add checkpoint-50
Browse files
checkpoint-50/adapter_config.json
ADDED
|
@@ -0,0 +1,46 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"alora_invocation_tokens": null,
|
| 3 |
+
"alpha_pattern": {},
|
| 4 |
+
"arrow_config": null,
|
| 5 |
+
"auto_mapping": null,
|
| 6 |
+
"base_model_name_or_path": "allenai/Olmo-3.1-32B-Instruct-SFT",
|
| 7 |
+
"bias": "none",
|
| 8 |
+
"corda_config": null,
|
| 9 |
+
"ensure_weight_tying": false,
|
| 10 |
+
"eva_config": null,
|
| 11 |
+
"exclude_modules": null,
|
| 12 |
+
"fan_in_fan_out": false,
|
| 13 |
+
"inference_mode": true,
|
| 14 |
+
"init_lora_weights": true,
|
| 15 |
+
"layer_replication": null,
|
| 16 |
+
"layers_pattern": null,
|
| 17 |
+
"layers_to_transform": null,
|
| 18 |
+
"loftq_config": {},
|
| 19 |
+
"lora_alpha": 32,
|
| 20 |
+
"lora_bias": false,
|
| 21 |
+
"lora_dropout": 0.0,
|
| 22 |
+
"megatron_config": null,
|
| 23 |
+
"megatron_core": "megatron.core",
|
| 24 |
+
"modules_to_save": null,
|
| 25 |
+
"peft_type": "LORA",
|
| 26 |
+
"peft_version": "0.18.1",
|
| 27 |
+
"qalora_group_size": 16,
|
| 28 |
+
"r": 32,
|
| 29 |
+
"rank_pattern": {},
|
| 30 |
+
"revision": null,
|
| 31 |
+
"target_modules": [
|
| 32 |
+
"down_proj",
|
| 33 |
+
"k_proj",
|
| 34 |
+
"v_proj",
|
| 35 |
+
"gate_proj",
|
| 36 |
+
"o_proj",
|
| 37 |
+
"q_proj",
|
| 38 |
+
"up_proj"
|
| 39 |
+
],
|
| 40 |
+
"target_parameters": null,
|
| 41 |
+
"task_type": "CAUSAL_LM",
|
| 42 |
+
"trainable_token_indices": null,
|
| 43 |
+
"use_dora": false,
|
| 44 |
+
"use_qalora": false,
|
| 45 |
+
"use_rslora": false
|
| 46 |
+
}
|
checkpoint-50/adapter_model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e94f20e45e7874ae72b08d3253a3d48a7b95cec52edd33762747bea066f62f0c
|
| 3 |
+
size 536992880
|
checkpoint-50/chat_template.jinja
ADDED
|
@@ -0,0 +1,16 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{%- set has_system = messages|selectattr('role', 'equalto', 'system')|list|length > 0 -%}{%- if not has_system -%}{{- '<|im_start|>system
|
| 2 |
+
You are a helpful function-calling AI assistant. ' -}}{%- if tools is none or (tools | length) == 0 -%}{{- 'You do not currently have access to any functions. <functions></functions><|im_end|>
|
| 3 |
+
' -}}{%- else -%}{{- 'You are provided with function signatures within <functions></functions> XML tags. You may call one or more functions to assist with the user query. Output any function calls within <function_calls></function_calls> XML tags. Do not make assumptions about what values to plug into functions.' -}}{{- '<functions>' -}}{{- tools | tojson -}}{{- '</functions><|im_end|>
|
| 4 |
+
' -}}{%- endif -%}{%- endif -%}{%- for message in messages -%}{%- if message['role'] == 'system' -%}{{- '<|im_start|>system
|
| 5 |
+
' + message['content'] -}}{%- if tools is not none -%}{{- '<functions>' -}}{{- tools | tojson -}}{{- '</functions>' -}}{%- elif message.get('functions', none) is not none -%}{{- ' <functions>' + message['functions'] + '</functions>' -}}{%- endif -%}{{- '<|im_end|>
|
| 6 |
+
' -}}{%- elif message['role'] == 'user' -%}{{- '<|im_start|>user
|
| 7 |
+
' + message['content'] + '<|im_end|>
|
| 8 |
+
' -}}{%- elif message['role'] == 'assistant' -%}{{- '<|im_start|>assistant
|
| 9 |
+
' -}}{%- if message.get('content', none) is not none -%}{{- message['content'] -}}{%- endif -%}{%- if message.get('function_calls', none) is not none -%}{{- '<function_calls>' + message['function_calls'] + '</function_calls>' -}}{% elif message.get('tool_calls', none) is not none %}{{- '<function_calls>' -}}{%- for tool_call in message['tool_calls'] %}{%- if tool_call is mapping and tool_call.get('function', none) is not none %}{%- set args = tool_call['function']['arguments'] -%}{%- set ns = namespace(arguments_list=[]) -%}{%- for key, value in args.items() -%}{%- set ns.arguments_list = ns.arguments_list + [key ~ '=' ~ (value | tojson)] -%}{%- endfor -%}{%- set arguments = ns.arguments_list | join(', ') -%}{{- tool_call['function']['name'] + '(' + arguments + ')' -}}{%- if not loop.last -%}{{ '
|
| 10 |
+
' }}{%- endif -%}{% else %}{{- tool_call -}}{%- endif %}{%- endfor %}{{- '</function_calls>' -}}{%- endif -%}{%- if not loop.last -%}{{- '<|im_end|>' + '
|
| 11 |
+
' -}}{%- else -%}{{- eos_token -}}{%- endif -%}{%- elif message['role'] == 'environment' -%}{{- '<|im_start|>environment
|
| 12 |
+
' + message['content'] + '<|im_end|>
|
| 13 |
+
' -}}{%- elif message['role'] == 'tool' -%}{{- '<|im_start|>environment
|
| 14 |
+
' + message['content'] + '<|im_end|>
|
| 15 |
+
' -}}{%- endif -%}{%- if loop.last and add_generation_prompt -%}{{- '<|im_start|>assistant
|
| 16 |
+
' -}}{%- endif -%}{%- endfor -%}
|
checkpoint-50/tokenizer.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
checkpoint-50/tokenizer_config.json
ADDED
|
@@ -0,0 +1,14 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"add_prefix_space": false,
|
| 3 |
+
"backend": "tokenizers",
|
| 4 |
+
"bos_token": "<|endoftext|>",
|
| 5 |
+
"clean_up_tokenization_spaces": false,
|
| 6 |
+
"eos_token": "<|endoftext|>",
|
| 7 |
+
"is_local": false,
|
| 8 |
+
"model_max_length": 65536,
|
| 9 |
+
"pad_token": "<|pad|>",
|
| 10 |
+
"padding_side": "left",
|
| 11 |
+
"tokenizer_class": "TokenizersBackend",
|
| 12 |
+
"truncation_side": "left",
|
| 13 |
+
"unk_token": "<|endoftext|>"
|
| 14 |
+
}
|
checkpoint-50/trainer_state.json
ADDED
|
@@ -0,0 +1,2234 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_global_step": null,
|
| 3 |
+
"best_metric": null,
|
| 4 |
+
"best_model_checkpoint": null,
|
| 5 |
+
"epoch": 0.12315270935960591,
|
| 6 |
+
"eval_steps": 500,
|
| 7 |
+
"global_step": 50,
|
| 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": 0.078125,
|
| 19 |
+
"completions/max_length": 4096.0,
|
| 20 |
+
"completions/max_terminated_length": 4085.0,
|
| 21 |
+
"completions/mean_length": 1251.234375,
|
| 22 |
+
"completions/mean_terminated_length": 1139.6949462890625,
|
| 23 |
+
"completions/min_length": 33.0,
|
| 24 |
+
"completions/min_terminated_length": 33.0,
|
| 25 |
+
"degenerate_groups_filtered": 0.0,
|
| 26 |
+
"entropy": 0.7702056467533112,
|
| 27 |
+
"entropy/code": 0.6311403214931488,
|
| 28 |
+
"entropy/thinking": 1.2422278225421906,
|
| 29 |
+
"epoch": 0.0024630541871921183,
|
| 30 |
+
"frac_reward_zero_std": 0.0,
|
| 31 |
+
"grad_norm": 0.013573926092613244,
|
| 32 |
+
"kl": 0.0,
|
| 33 |
+
"kl/code_mean": 0.0,
|
| 34 |
+
"kl/code_sum": 0.0,
|
| 35 |
+
"kl/thinking_mean": 0.0,
|
| 36 |
+
"kl/thinking_sum": 0.0,
|
| 37 |
+
"kl/total_loss_contribution": 0.0,
|
| 38 |
+
"kl_mask/effective_frac": 1.0,
|
| 39 |
+
"kl_mask/fallback_rate": 0.375,
|
| 40 |
+
"learning_rate": 0.0,
|
| 41 |
+
"loss": -0.06258917599916458,
|
| 42 |
+
"num_tokens": 165375.0,
|
| 43 |
+
"reward": 0.8046875,
|
| 44 |
+
"reward_std": 0.8327006697654724,
|
| 45 |
+
"rewards/reward_func/mean": 0.08940972222222222,
|
| 46 |
+
"rewards/reward_func/std": 0.14380671580632529,
|
| 47 |
+
"sampling/importance_sampling_ratio/max": 2.998521327972412,
|
| 48 |
+
"sampling/importance_sampling_ratio/mean": 0.9399385452270508,
|
| 49 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 50 |
+
"sampling/sampling_logp_difference/max": 14.430514335632324,
|
| 51 |
+
"sampling/sampling_logp_difference/mean": 0.22636429965496063,
|
| 52 |
+
"seg/thinking_frac": 0.23930630087852478,
|
| 53 |
+
"step": 1,
|
| 54 |
+
"step_time": 193.109430239012
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"clip_ratio/high_max": 0.0,
|
| 58 |
+
"clip_ratio/high_mean": 0.0,
|
| 59 |
+
"clip_ratio/low_mean": 0.0,
|
| 60 |
+
"clip_ratio/low_min": 0.0,
|
| 61 |
+
"clip_ratio/region_mean": 0.0,
|
| 62 |
+
"completions/clipped_ratio": 0.03125,
|
| 63 |
+
"completions/max_length": 4096.0,
|
| 64 |
+
"completions/max_terminated_length": 3068.0,
|
| 65 |
+
"completions/mean_length": 709.1875,
|
| 66 |
+
"completions/mean_terminated_length": 643.51611328125,
|
| 67 |
+
"completions/min_length": 22.0,
|
| 68 |
+
"completions/min_terminated_length": 22.0,
|
| 69 |
+
"degenerate_groups_filtered": 0.0,
|
| 70 |
+
"entropy": 0.7780245691537857,
|
| 71 |
+
"entropy/code": 0.5997304469347,
|
| 72 |
+
"entropy/thinking": 1.2778899669647217,
|
| 73 |
+
"epoch": 0.0049261083743842365,
|
| 74 |
+
"frac_reward_zero_std": 0.0,
|
| 75 |
+
"grad_norm": 0.0072723356697708234,
|
| 76 |
+
"kl": 0.0,
|
| 77 |
+
"kl/code_mean": 0.0,
|
| 78 |
+
"kl/code_sum": 0.0,
|
| 79 |
+
"kl/thinking_mean": 0.0,
|
| 80 |
+
"kl/thinking_sum": 0.0,
|
| 81 |
+
"kl/total_loss_contribution": 0.0,
|
| 82 |
+
"kl_mask/effective_frac": 1.0,
|
| 83 |
+
"kl_mask/fallback_rate": 0.375,
|
| 84 |
+
"learning_rate": 1e-05,
|
| 85 |
+
"loss": -0.07276922464370728,
|
| 86 |
+
"num_tokens": 281787.0,
|
| 87 |
+
"reward": 0.8359375,
|
| 88 |
+
"reward_std": 0.586249053478241,
|
| 89 |
+
"rewards/reward_func/mean": 0.09288194444444445,
|
| 90 |
+
"rewards/reward_func/std": 0.08260986540052626,
|
| 91 |
+
"sampling/importance_sampling_ratio/max": 2.994373321533203,
|
| 92 |
+
"sampling/importance_sampling_ratio/mean": 0.9549344182014465,
|
| 93 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 94 |
+
"sampling/sampling_logp_difference/max": 12.530097961425781,
|
| 95 |
+
"sampling/sampling_logp_difference/mean": 0.19683969020843506,
|
| 96 |
+
"seg/thinking_frac": 0.277571365237236,
|
| 97 |
+
"step": 2,
|
| 98 |
+
"step_time": 101.29505116201472
|
| 99 |
+
},
|
| 100 |
+
{
|
| 101 |
+
"clip_ratio/high_max": 0.0,
|
| 102 |
+
"clip_ratio/high_mean": 0.0,
|
| 103 |
+
"clip_ratio/low_mean": 0.0,
|
| 104 |
+
"clip_ratio/low_min": 0.0,
|
| 105 |
+
"clip_ratio/region_mean": 0.0,
|
| 106 |
+
"completions/clipped_ratio": 0.0625,
|
| 107 |
+
"completions/max_length": 4096.0,
|
| 108 |
+
"completions/max_terminated_length": 3639.0,
|
| 109 |
+
"completions/mean_length": 1161.890625,
|
| 110 |
+
"completions/mean_terminated_length": 1064.8167724609375,
|
| 111 |
+
"completions/min_length": 27.0,
|
| 112 |
+
"completions/min_terminated_length": 71.0,
|
| 113 |
+
"degenerate_groups_filtered": 0.0,
|
| 114 |
+
"entropy": 0.665709063410759,
|
| 115 |
+
"entropy/code": 0.5846218094229698,
|
| 116 |
+
"entropy/thinking": 1.0504512786865234,
|
| 117 |
+
"epoch": 0.007389162561576354,
|
| 118 |
+
"frac_reward_zero_std": 0.0,
|
| 119 |
+
"grad_norm": 0.0069561315311969085,
|
| 120 |
+
"kl": 0.00022882659686729312,
|
| 121 |
+
"kl/code_mean": 0.00020637200213968754,
|
| 122 |
+
"kl/code_sum": 0.746826171875,
|
| 123 |
+
"kl/thinking_mean": 0.0003513325864332728,
|
| 124 |
+
"kl/thinking_sum": 0.18701171875,
|
| 125 |
+
"kl/total_loss_contribution": 4.5765317508994485e-06,
|
| 126 |
+
"kl_mask/effective_frac": 1.0,
|
| 127 |
+
"kl_mask/fallback_rate": 0.6875,
|
| 128 |
+
"learning_rate": 2e-05,
|
| 129 |
+
"loss": -0.058667171746492386,
|
| 130 |
+
"num_tokens": 441812.0,
|
| 131 |
+
"reward": 0.59765625,
|
| 132 |
+
"reward_std": 0.7554263472557068,
|
| 133 |
+
"rewards/reward_func/mean": 0.06640625,
|
| 134 |
+
"rewards/reward_func/std": 0.13972595002916124,
|
| 135 |
+
"sampling/importance_sampling_ratio/max": 2.997058868408203,
|
| 136 |
+
"sampling/importance_sampling_ratio/mean": 0.9472224712371826,
|
| 137 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 138 |
+
"sampling/sampling_logp_difference/max": 15.625813484191895,
|
| 139 |
+
"sampling/sampling_logp_difference/mean": 0.20085622370243073,
|
| 140 |
+
"seg/thinking_frac": 0.1202197466045618,
|
| 141 |
+
"step": 3,
|
| 142 |
+
"step_time": 99.6158953839913
|
| 143 |
+
},
|
| 144 |
+
{
|
| 145 |
+
"clip_ratio/high_max": 0.0,
|
| 146 |
+
"clip_ratio/high_mean": 0.0,
|
| 147 |
+
"clip_ratio/low_mean": 0.0,
|
| 148 |
+
"clip_ratio/low_min": 0.0,
|
| 149 |
+
"clip_ratio/region_mean": 0.0,
|
| 150 |
+
"completions/clipped_ratio": 0.0625,
|
| 151 |
+
"completions/max_length": 4096.0,
|
| 152 |
+
"completions/max_terminated_length": 3975.0,
|
| 153 |
+
"completions/mean_length": 1027.015625,
|
| 154 |
+
"completions/mean_terminated_length": 976.2167358398438,
|
| 155 |
+
"completions/min_length": 63.0,
|
| 156 |
+
"completions/min_terminated_length": 63.0,
|
| 157 |
+
"degenerate_groups_filtered": 0.0,
|
| 158 |
+
"entropy": 0.7792375385761261,
|
| 159 |
+
"entropy/code": 0.6018073260784149,
|
| 160 |
+
"entropy/thinking": 1.1589028090238571,
|
| 161 |
+
"epoch": 0.009852216748768473,
|
| 162 |
+
"frac_reward_zero_std": 0.0,
|
| 163 |
+
"grad_norm": 0.0073976111259691744,
|
| 164 |
+
"kl": 0.00026545999935478903,
|
| 165 |
+
"kl/code_mean": 0.00021759581431979313,
|
| 166 |
+
"kl/code_sum": 0.751220703125,
|
| 167 |
+
"kl/thinking_mean": 0.00034365309693384916,
|
| 168 |
+
"kl/thinking_sum": 0.265380859375,
|
| 169 |
+
"kl/total_loss_contribution": 5.309199991643254e-06,
|
| 170 |
+
"kl_mask/effective_frac": 1.0,
|
| 171 |
+
"kl_mask/fallback_rate": 0.75,
|
| 172 |
+
"learning_rate": 3e-05,
|
| 173 |
+
"loss": -0.08001244068145752,
|
| 174 |
+
"num_tokens": 588165.0,
|
| 175 |
+
"reward": 0.66796875,
|
| 176 |
+
"reward_std": 0.5808700919151306,
|
| 177 |
+
"rewards/reward_func/mean": 0.07421875,
|
| 178 |
+
"rewards/reward_func/std": 0.0801383869515525,
|
| 179 |
+
"sampling/importance_sampling_ratio/max": 2.9991674423217773,
|
| 180 |
+
"sampling/importance_sampling_ratio/mean": 0.9509478211402893,
|
| 181 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 182 |
+
"sampling/sampling_logp_difference/max": 13.950121879577637,
|
| 183 |
+
"sampling/sampling_logp_difference/mean": 0.21089348196983337,
|
| 184 |
+
"seg/thinking_frac": 0.24568463489413261,
|
| 185 |
+
"step": 4,
|
| 186 |
+
"step_time": 89.85415055201156
|
| 187 |
+
},
|
| 188 |
+
{
|
| 189 |
+
"clip_ratio/high_max": 0.0,
|
| 190 |
+
"clip_ratio/high_mean": 0.0,
|
| 191 |
+
"clip_ratio/low_mean": 0.0,
|
| 192 |
+
"clip_ratio/low_min": 0.0,
|
| 193 |
+
"clip_ratio/region_mean": 0.0,
|
| 194 |
+
"completions/clipped_ratio": 0.109375,
|
| 195 |
+
"completions/max_length": 4096.0,
|
| 196 |
+
"completions/max_terminated_length": 3713.0,
|
| 197 |
+
"completions/mean_length": 1195.8125,
|
| 198 |
+
"completions/mean_terminated_length": 1138.22802734375,
|
| 199 |
+
"completions/min_length": 114.0,
|
| 200 |
+
"completions/min_terminated_length": 114.0,
|
| 201 |
+
"degenerate_groups_filtered": 0.0,
|
| 202 |
+
"entropy": 0.6786182373762131,
|
| 203 |
+
"entropy/code": 0.5457415133714676,
|
| 204 |
+
"entropy/thinking": 1.1302461177110672,
|
| 205 |
+
"epoch": 0.012315270935960592,
|
| 206 |
+
"frac_reward_zero_std": 0.0,
|
| 207 |
+
"grad_norm": 0.008588841975023267,
|
| 208 |
+
"kl": 0.00021128138541826047,
|
| 209 |
+
"kl/code_mean": 0.0001765529959811829,
|
| 210 |
+
"kl/code_sum": 0.590087890625,
|
| 211 |
+
"kl/thinking_mean": 0.0003517654477036558,
|
| 212 |
+
"kl/thinking_sum": 0.259033203125,
|
| 213 |
+
"kl/total_loss_contribution": 4.225627776577312e-06,
|
| 214 |
+
"kl_mask/effective_frac": 1.0,
|
| 215 |
+
"kl_mask/fallback_rate": 0.5625,
|
| 216 |
+
"learning_rate": 4e-05,
|
| 217 |
+
"loss": -0.059298016130924225,
|
| 218 |
+
"num_tokens": 742697.0,
|
| 219 |
+
"reward": 0.55859375,
|
| 220 |
+
"reward_std": 0.5414233803749084,
|
| 221 |
+
"rewards/reward_func/mean": 0.062065972222222224,
|
| 222 |
+
"rewards/reward_func/std": 0.07304682334264119,
|
| 223 |
+
"sampling/importance_sampling_ratio/max": 2.9963865280151367,
|
| 224 |
+
"sampling/importance_sampling_ratio/mean": 0.9545855522155762,
|
| 225 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 226 |
+
"sampling/sampling_logp_difference/max": 17.89510154724121,
|
| 227 |
+
"sampling/sampling_logp_difference/mean": 0.18367084860801697,
|
| 228 |
+
"seg/thinking_frac": 0.19514773413538933,
|
| 229 |
+
"step": 5,
|
| 230 |
+
"step_time": 102.79374916400411
|
| 231 |
+
},
|
| 232 |
+
{
|
| 233 |
+
"clip_ratio/high_max": 0.0,
|
| 234 |
+
"clip_ratio/high_mean": 0.0,
|
| 235 |
+
"clip_ratio/low_mean": 0.0,
|
| 236 |
+
"clip_ratio/low_min": 0.0,
|
| 237 |
+
"clip_ratio/region_mean": 0.0,
|
| 238 |
+
"completions/clipped_ratio": 0.078125,
|
| 239 |
+
"completions/max_length": 4096.0,
|
| 240 |
+
"completions/max_terminated_length": 3556.0,
|
| 241 |
+
"completions/mean_length": 1217.296875,
|
| 242 |
+
"completions/mean_terminated_length": 1015.322021484375,
|
| 243 |
+
"completions/min_length": 120.0,
|
| 244 |
+
"completions/min_terminated_length": 120.0,
|
| 245 |
+
"degenerate_groups_filtered": 0.0,
|
| 246 |
+
"entropy": 0.6884773075580597,
|
| 247 |
+
"entropy/code": 0.5450494289398193,
|
| 248 |
+
"entropy/thinking": 1.271912395954132,
|
| 249 |
+
"epoch": 0.014778325123152709,
|
| 250 |
+
"frac_reward_zero_std": 0.0,
|
| 251 |
+
"grad_norm": 0.005020993831263842,
|
| 252 |
+
"kl": 0.000222868773562368,
|
| 253 |
+
"kl/code_mean": 0.00019304524539620616,
|
| 254 |
+
"kl/code_sum": 0.654541015625,
|
| 255 |
+
"kl/thinking_mean": 0.0003449782379902899,
|
| 256 |
+
"kl/thinking_sum": 0.23876953125,
|
| 257 |
+
"kl/total_loss_contribution": 4.457375553101883e-06,
|
| 258 |
+
"kl_mask/effective_frac": 1.0,
|
| 259 |
+
"kl_mask/fallback_rate": 0.3125,
|
| 260 |
+
"learning_rate": 5e-05,
|
| 261 |
+
"loss": -0.045627061277627945,
|
| 262 |
+
"num_tokens": 902812.0,
|
| 263 |
+
"reward": 0.8515625,
|
| 264 |
+
"reward_std": 0.4873778820037842,
|
| 265 |
+
"rewards/reward_func/mean": 0.09461805555555555,
|
| 266 |
+
"rewards/reward_func/std": 0.07031723029083675,
|
| 267 |
+
"sampling/importance_sampling_ratio/max": 2.9967851638793945,
|
| 268 |
+
"sampling/importance_sampling_ratio/mean": 0.9538474678993225,
|
| 269 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 270 |
+
"sampling/sampling_logp_difference/max": 11.532113075256348,
|
| 271 |
+
"sampling/sampling_logp_difference/mean": 0.18934644758701324,
|
| 272 |
+
"seg/thinking_frac": 0.19443420693278313,
|
| 273 |
+
"step": 6,
|
| 274 |
+
"step_time": 95.69083843295812
|
| 275 |
+
},
|
| 276 |
+
{
|
| 277 |
+
"clip_ratio/high_max": 0.0,
|
| 278 |
+
"clip_ratio/high_mean": 0.0,
|
| 279 |
+
"clip_ratio/low_mean": 0.0,
|
| 280 |
+
"clip_ratio/low_min": 0.0,
|
| 281 |
+
"clip_ratio/region_mean": 0.0,
|
| 282 |
+
"completions/clipped_ratio": 0.109375,
|
| 283 |
+
"completions/max_length": 4096.0,
|
| 284 |
+
"completions/max_terminated_length": 3385.0,
|
| 285 |
+
"completions/mean_length": 1335.6875,
|
| 286 |
+
"completions/mean_terminated_length": 1063.0701904296875,
|
| 287 |
+
"completions/min_length": 171.0,
|
| 288 |
+
"completions/min_terminated_length": 171.0,
|
| 289 |
+
"degenerate_groups_filtered": 0.0,
|
| 290 |
+
"entropy": 0.6550470441579819,
|
| 291 |
+
"entropy/code": 0.5024034604430199,
|
| 292 |
+
"entropy/thinking": 1.1513627171516418,
|
| 293 |
+
"epoch": 0.017241379310344827,
|
| 294 |
+
"frac_reward_zero_std": 0.0,
|
| 295 |
+
"grad_norm": 0.004672486046062607,
|
| 296 |
+
"kl": 0.0003102169794146903,
|
| 297 |
+
"kl/code_mean": 0.00023193488596007228,
|
| 298 |
+
"kl/code_sum": 0.65185546875,
|
| 299 |
+
"kl/thinking_mean": 0.0005790923314634711,
|
| 300 |
+
"kl/thinking_sum": 0.46826171875,
|
| 301 |
+
"kl/total_loss_contribution": 6.204339683790749e-06,
|
| 302 |
+
"kl_mask/effective_frac": 1.0,
|
| 303 |
+
"kl_mask/fallback_rate": 0.25,
|
| 304 |
+
"learning_rate": 4.999920929719799e-05,
|
| 305 |
+
"loss": -0.01926868036389351,
|
| 306 |
+
"num_tokens": 1068200.0,
|
| 307 |
+
"reward": 0.7890625,
|
| 308 |
+
"reward_std": 0.556738555431366,
|
| 309 |
+
"rewards/reward_func/mean": 0.0876736111111111,
|
| 310 |
+
"rewards/reward_func/std": 0.10424729684988658,
|
| 311 |
+
"sampling/importance_sampling_ratio/max": 2.9987099170684814,
|
| 312 |
+
"sampling/importance_sampling_ratio/mean": 0.9546715021133423,
|
| 313 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 314 |
+
"sampling/sampling_logp_difference/max": 12.85855770111084,
|
| 315 |
+
"sampling/sampling_logp_difference/mean": 0.1804715096950531,
|
| 316 |
+
"seg/thinking_frac": 0.22628148272633553,
|
| 317 |
+
"step": 7,
|
| 318 |
+
"step_time": 90.55077856796561
|
| 319 |
+
},
|
| 320 |
+
{
|
| 321 |
+
"clip_ratio/high_max": 0.0,
|
| 322 |
+
"clip_ratio/high_mean": 0.0,
|
| 323 |
+
"clip_ratio/low_mean": 0.0,
|
| 324 |
+
"clip_ratio/low_min": 0.0,
|
| 325 |
+
"clip_ratio/region_mean": 0.0,
|
| 326 |
+
"completions/clipped_ratio": 0.046875,
|
| 327 |
+
"completions/max_length": 4096.0,
|
| 328 |
+
"completions/max_terminated_length": 4033.0,
|
| 329 |
+
"completions/mean_length": 1152.546875,
|
| 330 |
+
"completions/mean_terminated_length": 1104.4261474609375,
|
| 331 |
+
"completions/min_length": 178.0,
|
| 332 |
+
"completions/min_terminated_length": 197.0,
|
| 333 |
+
"degenerate_groups_filtered": 0.0,
|
| 334 |
+
"entropy": 0.7468318492174149,
|
| 335 |
+
"entropy/code": 0.546456515789032,
|
| 336 |
+
"entropy/thinking": 1.3342507183551788,
|
| 337 |
+
"epoch": 0.019704433497536946,
|
| 338 |
+
"frac_reward_zero_std": 0.0,
|
| 339 |
+
"grad_norm": 0.006966918634559277,
|
| 340 |
+
"kl": 0.00033052685466827825,
|
| 341 |
+
"kl/code_mean": 0.00021720394579460844,
|
| 342 |
+
"kl/code_sum": 0.74853515625,
|
| 343 |
+
"kl/thinking_mean": 0.0007158305088523775,
|
| 344 |
+
"kl/thinking_sum": 0.69140625,
|
| 345 |
+
"kl/total_loss_contribution": 6.610537184315035e-06,
|
| 346 |
+
"kl_mask/effective_frac": 1.0,
|
| 347 |
+
"kl_mask/fallback_rate": 0.125,
|
| 348 |
+
"learning_rate": 4.999683723880881e-05,
|
| 349 |
+
"loss": -0.0853881910443306,
|
| 350 |
+
"num_tokens": 1217915.0,
|
| 351 |
+
"reward": 1.078125,
|
| 352 |
+
"reward_std": 0.7700688242912292,
|
| 353 |
+
"rewards/reward_func/mean": 0.11979166666666667,
|
| 354 |
+
"rewards/reward_func/std": 0.1717546267641915,
|
| 355 |
+
"sampling/importance_sampling_ratio/max": 2.9975016117095947,
|
| 356 |
+
"sampling/importance_sampling_ratio/mean": 0.9532677531242371,
|
| 357 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 358 |
+
"sampling/sampling_logp_difference/max": 11.485830307006836,
|
| 359 |
+
"sampling/sampling_logp_difference/mean": 0.18722301721572876,
|
| 360 |
+
"seg/thinking_frac": 0.24554069340229034,
|
| 361 |
+
"step": 8,
|
| 362 |
+
"step_time": 108.65152971699717
|
| 363 |
+
},
|
| 364 |
+
{
|
| 365 |
+
"clip_ratio/high_max": 0.0,
|
| 366 |
+
"clip_ratio/high_mean": 0.0,
|
| 367 |
+
"clip_ratio/low_mean": 0.0,
|
| 368 |
+
"clip_ratio/low_min": 0.0,
|
| 369 |
+
"clip_ratio/region_mean": 0.0,
|
| 370 |
+
"completions/clipped_ratio": 0.0625,
|
| 371 |
+
"completions/max_length": 4096.0,
|
| 372 |
+
"completions/max_terminated_length": 3642.0,
|
| 373 |
+
"completions/mean_length": 1393.640625,
|
| 374 |
+
"completions/mean_terminated_length": 1213.4833984375,
|
| 375 |
+
"completions/min_length": 204.0,
|
| 376 |
+
"completions/min_terminated_length": 204.0,
|
| 377 |
+
"degenerate_groups_filtered": 0.0,
|
| 378 |
+
"entropy": 0.6949516534805298,
|
| 379 |
+
"entropy/code": 0.5527757704257965,
|
| 380 |
+
"entropy/thinking": 1.2355611324310303,
|
| 381 |
+
"epoch": 0.022167487684729065,
|
| 382 |
+
"frac_reward_zero_std": 0.0,
|
| 383 |
+
"grad_norm": 0.005639985562132842,
|
| 384 |
+
"kl": 0.001052522158715874,
|
| 385 |
+
"kl/code_mean": 0.0010765275510493666,
|
| 386 |
+
"kl/code_sum": 3.391357421875,
|
| 387 |
+
"kl/thinking_mean": 0.000857135746628046,
|
| 388 |
+
"kl/thinking_sum": 0.643798828125,
|
| 389 |
+
"kl/total_loss_contribution": 2.1050443365311367e-05,
|
| 390 |
+
"kl_mask/effective_frac": 1.0,
|
| 391 |
+
"kl_mask/fallback_rate": 0.375,
|
| 392 |
+
"learning_rate": 4.9992883974879933e-05,
|
| 393 |
+
"loss": -0.016033891588449478,
|
| 394 |
+
"num_tokens": 1387044.0,
|
| 395 |
+
"reward": 0.83203125,
|
| 396 |
+
"reward_std": 0.6221038699150085,
|
| 397 |
+
"rewards/reward_func/mean": 0.09244791666666667,
|
| 398 |
+
"rewards/reward_func/std": 0.0869131088256836,
|
| 399 |
+
"sampling/importance_sampling_ratio/max": 2.9940664768218994,
|
| 400 |
+
"sampling/importance_sampling_ratio/mean": 0.9514634013175964,
|
| 401 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 402 |
+
"sampling/sampling_logp_difference/max": 12.674421310424805,
|
| 403 |
+
"sampling/sampling_logp_difference/mean": 0.19623836874961853,
|
| 404 |
+
"seg/thinking_frac": 0.20687565580010414,
|
| 405 |
+
"step": 9,
|
| 406 |
+
"step_time": 102.96673059702152
|
| 407 |
+
},
|
| 408 |
+
{
|
| 409 |
+
"clip_ratio/high_max": 0.0,
|
| 410 |
+
"clip_ratio/high_mean": 0.0,
|
| 411 |
+
"clip_ratio/low_mean": 0.0,
|
| 412 |
+
"clip_ratio/low_min": 0.0,
|
| 413 |
+
"clip_ratio/region_mean": 0.0,
|
| 414 |
+
"completions/clipped_ratio": 0.046875,
|
| 415 |
+
"completions/max_length": 4096.0,
|
| 416 |
+
"completions/max_terminated_length": 3533.0,
|
| 417 |
+
"completions/mean_length": 1321.40625,
|
| 418 |
+
"completions/mean_terminated_length": 1184.9508056640625,
|
| 419 |
+
"completions/min_length": 56.0,
|
| 420 |
+
"completions/min_terminated_length": 56.0,
|
| 421 |
+
"degenerate_groups_filtered": 0.0,
|
| 422 |
+
"entropy": 0.706970289349556,
|
| 423 |
+
"entropy/code": 0.5685221552848816,
|
| 424 |
+
"entropy/thinking": 1.2291285693645477,
|
| 425 |
+
"epoch": 0.024630541871921183,
|
| 426 |
+
"frac_reward_zero_std": 0.0,
|
| 427 |
+
"grad_norm": 0.010049425309084847,
|
| 428 |
+
"kl": 0.003098885623330716,
|
| 429 |
+
"kl/code_mean": 0.0034346847896813415,
|
| 430 |
+
"kl/code_sum": 11.960693359375,
|
| 431 |
+
"kl/thinking_mean": 0.0014323829964268953,
|
| 432 |
+
"kl/thinking_sum": 1.0888671875,
|
| 433 |
+
"kl/total_loss_contribution": 6.197771040206135e-05,
|
| 434 |
+
"kl_mask/effective_frac": 1.0,
|
| 435 |
+
"kl_mask/fallback_rate": 0.25,
|
| 436 |
+
"learning_rate": 4.998734975547991e-05,
|
| 437 |
+
"loss": -0.005550745874643326,
|
| 438 |
+
"num_tokens": 1546046.0,
|
| 439 |
+
"reward": 0.9140625,
|
| 440 |
+
"reward_std": 0.7335751056671143,
|
| 441 |
+
"rewards/reward_func/mean": 0.1015625,
|
| 442 |
+
"rewards/reward_func/std": 0.12709666623009575,
|
| 443 |
+
"sampling/importance_sampling_ratio/max": 2.998774290084839,
|
| 444 |
+
"sampling/importance_sampling_ratio/mean": 0.9479279518127441,
|
| 445 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 446 |
+
"sampling/sampling_logp_difference/max": 11.668327331542969,
|
| 447 |
+
"sampling/sampling_logp_difference/mean": 0.20149840414524078,
|
| 448 |
+
"seg/thinking_frac": 0.1998237892985344,
|
| 449 |
+
"step": 10,
|
| 450 |
+
"step_time": 100.57014895303291
|
| 451 |
+
},
|
| 452 |
+
{
|
| 453 |
+
"clip_ratio/high_max": 0.0,
|
| 454 |
+
"clip_ratio/high_mean": 0.0,
|
| 455 |
+
"clip_ratio/low_mean": 0.0,
|
| 456 |
+
"clip_ratio/low_min": 0.0,
|
| 457 |
+
"clip_ratio/region_mean": 0.0,
|
| 458 |
+
"completions/clipped_ratio": 0.046875,
|
| 459 |
+
"completions/max_length": 4096.0,
|
| 460 |
+
"completions/max_terminated_length": 2309.0,
|
| 461 |
+
"completions/mean_length": 884.609375,
|
| 462 |
+
"completions/mean_terminated_length": 770.6065063476562,
|
| 463 |
+
"completions/min_length": 253.0,
|
| 464 |
+
"completions/min_terminated_length": 253.0,
|
| 465 |
+
"degenerate_groups_filtered": 0.0,
|
| 466 |
+
"entropy": 0.7739142775535583,
|
| 467 |
+
"entropy/code": 0.504700556397438,
|
| 468 |
+
"entropy/thinking": 1.1482309997081757,
|
| 469 |
+
"epoch": 0.027093596059113302,
|
| 470 |
+
"frac_reward_zero_std": 0.25,
|
| 471 |
+
"grad_norm": 0.006082433765504613,
|
| 472 |
+
"kl": 0.001316279056482017,
|
| 473 |
+
"kl/code_mean": 0.0010569319274509326,
|
| 474 |
+
"kl/code_sum": 3.104736328125,
|
| 475 |
+
"kl/thinking_mean": 0.0011445207201177254,
|
| 476 |
+
"kl/thinking_sum": 1.16162109375,
|
| 477 |
+
"kl/total_loss_contribution": 2.6325580620323308e-05,
|
| 478 |
+
"kl_mask/effective_frac": 1.0,
|
| 479 |
+
"kl_mask/fallback_rate": 0.0625,
|
| 480 |
+
"learning_rate": 4.9980234930682546e-05,
|
| 481 |
+
"loss": 0.006589197088032961,
|
| 482 |
+
"num_tokens": 1686501.0,
|
| 483 |
+
"reward": 0.8671875,
|
| 484 |
+
"reward_std": 0.43179386854171753,
|
| 485 |
+
"rewards/reward_func/mean": 0.09635416666666667,
|
| 486 |
+
"rewards/reward_func/std": 0.06402202943960826,
|
| 487 |
+
"sampling/importance_sampling_ratio/max": 2.9949393272399902,
|
| 488 |
+
"sampling/importance_sampling_ratio/mean": 0.9484283924102783,
|
| 489 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 490 |
+
"sampling/sampling_logp_difference/max": 12.686782836914062,
|
| 491 |
+
"sampling/sampling_logp_difference/mean": 0.21330925822257996,
|
| 492 |
+
"seg/thinking_frac": 0.3824174106121063,
|
| 493 |
+
"step": 11,
|
| 494 |
+
"step_time": 98.66246142203454
|
| 495 |
+
},
|
| 496 |
+
{
|
| 497 |
+
"clip_ratio/high_max": 0.0,
|
| 498 |
+
"clip_ratio/high_mean": 0.0,
|
| 499 |
+
"clip_ratio/low_mean": 0.0,
|
| 500 |
+
"clip_ratio/low_min": 0.0,
|
| 501 |
+
"clip_ratio/region_mean": 0.0,
|
| 502 |
+
"completions/clipped_ratio": 0.046875,
|
| 503 |
+
"completions/max_length": 4096.0,
|
| 504 |
+
"completions/max_terminated_length": 3352.0,
|
| 505 |
+
"completions/mean_length": 1388.96875,
|
| 506 |
+
"completions/mean_terminated_length": 1308.9508056640625,
|
| 507 |
+
"completions/min_length": 271.0,
|
| 508 |
+
"completions/min_terminated_length": 271.0,
|
| 509 |
+
"degenerate_groups_filtered": 0.0,
|
| 510 |
+
"entropy": 0.7141824215650558,
|
| 511 |
+
"entropy/code": 0.5431285202503204,
|
| 512 |
+
"entropy/thinking": 1.3165772259235382,
|
| 513 |
+
"epoch": 0.029556650246305417,
|
| 514 |
+
"frac_reward_zero_std": 0.0,
|
| 515 |
+
"grad_norm": 0.004133542848468609,
|
| 516 |
+
"kl": 0.0004349540249677375,
|
| 517 |
+
"kl/code_mean": 0.0002485945042280946,
|
| 518 |
+
"kl/code_sum": 1.0,
|
| 519 |
+
"kl/thinking_mean": 0.0010879361361730844,
|
| 520 |
+
"kl/thinking_sum": 1.05615234375,
|
| 521 |
+
"kl/total_loss_contribution": 8.699080353835598e-06,
|
| 522 |
+
"kl_mask/effective_frac": 1.0,
|
| 523 |
+
"kl_mask/fallback_rate": 0.125,
|
| 524 |
+
"learning_rate": 4.997153995054481e-05,
|
| 525 |
+
"loss": -0.0025130638387054205,
|
| 526 |
+
"num_tokens": 1863987.0,
|
| 527 |
+
"reward": 1.015625,
|
| 528 |
+
"reward_std": 0.4753340482711792,
|
| 529 |
+
"rewards/reward_func/mean": 0.11284722222222222,
|
| 530 |
+
"rewards/reward_func/std": 0.06982019543647766,
|
| 531 |
+
"sampling/importance_sampling_ratio/max": 2.995980739593506,
|
| 532 |
+
"sampling/importance_sampling_ratio/mean": 0.944504976272583,
|
| 533 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 534 |
+
"sampling/sampling_logp_difference/max": 12.43104076385498,
|
| 535 |
+
"sampling/sampling_logp_difference/mean": 0.2099040448665619,
|
| 536 |
+
"seg/thinking_frac": 0.22088434547185898,
|
| 537 |
+
"step": 12,
|
| 538 |
+
"step_time": 115.35374042004696
|
| 539 |
+
},
|
| 540 |
+
{
|
| 541 |
+
"clip_ratio/high_max": 0.0,
|
| 542 |
+
"clip_ratio/high_mean": 0.0,
|
| 543 |
+
"clip_ratio/low_mean": 0.0,
|
| 544 |
+
"clip_ratio/low_min": 0.0,
|
| 545 |
+
"clip_ratio/region_mean": 0.0,
|
| 546 |
+
"completions/clipped_ratio": 0.09375,
|
| 547 |
+
"completions/max_length": 4096.0,
|
| 548 |
+
"completions/max_terminated_length": 2295.0,
|
| 549 |
+
"completions/mean_length": 1101.296875,
|
| 550 |
+
"completions/mean_terminated_length": 842.5344848632812,
|
| 551 |
+
"completions/min_length": 59.0,
|
| 552 |
+
"completions/min_terminated_length": 59.0,
|
| 553 |
+
"degenerate_groups_filtered": 0.0,
|
| 554 |
+
"entropy": 0.698951929807663,
|
| 555 |
+
"entropy/code": 0.4811418578028679,
|
| 556 |
+
"entropy/thinking": 1.2490146160125732,
|
| 557 |
+
"epoch": 0.03201970443349754,
|
| 558 |
+
"frac_reward_zero_std": 0.0,
|
| 559 |
+
"grad_norm": 0.024108448275232883,
|
| 560 |
+
"kl": 0.04201257172826445,
|
| 561 |
+
"kl/code_mean": 0.05459589597012382,
|
| 562 |
+
"kl/code_sum": 165.015869140625,
|
| 563 |
+
"kl/thinking_mean": 0.0009699366055428982,
|
| 564 |
+
"kl/thinking_sum": 0.787841796875,
|
| 565 |
+
"kl/total_loss_contribution": 0.0008402513781220478,
|
| 566 |
+
"kl_mask/effective_frac": 1.0,
|
| 567 |
+
"kl_mask/fallback_rate": 0.125,
|
| 568 |
+
"learning_rate": 4.996126536507831e-05,
|
| 569 |
+
"loss": 0.004714315757155418,
|
| 570 |
+
"num_tokens": 2009414.0,
|
| 571 |
+
"reward": 1.1015625,
|
| 572 |
+
"reward_std": 0.8318066596984863,
|
| 573 |
+
"rewards/reward_func/mean": 0.12239583333333333,
|
| 574 |
+
"rewards/reward_func/std": 0.12802670896053314,
|
| 575 |
+
"sampling/importance_sampling_ratio/max": 2.99251651763916,
|
| 576 |
+
"sampling/importance_sampling_ratio/mean": 0.9603347778320312,
|
| 577 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 578 |
+
"sampling/sampling_logp_difference/max": 10.050997734069824,
|
| 579 |
+
"sampling/sampling_logp_difference/mean": 0.17832711338996887,
|
| 580 |
+
"seg/thinking_frac": 0.285446472465992,
|
| 581 |
+
"step": 13,
|
| 582 |
+
"step_time": 99.54845813498832
|
| 583 |
+
},
|
| 584 |
+
{
|
| 585 |
+
"clip_ratio/high_max": 0.0,
|
| 586 |
+
"clip_ratio/high_mean": 0.0,
|
| 587 |
+
"clip_ratio/low_mean": 0.0,
|
| 588 |
+
"clip_ratio/low_min": 0.0,
|
| 589 |
+
"clip_ratio/region_mean": 0.0,
|
| 590 |
+
"completions/clipped_ratio": 0.03125,
|
| 591 |
+
"completions/max_length": 4096.0,
|
| 592 |
+
"completions/max_terminated_length": 2216.0,
|
| 593 |
+
"completions/mean_length": 914.296875,
|
| 594 |
+
"completions/mean_terminated_length": 811.6612548828125,
|
| 595 |
+
"completions/min_length": 101.0,
|
| 596 |
+
"completions/min_terminated_length": 101.0,
|
| 597 |
+
"degenerate_groups_filtered": 0.0,
|
| 598 |
+
"entropy": 0.7749249637126923,
|
| 599 |
+
"entropy/code": 0.5825137197971344,
|
| 600 |
+
"entropy/thinking": 1.2152594327926636,
|
| 601 |
+
"epoch": 0.034482758620689655,
|
| 602 |
+
"frac_reward_zero_std": 0.0,
|
| 603 |
+
"grad_norm": 0.006546118059693103,
|
| 604 |
+
"kl": 0.000679801290971227,
|
| 605 |
+
"kl/code_mean": 0.0003832913644146174,
|
| 606 |
+
"kl/code_sum": 0.809814453125,
|
| 607 |
+
"kl/thinking_mean": 0.0013703204749617726,
|
| 608 |
+
"kl/thinking_sum": 1.187255859375,
|
| 609 |
+
"kl/total_loss_contribution": 1.3596025382867083e-05,
|
| 610 |
+
"kl_mask/effective_frac": 1.0,
|
| 611 |
+
"kl_mask/fallback_rate": 0.25,
|
| 612 |
+
"learning_rate": 4.9949411824214534e-05,
|
| 613 |
+
"loss": 0.0027569299563765526,
|
| 614 |
+
"num_tokens": 2146825.0,
|
| 615 |
+
"reward": 0.87890625,
|
| 616 |
+
"reward_std": 0.46927881240844727,
|
| 617 |
+
"rewards/reward_func/mean": 0.09765625,
|
| 618 |
+
"rewards/reward_func/std": 0.06829224858019087,
|
| 619 |
+
"sampling/importance_sampling_ratio/max": 2.9998490810394287,
|
| 620 |
+
"sampling/importance_sampling_ratio/mean": 0.9546762704849243,
|
| 621 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 622 |
+
"sampling/sampling_logp_difference/max": 12.362849235534668,
|
| 623 |
+
"sampling/sampling_logp_difference/mean": 0.2001952826976776,
|
| 624 |
+
"seg/thinking_frac": 0.30725786089897156,
|
| 625 |
+
"step": 14,
|
| 626 |
+
"step_time": 100.15386100800242
|
| 627 |
+
},
|
| 628 |
+
{
|
| 629 |
+
"clip_ratio/high_max": 0.0,
|
| 630 |
+
"clip_ratio/high_mean": 0.0,
|
| 631 |
+
"clip_ratio/low_mean": 0.0,
|
| 632 |
+
"clip_ratio/low_min": 0.0,
|
| 633 |
+
"clip_ratio/region_mean": 0.0,
|
| 634 |
+
"completions/clipped_ratio": 0.046875,
|
| 635 |
+
"completions/max_length": 4096.0,
|
| 636 |
+
"completions/max_terminated_length": 3614.0,
|
| 637 |
+
"completions/mean_length": 1309.53125,
|
| 638 |
+
"completions/mean_terminated_length": 1172.49169921875,
|
| 639 |
+
"completions/min_length": 258.0,
|
| 640 |
+
"completions/min_terminated_length": 258.0,
|
| 641 |
+
"degenerate_groups_filtered": 0.0,
|
| 642 |
+
"entropy": 0.670013040304184,
|
| 643 |
+
"entropy/code": 0.5221169739961624,
|
| 644 |
+
"entropy/thinking": 1.2216048538684845,
|
| 645 |
+
"epoch": 0.03694581280788178,
|
| 646 |
+
"frac_reward_zero_std": 0.0,
|
| 647 |
+
"grad_norm": 0.00457610185156593,
|
| 648 |
+
"kl": 0.0006424650200642645,
|
| 649 |
+
"kl/code_mean": 0.00036872723285341635,
|
| 650 |
+
"kl/code_sum": 1.193359375,
|
| 651 |
+
"kl/thinking_mean": 0.001521107944427058,
|
| 652 |
+
"kl/thinking_sum": 1.360107421875,
|
| 653 |
+
"kl/total_loss_contribution": 1.2849299992012675e-05,
|
| 654 |
+
"kl_mask/effective_frac": 1.0,
|
| 655 |
+
"kl_mask/fallback_rate": 0.25,
|
| 656 |
+
"learning_rate": 4.9935980077763706e-05,
|
| 657 |
+
"loss": 0.008143404498696327,
|
| 658 |
+
"num_tokens": 2311867.0,
|
| 659 |
+
"reward": 0.8828125,
|
| 660 |
+
"reward_std": 0.42717406153678894,
|
| 661 |
+
"rewards/reward_func/mean": 0.09809027777777778,
|
| 662 |
+
"rewards/reward_func/std": 0.06239571670691172,
|
| 663 |
+
"sampling/importance_sampling_ratio/max": 2.997850179672241,
|
| 664 |
+
"sampling/importance_sampling_ratio/mean": 0.9479074478149414,
|
| 665 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 666 |
+
"sampling/sampling_logp_difference/max": 13.96391773223877,
|
| 667 |
+
"sampling/sampling_logp_difference/mean": 0.19878119230270386,
|
| 668 |
+
"seg/thinking_frac": 0.22116510942578316,
|
| 669 |
+
"step": 15,
|
| 670 |
+
"step_time": 104.38469738399726
|
| 671 |
+
},
|
| 672 |
+
{
|
| 673 |
+
"clip_ratio/high_max": 0.0,
|
| 674 |
+
"clip_ratio/high_mean": 0.0,
|
| 675 |
+
"clip_ratio/low_mean": 0.0,
|
| 676 |
+
"clip_ratio/low_min": 0.0,
|
| 677 |
+
"clip_ratio/region_mean": 0.0,
|
| 678 |
+
"completions/clipped_ratio": 0.09375,
|
| 679 |
+
"completions/max_length": 4096.0,
|
| 680 |
+
"completions/max_terminated_length": 3530.0,
|
| 681 |
+
"completions/mean_length": 1238.90625,
|
| 682 |
+
"completions/mean_terminated_length": 1061.2930908203125,
|
| 683 |
+
"completions/min_length": 210.0,
|
| 684 |
+
"completions/min_terminated_length": 210.0,
|
| 685 |
+
"degenerate_groups_filtered": 0.0,
|
| 686 |
+
"entropy": 0.7045484632253647,
|
| 687 |
+
"entropy/code": 0.5236108005046844,
|
| 688 |
+
"entropy/thinking": 1.0629068464040756,
|
| 689 |
+
"epoch": 0.03940886699507389,
|
| 690 |
+
"frac_reward_zero_std": 0.0,
|
| 691 |
+
"grad_norm": 0.0040434402688446875,
|
| 692 |
+
"kl": 0.0007065868558129296,
|
| 693 |
+
"kl/code_mean": 0.0003540540055837482,
|
| 694 |
+
"kl/code_sum": 0.94384765625,
|
| 695 |
+
"kl/thinking_mean": 0.0015548540104646236,
|
| 696 |
+
"kl/thinking_sum": 1.47021484375,
|
| 697 |
+
"kl/total_loss_contribution": 1.41317366342264e-05,
|
| 698 |
+
"kl_mask/effective_frac": 1.0,
|
| 699 |
+
"kl_mask/fallback_rate": 0.0625,
|
| 700 |
+
"learning_rate": 4.99209709753674e-05,
|
| 701 |
+
"loss": -0.01762321963906288,
|
| 702 |
+
"num_tokens": 2470421.0,
|
| 703 |
+
"reward": 0.94921875,
|
| 704 |
+
"reward_std": 0.40868473052978516,
|
| 705 |
+
"rewards/reward_func/mean": 0.10546875,
|
| 706 |
+
"rewards/reward_func/std": 0.06101351810826196,
|
| 707 |
+
"sampling/importance_sampling_ratio/max": 2.994006395339966,
|
| 708 |
+
"sampling/importance_sampling_ratio/mean": 0.9534869194030762,
|
| 709 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 710 |
+
"sampling/sampling_logp_difference/max": 11.280420303344727,
|
| 711 |
+
"sampling/sampling_logp_difference/mean": 0.1967828869819641,
|
| 712 |
+
"seg/thinking_frac": 0.2788471654057503,
|
| 713 |
+
"step": 16,
|
| 714 |
+
"step_time": 104.19931473597535
|
| 715 |
+
},
|
| 716 |
+
{
|
| 717 |
+
"clip_ratio/high_max": 0.0,
|
| 718 |
+
"clip_ratio/high_mean": 0.0,
|
| 719 |
+
"clip_ratio/low_mean": 0.0,
|
| 720 |
+
"clip_ratio/low_min": 0.0,
|
| 721 |
+
"clip_ratio/region_mean": 0.0,
|
| 722 |
+
"completions/clipped_ratio": 0.015625,
|
| 723 |
+
"completions/max_length": 4096.0,
|
| 724 |
+
"completions/max_terminated_length": 2484.0,
|
| 725 |
+
"completions/mean_length": 833.859375,
|
| 726 |
+
"completions/mean_terminated_length": 782.0794067382812,
|
| 727 |
+
"completions/min_length": 235.0,
|
| 728 |
+
"completions/min_terminated_length": 235.0,
|
| 729 |
+
"degenerate_groups_filtered": 0.0,
|
| 730 |
+
"entropy": 0.6594647169113159,
|
| 731 |
+
"entropy/code": 0.46378815174102783,
|
| 732 |
+
"entropy/thinking": 1.1030372232198715,
|
| 733 |
+
"epoch": 0.04187192118226601,
|
| 734 |
+
"frac_reward_zero_std": 0.0,
|
| 735 |
+
"grad_norm": 0.004534186381120654,
|
| 736 |
+
"kl": 0.0007302999729290605,
|
| 737 |
+
"kl/code_mean": 0.0006507219077320769,
|
| 738 |
+
"kl/code_sum": 1.56005859375,
|
| 739 |
+
"kl/thinking_mean": 0.0008526231104042381,
|
| 740 |
+
"kl/thinking_sum": 0.711181640625,
|
| 741 |
+
"kl/total_loss_contribution": 1.460599924030248e-05,
|
| 742 |
+
"kl_mask/effective_frac": 1.0,
|
| 743 |
+
"kl_mask/fallback_rate": 0.125,
|
| 744 |
+
"learning_rate": 4.9904385466444756e-05,
|
| 745 |
+
"loss": -0.006497188005596399,
|
| 746 |
+
"num_tokens": 2591468.0,
|
| 747 |
+
"reward": 0.9375,
|
| 748 |
+
"reward_std": 0.44095855951309204,
|
| 749 |
+
"rewards/reward_func/mean": 0.10416666666666667,
|
| 750 |
+
"rewards/reward_func/std": 0.06500735216670567,
|
| 751 |
+
"sampling/importance_sampling_ratio/max": 2.9962902069091797,
|
| 752 |
+
"sampling/importance_sampling_ratio/mean": 0.963579535484314,
|
| 753 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 754 |
+
"sampling/sampling_logp_difference/max": 11.426942825317383,
|
| 755 |
+
"sampling/sampling_logp_difference/mean": 0.16766537725925446,
|
| 756 |
+
"seg/thinking_frac": 0.3021527975797653,
|
| 757 |
+
"step": 17,
|
| 758 |
+
"step_time": 92.71895237697754
|
| 759 |
+
},
|
| 760 |
+
{
|
| 761 |
+
"clip_ratio/high_max": 0.0,
|
| 762 |
+
"clip_ratio/high_mean": 0.0,
|
| 763 |
+
"clip_ratio/low_mean": 0.0,
|
| 764 |
+
"clip_ratio/low_min": 0.0,
|
| 765 |
+
"clip_ratio/region_mean": 0.0,
|
| 766 |
+
"completions/clipped_ratio": 0.046875,
|
| 767 |
+
"completions/max_length": 4096.0,
|
| 768 |
+
"completions/max_terminated_length": 3440.0,
|
| 769 |
+
"completions/mean_length": 1285.203125,
|
| 770 |
+
"completions/mean_terminated_length": 1200.3934326171875,
|
| 771 |
+
"completions/min_length": 123.0,
|
| 772 |
+
"completions/min_terminated_length": 123.0,
|
| 773 |
+
"degenerate_groups_filtered": 0.0,
|
| 774 |
+
"entropy": 0.709451287984848,
|
| 775 |
+
"entropy/code": 0.5227500647306442,
|
| 776 |
+
"entropy/thinking": 1.2196418046951294,
|
| 777 |
+
"epoch": 0.04433497536945813,
|
| 778 |
+
"frac_reward_zero_std": 0.0,
|
| 779 |
+
"grad_norm": 0.003751340433878426,
|
| 780 |
+
"kl": 0.0007622432458447292,
|
| 781 |
+
"kl/code_mean": 0.0005503558422788046,
|
| 782 |
+
"kl/code_sum": 1.616943359375,
|
| 783 |
+
"kl/thinking_mean": 0.0017978230316657573,
|
| 784 |
+
"kl/thinking_sum": 1.528076171875,
|
| 785 |
+
"kl/total_loss_contribution": 1.5244864925989532e-05,
|
| 786 |
+
"kl_mask/effective_frac": 1.0,
|
| 787 |
+
"kl_mask/fallback_rate": 0.125,
|
| 788 |
+
"learning_rate": 4.988622460013245e-05,
|
| 789 |
+
"loss": -0.028508899733424187,
|
| 790 |
+
"num_tokens": 2750249.0,
|
| 791 |
+
"reward": 0.96875,
|
| 792 |
+
"reward_std": 0.4787135720252991,
|
| 793 |
+
"rewards/reward_func/mean": 0.1076388888888889,
|
| 794 |
+
"rewards/reward_func/std": 0.06990587711334229,
|
| 795 |
+
"sampling/importance_sampling_ratio/max": 2.9972617626190186,
|
| 796 |
+
"sampling/importance_sampling_ratio/mean": 0.9533290863037109,
|
| 797 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 798 |
+
"sampling/sampling_logp_difference/max": 15.344735145568848,
|
| 799 |
+
"sampling/sampling_logp_difference/mean": 0.19208058714866638,
|
| 800 |
+
"seg/thinking_frac": 0.29232197254896164,
|
| 801 |
+
"step": 18,
|
| 802 |
+
"step_time": 100.43533309493796
|
| 803 |
+
},
|
| 804 |
+
{
|
| 805 |
+
"clip_ratio/high_max": 0.0,
|
| 806 |
+
"clip_ratio/high_mean": 0.0,
|
| 807 |
+
"clip_ratio/low_mean": 0.0,
|
| 808 |
+
"clip_ratio/low_min": 0.0,
|
| 809 |
+
"clip_ratio/region_mean": 0.0,
|
| 810 |
+
"completions/clipped_ratio": 0.09375,
|
| 811 |
+
"completions/max_length": 4096.0,
|
| 812 |
+
"completions/max_terminated_length": 3578.0,
|
| 813 |
+
"completions/mean_length": 1137.0,
|
| 814 |
+
"completions/mean_terminated_length": 972.0689697265625,
|
| 815 |
+
"completions/min_length": 1.0,
|
| 816 |
+
"completions/min_terminated_length": 18.0,
|
| 817 |
+
"degenerate_groups_filtered": 0.0,
|
| 818 |
+
"entropy": 0.6891960501670837,
|
| 819 |
+
"entropy/code": 0.5047332718968391,
|
| 820 |
+
"entropy/thinking": 1.2450141310691833,
|
| 821 |
+
"epoch": 0.046798029556650245,
|
| 822 |
+
"frac_reward_zero_std": 0.0,
|
| 823 |
+
"grad_norm": 0.004762653573544255,
|
| 824 |
+
"kl": 0.000925938191358,
|
| 825 |
+
"kl/code_mean": 0.0011078362440457568,
|
| 826 |
+
"kl/code_sum": 1.1904296875,
|
| 827 |
+
"kl/thinking_mean": 0.001560449309181422,
|
| 828 |
+
"kl/thinking_sum": 0.908447265625,
|
| 829 |
+
"kl/total_loss_contribution": 1.8518763226893498e-05,
|
| 830 |
+
"kl_mask/effective_frac": 1.0,
|
| 831 |
+
"kl_mask/fallback_rate": 0.3125,
|
| 832 |
+
"learning_rate": 4.986648952521831e-05,
|
| 833 |
+
"loss": -0.05186585336923599,
|
| 834 |
+
"num_tokens": 2903529.0,
|
| 835 |
+
"reward": 0.91015625,
|
| 836 |
+
"reward_std": 0.5831074118614197,
|
| 837 |
+
"rewards/reward_func/mean": 0.10112847222222222,
|
| 838 |
+
"rewards/reward_func/std": 0.11483020583788554,
|
| 839 |
+
"sampling/importance_sampling_ratio/max": 2.996598720550537,
|
| 840 |
+
"sampling/importance_sampling_ratio/mean": 0.9616074562072754,
|
| 841 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 842 |
+
"sampling/sampling_logp_difference/max": 12.562503814697266,
|
| 843 |
+
"sampling/sampling_logp_difference/mean": 0.16846302151679993,
|
| 844 |
+
"seg/thinking_frac": 0.25528161600232124,
|
| 845 |
+
"step": 19,
|
| 846 |
+
"step_time": 103.13669178399141
|
| 847 |
+
},
|
| 848 |
+
{
|
| 849 |
+
"clip_ratio/high_max": 0.0,
|
| 850 |
+
"clip_ratio/high_mean": 0.0,
|
| 851 |
+
"clip_ratio/low_mean": 0.0,
|
| 852 |
+
"clip_ratio/low_min": 0.0,
|
| 853 |
+
"clip_ratio/region_mean": 0.0,
|
| 854 |
+
"completions/clipped_ratio": 0.078125,
|
| 855 |
+
"completions/max_length": 4096.0,
|
| 856 |
+
"completions/max_terminated_length": 3802.0,
|
| 857 |
+
"completions/mean_length": 1158.625,
|
| 858 |
+
"completions/mean_terminated_length": 909.6949462890625,
|
| 859 |
+
"completions/min_length": 70.0,
|
| 860 |
+
"completions/min_terminated_length": 70.0,
|
| 861 |
+
"degenerate_groups_filtered": 0.0,
|
| 862 |
+
"entropy": 0.5834990069270134,
|
| 863 |
+
"entropy/code": 0.40924975275993347,
|
| 864 |
+
"entropy/thinking": 1.1035513877868652,
|
| 865 |
+
"epoch": 0.04926108374384237,
|
| 866 |
+
"frac_reward_zero_std": 0.0,
|
| 867 |
+
"grad_norm": 0.014197001474273144,
|
| 868 |
+
"kl": 0.0012770651519531384,
|
| 869 |
+
"kl/code_mean": 0.0014194050017977133,
|
| 870 |
+
"kl/code_sum": 3.10009765625,
|
| 871 |
+
"kl/thinking_mean": 0.0009059785661520436,
|
| 872 |
+
"kl/thinking_sum": 0.7353515625,
|
| 873 |
+
"kl/total_loss_contribution": 2.554130196585902e-05,
|
| 874 |
+
"kl_mask/effective_frac": 1.0,
|
| 875 |
+
"kl_mask/fallback_rate": 0.125,
|
| 876 |
+
"learning_rate": 4.9845181490068655e-05,
|
| 877 |
+
"loss": -0.17723286151885986,
|
| 878 |
+
"num_tokens": 3069633.0,
|
| 879 |
+
"reward": 0.99609375,
|
| 880 |
+
"reward_std": 0.8339190483093262,
|
| 881 |
+
"rewards/reward_func/mean": 0.11067708333333333,
|
| 882 |
+
"rewards/reward_func/std": 0.15295158161057365,
|
| 883 |
+
"sampling/importance_sampling_ratio/max": 2.999572277069092,
|
| 884 |
+
"sampling/importance_sampling_ratio/mean": 0.9564659595489502,
|
| 885 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 886 |
+
"sampling/sampling_logp_difference/max": 10.428086280822754,
|
| 887 |
+
"sampling/sampling_logp_difference/mean": 0.17278142273426056,
|
| 888 |
+
"seg/thinking_frac": 0.2584066540002823,
|
| 889 |
+
"step": 20,
|
| 890 |
+
"step_time": 107.87285372699262
|
| 891 |
+
},
|
| 892 |
+
{
|
| 893 |
+
"clip_ratio/high_max": 0.0,
|
| 894 |
+
"clip_ratio/high_mean": 0.0,
|
| 895 |
+
"clip_ratio/low_mean": 0.0,
|
| 896 |
+
"clip_ratio/low_min": 0.0,
|
| 897 |
+
"clip_ratio/region_mean": 0.0,
|
| 898 |
+
"completions/clipped_ratio": 0.0625,
|
| 899 |
+
"completions/max_length": 4096.0,
|
| 900 |
+
"completions/max_terminated_length": 4089.0,
|
| 901 |
+
"completions/mean_length": 1567.203125,
|
| 902 |
+
"completions/mean_terminated_length": 1445.716796875,
|
| 903 |
+
"completions/min_length": 184.0,
|
| 904 |
+
"completions/min_terminated_length": 184.0,
|
| 905 |
+
"degenerate_groups_filtered": 0.0,
|
| 906 |
+
"entropy": 0.7715802490711212,
|
| 907 |
+
"entropy/code": 0.6016347259283066,
|
| 908 |
+
"entropy/thinking": 1.316393405199051,
|
| 909 |
+
"epoch": 0.05172413793103448,
|
| 910 |
+
"frac_reward_zero_std": 0.25,
|
| 911 |
+
"grad_norm": 0.0034036483986607854,
|
| 912 |
+
"kl": 0.0006484113109763712,
|
| 913 |
+
"kl/code_mean": 0.0004752305685542524,
|
| 914 |
+
"kl/code_sum": 1.96435546875,
|
| 915 |
+
"kl/thinking_mean": 0.001203801337396726,
|
| 916 |
+
"kl/thinking_sum": 1.367919921875,
|
| 917 |
+
"kl/total_loss_contribution": 1.2968225519216503e-05,
|
| 918 |
+
"kl_mask/effective_frac": 1.0,
|
| 919 |
+
"kl_mask/fallback_rate": 0.375,
|
| 920 |
+
"learning_rate": 4.982230184254933e-05,
|
| 921 |
+
"loss": -0.03974352777004242,
|
| 922 |
+
"num_tokens": 3247614.0,
|
| 923 |
+
"reward": 0.97265625,
|
| 924 |
+
"reward_std": 0.44346487522125244,
|
| 925 |
+
"rewards/reward_func/mean": 0.10807291666666667,
|
| 926 |
+
"rewards/reward_func/std": 0.06546966897116767,
|
| 927 |
+
"sampling/importance_sampling_ratio/max": 2.9979422092437744,
|
| 928 |
+
"sampling/importance_sampling_ratio/mean": 0.9420000314712524,
|
| 929 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 930 |
+
"sampling/sampling_logp_difference/max": 20.839807510375977,
|
| 931 |
+
"sampling/sampling_logp_difference/mean": 0.22293400764465332,
|
| 932 |
+
"seg/thinking_frac": 0.2308126501739025,
|
| 933 |
+
"step": 21,
|
| 934 |
+
"step_time": 102.0408651669859
|
| 935 |
+
},
|
| 936 |
+
{
|
| 937 |
+
"clip_ratio/high_max": 0.0,
|
| 938 |
+
"clip_ratio/high_mean": 0.0,
|
| 939 |
+
"clip_ratio/low_mean": 0.0,
|
| 940 |
+
"clip_ratio/low_min": 0.0,
|
| 941 |
+
"clip_ratio/region_mean": 0.0,
|
| 942 |
+
"completions/clipped_ratio": 0.0625,
|
| 943 |
+
"completions/max_length": 4096.0,
|
| 944 |
+
"completions/max_terminated_length": 3939.0,
|
| 945 |
+
"completions/mean_length": 1351.359375,
|
| 946 |
+
"completions/mean_terminated_length": 1196.11669921875,
|
| 947 |
+
"completions/min_length": 312.0,
|
| 948 |
+
"completions/min_terminated_length": 312.0,
|
| 949 |
+
"degenerate_groups_filtered": 0.0,
|
| 950 |
+
"entropy": 0.6702008694410324,
|
| 951 |
+
"entropy/code": 0.529250793159008,
|
| 952 |
+
"entropy/thinking": 1.2768133580684662,
|
| 953 |
+
"epoch": 0.054187192118226604,
|
| 954 |
+
"frac_reward_zero_std": 0.0,
|
| 955 |
+
"grad_norm": 0.004740578989128821,
|
| 956 |
+
"kl": 0.0008713168354006484,
|
| 957 |
+
"kl/code_mean": 0.0008001201858860441,
|
| 958 |
+
"kl/code_sum": 2.806884765625,
|
| 959 |
+
"kl/thinking_mean": 0.0012732007016893476,
|
| 960 |
+
"kl/thinking_sum": 1.037109375,
|
| 961 |
+
"kl/total_loss_contribution": 1.742633639878477e-05,
|
| 962 |
+
"kl_mask/effective_frac": 1.0,
|
| 963 |
+
"kl_mask/fallback_rate": 0.25,
|
| 964 |
+
"learning_rate": 4.9797852029940475e-05,
|
| 965 |
+
"loss": -0.008679582737386227,
|
| 966 |
+
"num_tokens": 3410357.0,
|
| 967 |
+
"reward": 1.0390625,
|
| 968 |
+
"reward_std": 0.6743879318237305,
|
| 969 |
+
"rewards/reward_func/mean": 0.1154513888888889,
|
| 970 |
+
"rewards/reward_func/std": 0.14854909976323447,
|
| 971 |
+
"sampling/importance_sampling_ratio/max": 2.999838352203369,
|
| 972 |
+
"sampling/importance_sampling_ratio/mean": 0.9545325636863708,
|
| 973 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 974 |
+
"sampling/sampling_logp_difference/max": 12.618735313415527,
|
| 975 |
+
"sampling/sampling_logp_difference/mean": 0.19043642282485962,
|
| 976 |
+
"seg/thinking_frac": 0.19356486573815346,
|
| 977 |
+
"step": 22,
|
| 978 |
+
"step_time": 102.6878371280036
|
| 979 |
+
},
|
| 980 |
+
{
|
| 981 |
+
"clip_ratio/high_max": 0.0,
|
| 982 |
+
"clip_ratio/high_mean": 0.0,
|
| 983 |
+
"clip_ratio/low_mean": 0.0,
|
| 984 |
+
"clip_ratio/low_min": 0.0,
|
| 985 |
+
"clip_ratio/region_mean": 0.0,
|
| 986 |
+
"completions/clipped_ratio": 0.046875,
|
| 987 |
+
"completions/max_length": 4096.0,
|
| 988 |
+
"completions/max_terminated_length": 3454.0,
|
| 989 |
+
"completions/mean_length": 1112.84375,
|
| 990 |
+
"completions/mean_terminated_length": 1019.5409545898438,
|
| 991 |
+
"completions/min_length": 327.0,
|
| 992 |
+
"completions/min_terminated_length": 327.0,
|
| 993 |
+
"degenerate_groups_filtered": 0.0,
|
| 994 |
+
"entropy": 0.6103805005550385,
|
| 995 |
+
"entropy/code": 0.4392230287194252,
|
| 996 |
+
"entropy/thinking": 1.1203276216983795,
|
| 997 |
+
"epoch": 0.05665024630541872,
|
| 998 |
+
"frac_reward_zero_std": 0.0,
|
| 999 |
+
"grad_norm": 0.0034154360430670018,
|
| 1000 |
+
"kl": 0.0007456812891177833,
|
| 1001 |
+
"kl/code_mean": 0.0007122096649254672,
|
| 1002 |
+
"kl/code_sum": 2.15283203125,
|
| 1003 |
+
"kl/thinking_mean": 0.0008380540530197322,
|
| 1004 |
+
"kl/thinking_sum": 0.7138671875,
|
| 1005 |
+
"kl/total_loss_contribution": 1.4913624909240752e-05,
|
| 1006 |
+
"kl_mask/effective_frac": 1.0,
|
| 1007 |
+
"kl_mask/fallback_rate": 0.125,
|
| 1008 |
+
"learning_rate": 4.9771833598844874e-05,
|
| 1009 |
+
"loss": -0.05771990492939949,
|
| 1010 |
+
"num_tokens": 3552395.0,
|
| 1011 |
+
"reward": 0.88671875,
|
| 1012 |
+
"reward_std": 0.3881419003009796,
|
| 1013 |
+
"rewards/reward_func/mean": 0.09852430555555555,
|
| 1014 |
+
"rewards/reward_func/std": 0.056457738081614174,
|
| 1015 |
+
"sampling/importance_sampling_ratio/max": 2.9998326301574707,
|
| 1016 |
+
"sampling/importance_sampling_ratio/mean": 0.9591754674911499,
|
| 1017 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 1018 |
+
"sampling/sampling_logp_difference/max": 16.235795974731445,
|
| 1019 |
+
"sampling/sampling_logp_difference/mean": 0.1700170934200287,
|
| 1020 |
+
"seg/thinking_frac": 0.25201668962836266,
|
| 1021 |
+
"step": 23,
|
| 1022 |
+
"step_time": 104.56320369199966
|
| 1023 |
+
},
|
| 1024 |
+
{
|
| 1025 |
+
"clip_ratio/high_max": 0.0,
|
| 1026 |
+
"clip_ratio/high_mean": 0.0,
|
| 1027 |
+
"clip_ratio/low_mean": 0.0,
|
| 1028 |
+
"clip_ratio/low_min": 0.0,
|
| 1029 |
+
"clip_ratio/region_mean": 0.0,
|
| 1030 |
+
"completions/clipped_ratio": 0.078125,
|
| 1031 |
+
"completions/max_length": 4096.0,
|
| 1032 |
+
"completions/max_terminated_length": 2894.0,
|
| 1033 |
+
"completions/mean_length": 1068.375,
|
| 1034 |
+
"completions/mean_terminated_length": 940.2373046875,
|
| 1035 |
+
"completions/min_length": 120.0,
|
| 1036 |
+
"completions/min_terminated_length": 177.0,
|
| 1037 |
+
"degenerate_groups_filtered": 0.0,
|
| 1038 |
+
"entropy": 0.7939669191837311,
|
| 1039 |
+
"entropy/code": 0.5714405104517937,
|
| 1040 |
+
"entropy/thinking": 1.1087244749069214,
|
| 1041 |
+
"epoch": 0.059113300492610835,
|
| 1042 |
+
"frac_reward_zero_std": 0.0,
|
| 1043 |
+
"grad_norm": 0.0058141401094253895,
|
| 1044 |
+
"kl": 0.0008775486930971965,
|
| 1045 |
+
"kl/code_mean": 0.0007419007160933688,
|
| 1046 |
+
"kl/code_sum": 1.62353515625,
|
| 1047 |
+
"kl/thinking_mean": 0.0010127306159120053,
|
| 1048 |
+
"kl/thinking_sum": 1.08203125,
|
| 1049 |
+
"kl/total_loss_contribution": 1.755097309796838e-05,
|
| 1050 |
+
"kl_mask/effective_frac": 1.0,
|
| 1051 |
+
"kl_mask/fallback_rate": 0.125,
|
| 1052 |
+
"learning_rate": 4.974424819509027e-05,
|
| 1053 |
+
"loss": 0.012094950303435326,
|
| 1054 |
+
"num_tokens": 3697075.0,
|
| 1055 |
+
"reward": 0.890625,
|
| 1056 |
+
"reward_std": 0.419927716255188,
|
| 1057 |
+
"rewards/reward_func/mean": 0.09895833333333333,
|
| 1058 |
+
"rewards/reward_func/std": 0.061526068382793,
|
| 1059 |
+
"sampling/importance_sampling_ratio/max": 2.9988396167755127,
|
| 1060 |
+
"sampling/importance_sampling_ratio/mean": 0.9510908722877502,
|
| 1061 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 1062 |
+
"sampling/sampling_logp_difference/max": 13.990053176879883,
|
| 1063 |
+
"sampling/sampling_logp_difference/mean": 0.20467409491539001,
|
| 1064 |
+
"seg/thinking_frac": 0.37359438091516495,
|
| 1065 |
+
"step": 24,
|
| 1066 |
+
"step_time": 100.41657635298907
|
| 1067 |
+
},
|
| 1068 |
+
{
|
| 1069 |
+
"clip_ratio/high_max": 0.0,
|
| 1070 |
+
"clip_ratio/high_mean": 0.0,
|
| 1071 |
+
"clip_ratio/low_mean": 0.0,
|
| 1072 |
+
"clip_ratio/low_min": 0.0,
|
| 1073 |
+
"clip_ratio/region_mean": 0.0,
|
| 1074 |
+
"completions/clipped_ratio": 0.015625,
|
| 1075 |
+
"completions/max_length": 3718.0,
|
| 1076 |
+
"completions/max_terminated_length": 3718.0,
|
| 1077 |
+
"completions/mean_length": 933.828125,
|
| 1078 |
+
"completions/mean_terminated_length": 916.0000610351562,
|
| 1079 |
+
"completions/min_length": 146.0,
|
| 1080 |
+
"completions/min_terminated_length": 146.0,
|
| 1081 |
+
"degenerate_groups_filtered": 0.0,
|
| 1082 |
+
"entropy": 0.7235542684793472,
|
| 1083 |
+
"entropy/code": 0.5413009449839592,
|
| 1084 |
+
"entropy/thinking": 1.1473042964935303,
|
| 1085 |
+
"epoch": 0.06157635467980296,
|
| 1086 |
+
"frac_reward_zero_std": 0.0,
|
| 1087 |
+
"grad_norm": 0.011043090894340931,
|
| 1088 |
+
"kl": 0.0011932639026781544,
|
| 1089 |
+
"kl/code_mean": 0.0011964660079684108,
|
| 1090 |
+
"kl/code_sum": 3.124267578125,
|
| 1091 |
+
"kl/thinking_mean": 0.0011966245947405696,
|
| 1092 |
+
"kl/thinking_sum": 1.034423828125,
|
| 1093 |
+
"kl/total_loss_contribution": 2.386527808084793e-05,
|
| 1094 |
+
"kl_mask/effective_frac": 1.0,
|
| 1095 |
+
"kl_mask/fallback_rate": 0.1875,
|
| 1096 |
+
"learning_rate": 4.971509756362513e-05,
|
| 1097 |
+
"loss": -0.037315864115953445,
|
| 1098 |
+
"num_tokens": 3838648.0,
|
| 1099 |
+
"reward": 0.91796875,
|
| 1100 |
+
"reward_std": 0.7113795876502991,
|
| 1101 |
+
"rewards/reward_func/mean": 0.10199652777777778,
|
| 1102 |
+
"rewards/reward_func/std": 0.12596790161397722,
|
| 1103 |
+
"sampling/importance_sampling_ratio/max": 2.99892520904541,
|
| 1104 |
+
"sampling/importance_sampling_ratio/mean": 0.9457910656929016,
|
| 1105 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 1106 |
+
"sampling/sampling_logp_difference/max": 11.182188034057617,
|
| 1107 |
+
"sampling/sampling_logp_difference/mean": 0.20948098599910736,
|
| 1108 |
+
"seg/thinking_frac": 0.2923164591193199,
|
| 1109 |
+
"step": 25,
|
| 1110 |
+
"step_time": 100.6090605330246
|
| 1111 |
+
},
|
| 1112 |
+
{
|
| 1113 |
+
"clip_ratio/high_max": 0.0,
|
| 1114 |
+
"clip_ratio/high_mean": 0.0,
|
| 1115 |
+
"clip_ratio/low_mean": 0.0,
|
| 1116 |
+
"clip_ratio/low_min": 0.0,
|
| 1117 |
+
"clip_ratio/region_mean": 0.0,
|
| 1118 |
+
"completions/clipped_ratio": 0.03125,
|
| 1119 |
+
"completions/max_length": 4096.0,
|
| 1120 |
+
"completions/max_terminated_length": 3544.0,
|
| 1121 |
+
"completions/mean_length": 1124.53125,
|
| 1122 |
+
"completions/mean_terminated_length": 1028.6773681640625,
|
| 1123 |
+
"completions/min_length": 265.0,
|
| 1124 |
+
"completions/min_terminated_length": 265.0,
|
| 1125 |
+
"degenerate_groups_filtered": 0.0,
|
| 1126 |
+
"entropy": 0.667577400803566,
|
| 1127 |
+
"entropy/code": 0.5010233372449875,
|
| 1128 |
+
"entropy/thinking": 1.2092621326446533,
|
| 1129 |
+
"epoch": 0.06403940886699508,
|
| 1130 |
+
"frac_reward_zero_std": 0.0,
|
| 1131 |
+
"grad_norm": 0.003569919021317598,
|
| 1132 |
+
"kl": 0.0010375405545346439,
|
| 1133 |
+
"kl/code_mean": 0.0009067944920388982,
|
| 1134 |
+
"kl/code_sum": 3.190185546875,
|
| 1135 |
+
"kl/thinking_mean": 0.001302398246480152,
|
| 1136 |
+
"kl/thinking_sum": 1.15185546875,
|
| 1137 |
+
"kl/total_loss_contribution": 2.0750810790559626e-05,
|
| 1138 |
+
"kl_mask/effective_frac": 1.0,
|
| 1139 |
+
"kl_mask/fallback_rate": 0.1875,
|
| 1140 |
+
"learning_rate": 4.968438354840834e-05,
|
| 1141 |
+
"loss": -0.020396601408720016,
|
| 1142 |
+
"num_tokens": 3986106.0,
|
| 1143 |
+
"reward": 1.05859375,
|
| 1144 |
+
"reward_std": 0.515132486820221,
|
| 1145 |
+
"rewards/reward_func/mean": 0.11762152777777778,
|
| 1146 |
+
"rewards/reward_func/std": 0.11036739084455702,
|
| 1147 |
+
"sampling/importance_sampling_ratio/max": 2.998187303543091,
|
| 1148 |
+
"sampling/importance_sampling_ratio/mean": 0.9533237814903259,
|
| 1149 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 1150 |
+
"sampling/sampling_logp_difference/max": 16.24921226501465,
|
| 1151 |
+
"sampling/sampling_logp_difference/mean": 0.19513137638568878,
|
| 1152 |
+
"seg/thinking_frac": 0.24044565111398697,
|
| 1153 |
+
"step": 26,
|
| 1154 |
+
"step_time": 102.36693953498616
|
| 1155 |
+
},
|
| 1156 |
+
{
|
| 1157 |
+
"clip_ratio/high_max": 0.0,
|
| 1158 |
+
"clip_ratio/high_mean": 0.0,
|
| 1159 |
+
"clip_ratio/low_mean": 0.0,
|
| 1160 |
+
"clip_ratio/low_min": 0.0,
|
| 1161 |
+
"clip_ratio/region_mean": 0.0,
|
| 1162 |
+
"completions/clipped_ratio": 0.015625,
|
| 1163 |
+
"completions/max_length": 4096.0,
|
| 1164 |
+
"completions/max_terminated_length": 4069.0,
|
| 1165 |
+
"completions/mean_length": 1156.78125,
|
| 1166 |
+
"completions/mean_terminated_length": 1110.1270751953125,
|
| 1167 |
+
"completions/min_length": 39.0,
|
| 1168 |
+
"completions/min_terminated_length": 39.0,
|
| 1169 |
+
"degenerate_groups_filtered": 0.0,
|
| 1170 |
+
"entropy": 0.6581976562738419,
|
| 1171 |
+
"entropy/code": 0.46350903809070587,
|
| 1172 |
+
"entropy/thinking": 1.2082424759864807,
|
| 1173 |
+
"epoch": 0.0665024630541872,
|
| 1174 |
+
"frac_reward_zero_std": 0.0,
|
| 1175 |
+
"grad_norm": 0.007922139480936617,
|
| 1176 |
+
"kl": 0.001241300575202331,
|
| 1177 |
+
"kl/code_mean": 0.0011370833672117442,
|
| 1178 |
+
"kl/code_sum": 3.486083984375,
|
| 1179 |
+
"kl/thinking_mean": 0.0014678456645924598,
|
| 1180 |
+
"kl/thinking_sum": 1.35498046875,
|
| 1181 |
+
"kl/total_loss_contribution": 2.482600939401891e-05,
|
| 1182 |
+
"kl_mask/effective_frac": 1.0,
|
| 1183 |
+
"kl_mask/fallback_rate": 0.25,
|
| 1184 |
+
"learning_rate": 4.965210809229253e-05,
|
| 1185 |
+
"loss": 0.041888486593961716,
|
| 1186 |
+
"num_tokens": 4141164.0,
|
| 1187 |
+
"reward": 0.953125,
|
| 1188 |
+
"reward_std": 0.5924039483070374,
|
| 1189 |
+
"rewards/reward_func/mean": 0.10590277777777778,
|
| 1190 |
+
"rewards/reward_func/std": 0.10758823321925269,
|
| 1191 |
+
"sampling/importance_sampling_ratio/max": 2.9960408210754395,
|
| 1192 |
+
"sampling/importance_sampling_ratio/mean": 0.9536508917808533,
|
| 1193 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 1194 |
+
"sampling/sampling_logp_difference/max": 11.518455505371094,
|
| 1195 |
+
"sampling/sampling_logp_difference/mean": 0.19073748588562012,
|
| 1196 |
+
"seg/thinking_frac": 0.2633824087679386,
|
| 1197 |
+
"step": 27,
|
| 1198 |
+
"step_time": 103.5348206210183
|
| 1199 |
+
},
|
| 1200 |
+
{
|
| 1201 |
+
"clip_ratio/high_max": 0.0,
|
| 1202 |
+
"clip_ratio/high_mean": 0.0,
|
| 1203 |
+
"clip_ratio/low_mean": 0.0,
|
| 1204 |
+
"clip_ratio/low_min": 0.0,
|
| 1205 |
+
"clip_ratio/region_mean": 0.0,
|
| 1206 |
+
"completions/clipped_ratio": 0.015625,
|
| 1207 |
+
"completions/max_length": 4096.0,
|
| 1208 |
+
"completions/max_terminated_length": 3945.0,
|
| 1209 |
+
"completions/mean_length": 1068.734375,
|
| 1210 |
+
"completions/mean_terminated_length": 1020.6826171875,
|
| 1211 |
+
"completions/min_length": 207.0,
|
| 1212 |
+
"completions/min_terminated_length": 207.0,
|
| 1213 |
+
"degenerate_groups_filtered": 0.0,
|
| 1214 |
+
"entropy": 0.8108957409858704,
|
| 1215 |
+
"entropy/code": 0.590365543961525,
|
| 1216 |
+
"entropy/thinking": 1.3190259039402008,
|
| 1217 |
+
"epoch": 0.06896551724137931,
|
| 1218 |
+
"frac_reward_zero_std": 0.0,
|
| 1219 |
+
"grad_norm": 0.0049200703237446694,
|
| 1220 |
+
"kl": 0.0009075804700842127,
|
| 1221 |
+
"kl/code_mean": 0.0007819266465958208,
|
| 1222 |
+
"kl/code_sum": 1.9453125,
|
| 1223 |
+
"kl/thinking_mean": 0.0012515032722149044,
|
| 1224 |
+
"kl/thinking_sum": 1.291259765625,
|
| 1225 |
+
"kl/total_loss_contribution": 1.8151608855987433e-05,
|
| 1226 |
+
"kl_mask/effective_frac": 1.0,
|
| 1227 |
+
"kl_mask/fallback_rate": 0.0625,
|
| 1228 |
+
"learning_rate": 4.961827323690118e-05,
|
| 1229 |
+
"loss": 0.015619301237165928,
|
| 1230 |
+
"num_tokens": 4304731.0,
|
| 1231 |
+
"reward": 1.0703125,
|
| 1232 |
+
"reward_std": 0.4909271001815796,
|
| 1233 |
+
"rewards/reward_func/mean": 0.1189236111111111,
|
| 1234 |
+
"rewards/reward_func/std": 0.07141293750868903,
|
| 1235 |
+
"sampling/importance_sampling_ratio/max": 2.9930286407470703,
|
| 1236 |
+
"sampling/importance_sampling_ratio/mean": 0.9434918165206909,
|
| 1237 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 1238 |
+
"sampling/sampling_logp_difference/max": 10.701640129089355,
|
| 1239 |
+
"sampling/sampling_logp_difference/mean": 0.22357606887817383,
|
| 1240 |
+
"seg/thinking_frac": 0.3074530102312565,
|
| 1241 |
+
"step": 28,
|
| 1242 |
+
"step_time": 107.61345914195408
|
| 1243 |
+
},
|
| 1244 |
+
{
|
| 1245 |
+
"clip_ratio/high_max": 0.0,
|
| 1246 |
+
"clip_ratio/high_mean": 0.0,
|
| 1247 |
+
"clip_ratio/low_mean": 0.0,
|
| 1248 |
+
"clip_ratio/low_min": 0.0,
|
| 1249 |
+
"clip_ratio/region_mean": 0.0,
|
| 1250 |
+
"completions/clipped_ratio": 0.046875,
|
| 1251 |
+
"completions/max_length": 4096.0,
|
| 1252 |
+
"completions/max_terminated_length": 2966.0,
|
| 1253 |
+
"completions/mean_length": 967.75,
|
| 1254 |
+
"completions/mean_terminated_length": 867.3770141601562,
|
| 1255 |
+
"completions/min_length": 60.0,
|
| 1256 |
+
"completions/min_terminated_length": 60.0,
|
| 1257 |
+
"degenerate_groups_filtered": 0.0,
|
| 1258 |
+
"entropy": 0.871458038687706,
|
| 1259 |
+
"entropy/code": 0.7086656093597412,
|
| 1260 |
+
"entropy/thinking": 1.070669338107109,
|
| 1261 |
+
"epoch": 0.07142857142857142,
|
| 1262 |
+
"frac_reward_zero_std": 0.0,
|
| 1263 |
+
"grad_norm": 0.013816576213839553,
|
| 1264 |
+
"kl": 0.0023918055812828243,
|
| 1265 |
+
"kl/code_mean": 0.0024392873310716823,
|
| 1266 |
+
"kl/code_sum": 5.06689453125,
|
| 1267 |
+
"kl/thinking_mean": 0.001628782250918448,
|
| 1268 |
+
"kl/thinking_sum": 1.16357421875,
|
| 1269 |
+
"kl/total_loss_contribution": 4.783611120728892e-05,
|
| 1270 |
+
"kl_mask/effective_frac": 1.0,
|
| 1271 |
+
"kl_mask/fallback_rate": 0.0625,
|
| 1272 |
+
"learning_rate": 4.9582881122499494e-05,
|
| 1273 |
+
"loss": -0.07957114279270172,
|
| 1274 |
+
"num_tokens": 4447611.0,
|
| 1275 |
+
"reward": 0.95703125,
|
| 1276 |
+
"reward_std": 0.9014480113983154,
|
| 1277 |
+
"rewards/reward_func/mean": 0.10633680555555555,
|
| 1278 |
+
"rewards/reward_func/std": 0.16258245706558228,
|
| 1279 |
+
"sampling/importance_sampling_ratio/max": 2.9966330528259277,
|
| 1280 |
+
"sampling/importance_sampling_ratio/mean": 0.9504553079605103,
|
| 1281 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 1282 |
+
"sampling/sampling_logp_difference/max": 13.290483474731445,
|
| 1283 |
+
"sampling/sampling_logp_difference/mean": 0.20878449082374573,
|
| 1284 |
+
"seg/thinking_frac": 0.2664119228720665,
|
| 1285 |
+
"step": 29,
|
| 1286 |
+
"step_time": 99.54375722302939
|
| 1287 |
+
},
|
| 1288 |
+
{
|
| 1289 |
+
"clip_ratio/high_max": 0.0,
|
| 1290 |
+
"clip_ratio/high_mean": 0.0,
|
| 1291 |
+
"clip_ratio/low_mean": 0.0,
|
| 1292 |
+
"clip_ratio/low_min": 0.0,
|
| 1293 |
+
"clip_ratio/region_mean": 0.0,
|
| 1294 |
+
"completions/clipped_ratio": 0.015625,
|
| 1295 |
+
"completions/max_length": 3684.0,
|
| 1296 |
+
"completions/max_terminated_length": 3684.0,
|
| 1297 |
+
"completions/mean_length": 1079.546875,
|
| 1298 |
+
"completions/mean_terminated_length": 1076.3175048828125,
|
| 1299 |
+
"completions/min_length": 265.0,
|
| 1300 |
+
"completions/min_terminated_length": 265.0,
|
| 1301 |
+
"degenerate_groups_filtered": 0.0,
|
| 1302 |
+
"entropy": 0.6870646774768829,
|
| 1303 |
+
"entropy/code": 0.5111498460173607,
|
| 1304 |
+
"entropy/thinking": 1.2974431812763214,
|
| 1305 |
+
"epoch": 0.07389162561576355,
|
| 1306 |
+
"frac_reward_zero_std": 0.25,
|
| 1307 |
+
"grad_norm": 0.0030894800720645773,
|
| 1308 |
+
"kl": 0.001190918919746764,
|
| 1309 |
+
"kl/code_mean": 0.0009750917670316994,
|
| 1310 |
+
"kl/code_sum": 3.854248046875,
|
| 1311 |
+
"kl/thinking_mean": 0.0021774407650809735,
|
| 1312 |
+
"kl/thinking_sum": 1.533935546875,
|
| 1313 |
+
"kl/total_loss_contribution": 2.381837794018793e-05,
|
| 1314 |
+
"kl_mask/effective_frac": 1.0,
|
| 1315 |
+
"kl_mask/fallback_rate": 0.125,
|
| 1316 |
+
"learning_rate": 4.954593398785899e-05,
|
| 1317 |
+
"loss": -0.02401747554540634,
|
| 1318 |
+
"num_tokens": 4613966.0,
|
| 1319 |
+
"reward": 1.1015625,
|
| 1320 |
+
"reward_std": 0.5358548164367676,
|
| 1321 |
+
"rewards/reward_func/mean": 0.12239583333333333,
|
| 1322 |
+
"rewards/reward_func/std": 0.11644800007343292,
|
| 1323 |
+
"sampling/importance_sampling_ratio/max": 2.99902606010437,
|
| 1324 |
+
"sampling/importance_sampling_ratio/mean": 0.9430747032165527,
|
| 1325 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 1326 |
+
"sampling/sampling_logp_difference/max": 14.433944702148438,
|
| 1327 |
+
"sampling/sampling_logp_difference/mean": 0.21299226582050323,
|
| 1328 |
+
"seg/thinking_frac": 0.22889656573534012,
|
| 1329 |
+
"step": 30,
|
| 1330 |
+
"step_time": 84.51256076604477
|
| 1331 |
+
},
|
| 1332 |
+
{
|
| 1333 |
+
"clip_ratio/high_max": 0.0,
|
| 1334 |
+
"clip_ratio/high_mean": 0.0,
|
| 1335 |
+
"clip_ratio/low_mean": 0.0,
|
| 1336 |
+
"clip_ratio/low_min": 0.0,
|
| 1337 |
+
"clip_ratio/region_mean": 0.0,
|
| 1338 |
+
"completions/clipped_ratio": 0.03125,
|
| 1339 |
+
"completions/max_length": 4096.0,
|
| 1340 |
+
"completions/max_terminated_length": 3536.0,
|
| 1341 |
+
"completions/mean_length": 1114.515625,
|
| 1342 |
+
"completions/mean_terminated_length": 1018.3386840820312,
|
| 1343 |
+
"completions/min_length": 123.0,
|
| 1344 |
+
"completions/min_terminated_length": 123.0,
|
| 1345 |
+
"degenerate_groups_filtered": 0.0,
|
| 1346 |
+
"entropy": 0.7102625817060471,
|
| 1347 |
+
"entropy/code": 0.46760451048612595,
|
| 1348 |
+
"entropy/thinking": 1.2981998324394226,
|
| 1349 |
+
"epoch": 0.07635467980295567,
|
| 1350 |
+
"frac_reward_zero_std": 0.0,
|
| 1351 |
+
"grad_norm": 0.012252487483686148,
|
| 1352 |
+
"kl": 0.0014072840858716518,
|
| 1353 |
+
"kl/code_mean": 0.0012857891560997814,
|
| 1354 |
+
"kl/code_sum": 4.636962890625,
|
| 1355 |
+
"kl/thinking_mean": 0.0017640646256040782,
|
| 1356 |
+
"kl/thinking_sum": 1.775146484375,
|
| 1357 |
+
"kl/total_loss_contribution": 2.814568142639473e-05,
|
| 1358 |
+
"kl_mask/effective_frac": 1.0,
|
| 1359 |
+
"kl_mask/fallback_rate": 0.0625,
|
| 1360 |
+
"learning_rate": 4.950743417011591e-05,
|
| 1361 |
+
"loss": -0.01380426436662674,
|
| 1362 |
+
"num_tokens": 4774575.0,
|
| 1363 |
+
"reward": 1.14453125,
|
| 1364 |
+
"reward_std": 0.8542664647102356,
|
| 1365 |
+
"rewards/reward_func/mean": 0.1271701388888889,
|
| 1366 |
+
"rewards/reward_func/std": 0.15118403567208183,
|
| 1367 |
+
"sampling/importance_sampling_ratio/max": 2.9944846630096436,
|
| 1368 |
+
"sampling/importance_sampling_ratio/mean": 0.9515092372894287,
|
| 1369 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 1370 |
+
"sampling/sampling_logp_difference/max": 13.624022483825684,
|
| 1371 |
+
"sampling/sampling_logp_difference/mean": 0.20099006593227386,
|
| 1372 |
+
"seg/thinking_frac": 0.29684875905513763,
|
| 1373 |
+
"step": 31,
|
| 1374 |
+
"step_time": 105.35778426099569
|
| 1375 |
+
},
|
| 1376 |
+
{
|
| 1377 |
+
"clip_ratio/high_max": 0.0,
|
| 1378 |
+
"clip_ratio/high_mean": 0.0,
|
| 1379 |
+
"clip_ratio/low_mean": 0.0,
|
| 1380 |
+
"clip_ratio/low_min": 0.0,
|
| 1381 |
+
"clip_ratio/region_mean": 0.0,
|
| 1382 |
+
"completions/clipped_ratio": 0.0,
|
| 1383 |
+
"completions/max_length": 3381.0,
|
| 1384 |
+
"completions/max_terminated_length": 3381.0,
|
| 1385 |
+
"completions/mean_length": 937.4375,
|
| 1386 |
+
"completions/mean_terminated_length": 937.4375,
|
| 1387 |
+
"completions/min_length": 127.0,
|
| 1388 |
+
"completions/min_terminated_length": 127.0,
|
| 1389 |
+
"degenerate_groups_filtered": 0.0,
|
| 1390 |
+
"entropy": 0.6670823246240616,
|
| 1391 |
+
"entropy/code": 0.4919791743159294,
|
| 1392 |
+
"entropy/thinking": 1.0447155684232712,
|
| 1393 |
+
"epoch": 0.07881773399014778,
|
| 1394 |
+
"frac_reward_zero_std": 0.25,
|
| 1395 |
+
"grad_norm": 0.0053964379233528625,
|
| 1396 |
+
"kl": 0.0011635527189355344,
|
| 1397 |
+
"kl/code_mean": 0.000987008897936903,
|
| 1398 |
+
"kl/code_sum": 2.8154296875,
|
| 1399 |
+
"kl/thinking_mean": 0.0014957001258153468,
|
| 1400 |
+
"kl/thinking_sum": 1.298583984375,
|
| 1401 |
+
"kl/total_loss_contribution": 2.32710540331027e-05,
|
| 1402 |
+
"kl_mask/effective_frac": 1.0,
|
| 1403 |
+
"kl_mask/fallback_rate": 0.0625,
|
| 1404 |
+
"learning_rate": 4.946738410462334e-05,
|
| 1405 |
+
"loss": -0.017866453155875206,
|
| 1406 |
+
"num_tokens": 4900395.0,
|
| 1407 |
+
"reward": 0.9609375,
|
| 1408 |
+
"reward_std": 0.3914227783679962,
|
| 1409 |
+
"rewards/reward_func/mean": 0.10677083333333333,
|
| 1410 |
+
"rewards/reward_func/std": 0.05869032442569733,
|
| 1411 |
+
"sampling/importance_sampling_ratio/max": 2.988154888153076,
|
| 1412 |
+
"sampling/importance_sampling_ratio/mean": 0.9597394466400146,
|
| 1413 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 1414 |
+
"sampling/sampling_logp_difference/max": 17.707983016967773,
|
| 1415 |
+
"sampling/sampling_logp_difference/mean": 0.18128016591072083,
|
| 1416 |
+
"seg/thinking_frac": 0.31788327544927597,
|
| 1417 |
+
"step": 32,
|
| 1418 |
+
"step_time": 83.28873698800453
|
| 1419 |
+
},
|
| 1420 |
+
{
|
| 1421 |
+
"clip_ratio/high_max": 0.0,
|
| 1422 |
+
"clip_ratio/high_mean": 0.0,
|
| 1423 |
+
"clip_ratio/low_mean": 0.0,
|
| 1424 |
+
"clip_ratio/low_min": 0.0,
|
| 1425 |
+
"clip_ratio/region_mean": 0.0,
|
| 1426 |
+
"completions/clipped_ratio": 0.0625,
|
| 1427 |
+
"completions/max_length": 4096.0,
|
| 1428 |
+
"completions/max_terminated_length": 3832.0,
|
| 1429 |
+
"completions/mean_length": 1276.671875,
|
| 1430 |
+
"completions/mean_terminated_length": 1147.416748046875,
|
| 1431 |
+
"completions/min_length": 218.0,
|
| 1432 |
+
"completions/min_terminated_length": 218.0,
|
| 1433 |
+
"degenerate_groups_filtered": 0.0,
|
| 1434 |
+
"entropy": 0.6430596709251404,
|
| 1435 |
+
"entropy/code": 0.46561597287654877,
|
| 1436 |
+
"entropy/thinking": 1.276490867137909,
|
| 1437 |
+
"epoch": 0.0812807881773399,
|
| 1438 |
+
"frac_reward_zero_std": 0.0,
|
| 1439 |
+
"grad_norm": 0.007146024236238484,
|
| 1440 |
+
"kl": 0.0008532363426638767,
|
| 1441 |
+
"kl/code_mean": 0.0006670264483545907,
|
| 1442 |
+
"kl/code_sum": 2.412841796875,
|
| 1443 |
+
"kl/thinking_mean": 0.001475480239605531,
|
| 1444 |
+
"kl/thinking_sum": 1.317626953125,
|
| 1445 |
+
"kl/total_loss_contribution": 1.7064726307580713e-05,
|
| 1446 |
+
"kl_mask/effective_frac": 1.0,
|
| 1447 |
+
"kl_mask/fallback_rate": 0.1875,
|
| 1448 |
+
"learning_rate": 4.942578632479721e-05,
|
| 1449 |
+
"loss": 0.016947615891695023,
|
| 1450 |
+
"num_tokens": 5060278.0,
|
| 1451 |
+
"reward": 1.1015625,
|
| 1452 |
+
"reward_std": 0.6435250639915466,
|
| 1453 |
+
"rewards/reward_func/mean": 0.12239583333333333,
|
| 1454 |
+
"rewards/reward_func/std": 0.14793591698010763,
|
| 1455 |
+
"sampling/importance_sampling_ratio/max": 2.9968602657318115,
|
| 1456 |
+
"sampling/importance_sampling_ratio/mean": 0.956814706325531,
|
| 1457 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 1458 |
+
"sampling/sampling_logp_difference/max": 14.568519592285156,
|
| 1459 |
+
"sampling/sampling_logp_difference/mean": 0.1785118281841278,
|
| 1460 |
+
"seg/thinking_frac": 0.22179431468248367,
|
| 1461 |
+
"step": 33,
|
| 1462 |
+
"step_time": 110.00213376397733
|
| 1463 |
+
},
|
| 1464 |
+
{
|
| 1465 |
+
"clip_ratio/high_max": 0.0,
|
| 1466 |
+
"clip_ratio/high_mean": 0.0,
|
| 1467 |
+
"clip_ratio/low_mean": 0.0,
|
| 1468 |
+
"clip_ratio/low_min": 0.0,
|
| 1469 |
+
"clip_ratio/region_mean": 0.0,
|
| 1470 |
+
"completions/clipped_ratio": 0.03125,
|
| 1471 |
+
"completions/max_length": 4096.0,
|
| 1472 |
+
"completions/max_terminated_length": 3753.0,
|
| 1473 |
+
"completions/mean_length": 1193.671875,
|
| 1474 |
+
"completions/mean_terminated_length": 1100.04833984375,
|
| 1475 |
+
"completions/min_length": 414.0,
|
| 1476 |
+
"completions/min_terminated_length": 414.0,
|
| 1477 |
+
"degenerate_groups_filtered": 1.0,
|
| 1478 |
+
"entropy": 0.6992127895355225,
|
| 1479 |
+
"entropy/code": 0.5307376012206078,
|
| 1480 |
+
"entropy/thinking": 1.2592804431915283,
|
| 1481 |
+
"epoch": 0.08374384236453201,
|
| 1482 |
+
"frac_reward_zero_std": 0.25,
|
| 1483 |
+
"grad_norm": 0.003208133608567953,
|
| 1484 |
+
"kl": 0.0011674772249534726,
|
| 1485 |
+
"kl/code_mean": 0.0009661648218752816,
|
| 1486 |
+
"kl/code_sum": 3.269775390625,
|
| 1487 |
+
"kl/thinking_mean": 0.0018231090507470071,
|
| 1488 |
+
"kl/thinking_sum": 1.64013671875,
|
| 1489 |
+
"kl/total_loss_contribution": 2.3349544335360406e-05,
|
| 1490 |
+
"kl_mask/effective_frac": 1.0,
|
| 1491 |
+
"kl_mask/fallback_rate": 0.0,
|
| 1492 |
+
"learning_rate": 4.9382643461956e-05,
|
| 1493 |
+
"loss": -0.02225431613624096,
|
| 1494 |
+
"num_tokens": 5218913.0,
|
| 1495 |
+
"reward": 1.046875,
|
| 1496 |
+
"reward_std": 0.4997519254684448,
|
| 1497 |
+
"rewards/reward_func/mean": 0.11631944444444445,
|
| 1498 |
+
"rewards/reward_func/std": 0.1087945087088479,
|
| 1499 |
+
"sampling/importance_sampling_ratio/max": 2.9990532398223877,
|
| 1500 |
+
"sampling/importance_sampling_ratio/mean": 0.9516765475273132,
|
| 1501 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 1502 |
+
"sampling/sampling_logp_difference/max": 14.686995506286621,
|
| 1503 |
+
"sampling/sampling_logp_difference/mean": 0.20307716727256775,
|
| 1504 |
+
"seg/thinking_frac": 0.23250583186745644,
|
| 1505 |
+
"step": 34,
|
| 1506 |
+
"step_time": 98.96413004602073
|
| 1507 |
+
},
|
| 1508 |
+
{
|
| 1509 |
+
"clip_ratio/high_max": 0.0,
|
| 1510 |
+
"clip_ratio/high_mean": 0.0,
|
| 1511 |
+
"clip_ratio/low_mean": 0.0,
|
| 1512 |
+
"clip_ratio/low_min": 0.0,
|
| 1513 |
+
"clip_ratio/region_mean": 0.0,
|
| 1514 |
+
"completions/clipped_ratio": 0.03125,
|
| 1515 |
+
"completions/max_length": 4096.0,
|
| 1516 |
+
"completions/max_terminated_length": 3784.0,
|
| 1517 |
+
"completions/mean_length": 1225.0,
|
| 1518 |
+
"completions/mean_terminated_length": 1132.3870849609375,
|
| 1519 |
+
"completions/min_length": 243.0,
|
| 1520 |
+
"completions/min_terminated_length": 243.0,
|
| 1521 |
+
"degenerate_groups_filtered": 0.0,
|
| 1522 |
+
"entropy": 0.6364438235759735,
|
| 1523 |
+
"entropy/code": 0.42579710483551025,
|
| 1524 |
+
"entropy/thinking": 1.2227859497070312,
|
| 1525 |
+
"epoch": 0.08620689655172414,
|
| 1526 |
+
"frac_reward_zero_std": 0.0,
|
| 1527 |
+
"grad_norm": 0.008652474030153076,
|
| 1528 |
+
"kl": 0.0011208790092496201,
|
| 1529 |
+
"kl/code_mean": 0.0007837592929718085,
|
| 1530 |
+
"kl/code_sum": 2.531982421875,
|
| 1531 |
+
"kl/thinking_mean": 0.0019486932724248618,
|
| 1532 |
+
"kl/thinking_sum": 1.9189453125,
|
| 1533 |
+
"kl/total_loss_contribution": 2.2417580112232827e-05,
|
| 1534 |
+
"kl_mask/effective_frac": 1.0,
|
| 1535 |
+
"kl_mask/fallback_rate": 0.0,
|
| 1536 |
+
"learning_rate": 4.933795824515431e-05,
|
| 1537 |
+
"loss": 0.01793116144835949,
|
| 1538 |
+
"num_tokens": 5371729.0,
|
| 1539 |
+
"reward": 1.140625,
|
| 1540 |
+
"reward_std": 0.6137883067131042,
|
| 1541 |
+
"rewards/reward_func/mean": 0.1267361111111111,
|
| 1542 |
+
"rewards/reward_func/std": 0.10891397297382355,
|
| 1543 |
+
"sampling/importance_sampling_ratio/max": 2.9971060752868652,
|
| 1544 |
+
"sampling/importance_sampling_ratio/mean": 0.9612535834312439,
|
| 1545 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 1546 |
+
"sampling/sampling_logp_difference/max": 16.282615661621094,
|
| 1547 |
+
"sampling/sampling_logp_difference/mean": 0.16757208108901978,
|
| 1548 |
+
"seg/thinking_frac": 0.275283794850111,
|
| 1549 |
+
"step": 35,
|
| 1550 |
+
"step_time": 99.7352547509945
|
| 1551 |
+
},
|
| 1552 |
+
{
|
| 1553 |
+
"clip_ratio/high_max": 0.0,
|
| 1554 |
+
"clip_ratio/high_mean": 0.0,
|
| 1555 |
+
"clip_ratio/low_mean": 0.0,
|
| 1556 |
+
"clip_ratio/low_min": 0.0,
|
| 1557 |
+
"clip_ratio/region_mean": 0.0,
|
| 1558 |
+
"completions/clipped_ratio": 0.078125,
|
| 1559 |
+
"completions/max_length": 4096.0,
|
| 1560 |
+
"completions/max_terminated_length": 3742.0,
|
| 1561 |
+
"completions/mean_length": 1003.171875,
|
| 1562 |
+
"completions/mean_terminated_length": 741.0678100585938,
|
| 1563 |
+
"completions/min_length": 75.0,
|
| 1564 |
+
"completions/min_terminated_length": 75.0,
|
| 1565 |
+
"degenerate_groups_filtered": 0.0,
|
| 1566 |
+
"entropy": 0.772294670343399,
|
| 1567 |
+
"entropy/code": 0.5195827931165695,
|
| 1568 |
+
"entropy/thinking": 1.1510973572731018,
|
| 1569 |
+
"epoch": 0.08866995073891626,
|
| 1570 |
+
"frac_reward_zero_std": 0.0,
|
| 1571 |
+
"grad_norm": 0.004607290709564949,
|
| 1572 |
+
"kl": 0.002750989282503724,
|
| 1573 |
+
"kl/code_mean": 0.002863584115402773,
|
| 1574 |
+
"kl/code_sum": 4.703125,
|
| 1575 |
+
"kl/thinking_mean": 0.002290709118824452,
|
| 1576 |
+
"kl/thinking_sum": 1.541259765625,
|
| 1577 |
+
"kl/total_loss_contribution": 5.5019785577314906e-05,
|
| 1578 |
+
"kl_mask/effective_frac": 1.0,
|
| 1579 |
+
"kl_mask/fallback_rate": 0.3125,
|
| 1580 |
+
"learning_rate": 4.929173350101025e-05,
|
| 1581 |
+
"loss": -0.019371718168258667,
|
| 1582 |
+
"num_tokens": 5521068.0,
|
| 1583 |
+
"reward": 0.8828125,
|
| 1584 |
+
"reward_std": 0.4008145332336426,
|
| 1585 |
+
"rewards/reward_func/mean": 0.09809027777777778,
|
| 1586 |
+
"rewards/reward_func/std": 0.058744521604643926,
|
| 1587 |
+
"sampling/importance_sampling_ratio/max": 2.999795436859131,
|
| 1588 |
+
"sampling/importance_sampling_ratio/mean": 0.9497913122177124,
|
| 1589 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 1590 |
+
"sampling/sampling_logp_difference/max": 18.12101936340332,
|
| 1591 |
+
"sampling/sampling_logp_difference/mean": 0.2026529312133789,
|
| 1592 |
+
"seg/thinking_frac": 0.3710649535059929,
|
| 1593 |
+
"step": 36,
|
| 1594 |
+
"step_time": 107.20923037303146
|
| 1595 |
+
},
|
| 1596 |
+
{
|
| 1597 |
+
"clip_ratio/high_max": 0.0,
|
| 1598 |
+
"clip_ratio/high_mean": 0.0,
|
| 1599 |
+
"clip_ratio/low_mean": 0.0,
|
| 1600 |
+
"clip_ratio/low_min": 0.0,
|
| 1601 |
+
"clip_ratio/region_mean": 0.0,
|
| 1602 |
+
"completions/clipped_ratio": 0.015625,
|
| 1603 |
+
"completions/max_length": 4096.0,
|
| 1604 |
+
"completions/max_terminated_length": 3582.0,
|
| 1605 |
+
"completions/mean_length": 1005.8125,
|
| 1606 |
+
"completions/mean_terminated_length": 956.761962890625,
|
| 1607 |
+
"completions/min_length": 262.0,
|
| 1608 |
+
"completions/min_terminated_length": 262.0,
|
| 1609 |
+
"degenerate_groups_filtered": 0.0,
|
| 1610 |
+
"entropy": 0.6745053678750992,
|
| 1611 |
+
"entropy/code": 0.4715415760874748,
|
| 1612 |
+
"entropy/thinking": 1.1394010484218597,
|
| 1613 |
+
"epoch": 0.09113300492610837,
|
| 1614 |
+
"frac_reward_zero_std": 0.0,
|
| 1615 |
+
"grad_norm": 0.011330490203574104,
|
| 1616 |
+
"kl": 0.0022949971025809646,
|
| 1617 |
+
"kl/code_mean": 0.0021878597763134167,
|
| 1618 |
+
"kl/code_sum": 7.55615234375,
|
| 1619 |
+
"kl/thinking_mean": 0.0023160133277997375,
|
| 1620 |
+
"kl/thinking_sum": 2.461669921875,
|
| 1621 |
+
"kl/total_loss_contribution": 4.589994068737724e-05,
|
| 1622 |
+
"kl_mask/effective_frac": 1.0,
|
| 1623 |
+
"kl_mask/fallback_rate": 0.0625,
|
| 1624 |
+
"learning_rate": 4.924397215352658e-05,
|
| 1625 |
+
"loss": -0.04807479679584503,
|
| 1626 |
+
"num_tokens": 5666368.0,
|
| 1627 |
+
"reward": 1.109375,
|
| 1628 |
+
"reward_std": 0.7504958510398865,
|
| 1629 |
+
"rewards/reward_func/mean": 0.1232638888888889,
|
| 1630 |
+
"rewards/reward_func/std": 0.12009341849221124,
|
| 1631 |
+
"sampling/importance_sampling_ratio/max": 2.997396945953369,
|
| 1632 |
+
"sampling/importance_sampling_ratio/mean": 0.9514970779418945,
|
| 1633 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 1634 |
+
"sampling/sampling_logp_difference/max": 14.73610782623291,
|
| 1635 |
+
"sampling/sampling_logp_difference/mean": 0.19920861721038818,
|
| 1636 |
+
"seg/thinking_frac": 0.30746858566999435,
|
| 1637 |
+
"step": 37,
|
| 1638 |
+
"step_time": 95.67100700404262
|
| 1639 |
+
},
|
| 1640 |
+
{
|
| 1641 |
+
"clip_ratio/high_max": 0.0,
|
| 1642 |
+
"clip_ratio/high_mean": 0.0,
|
| 1643 |
+
"clip_ratio/low_mean": 0.0,
|
| 1644 |
+
"clip_ratio/low_min": 0.0,
|
| 1645 |
+
"clip_ratio/region_mean": 0.0,
|
| 1646 |
+
"completions/clipped_ratio": 0.046875,
|
| 1647 |
+
"completions/max_length": 4096.0,
|
| 1648 |
+
"completions/max_terminated_length": 3641.0,
|
| 1649 |
+
"completions/mean_length": 760.890625,
|
| 1650 |
+
"completions/mean_terminated_length": 727.5901489257812,
|
| 1651 |
+
"completions/min_length": 1.0,
|
| 1652 |
+
"completions/min_terminated_length": 209.0,
|
| 1653 |
+
"degenerate_groups_filtered": 0.0,
|
| 1654 |
+
"entropy": 0.7481428235769272,
|
| 1655 |
+
"entropy/code": 0.4826525151729584,
|
| 1656 |
+
"entropy/thinking": 1.211728423833847,
|
| 1657 |
+
"epoch": 0.09359605911330049,
|
| 1658 |
+
"frac_reward_zero_std": 0.0,
|
| 1659 |
+
"grad_norm": 0.0024050915089118274,
|
| 1660 |
+
"kl": 0.001688124582869932,
|
| 1661 |
+
"kl/code_mean": 0.0009344722639070824,
|
| 1662 |
+
"kl/code_sum": 1.799560546875,
|
| 1663 |
+
"kl/thinking_mean": 0.0029591874917969108,
|
| 1664 |
+
"kl/thinking_sum": 2.53515625,
|
| 1665 |
+
"kl/total_loss_contribution": 3.3762489692890085e-05,
|
| 1666 |
+
"kl_mask/effective_frac": 1.0,
|
| 1667 |
+
"kl_mask/fallback_rate": 0.0625,
|
| 1668 |
+
"learning_rate": 4.91946772239058e-05,
|
| 1669 |
+
"loss": -0.048725854605436325,
|
| 1670 |
+
"num_tokens": 5786841.0,
|
| 1671 |
+
"reward": 0.953125,
|
| 1672 |
+
"reward_std": 0.301698237657547,
|
| 1673 |
+
"rewards/reward_func/mean": 0.10590277777777778,
|
| 1674 |
+
"rewards/reward_func/std": 0.045615961154301964,
|
| 1675 |
+
"sampling/importance_sampling_ratio/max": 2.993288993835449,
|
| 1676 |
+
"sampling/importance_sampling_ratio/mean": 0.9611358642578125,
|
| 1677 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 1678 |
+
"sampling/sampling_logp_difference/max": 10.59887981414795,
|
| 1679 |
+
"sampling/sampling_logp_difference/mean": 0.18745650351047516,
|
| 1680 |
+
"seg/thinking_frac": 0.373220257461071,
|
| 1681 |
+
"step": 38,
|
| 1682 |
+
"step_time": 103.41730736198951
|
| 1683 |
+
},
|
| 1684 |
+
{
|
| 1685 |
+
"clip_ratio/high_max": 0.0,
|
| 1686 |
+
"clip_ratio/high_mean": 0.0,
|
| 1687 |
+
"clip_ratio/low_mean": 0.0,
|
| 1688 |
+
"clip_ratio/low_min": 0.0,
|
| 1689 |
+
"clip_ratio/region_mean": 0.0,
|
| 1690 |
+
"completions/clipped_ratio": 0.125,
|
| 1691 |
+
"completions/max_length": 4096.0,
|
| 1692 |
+
"completions/max_terminated_length": 4089.0,
|
| 1693 |
+
"completions/mean_length": 1673.765625,
|
| 1694 |
+
"completions/mean_terminated_length": 1377.0001220703125,
|
| 1695 |
+
"completions/min_length": 213.0,
|
| 1696 |
+
"completions/min_terminated_length": 213.0,
|
| 1697 |
+
"degenerate_groups_filtered": 0.0,
|
| 1698 |
+
"entropy": 0.7384326457977295,
|
| 1699 |
+
"entropy/code": 0.5060838386416435,
|
| 1700 |
+
"entropy/thinking": 1.2186180353164673,
|
| 1701 |
+
"epoch": 0.0960591133004926,
|
| 1702 |
+
"frac_reward_zero_std": 0.0,
|
| 1703 |
+
"grad_norm": 0.001178840832890337,
|
| 1704 |
+
"kl": 0.0009198132465826347,
|
| 1705 |
+
"kl/code_mean": 0.0005843743201694451,
|
| 1706 |
+
"kl/code_sum": 1.921142578125,
|
| 1707 |
+
"kl/thinking_mean": 0.0017454991466365755,
|
| 1708 |
+
"kl/thinking_sum": 2.378173828125,
|
| 1709 |
+
"kl/total_loss_contribution": 1.8396264749753755e-05,
|
| 1710 |
+
"kl_mask/effective_frac": 1.0,
|
| 1711 |
+
"kl_mask/fallback_rate": 0.1875,
|
| 1712 |
+
"learning_rate": 4.9143851830359044e-05,
|
| 1713 |
+
"loss": -0.008239601738750935,
|
| 1714 |
+
"num_tokens": 5968218.0,
|
| 1715 |
+
"reward": 1.0390625,
|
| 1716 |
+
"reward_std": 0.3513543903827667,
|
| 1717 |
+
"rewards/reward_func/mean": 0.1154513888888889,
|
| 1718 |
+
"rewards/reward_func/std": 0.05332859853903452,
|
| 1719 |
+
"sampling/importance_sampling_ratio/max": 2.9985454082489014,
|
| 1720 |
+
"sampling/importance_sampling_ratio/mean": 0.9483808279037476,
|
| 1721 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 1722 |
+
"sampling/sampling_logp_difference/max": 14.2007417678833,
|
| 1723 |
+
"sampling/sampling_logp_difference/mean": 0.2109224498271942,
|
| 1724 |
+
"seg/thinking_frac": 0.3298206031322479,
|
| 1725 |
+
"step": 39,
|
| 1726 |
+
"step_time": 109.65771615295671
|
| 1727 |
+
},
|
| 1728 |
+
{
|
| 1729 |
+
"clip_ratio/high_max": 0.0,
|
| 1730 |
+
"clip_ratio/high_mean": 0.0,
|
| 1731 |
+
"clip_ratio/low_mean": 0.0,
|
| 1732 |
+
"clip_ratio/low_min": 0.0,
|
| 1733 |
+
"clip_ratio/region_mean": 0.0,
|
| 1734 |
+
"completions/clipped_ratio": 0.03125,
|
| 1735 |
+
"completions/max_length": 4096.0,
|
| 1736 |
+
"completions/max_terminated_length": 3348.0,
|
| 1737 |
+
"completions/mean_length": 900.703125,
|
| 1738 |
+
"completions/mean_terminated_length": 797.6290283203125,
|
| 1739 |
+
"completions/min_length": 285.0,
|
| 1740 |
+
"completions/min_terminated_length": 285.0,
|
| 1741 |
+
"degenerate_groups_filtered": 0.0,
|
| 1742 |
+
"entropy": 0.7858085781335831,
|
| 1743 |
+
"entropy/code": 0.5602374970912933,
|
| 1744 |
+
"entropy/thinking": 1.2263537347316742,
|
| 1745 |
+
"epoch": 0.09852216748768473,
|
| 1746 |
+
"frac_reward_zero_std": 0.25,
|
| 1747 |
+
"grad_norm": 0.005010977073307075,
|
| 1748 |
+
"kl": 0.0025712139322422445,
|
| 1749 |
+
"kl/code_mean": 0.00229311446310021,
|
| 1750 |
+
"kl/code_sum": 5.88916015625,
|
| 1751 |
+
"kl/thinking_mean": 0.0025739773991517723,
|
| 1752 |
+
"kl/thinking_sum": 2.38818359375,
|
| 1753 |
+
"kl/total_loss_contribution": 5.1424278353806585e-05,
|
| 1754 |
+
"kl_mask/effective_frac": 1.0,
|
| 1755 |
+
"kl_mask/fallback_rate": 0.125,
|
| 1756 |
+
"learning_rate": 4.9091499187908784e-05,
|
| 1757 |
+
"loss": 0.009397903457283974,
|
| 1758 |
+
"num_tokens": 6099655.0,
|
| 1759 |
+
"reward": 1.125,
|
| 1760 |
+
"reward_std": 0.4586038291454315,
|
| 1761 |
+
"rewards/reward_func/mean": 0.125,
|
| 1762 |
+
"rewards/reward_func/std": 0.06795731021298303,
|
| 1763 |
+
"sampling/importance_sampling_ratio/max": 2.9956657886505127,
|
| 1764 |
+
"sampling/importance_sampling_ratio/mean": 0.9564260244369507,
|
| 1765 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 1766 |
+
"sampling/sampling_logp_difference/max": 16.499637603759766,
|
| 1767 |
+
"sampling/sampling_logp_difference/mean": 0.20195510983467102,
|
| 1768 |
+
"seg/thinking_frac": 0.3272581100463867,
|
| 1769 |
+
"step": 40,
|
| 1770 |
+
"step_time": 95.523574277031
|
| 1771 |
+
},
|
| 1772 |
+
{
|
| 1773 |
+
"clip_ratio/high_max": 0.0,
|
| 1774 |
+
"clip_ratio/high_mean": 0.0,
|
| 1775 |
+
"clip_ratio/low_mean": 0.0,
|
| 1776 |
+
"clip_ratio/low_min": 0.0,
|
| 1777 |
+
"clip_ratio/region_mean": 0.0,
|
| 1778 |
+
"completions/clipped_ratio": 0.078125,
|
| 1779 |
+
"completions/max_length": 4096.0,
|
| 1780 |
+
"completions/max_terminated_length": 2368.0,
|
| 1781 |
+
"completions/mean_length": 1074.375,
|
| 1782 |
+
"completions/mean_terminated_length": 818.3051147460938,
|
| 1783 |
+
"completions/min_length": 147.0,
|
| 1784 |
+
"completions/min_terminated_length": 147.0,
|
| 1785 |
+
"degenerate_groups_filtered": 0.0,
|
| 1786 |
+
"entropy": 0.66144859790802,
|
| 1787 |
+
"entropy/code": 0.4168987274169922,
|
| 1788 |
+
"entropy/thinking": 1.1096002161502838,
|
| 1789 |
+
"epoch": 0.10098522167487685,
|
| 1790 |
+
"frac_reward_zero_std": 0.0,
|
| 1791 |
+
"grad_norm": 0.0010666028423866633,
|
| 1792 |
+
"kl": 0.0013391458196565509,
|
| 1793 |
+
"kl/code_mean": 0.0006703727121930569,
|
| 1794 |
+
"kl/code_sum": 1.4189453125,
|
| 1795 |
+
"kl/thinking_mean": 0.002626303758006543,
|
| 1796 |
+
"kl/thinking_sum": 2.52783203125,
|
| 1797 |
+
"kl/total_loss_contribution": 2.6782916393131018e-05,
|
| 1798 |
+
"kl_mask/effective_frac": 1.0,
|
| 1799 |
+
"kl_mask/fallback_rate": 0.0,
|
| 1800 |
+
"learning_rate": 4.903762260818551e-05,
|
| 1801 |
+
"loss": -0.03096545860171318,
|
| 1802 |
+
"num_tokens": 6235119.0,
|
| 1803 |
+
"reward": 1.03515625,
|
| 1804 |
+
"reward_std": 0.32692405581474304,
|
| 1805 |
+
"rewards/reward_func/mean": 0.1150173611111111,
|
| 1806 |
+
"rewards/reward_func/std": 0.04827603366639879,
|
| 1807 |
+
"sampling/importance_sampling_ratio/max": 2.9969332218170166,
|
| 1808 |
+
"sampling/importance_sampling_ratio/mean": 0.9614675045013428,
|
| 1809 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 1810 |
+
"sampling/sampling_logp_difference/max": 16.749778747558594,
|
| 1811 |
+
"sampling/sampling_logp_difference/mean": 0.18035539984703064,
|
| 1812 |
+
"seg/thinking_frac": 0.36044415831565857,
|
| 1813 |
+
"step": 41,
|
| 1814 |
+
"step_time": 97.57413268997334
|
| 1815 |
+
},
|
| 1816 |
+
{
|
| 1817 |
+
"clip_ratio/high_max": 0.0,
|
| 1818 |
+
"clip_ratio/high_mean": 0.0,
|
| 1819 |
+
"clip_ratio/low_mean": 0.0,
|
| 1820 |
+
"clip_ratio/low_min": 0.0,
|
| 1821 |
+
"clip_ratio/region_mean": 0.0,
|
| 1822 |
+
"completions/clipped_ratio": 0.046875,
|
| 1823 |
+
"completions/max_length": 4096.0,
|
| 1824 |
+
"completions/max_terminated_length": 3899.0,
|
| 1825 |
+
"completions/mean_length": 965.5,
|
| 1826 |
+
"completions/mean_terminated_length": 835.7048950195312,
|
| 1827 |
+
"completions/min_length": 118.0,
|
| 1828 |
+
"completions/min_terminated_length": 118.0,
|
| 1829 |
+
"degenerate_groups_filtered": 1.0,
|
| 1830 |
+
"entropy": 0.737716794013977,
|
| 1831 |
+
"entropy/code": 0.4643196240067482,
|
| 1832 |
+
"entropy/thinking": 1.1552053391933441,
|
| 1833 |
+
"epoch": 0.10344827586206896,
|
| 1834 |
+
"frac_reward_zero_std": 0.5,
|
| 1835 |
+
"grad_norm": 0.000791673132340617,
|
| 1836 |
+
"kl": 0.0013449728430714458,
|
| 1837 |
+
"kl/code_mean": 0.000663517217617482,
|
| 1838 |
+
"kl/code_sum": 1.628173828125,
|
| 1839 |
+
"kl/thinking_mean": 0.002538877772167325,
|
| 1840 |
+
"kl/thinking_sum": 2.2998046875,
|
| 1841 |
+
"kl/total_loss_contribution": 2.6899457225226797e-05,
|
| 1842 |
+
"kl_mask/effective_frac": 1.0,
|
| 1843 |
+
"kl_mask/fallback_rate": 0.0,
|
| 1844 |
+
"learning_rate": 4.8982225499218236e-05,
|
| 1845 |
+
"loss": -0.01895085535943508,
|
| 1846 |
+
"num_tokens": 6371711.0,
|
| 1847 |
+
"reward": 1.04296875,
|
| 1848 |
+
"reward_std": 0.2801455557346344,
|
| 1849 |
+
"rewards/reward_func/mean": 0.11588541666666667,
|
| 1850 |
+
"rewards/reward_func/std": 0.04137110710144043,
|
| 1851 |
+
"sampling/importance_sampling_ratio/max": 2.997574806213379,
|
| 1852 |
+
"sampling/importance_sampling_ratio/mean": 0.9584183096885681,
|
| 1853 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 1854 |
+
"sampling/sampling_logp_difference/max": 12.83202075958252,
|
| 1855 |
+
"sampling/sampling_logp_difference/mean": 0.19598975777626038,
|
| 1856 |
+
"seg/thinking_frac": 0.39801809191703796,
|
| 1857 |
+
"step": 42,
|
| 1858 |
+
"step_time": 102.39049553798395
|
| 1859 |
+
},
|
| 1860 |
+
{
|
| 1861 |
+
"clip_ratio/high_max": 0.0,
|
| 1862 |
+
"clip_ratio/high_mean": 0.0,
|
| 1863 |
+
"clip_ratio/low_mean": 0.0,
|
| 1864 |
+
"clip_ratio/low_min": 0.0,
|
| 1865 |
+
"clip_ratio/region_mean": 0.0,
|
| 1866 |
+
"completions/clipped_ratio": 0.109375,
|
| 1867 |
+
"completions/max_length": 4096.0,
|
| 1868 |
+
"completions/max_terminated_length": 3971.0,
|
| 1869 |
+
"completions/mean_length": 1370.34375,
|
| 1870 |
+
"completions/mean_terminated_length": 1088.5789794921875,
|
| 1871 |
+
"completions/min_length": 232.0,
|
| 1872 |
+
"completions/min_terminated_length": 232.0,
|
| 1873 |
+
"degenerate_groups_filtered": 0.0,
|
| 1874 |
+
"entropy": 0.6188477873802185,
|
| 1875 |
+
"entropy/code": 0.4291503205895424,
|
| 1876 |
+
"entropy/thinking": 1.114559531211853,
|
| 1877 |
+
"epoch": 0.10591133004926108,
|
| 1878 |
+
"frac_reward_zero_std": 0.25,
|
| 1879 |
+
"grad_norm": 0.0021248442768718057,
|
| 1880 |
+
"kl": 0.002251649391837418,
|
| 1881 |
+
"kl/code_mean": 0.002252957478049211,
|
| 1882 |
+
"kl/code_sum": 6.204833984375,
|
| 1883 |
+
"kl/thinking_mean": 0.0026337408926337957,
|
| 1884 |
+
"kl/thinking_sum": 2.3251953125,
|
| 1885 |
+
"kl/total_loss_contribution": 4.5032985781290336e-05,
|
| 1886 |
+
"kl_mask/effective_frac": 1.0,
|
| 1887 |
+
"kl_mask/fallback_rate": 0.25,
|
| 1888 |
+
"learning_rate": 4.892531136521889e-05,
|
| 1889 |
+
"loss": -0.006458385847508907,
|
| 1890 |
+
"num_tokens": 6530117.0,
|
| 1891 |
+
"reward": 1.0390625,
|
| 1892 |
+
"reward_std": 0.35972529649734497,
|
| 1893 |
+
"rewards/reward_func/mean": 0.1154513888888889,
|
| 1894 |
+
"rewards/reward_func/std": 0.055034501685036555,
|
| 1895 |
+
"sampling/importance_sampling_ratio/max": 2.999521017074585,
|
| 1896 |
+
"sampling/importance_sampling_ratio/mean": 0.9552509784698486,
|
| 1897 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 1898 |
+
"sampling/sampling_logp_difference/max": 14.135897636413574,
|
| 1899 |
+
"sampling/sampling_logp_difference/mean": 0.18394148349761963,
|
| 1900 |
+
"seg/thinking_frac": 0.2860334441065788,
|
| 1901 |
+
"step": 43,
|
| 1902 |
+
"step_time": 102.05549815899576
|
| 1903 |
+
},
|
| 1904 |
+
{
|
| 1905 |
+
"clip_ratio/high_max": 0.0,
|
| 1906 |
+
"clip_ratio/high_mean": 0.0,
|
| 1907 |
+
"clip_ratio/low_mean": 0.0,
|
| 1908 |
+
"clip_ratio/low_min": 0.0,
|
| 1909 |
+
"clip_ratio/region_mean": 0.0,
|
| 1910 |
+
"completions/clipped_ratio": 0.015625,
|
| 1911 |
+
"completions/max_length": 4096.0,
|
| 1912 |
+
"completions/max_terminated_length": 3881.0,
|
| 1913 |
+
"completions/mean_length": 994.5,
|
| 1914 |
+
"completions/mean_terminated_length": 945.2698974609375,
|
| 1915 |
+
"completions/min_length": 179.0,
|
| 1916 |
+
"completions/min_terminated_length": 179.0,
|
| 1917 |
+
"degenerate_groups_filtered": 1.0,
|
| 1918 |
+
"entropy": 0.7200822979211807,
|
| 1919 |
+
"entropy/code": 0.5478549301624298,
|
| 1920 |
+
"entropy/thinking": 1.1399975419044495,
|
| 1921 |
+
"epoch": 0.10837438423645321,
|
| 1922 |
+
"frac_reward_zero_std": 0.5,
|
| 1923 |
+
"grad_norm": 0.01204892174396934,
|
| 1924 |
+
"kl": 0.001926325203385204,
|
| 1925 |
+
"kl/code_mean": 0.0015991431137081236,
|
| 1926 |
+
"kl/code_sum": 4.4765625,
|
| 1927 |
+
"kl/thinking_mean": 0.0027728145360015333,
|
| 1928 |
+
"kl/thinking_sum": 2.69921875,
|
| 1929 |
+
"kl/total_loss_contribution": 3.852650297631044e-05,
|
| 1930 |
+
"kl_mask/effective_frac": 1.0,
|
| 1931 |
+
"kl_mask/fallback_rate": 0.0,
|
| 1932 |
+
"learning_rate": 4.88668838063607e-05,
|
| 1933 |
+
"loss": -0.03742249682545662,
|
| 1934 |
+
"num_tokens": 6667221.0,
|
| 1935 |
+
"reward": 1.203125,
|
| 1936 |
+
"reward_std": 0.8543481230735779,
|
| 1937 |
+
"rewards/reward_func/mean": 0.13368055555555555,
|
| 1938 |
+
"rewards/reward_func/std": 0.15026381446255577,
|
| 1939 |
+
"sampling/importance_sampling_ratio/max": 2.9996910095214844,
|
| 1940 |
+
"sampling/importance_sampling_ratio/mean": 0.950646162033081,
|
| 1941 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 1942 |
+
"sampling/sampling_logp_difference/max": 10.937347412109375,
|
| 1943 |
+
"sampling/sampling_logp_difference/mean": 0.20487606525421143,
|
| 1944 |
+
"seg/thinking_frac": 0.29649974405765533,
|
| 1945 |
+
"step": 44,
|
| 1946 |
+
"step_time": 95.65633125195745
|
| 1947 |
+
},
|
| 1948 |
+
{
|
| 1949 |
+
"clip_ratio/high_max": 0.0,
|
| 1950 |
+
"clip_ratio/high_mean": 0.0,
|
| 1951 |
+
"clip_ratio/low_mean": 0.0,
|
| 1952 |
+
"clip_ratio/low_min": 0.0,
|
| 1953 |
+
"clip_ratio/region_mean": 0.0,
|
| 1954 |
+
"completions/clipped_ratio": 0.046875,
|
| 1955 |
+
"completions/max_length": 4096.0,
|
| 1956 |
+
"completions/max_terminated_length": 3876.0,
|
| 1957 |
+
"completions/mean_length": 972.171875,
|
| 1958 |
+
"completions/mean_terminated_length": 870.6065063476562,
|
| 1959 |
+
"completions/min_length": 141.0,
|
| 1960 |
+
"completions/min_terminated_length": 141.0,
|
| 1961 |
+
"degenerate_groups_filtered": 0.0,
|
| 1962 |
+
"entropy": 0.6463710367679596,
|
| 1963 |
+
"entropy/code": 0.42478586733341217,
|
| 1964 |
+
"entropy/thinking": 1.0746629536151886,
|
| 1965 |
+
"epoch": 0.11083743842364532,
|
| 1966 |
+
"frac_reward_zero_std": 0.25,
|
| 1967 |
+
"grad_norm": 0.001943319993295526,
|
| 1968 |
+
"kl": 0.00394561214488931,
|
| 1969 |
+
"kl/code_mean": 0.004201330084470101,
|
| 1970 |
+
"kl/code_sum": 9.673095703125,
|
| 1971 |
+
"kl/thinking_mean": 0.0033420651452615857,
|
| 1972 |
+
"kl/thinking_sum": 2.7900390625,
|
| 1973 |
+
"kl/total_loss_contribution": 7.891224231570959e-05,
|
| 1974 |
+
"kl_mask/effective_frac": 1.0,
|
| 1975 |
+
"kl_mask/fallback_rate": 0.125,
|
| 1976 |
+
"learning_rate": 4.880694651855042e-05,
|
| 1977 |
+
"loss": -0.014538422226905823,
|
| 1978 |
+
"num_tokens": 6798848.0,
|
| 1979 |
+
"reward": 1.0,
|
| 1980 |
+
"reward_std": 0.28867512941360474,
|
| 1981 |
+
"rewards/reward_func/mean": 0.1111111111111111,
|
| 1982 |
+
"rewards/reward_func/std": 0.04400258925226,
|
| 1983 |
+
"sampling/importance_sampling_ratio/max": 2.9976205825805664,
|
| 1984 |
+
"sampling/importance_sampling_ratio/mean": 0.9668350219726562,
|
| 1985 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 1986 |
+
"sampling/sampling_logp_difference/max": 11.104084968566895,
|
| 1987 |
+
"sampling/sampling_logp_difference/mean": 0.16535790264606476,
|
| 1988 |
+
"seg/thinking_frac": 0.352353997528553,
|
| 1989 |
+
"step": 45,
|
| 1990 |
+
"step_time": 104.77708191899
|
| 1991 |
+
},
|
| 1992 |
+
{
|
| 1993 |
+
"clip_ratio/high_max": 0.0,
|
| 1994 |
+
"clip_ratio/high_mean": 0.0,
|
| 1995 |
+
"clip_ratio/low_mean": 0.0,
|
| 1996 |
+
"clip_ratio/low_min": 0.0,
|
| 1997 |
+
"clip_ratio/region_mean": 0.0,
|
| 1998 |
+
"completions/clipped_ratio": 0.09375,
|
| 1999 |
+
"completions/max_length": 4096.0,
|
| 2000 |
+
"completions/max_terminated_length": 2077.0,
|
| 2001 |
+
"completions/mean_length": 1209.734375,
|
| 2002 |
+
"completions/mean_terminated_length": 943.8965454101562,
|
| 2003 |
+
"completions/min_length": 169.0,
|
| 2004 |
+
"completions/min_terminated_length": 169.0,
|
| 2005 |
+
"degenerate_groups_filtered": 0.0,
|
| 2006 |
+
"entropy": 0.6656284928321838,
|
| 2007 |
+
"entropy/code": 0.415253646671772,
|
| 2008 |
+
"entropy/thinking": 1.2196472883224487,
|
| 2009 |
+
"epoch": 0.11330049261083744,
|
| 2010 |
+
"frac_reward_zero_std": 0.5,
|
| 2011 |
+
"grad_norm": 0.004176968470321685,
|
| 2012 |
+
"kl": 0.0015118622104637325,
|
| 2013 |
+
"kl/code_mean": 0.0007940270443214104,
|
| 2014 |
+
"kl/code_sum": 1.9609375,
|
| 2015 |
+
"kl/thinking_mean": 0.0032060034573078156,
|
| 2016 |
+
"kl/thinking_sum": 3.25537109375,
|
| 2017 |
+
"kl/total_loss_contribution": 3.0237242754083127e-05,
|
| 2018 |
+
"kl_mask/effective_frac": 1.0,
|
| 2019 |
+
"kl_mask/fallback_rate": 0.125,
|
| 2020 |
+
"learning_rate": 4.874550329319457e-05,
|
| 2021 |
+
"loss": 0.002406604588031769,
|
| 2022 |
+
"num_tokens": 6957615.0,
|
| 2023 |
+
"reward": 1.203125,
|
| 2024 |
+
"reward_std": 0.6422230005264282,
|
| 2025 |
+
"rewards/reward_func/mean": 0.13368055555555555,
|
| 2026 |
+
"rewards/reward_func/std": 0.10287589993741778,
|
| 2027 |
+
"sampling/importance_sampling_ratio/max": 2.9969844818115234,
|
| 2028 |
+
"sampling/importance_sampling_ratio/mean": 0.9560835361480713,
|
| 2029 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 2030 |
+
"sampling/sampling_logp_difference/max": 13.374930381774902,
|
| 2031 |
+
"sampling/sampling_logp_difference/mean": 0.1940203607082367,
|
| 2032 |
+
"seg/thinking_frac": 0.3170151486992836,
|
| 2033 |
+
"step": 46,
|
| 2034 |
+
"step_time": 104.4089871479664
|
| 2035 |
+
},
|
| 2036 |
+
{
|
| 2037 |
+
"clip_ratio/high_max": 0.0,
|
| 2038 |
+
"clip_ratio/high_mean": 0.0,
|
| 2039 |
+
"clip_ratio/low_mean": 0.0,
|
| 2040 |
+
"clip_ratio/low_min": 0.0,
|
| 2041 |
+
"clip_ratio/region_mean": 0.0,
|
| 2042 |
+
"completions/clipped_ratio": 0.046875,
|
| 2043 |
+
"completions/max_length": 4096.0,
|
| 2044 |
+
"completions/max_terminated_length": 3470.0,
|
| 2045 |
+
"completions/mean_length": 1165.484375,
|
| 2046 |
+
"completions/mean_terminated_length": 1021.360595703125,
|
| 2047 |
+
"completions/min_length": 123.0,
|
| 2048 |
+
"completions/min_terminated_length": 123.0,
|
| 2049 |
+
"degenerate_groups_filtered": 0.0,
|
| 2050 |
+
"entropy": 0.7201135754585266,
|
| 2051 |
+
"entropy/code": 0.5237086266279221,
|
| 2052 |
+
"entropy/thinking": 1.1887874007225037,
|
| 2053 |
+
"epoch": 0.11576354679802955,
|
| 2054 |
+
"frac_reward_zero_std": 0.0,
|
| 2055 |
+
"grad_norm": 0.0022398384269132483,
|
| 2056 |
+
"kl": 0.011335016140947118,
|
| 2057 |
+
"kl/code_mean": 0.017846462986199185,
|
| 2058 |
+
"kl/code_sum": 24.498779296875,
|
| 2059 |
+
"kl/thinking_mean": 0.0037020700401626527,
|
| 2060 |
+
"kl/thinking_sum": 3.77587890625,
|
| 2061 |
+
"kl/total_loss_contribution": 0.0002267002982989652,
|
| 2062 |
+
"kl_mask/effective_frac": 1.0,
|
| 2063 |
+
"kl_mask/fallback_rate": 0.125,
|
| 2064 |
+
"learning_rate": 4.868255801695958e-05,
|
| 2065 |
+
"loss": -0.02131621539592743,
|
| 2066 |
+
"num_tokens": 7122974.0,
|
| 2067 |
+
"reward": 1.1015625,
|
| 2068 |
+
"reward_std": 0.4380665123462677,
|
| 2069 |
+
"rewards/reward_func/mean": 0.12239583333333333,
|
| 2070 |
+
"rewards/reward_func/std": 0.09901134835349189,
|
| 2071 |
+
"sampling/importance_sampling_ratio/max": 2.997438907623291,
|
| 2072 |
+
"sampling/importance_sampling_ratio/mean": 0.9448568224906921,
|
| 2073 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 2074 |
+
"sampling/sampling_logp_difference/max": 14.779375076293945,
|
| 2075 |
+
"sampling/sampling_logp_difference/mean": 0.22430233657360077,
|
| 2076 |
+
"seg/thinking_frac": 0.30096931755542755,
|
| 2077 |
+
"step": 47,
|
| 2078 |
+
"step_time": 104.02336804801598
|
| 2079 |
+
},
|
| 2080 |
+
{
|
| 2081 |
+
"clip_ratio/high_max": 0.0,
|
| 2082 |
+
"clip_ratio/high_mean": 0.0,
|
| 2083 |
+
"clip_ratio/low_mean": 0.0,
|
| 2084 |
+
"clip_ratio/low_min": 0.0,
|
| 2085 |
+
"clip_ratio/region_mean": 0.0,
|
| 2086 |
+
"completions/clipped_ratio": 0.078125,
|
| 2087 |
+
"completions/max_length": 4096.0,
|
| 2088 |
+
"completions/max_terminated_length": 3476.0,
|
| 2089 |
+
"completions/mean_length": 1334.65625,
|
| 2090 |
+
"completions/mean_terminated_length": 1119.83056640625,
|
| 2091 |
+
"completions/min_length": 217.0,
|
| 2092 |
+
"completions/min_terminated_length": 217.0,
|
| 2093 |
+
"degenerate_groups_filtered": 0.0,
|
| 2094 |
+
"entropy": 0.7386683970689774,
|
| 2095 |
+
"entropy/code": 0.5181600898504257,
|
| 2096 |
+
"entropy/thinking": 1.1625761687755585,
|
| 2097 |
+
"epoch": 0.11822660098522167,
|
| 2098 |
+
"frac_reward_zero_std": 0.0,
|
| 2099 |
+
"grad_norm": 0.006840694943128053,
|
| 2100 |
+
"kl": 0.0018921646988019347,
|
| 2101 |
+
"kl/code_mean": 0.0011455350613687187,
|
| 2102 |
+
"kl/code_sum": 3.187744140625,
|
| 2103 |
+
"kl/thinking_mean": 0.0037745186127722263,
|
| 2104 |
+
"kl/thinking_sum": 4.132080078125,
|
| 2105 |
+
"kl/total_loss_contribution": 3.784329237532802e-05,
|
| 2106 |
+
"kl_mask/effective_frac": 1.0,
|
| 2107 |
+
"kl_mask/fallback_rate": 0.0,
|
| 2108 |
+
"learning_rate": 4.861811467152596e-05,
|
| 2109 |
+
"loss": -0.004604513756930828,
|
| 2110 |
+
"num_tokens": 7290776.0,
|
| 2111 |
+
"reward": 1.390625,
|
| 2112 |
+
"reward_std": 0.7726410627365112,
|
| 2113 |
+
"rewards/reward_func/mean": 0.1545138888888889,
|
| 2114 |
+
"rewards/reward_func/std": 0.1219082838959164,
|
| 2115 |
+
"sampling/importance_sampling_ratio/max": 2.9972431659698486,
|
| 2116 |
+
"sampling/importance_sampling_ratio/mean": 0.9446996450424194,
|
| 2117 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 2118 |
+
"sampling/sampling_logp_difference/max": 14.735367774963379,
|
| 2119 |
+
"sampling/sampling_logp_difference/mean": 0.22407299280166626,
|
| 2120 |
+
"seg/thinking_frac": 0.33292485028505325,
|
| 2121 |
+
"step": 48,
|
| 2122 |
+
"step_time": 108.9517350809474
|
| 2123 |
+
},
|
| 2124 |
+
{
|
| 2125 |
+
"clip_ratio/high_max": 0.0,
|
| 2126 |
+
"clip_ratio/high_mean": 0.0,
|
| 2127 |
+
"clip_ratio/low_mean": 0.0,
|
| 2128 |
+
"clip_ratio/low_min": 0.0,
|
| 2129 |
+
"clip_ratio/region_mean": 0.0,
|
| 2130 |
+
"completions/clipped_ratio": 0.0,
|
| 2131 |
+
"completions/max_length": 3980.0,
|
| 2132 |
+
"completions/max_terminated_length": 3980.0,
|
| 2133 |
+
"completions/mean_length": 930.25,
|
| 2134 |
+
"completions/mean_terminated_length": 930.25,
|
| 2135 |
+
"completions/min_length": 81.0,
|
| 2136 |
+
"completions/min_terminated_length": 81.0,
|
| 2137 |
+
"degenerate_groups_filtered": 0.0,
|
| 2138 |
+
"entropy": 0.8246443569660187,
|
| 2139 |
+
"entropy/code": 0.5987953245639801,
|
| 2140 |
+
"entropy/thinking": 1.174817144870758,
|
| 2141 |
+
"epoch": 0.1206896551724138,
|
| 2142 |
+
"frac_reward_zero_std": 0.5,
|
| 2143 |
+
"grad_norm": 0.003189835114732151,
|
| 2144 |
+
"kl": 0.002053185016848147,
|
| 2145 |
+
"kl/code_mean": 0.001470751318265684,
|
| 2146 |
+
"kl/code_sum": 4.21240234375,
|
| 2147 |
+
"kl/thinking_mean": 0.00297042028978467,
|
| 2148 |
+
"kl/thinking_sum": 3.513427734375,
|
| 2149 |
+
"kl/total_loss_contribution": 4.1063699427468237e-05,
|
| 2150 |
+
"kl_mask/effective_frac": 1.0,
|
| 2151 |
+
"kl_mask/fallback_rate": 0.1875,
|
| 2152 |
+
"learning_rate": 4.855217733333641e-05,
|
| 2153 |
+
"loss": 0.037618108093738556,
|
| 2154 |
+
"num_tokens": 7426344.0,
|
| 2155 |
+
"reward": 1.03515625,
|
| 2156 |
+
"reward_std": 0.3359043300151825,
|
| 2157 |
+
"rewards/reward_func/mean": 0.1150173611111111,
|
| 2158 |
+
"rewards/reward_func/std": 0.050191783242755465,
|
| 2159 |
+
"sampling/importance_sampling_ratio/max": 2.99593448638916,
|
| 2160 |
+
"sampling/importance_sampling_ratio/mean": 0.9514061808586121,
|
| 2161 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 2162 |
+
"sampling/sampling_logp_difference/max": 12.402661323547363,
|
| 2163 |
+
"sampling/sampling_logp_difference/mean": 0.21827572584152222,
|
| 2164 |
+
"seg/thinking_frac": 0.36690400540828705,
|
| 2165 |
+
"step": 49,
|
| 2166 |
+
"step_time": 93.88064169703284
|
| 2167 |
+
},
|
| 2168 |
+
{
|
| 2169 |
+
"clip_ratio/high_max": 0.0,
|
| 2170 |
+
"clip_ratio/high_mean": 0.0,
|
| 2171 |
+
"clip_ratio/low_mean": 0.0,
|
| 2172 |
+
"clip_ratio/low_min": 0.0,
|
| 2173 |
+
"clip_ratio/region_mean": 0.0,
|
| 2174 |
+
"completions/clipped_ratio": 0.09375,
|
| 2175 |
+
"completions/max_length": 4096.0,
|
| 2176 |
+
"completions/max_terminated_length": 2823.0,
|
| 2177 |
+
"completions/mean_length": 1351.3125,
|
| 2178 |
+
"completions/mean_terminated_length": 1101.7413330078125,
|
| 2179 |
+
"completions/min_length": 172.0,
|
| 2180 |
+
"completions/min_terminated_length": 172.0,
|
| 2181 |
+
"degenerate_groups_filtered": 0.0,
|
| 2182 |
+
"entropy": 0.6817112267017365,
|
| 2183 |
+
"entropy/code": 0.4861925318837166,
|
| 2184 |
+
"entropy/thinking": 1.209920734167099,
|
| 2185 |
+
"epoch": 0.12315270935960591,
|
| 2186 |
+
"frac_reward_zero_std": 0.25,
|
| 2187 |
+
"grad_norm": 0.002430456129926988,
|
| 2188 |
+
"kl": 0.0027986175555270165,
|
| 2189 |
+
"kl/code_mean": 0.002199954426032491,
|
| 2190 |
+
"kl/code_sum": 6.419189453125,
|
| 2191 |
+
"kl/thinking_mean": 0.0038550299941562116,
|
| 2192 |
+
"kl/thinking_sum": 3.7412109375,
|
| 2193 |
+
"kl/total_loss_contribution": 5.597234803644824e-05,
|
| 2194 |
+
"kl_mask/effective_frac": 1.0,
|
| 2195 |
+
"kl_mask/fallback_rate": 0.0,
|
| 2196 |
+
"learning_rate": 4.848475017333797e-05,
|
| 2197 |
+
"loss": 0.0033755588810890913,
|
| 2198 |
+
"num_tokens": 7588716.0,
|
| 2199 |
+
"reward": 1.09375,
|
| 2200 |
+
"reward_std": 0.41187721490859985,
|
| 2201 |
+
"rewards/reward_func/mean": 0.12152777777777778,
|
| 2202 |
+
"rewards/reward_func/std": 0.05926263497935401,
|
| 2203 |
+
"sampling/importance_sampling_ratio/max": 2.998502016067505,
|
| 2204 |
+
"sampling/importance_sampling_ratio/mean": 0.9491722583770752,
|
| 2205 |
+
"sampling/importance_sampling_ratio/min": 0.0,
|
| 2206 |
+
"sampling/sampling_logp_difference/max": 16.49994468688965,
|
| 2207 |
+
"sampling/sampling_logp_difference/mean": 0.20829011499881744,
|
| 2208 |
+
"seg/thinking_frac": 0.2652019150555134,
|
| 2209 |
+
"step": 50,
|
| 2210 |
+
"step_time": 105.45793682098156
|
| 2211 |
+
}
|
| 2212 |
+
],
|
| 2213 |
+
"logging_steps": 1,
|
| 2214 |
+
"max_steps": 400,
|
| 2215 |
+
"num_input_tokens_seen": 7588716,
|
| 2216 |
+
"num_train_epochs": 1,
|
| 2217 |
+
"save_steps": 10,
|
| 2218 |
+
"stateful_callbacks": {
|
| 2219 |
+
"TrainerControl": {
|
| 2220 |
+
"args": {
|
| 2221 |
+
"should_epoch_stop": false,
|
| 2222 |
+
"should_evaluate": false,
|
| 2223 |
+
"should_log": false,
|
| 2224 |
+
"should_save": true,
|
| 2225 |
+
"should_training_stop": false
|
| 2226 |
+
},
|
| 2227 |
+
"attributes": {}
|
| 2228 |
+
}
|
| 2229 |
+
},
|
| 2230 |
+
"total_flos": 0.0,
|
| 2231 |
+
"train_batch_size": 4,
|
| 2232 |
+
"trial_name": null,
|
| 2233 |
+
"trial_params": null
|
| 2234 |
+
}
|