ "rewards/AnswererRewardFunction/mean": 0.8309686183929443,
1642
+ "rewards/AnswererRewardFunction/std": 0.14328105747699738,
1643
+ "step": 89,
1644
+ "step_time": 7.7929221280001
1645
+ },
1646
+ {
1647
+ "clip_ratio/high_max": 0.0,
1648
+ "clip_ratio/high_mean": 0.0,
1649
+ "clip_ratio/low_mean": 0.0,
1650
+ "clip_ratio/low_min": 0.0,
1651
+ "clip_ratio/region_mean": 0.0,
1652
+ "entropy": 2.0398142337799072,
1653
+ "epoch": 2.25,
1654
+ "grad_norm": 0.04331798851490021,
1655
+ "kl": 0.19130492210388184,
1656
+ "learning_rate": 7.750000000000001e-07,
1657
+ "loss": 0.0004782622854690999,
1658
+ "step": 90,
1659
+ "step_time": 0.1110038219994749
1660
+ },
1661
+ {
1662
+ "clip_ratio/high_max": 0.0,
1663
+ "clip_ratio/high_mean": 0.0,
1664
+ "clip_ratio/low_mean": 0.0,
1665
+ "clip_ratio/low_min": 0.0,
1666
+ "clip_ratio/region_mean": 0.0,
1667
+ "entropy": 1.8197107315063477,
1668
+ "epoch": 2.275,
1669
+ "grad_norm": 0.0389009565114975,
1670
+ "kl": 0.1594439446926117,
1671
+ "learning_rate": 7.5e-07,
1672
+ "loss": 0.00039860987453721464,
1673
+ "step": 91,
1674
+ "step_time": 0.11234117200001492
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": 1.6061317920684814,
1683
+ "epoch": 2.3,
1684
+ "grad_norm": 0.035475604236125946,
1685
+ "kl": 0.16071225702762604,
1686
+ "learning_rate": 7.25e-07,
1687
+ "loss": 0.0004017806495539844,
1688
+ "step": 92,
1689
+ "step_time": 0.11047902500013151
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": 2.0506911277770996,
1705
+ "epoch": 2.325,
1706
+ "frac_reward_zero_std": 1.0,
1707
+ "grad_norm": 0.05084988474845886,
1708
+ "kl": 0.19656753540039062,
1709
+ "learning_rate": 7.000000000000001e-07,
1710
+ "loss": 0.0004914188175462186,
1711
+ "num_tokens": 241172.0,
1712
+ "reward": 0.8668582439422607,
1713
+ "reward_std": 0.1170896589756012,
1714
+ "rewards/AnswererRewardFunction/mean": 0.8668582439422607,
1715
+ "rewards/AnswererRewardFunction/std": 0.1170896664261818,
1716
+ "step": 93,
1717
+ "step_time": 7.777627965000647
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": 1.7454960346221924,
1726
+ "epoch": 2.35,
1727
+ "grad_norm": 0.0445859469473362,
1728
+ "kl": 0.18581417202949524,
1729
+ "learning_rate": 6.75e-07,
1730
+ "loss": 0.0004645354056265205,
1731
+ "step": 94,
1732
+ "step_time": 0.11115384400000039
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": 1.9964624643325806,
1741
+ "epoch": 2.375,
1742
+ "grad_norm": 0.03978120535612106,
1743
+ "kl": 0.1595752239227295,
1744
+ "learning_rate": 6.5e-07,
1745
+ "loss": 0.00039893804932944477,
1746
+ "step": 95,
1747
+ "step_time": 0.11099709100017208
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": 1.9201608896255493,
1756
+ "epoch": 2.4,
1757
+ "grad_norm": 0.04184085130691528,
1758
+ "kl": 0.16675138473510742,
1759
+ "learning_rate": 6.25e-07,
1760
+ "loss": 0.00041687843622639775,
1761
+ "step": 96,
1762
+ "step_time": 0.11661479099984717
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": 1.7013036012649536,
1778
+ "epoch": 2.425,
1779
+ "frac_reward_zero_std": 1.0,
1780
+ "grad_norm": 0.037217915058135986,
1781
+ "kl": 0.15892338752746582,
1782
+ "learning_rate": 6.000000000000001e-07,
1783
+ "loss": 0.0003973084385506809,
1784
+ "num_tokens": 251112.0,
1785
+ "reward": 0.8227691054344177,
1786
+ "reward_std": 0.11510815471410751,
1787
+ "rewards/AnswererRewardFunction/mean": 0.8227691054344177,
1788
+ "rewards/AnswererRewardFunction/std": 0.11510813981294632,
1789
+ "step": 97,
1790
+ "step_time": 7.6884349259999
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": 1.912070393562317,
1799
+ "epoch": 2.45,
1800
+ "grad_norm": 0.07283549755811691,
1801
+ "kl": 0.1705121099948883,
1802
+ "learning_rate": 5.750000000000001e-07,
1803
+ "loss": 0.00042628025403246284,
1804
+ "step": 98,
1805
+ "step_time": 0.11253656700046122
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": 1.7756853103637695,
1814
+ "epoch": 2.475,
1815
+ "grad_norm": 0.05578094348311424,
1816
+ "kl": 0.1911633312702179,
1817
+ "learning_rate": 5.499999999999999e-07,
1818
+ "loss": 0.00047790829557925463,
1819
+ "step": 99,
1820
+ "step_time": 0.11162257700016198
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": 1.8756659030914307,
1829
+ "epoch": 2.5,
1830
+ "grad_norm": 0.05396004766225815,
1831
+ "kl": 0.18036732077598572,
1832
+ "learning_rate": 5.25e-07,
1833
+ "loss": 0.0004509182763285935,
1834
+ "step": 100,
1835
+ "step_time": 0.11048615199979395
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": 1.766379952430725,
1851
+ "epoch": 2.525,
1852
+ "frac_reward_zero_std": 0.75,
1853
+ "grad_norm": 0.8942388296127319,
1854
+ "kl": 0.16869878768920898,
1855
+ "learning_rate": 5e-07,
1856
+ "loss": -0.03075210005044937,
1857
+ "num_tokens": 261108.0,
1858
+ "reward": 0.8209283351898193,
1859
+ "reward_std": 0.10294758528470993,
1860
+ "rewards/AnswererRewardFunction/mean": 0.8209283351898193,
1861
+ "rewards/AnswererRewardFunction/std": 0.10294758528470993,
1862
+ "step": 101,
1863
+ "step_time": 7.760543512999902
1864
+ },
1865
+ {
1866
+ "clip_ratio/high_max": 0.0009765625,
1867
+ "clip_ratio/high_mean": 0.0009765625,
1868
+ "clip_ratio/low_mean": 0.0,
1869
+ "clip_ratio/low_min": 0.0,
1870
+ "clip_ratio/region_mean": 0.0009765625,
1871
+ "entropy": 1.7685692310333252,
1872
+ "epoch": 2.55,
1873
+ "grad_norm": 0.9231233596801758,
1874
+ "kl": 0.14601315557956696,
1875
+ "learning_rate": 4.75e-07,
1876
+ "loss": -0.030733846127986908,
1877
+ "step": 102,
1878
+ "step_time": 0.11171397100042668
1879
+ },
1880
+ {
1881
+ "clip_ratio/high_max": 0.0,
1882
+ "clip_ratio/high_mean": 0.0,
1883
+ "clip_ratio/low_mean": 0.0,
1884
+ "clip_ratio/low_min": 0.0,
1885
+ "clip_ratio/region_mean": 0.0,
1886
+ "entropy": 1.8422847986221313,
1887
+ "epoch": 2.575,
1888
+ "grad_norm": 2.5536715984344482,
1889
+ "kl": 0.1598617136478424,
1890
+ "learning_rate": 4.5e-07,
1891
+ "loss": 0.09431644529104233,
1892
+ "step": 103,
1893
+ "step_time": 0.11270964300001651
1894
+ },
1895
+ {
1896
+ "clip_ratio/high_max": 0.001953125,
1897
+ "clip_ratio/high_mean": 0.001953125,
1898
+ "clip_ratio/low_mean": 0.0,
1899
+ "clip_ratio/low_min": 0.0,
1900
+ "clip_ratio/region_mean": 0.001953125,
1901
+ "entropy": 1.8202838897705078,
1902
+ "epoch": 2.6,
1903
+ "grad_norm": 0.9551231861114502,
1904
+ "kl": 0.15085619688034058,
1905
+ "learning_rate": 4.25e-07,
1906
+ "loss": -0.031034285202622414,
1907
+ "step": 104,
1908
+ "step_time": 0.11204244799955632
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": 1.7124009132385254,
1924
+ "epoch": 2.625,
1925
+ "frac_reward_zero_std": 1.0,
1926
+ "grad_norm": 0.038327667862176895,
1927
+ "kl": 0.16573426127433777,
1928
+ "learning_rate": 4e-07,
1929
+ "loss": 0.00041433563455939293,
1930
+ "num_tokens": 271148.0,
1931
+ "reward": 0.8883568644523621,
1932
+ "reward_std": 0.08731486648321152,
1933
+ "rewards/AnswererRewardFunction/mean": 0.8883568644523621,
1934
+ "rewards/AnswererRewardFunction/std": 0.08731484413146973,
1935
+ "step": 105,
1936
+ "step_time": 7.726755570999558
1937
+ },
1938
+ {
1939
+ "clip_ratio/high_max": 0.0,
1940
+ "clip_ratio/high_mean": 0.0,
1941
+ "clip_ratio/low_mean": 0.0,
1942
+ "clip_ratio/low_min": 0.0,
1943
+ "clip_ratio/region_mean": 0.0,
1944
+ "entropy": 1.7165004014968872,
1945
+ "epoch": 2.65,
1946
+ "grad_norm": 0.042729344218969345,
1947
+ "kl": 0.1634857952594757,
1948
+ "learning_rate": 3.75e-07,
1949
+ "loss": 0.0004087144916411489,
1950
+ "step": 106,
1951
+ "step_time": 0.11138214499987953
1952
+ },
1953
+ {
1954
+ "clip_ratio/high_max": 0.0,
1955
+ "clip_ratio/high_mean": 0.0,
1956
+ "clip_ratio/low_mean": 0.0,
1957
+ "clip_ratio/low_min": 0.0,
1958
+ "clip_ratio/region_mean": 0.0,
1959
+ "entropy": 1.5836355686187744,
1960
+ "epoch": 2.675,
1961
+ "grad_norm": 0.03519313782453537,
1962
+ "kl": 0.1412312090396881,
1963
+ "learning_rate": 3.5000000000000004e-07,
1964
+ "loss": 0.00035307800862938166,
1965
+ "step": 107,
1966
+ "step_time": 0.11289172600027086
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": 1.856842279434204,
1975
+ "epoch": 2.7,
1976
+ "grad_norm": 0.037977833300828934,
1977
+ "kl": 0.15611428022384644,
1978
+ "learning_rate": 3.25e-07,
1979
+ "loss": 0.00039028568426147103,
1980
+ "step": 108,
1981
+ "step_time": 0.11108121899997059
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": 1.7743322849273682,
1997
+ "epoch": 2.725,
1998
+ "frac_reward_zero_std": 1.0,
1999
+ "grad_norm": 0.038768816739320755,
2000
+ "kl": 0.16674529016017914,
2001
+ "learning_rate": 3.0000000000000004e-07,
2002
+ "loss": 0.0004168632149230689,
2003
+ "num_tokens": 281344.0,
2004
+ "reward": 0.9358674883842468,
2005
+ "reward_std": 0.0063578663393855095,
2006
+ "rewards/AnswererRewardFunction/mean": 0.9358674883842468,
2007
+ "rewards/AnswererRewardFunction/std": 0.006357857491821051,
2008
+ "step": 109,
2009
+ "step_time": 7.73822644899974
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": 1.5768123865127563,
2018
+ "epoch": 2.75,
2019
+ "grad_norm": 0.03679906204342842,
2020
+ "kl": 0.1504622995853424,
2021
+ "learning_rate": 2.7499999999999996e-07,
2022
+ "loss": 0.0003761557163670659,
2023
+ "step": 110,
2024
+ "step_time": 0.11385159400015255
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": 1.6607877016067505,
2033
+ "epoch": 2.775,
2034
+ "grad_norm": 0.03713918849825859,
2035
+ "kl": 0.16259214282035828,
2036
+ "learning_rate": 2.5e-07,
2037
+ "loss": 0.0004064803651999682,
2038
+ "step": 111,
2039
+ "step_time": 0.1118541819996608
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": 1.8108391761779785,
2048
+ "epoch": 2.8,
2049
+ "grad_norm": 0.05037199705839157,
2050
+ "kl": 0.1863050013780594,
2051
+ "learning_rate": 2.25e-07,
2052
+ "loss": 0.0004657625104300678,
2053
+ "step": 112,
2054
+ "step_time": 0.11186235300010594
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": 1.7566189765930176,
2070
+ "epoch": 2.825,
2071
+ "frac_reward_zero_std": 1.0,
2072
+ "grad_norm": 0.036518849432468414,
2073
+ "kl": 0.1620350182056427,
2074
+ "learning_rate": 2e-07,
2075
+ "loss": 0.00040508751408196986,
2076
+ "num_tokens": 291560.0,
2077
+ "reward": 0.8848026990890503,
2078
+ "reward_std": 0.08498957008123398,
2079
+ "rewards/AnswererRewardFunction/mean": 0.8848026990890503,
2080
+ "rewards/AnswererRewardFunction/std": 0.08498956263065338,
2081
+ "step": 113,
2082
+ "step_time": 7.7582261870002185
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": 1.7600432634353638,
2091
+ "epoch": 2.85,
2092
+ "grad_norm": 0.05634801462292671,
2093
+ "kl": 0.17256814241409302,
2094
+ "learning_rate": 1.7500000000000002e-07,
2095
+ "loss": 0.00043142036884091794,
2096
+ "step": 114,
2097
+ "step_time": 0.11197551500026748
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": 1.874969720840454,
2106
+ "epoch": 2.875,
2107
+ "grad_norm": 0.03801681101322174,
2108
+ "kl": 0.15792644023895264,
2109
+ "learning_rate": 1.5000000000000002e-07,
2110
+ "loss": 0.00039481610292568803,
2111
+ "step": 115,
2112
+ "step_time": 0.11017055599950254
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": 1.9454420804977417,
2121
+ "epoch": 2.9,
2122
+ "grad_norm": 0.03727322444319725,
2123
+ "kl": 0.1802506446838379,
2124
+ "learning_rate": 1.25e-07,
2125
+ "loss": 0.0004506265977397561,
2126
+ "step": 116,
2127
+ "step_time": 0.1150881510002364
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": 1.8203485012054443,
2143
+ "epoch": 2.925,
2144
+ "frac_reward_zero_std": 1.0,
2145
+ "grad_norm": 0.044991906732320786,
2146
+ "kl": 0.17885687947273254,
2147
+ "learning_rate": 1e-07,
2148
+ "loss": 0.00044714222894981503,
2149
+ "num_tokens": 301788.0,
2150
+ "reward": 0.9323133230209351,
2151
+ "reward_std": 0.0,
2152
+ "rewards/AnswererRewardFunction/mean": 0.9323133230209351,
2153
+ "rewards/AnswererRewardFunction/std": 0.0,
2154
+ "step": 117,
2155
+ "step_time": 7.785143383000104
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": 2.0509285926818848,
2164
+ "epoch": 2.95,
2165
+ "grad_norm": 0.05878576263785362,
2166
+ "kl": 0.1711733639240265,
2167
+ "learning_rate": 7.500000000000001e-08,
2168
+ "loss": 0.0004279334098100662,
2169
+ "step": 118,
2170
+ "step_time": 0.11581213099998422
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": 1.8962795734405518,
2179
+ "epoch": 2.975,
2180
+ "grad_norm": 0.07008994370698929,
2181
+ "kl": 0.17706063389778137,
2182
+ "learning_rate": 5e-08,
2183
+ "loss": 0.00044265156611800194,
2184
+ "step": 119,
2185
+ "step_time": 0.11132281099980901
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": 2.0024328231811523,
2194
+ "epoch": 3.0,
2195
+ "grad_norm": 0.06333177536725998,
2196
+ "kl": 0.18442605435848236,
2197
+ "learning_rate": 2.5e-08,
2198
+ "loss": 0.0004610651230905205,
2199
+ "step": 120,
2200
+ "step_time": 0.11119805700036522
2201
+ }
2202
+ ],
2203
+ "logging_steps": 1,
2204
+ "max_steps": 120,
2205
+ "num_input_tokens_seen": 301788,
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_004/answerer_train/checkpoint-120/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2d1ca30827c82daaee74796a91c1d74f1cbf0768a07df4d6cf7d927229c515d1
3
+ size 7249