Upload bf16 - Migration QLoRA DPO
Browse files- .gitattributes +1 -0
- README.md +69 -0
- chat_template.jinja +89 -0
- config.json +71 -0
- generation_config.json +13 -0
- model.safetensors +3 -0
- tokenizer.json +3 -0
- tokenizer_config.json +29 -0
- trainer_state.json +514 -0
.gitattributes
CHANGED
|
@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
|
| 33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
| 33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
| 36 |
+
tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
README.md
ADDED
|
@@ -0,0 +1,69 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
language:
|
| 3 |
+
- ja
|
| 4 |
+
- ko
|
| 5 |
+
- en
|
| 6 |
+
license: apache-2.0
|
| 7 |
+
base_model: Qwen/Qwen3.5-9B
|
| 8 |
+
tags:
|
| 9 |
+
- migration
|
| 10 |
+
- mainframe
|
| 11 |
+
- cobol
|
| 12 |
+
- jcl
|
| 13 |
+
- assembler
|
| 14 |
+
- qlora
|
| 15 |
+
- dpo
|
| 16 |
+
- ofkms
|
| 17 |
+
library_name: transformers
|
| 18 |
+
pipeline_tag: text-generation
|
| 19 |
+
---
|
| 20 |
+
|
| 21 |
+
# OFKMS Migration Design - Qwen3.5-9B DPO (bf16)
|
| 22 |
+
|
| 23 |
+
Mainframe migration design specialized model fine-tuned from Qwen3.5-9B.
|
| 24 |
+
|
| 25 |
+
## Model Description
|
| 26 |
+
|
| 27 |
+
This model is fine-tuned for **COBOL/JCL/Assembler migration design** tasks,
|
| 28 |
+
trained on TmaxSoft Japan's proprietary migration knowledge base.
|
| 29 |
+
|
| 30 |
+
- **Base Model**: Qwen3.5-9B
|
| 31 |
+
- **Fine-tuning**: QLoRA (DPO)
|
| 32 |
+
- **Training Data**: 1,288 SFT entries + 1,288 DPO pairs
|
| 33 |
+
- **Languages**: Japanese (primary), Korean, English
|
| 34 |
+
- **Variant**: bf16
|
| 35 |
+
|
| 36 |
+
## Training Details
|
| 37 |
+
|
| 38 |
+
- **Method**: QLoRA (rank=64, alpha=128)
|
| 39 |
+
- **Trainable params**: 174M / 8.4B (2.09%)
|
| 40 |
+
- **Epochs**: 3
|
| 41 |
+
- **Batch size**: 4 (gradient accumulation: 16, effective: 64)
|
| 42 |
+
- **Learning rate**: 2e-5 (cosine schedule)
|
| 43 |
+
- **Hardware**: NVIDIA A100 40GB
|
| 44 |
+
|
| 45 |
+
## Supported Tasks
|
| 46 |
+
|
| 47 |
+
- COBOL source pattern analysis and conversion rules
|
| 48 |
+
- JCL to OpenFrame JCL migration
|
| 49 |
+
- Assembler to C/OFASM migration
|
| 50 |
+
- Migration design document generation
|
| 51 |
+
- Error pattern diagnosis (ABEND codes, JES messages)
|
| 52 |
+
|
| 53 |
+
## Usage
|
| 54 |
+
|
| 55 |
+
```python
|
| 56 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
| 57 |
+
|
| 58 |
+
model = AutoModelForCausalLM.from_pretrained("jtmaxsoft/OFKMS-Migration-Qwen3.5-9B-DPO")
|
| 59 |
+
tokenizer = AutoTokenizer.from_pretrained("jtmaxsoft/OFKMS-Migration-Qwen3.5-9B-DPO")
|
| 60 |
+
|
| 61 |
+
prompt = "COBOL PERFORM statement OpenFrame migration pattern"
|
| 62 |
+
inputs = tokenizer(prompt, return_tensors="pt")
|
| 63 |
+
outputs = model.generate(**inputs, max_new_tokens=512)
|
| 64 |
+
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
|
| 65 |
+
```
|
| 66 |
+
|
| 67 |
+
## Organization
|
| 68 |
+
|
| 69 |
+
[TmaxSoft Japan](https://huggingface.co/jtmaxsoft)
|
chat_template.jinja
ADDED
|
@@ -0,0 +1,89 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{%- if tools %}
|
| 2 |
+
{{- '<|im_start|>system\n' }}
|
| 3 |
+
{%- if messages[0].role == 'system' %}
|
| 4 |
+
{{- messages[0].content + '\n\n' }}
|
| 5 |
+
{%- endif %}
|
| 6 |
+
{{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
|
| 7 |
+
{%- for tool in tools %}
|
| 8 |
+
{{- "\n" }}
|
| 9 |
+
{{- tool | tojson }}
|
| 10 |
+
{%- endfor %}
|
| 11 |
+
{{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
|
| 12 |
+
{%- else %}
|
| 13 |
+
{%- if messages[0].role == 'system' %}
|
| 14 |
+
{{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
|
| 15 |
+
{%- endif %}
|
| 16 |
+
{%- endif %}
|
| 17 |
+
{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
|
| 18 |
+
{%- for message in messages[::-1] %}
|
| 19 |
+
{%- set index = (messages|length - 1) - loop.index0 %}
|
| 20 |
+
{%- if ns.multi_step_tool and message.role == "user" and message.content is string and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
|
| 21 |
+
{%- set ns.multi_step_tool = false %}
|
| 22 |
+
{%- set ns.last_query_index = index %}
|
| 23 |
+
{%- endif %}
|
| 24 |
+
{%- endfor %}
|
| 25 |
+
{%- for message in messages %}
|
| 26 |
+
{%- if message.content is string %}
|
| 27 |
+
{%- set content = message.content %}
|
| 28 |
+
{%- else %}
|
| 29 |
+
{%- set content = '' %}
|
| 30 |
+
{%- endif %}
|
| 31 |
+
{%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
|
| 32 |
+
{{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
|
| 33 |
+
{%- elif message.role == "assistant" %}
|
| 34 |
+
{%- set reasoning_content = '' %}
|
| 35 |
+
{%- if message.reasoning_content is string %}
|
| 36 |
+
{%- set reasoning_content = message.reasoning_content %}
|
| 37 |
+
{%- else %}
|
| 38 |
+
{%- if '</think>' in content %}
|
| 39 |
+
{%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
|
| 40 |
+
{%- set content = content.split('</think>')[-1].lstrip('\n') %}
|
| 41 |
+
{%- endif %}
|
| 42 |
+
{%- endif %}
|
| 43 |
+
{%- if loop.index0 > ns.last_query_index %}
|
| 44 |
+
{%- if loop.last or (not loop.last and reasoning_content) %}
|
| 45 |
+
{{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
|
| 46 |
+
{%- else %}
|
| 47 |
+
{{- '<|im_start|>' + message.role + '\n' + content }}
|
| 48 |
+
{%- endif %}
|
| 49 |
+
{%- else %}
|
| 50 |
+
{{- '<|im_start|>' + message.role + '\n' + content }}
|
| 51 |
+
{%- endif %}
|
| 52 |
+
{%- if message.tool_calls %}
|
| 53 |
+
{%- for tool_call in message.tool_calls %}
|
| 54 |
+
{%- if (loop.first and content) or (not loop.first) %}
|
| 55 |
+
{{- '\n' }}
|
| 56 |
+
{%- endif %}
|
| 57 |
+
{%- if tool_call.function %}
|
| 58 |
+
{%- set tool_call = tool_call.function %}
|
| 59 |
+
{%- endif %}
|
| 60 |
+
{{- '<tool_call>\n{"name": "' }}
|
| 61 |
+
{{- tool_call.name }}
|
| 62 |
+
{{- '", "arguments": ' }}
|
| 63 |
+
{%- if tool_call.arguments is string %}
|
| 64 |
+
{{- tool_call.arguments }}
|
| 65 |
+
{%- else %}
|
| 66 |
+
{{- tool_call.arguments | tojson }}
|
| 67 |
+
{%- endif %}
|
| 68 |
+
{{- '}\n</tool_call>' }}
|
| 69 |
+
{%- endfor %}
|
| 70 |
+
{%- endif %}
|
| 71 |
+
{{- '<|im_end|>\n' }}
|
| 72 |
+
{%- elif message.role == "tool" %}
|
| 73 |
+
{%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
|
| 74 |
+
{{- '<|im_start|>user' }}
|
| 75 |
+
{%- endif %}
|
| 76 |
+
{{- '\n<tool_response>\n' }}
|
| 77 |
+
{{- content }}
|
| 78 |
+
{{- '\n</tool_response>' }}
|
| 79 |
+
{%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
|
| 80 |
+
{{- '<|im_end|>\n' }}
|
| 81 |
+
{%- endif %}
|
| 82 |
+
{%- endif %}
|
| 83 |
+
{%- endfor %}
|
| 84 |
+
{%- if add_generation_prompt %}
|
| 85 |
+
{{- '<|im_start|>assistant\n' }}
|
| 86 |
+
{%- if enable_thinking is defined and enable_thinking is false %}
|
| 87 |
+
{{- '<think>\n\n</think>\n\n' }}
|
| 88 |
+
{%- endif %}
|
| 89 |
+
{%- endif %}
|
config.json
ADDED
|
@@ -0,0 +1,71 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"Qwen3ForCausalLM"
|
| 4 |
+
],
|
| 5 |
+
"attention_bias": false,
|
| 6 |
+
"attention_dropout": 0.0,
|
| 7 |
+
"bos_token_id": 151643,
|
| 8 |
+
"dtype": "bfloat16",
|
| 9 |
+
"eos_token_id": 151645,
|
| 10 |
+
"head_dim": 128,
|
| 11 |
+
"hidden_act": "silu",
|
| 12 |
+
"hidden_size": 4096,
|
| 13 |
+
"initializer_range": 0.02,
|
| 14 |
+
"intermediate_size": 12288,
|
| 15 |
+
"layer_types": [
|
| 16 |
+
"full_attention",
|
| 17 |
+
"full_attention",
|
| 18 |
+
"full_attention",
|
| 19 |
+
"full_attention",
|
| 20 |
+
"full_attention",
|
| 21 |
+
"full_attention",
|
| 22 |
+
"full_attention",
|
| 23 |
+
"full_attention",
|
| 24 |
+
"full_attention",
|
| 25 |
+
"full_attention",
|
| 26 |
+
"full_attention",
|
| 27 |
+
"full_attention",
|
| 28 |
+
"full_attention",
|
| 29 |
+
"full_attention",
|
| 30 |
+
"full_attention",
|
| 31 |
+
"full_attention",
|
| 32 |
+
"full_attention",
|
| 33 |
+
"full_attention",
|
| 34 |
+
"full_attention",
|
| 35 |
+
"full_attention",
|
| 36 |
+
"full_attention",
|
| 37 |
+
"full_attention",
|
| 38 |
+
"full_attention",
|
| 39 |
+
"full_attention",
|
| 40 |
+
"full_attention",
|
| 41 |
+
"full_attention",
|
| 42 |
+
"full_attention",
|
| 43 |
+
"full_attention",
|
| 44 |
+
"full_attention",
|
| 45 |
+
"full_attention",
|
| 46 |
+
"full_attention",
|
| 47 |
+
"full_attention",
|
| 48 |
+
"full_attention",
|
| 49 |
+
"full_attention",
|
| 50 |
+
"full_attention",
|
| 51 |
+
"full_attention"
|
| 52 |
+
],
|
| 53 |
+
"max_position_embeddings": 40960,
|
| 54 |
+
"max_window_layers": 36,
|
| 55 |
+
"model_type": "qwen3",
|
| 56 |
+
"num_attention_heads": 32,
|
| 57 |
+
"num_hidden_layers": 36,
|
| 58 |
+
"num_key_value_heads": 8,
|
| 59 |
+
"pad_token_id": null,
|
| 60 |
+
"rms_norm_eps": 1e-06,
|
| 61 |
+
"rope_parameters": {
|
| 62 |
+
"rope_theta": 1000000,
|
| 63 |
+
"rope_type": "default"
|
| 64 |
+
},
|
| 65 |
+
"sliding_window": null,
|
| 66 |
+
"tie_word_embeddings": false,
|
| 67 |
+
"transformers_version": "5.3.0",
|
| 68 |
+
"use_cache": true,
|
| 69 |
+
"use_sliding_window": false,
|
| 70 |
+
"vocab_size": 151936
|
| 71 |
+
}
|
generation_config.json
ADDED
|
@@ -0,0 +1,13 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bos_token_id": 151643,
|
| 3 |
+
"do_sample": true,
|
| 4 |
+
"eos_token_id": [
|
| 5 |
+
151645,
|
| 6 |
+
151643
|
| 7 |
+
],
|
| 8 |
+
"pad_token_id": 151643,
|
| 9 |
+
"temperature": 0.6,
|
| 10 |
+
"top_k": 20,
|
| 11 |
+
"top_p": 0.95,
|
| 12 |
+
"transformers_version": "5.3.0"
|
| 13 |
+
}
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b13ab529495892871ef1e774d60099b3cc4ced5fd5e36116557309de23e3a4fd
|
| 3 |
+
size 16381517208
|
tokenizer.json
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:be75606093db2094d7cd20f3c2f385c212750648bd6ea4fb2bf507a6a4c55506
|
| 3 |
+
size 11422650
|
tokenizer_config.json
ADDED
|
@@ -0,0 +1,29 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"add_prefix_space": false,
|
| 3 |
+
"backend": "tokenizers",
|
| 4 |
+
"bos_token": null,
|
| 5 |
+
"clean_up_tokenization_spaces": false,
|
| 6 |
+
"eos_token": "<|im_end|>",
|
| 7 |
+
"errors": "replace",
|
| 8 |
+
"extra_special_tokens": [
|
| 9 |
+
"<|im_start|>",
|
| 10 |
+
"<|im_end|>",
|
| 11 |
+
"<|object_ref_start|>",
|
| 12 |
+
"<|object_ref_end|>",
|
| 13 |
+
"<|box_start|>",
|
| 14 |
+
"<|box_end|>",
|
| 15 |
+
"<|quad_start|>",
|
| 16 |
+
"<|quad_end|>",
|
| 17 |
+
"<|vision_start|>",
|
| 18 |
+
"<|vision_end|>",
|
| 19 |
+
"<|vision_pad|>",
|
| 20 |
+
"<|image_pad|>",
|
| 21 |
+
"<|video_pad|>"
|
| 22 |
+
],
|
| 23 |
+
"is_local": true,
|
| 24 |
+
"model_max_length": 131072,
|
| 25 |
+
"pad_token": "<|endoftext|>",
|
| 26 |
+
"split_special_tokens": false,
|
| 27 |
+
"tokenizer_class": "Qwen2Tokenizer",
|
| 28 |
+
"unk_token": null
|
| 29 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,514 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_global_step": null,
|
| 3 |
+
"best_metric": null,
|
| 4 |
+
"best_model_checkpoint": null,
|
| 5 |
+
"epoch": 2.0,
|
| 6 |
+
"eval_steps": 500,
|
| 7 |
+
"global_step": 322,
|
| 8 |
+
"is_hyper_param_search": false,
|
| 9 |
+
"is_local_process_zero": true,
|
| 10 |
+
"is_world_process_zero": true,
|
| 11 |
+
"log_history": [
|
| 12 |
+
{
|
| 13 |
+
"epoch": 0.062111801242236024,
|
| 14 |
+
"grad_norm": 2.031649589538574,
|
| 15 |
+
"learning_rate": 1.3636363636363636e-06,
|
| 16 |
+
"logits/chosen": -1.1706674098968506,
|
| 17 |
+
"logits/rejected": -1.0306828022003174,
|
| 18 |
+
"logps/chosen": -564.5164794921875,
|
| 19 |
+
"logps/rejected": -463.99603271484375,
|
| 20 |
+
"loss": 0.6916566371917725,
|
| 21 |
+
"rewards/accuracies": 0.4375,
|
| 22 |
+
"rewards/chosen": 0.010852223262190819,
|
| 23 |
+
"rewards/margins": 0.004624844063073397,
|
| 24 |
+
"rewards/rejected": 0.006227378733456135,
|
| 25 |
+
"step": 10
|
| 26 |
+
},
|
| 27 |
+
{
|
| 28 |
+
"epoch": 0.12422360248447205,
|
| 29 |
+
"grad_norm": 2.3162496089935303,
|
| 30 |
+
"learning_rate": 2.8787878787878793e-06,
|
| 31 |
+
"logits/chosen": -1.0546079874038696,
|
| 32 |
+
"logits/rejected": -0.9307195544242859,
|
| 33 |
+
"logps/chosen": -475.3399963378906,
|
| 34 |
+
"logps/rejected": -421.66693115234375,
|
| 35 |
+
"loss": 0.6910689353942872,
|
| 36 |
+
"rewards/accuracies": 0.5249999761581421,
|
| 37 |
+
"rewards/chosen": 0.020578766241669655,
|
| 38 |
+
"rewards/margins": 0.0066573722288012505,
|
| 39 |
+
"rewards/rejected": 0.013921394944190979,
|
| 40 |
+
"step": 20
|
| 41 |
+
},
|
| 42 |
+
{
|
| 43 |
+
"epoch": 0.18633540372670807,
|
| 44 |
+
"grad_norm": 2.447273015975952,
|
| 45 |
+
"learning_rate": 4.393939393939394e-06,
|
| 46 |
+
"logits/chosen": -1.1174654960632324,
|
| 47 |
+
"logits/rejected": -1.0690593719482422,
|
| 48 |
+
"logps/chosen": -606.117919921875,
|
| 49 |
+
"logps/rejected": -445.750244140625,
|
| 50 |
+
"loss": 0.6635114669799804,
|
| 51 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 52 |
+
"rewards/chosen": 0.11121845245361328,
|
| 53 |
+
"rewards/margins": 0.06357013434171677,
|
| 54 |
+
"rewards/rejected": 0.047648314386606216,
|
| 55 |
+
"step": 30
|
| 56 |
+
},
|
| 57 |
+
{
|
| 58 |
+
"epoch": 0.2484472049689441,
|
| 59 |
+
"grad_norm": 2.597184896469116,
|
| 60 |
+
"learning_rate": 4.994684273748881e-06,
|
| 61 |
+
"logits/chosen": -1.0191378593444824,
|
| 62 |
+
"logits/rejected": -1.0478084087371826,
|
| 63 |
+
"logps/chosen": -552.3709716796875,
|
| 64 |
+
"logps/rejected": -423.89892578125,
|
| 65 |
+
"loss": 0.6465987205505371,
|
| 66 |
+
"rewards/accuracies": 0.737500011920929,
|
| 67 |
+
"rewards/chosen": 0.2686358094215393,
|
| 68 |
+
"rewards/margins": 0.10441793501377106,
|
| 69 |
+
"rewards/rejected": 0.16421787440776825,
|
| 70 |
+
"step": 40
|
| 71 |
+
},
|
| 72 |
+
{
|
| 73 |
+
"epoch": 0.3105590062111801,
|
| 74 |
+
"grad_norm": 1.94443941116333,
|
| 75 |
+
"learning_rate": 4.962281107627828e-06,
|
| 76 |
+
"logits/chosen": -1.295013427734375,
|
| 77 |
+
"logits/rejected": -0.8611756563186646,
|
| 78 |
+
"logps/chosen": -545.5277709960938,
|
| 79 |
+
"logps/rejected": -374.797119140625,
|
| 80 |
+
"loss": 0.5243598937988281,
|
| 81 |
+
"rewards/accuracies": 0.887499988079071,
|
| 82 |
+
"rewards/chosen": 0.5271034836769104,
|
| 83 |
+
"rewards/margins": 0.4082818925380707,
|
| 84 |
+
"rewards/rejected": 0.11882160604000092,
|
| 85 |
+
"step": 50
|
| 86 |
+
},
|
| 87 |
+
{
|
| 88 |
+
"epoch": 0.37267080745341613,
|
| 89 |
+
"grad_norm": 1.8612565994262695,
|
| 90 |
+
"learning_rate": 4.900810019766151e-06,
|
| 91 |
+
"logits/chosen": -1.106858253479004,
|
| 92 |
+
"logits/rejected": -0.8895740509033203,
|
| 93 |
+
"logps/chosen": -574.3599853515625,
|
| 94 |
+
"logps/rejected": -398.66595458984375,
|
| 95 |
+
"loss": 0.46582393646240233,
|
| 96 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 97 |
+
"rewards/chosen": 1.0713833570480347,
|
| 98 |
+
"rewards/margins": 0.6883617639541626,
|
| 99 |
+
"rewards/rejected": 0.38302165269851685,
|
| 100 |
+
"step": 60
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"epoch": 0.43478260869565216,
|
| 104 |
+
"grad_norm": 2.610835313796997,
|
| 105 |
+
"learning_rate": 4.8109966936634885e-06,
|
| 106 |
+
"logits/chosen": -0.9745047688484192,
|
| 107 |
+
"logits/rejected": -0.8878741264343262,
|
| 108 |
+
"logps/chosen": -603.6141357421875,
|
| 109 |
+
"logps/rejected": -422.431396484375,
|
| 110 |
+
"loss": 0.3722702980041504,
|
| 111 |
+
"rewards/accuracies": 0.862500011920929,
|
| 112 |
+
"rewards/chosen": 1.5592485666275024,
|
| 113 |
+
"rewards/margins": 1.0986794233322144,
|
| 114 |
+
"rewards/rejected": 0.46056899428367615,
|
| 115 |
+
"step": 70
|
| 116 |
+
},
|
| 117 |
+
{
|
| 118 |
+
"epoch": 0.4968944099378882,
|
| 119 |
+
"grad_norm": 1.7999701499938965,
|
| 120 |
+
"learning_rate": 4.693901400921782e-06,
|
| 121 |
+
"logits/chosen": -0.9704159498214722,
|
| 122 |
+
"logits/rejected": -0.8450535535812378,
|
| 123 |
+
"logps/chosen": -521.1560668945312,
|
| 124 |
+
"logps/rejected": -403.3153076171875,
|
| 125 |
+
"loss": 0.36702446937561034,
|
| 126 |
+
"rewards/accuracies": 0.875,
|
| 127 |
+
"rewards/chosen": 1.462435245513916,
|
| 128 |
+
"rewards/margins": 1.3280738592147827,
|
| 129 |
+
"rewards/rejected": 0.13436131179332733,
|
| 130 |
+
"step": 80
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 0.5590062111801242,
|
| 134 |
+
"grad_norm": 1.2426642179489136,
|
| 135 |
+
"learning_rate": 4.550906484440495e-06,
|
| 136 |
+
"logits/chosen": -0.9953195452690125,
|
| 137 |
+
"logits/rejected": -0.8482304811477661,
|
| 138 |
+
"logps/chosen": -531.6632080078125,
|
| 139 |
+
"logps/rejected": -381.96258544921875,
|
| 140 |
+
"loss": 0.28965752124786376,
|
| 141 |
+
"rewards/accuracies": 0.9375,
|
| 142 |
+
"rewards/chosen": 1.4531519412994385,
|
| 143 |
+
"rewards/margins": 1.6451427936553955,
|
| 144 |
+
"rewards/rejected": -0.19199064373970032,
|
| 145 |
+
"step": 90
|
| 146 |
+
},
|
| 147 |
+
{
|
| 148 |
+
"epoch": 0.6211180124223602,
|
| 149 |
+
"grad_norm": 0.7566975951194763,
|
| 150 |
+
"learning_rate": 4.38370003946948e-06,
|
| 151 |
+
"logits/chosen": -1.0843112468719482,
|
| 152 |
+
"logits/rejected": -0.9646870493888855,
|
| 153 |
+
"logps/chosen": -541.9859619140625,
|
| 154 |
+
"logps/rejected": -427.3182067871094,
|
| 155 |
+
"loss": 0.2813025951385498,
|
| 156 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 157 |
+
"rewards/chosen": 1.0933747291564941,
|
| 158 |
+
"rewards/margins": 1.867444634437561,
|
| 159 |
+
"rewards/rejected": -0.7740699052810669,
|
| 160 |
+
"step": 100
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"epoch": 0.6832298136645962,
|
| 164 |
+
"grad_norm": 1.8626788854599,
|
| 165 |
+
"learning_rate": 4.194255985169259e-06,
|
| 166 |
+
"logits/chosen": -1.16153883934021,
|
| 167 |
+
"logits/rejected": -1.1651047468185425,
|
| 168 |
+
"logps/chosen": -602.457763671875,
|
| 169 |
+
"logps/rejected": -449.3597106933594,
|
| 170 |
+
"loss": 0.1985023260116577,
|
| 171 |
+
"rewards/accuracies": 0.949999988079071,
|
| 172 |
+
"rewards/chosen": 0.8817703127861023,
|
| 173 |
+
"rewards/margins": 2.40962815284729,
|
| 174 |
+
"rewards/rejected": -1.5278576612472534,
|
| 175 |
+
"step": 110
|
| 176 |
+
},
|
| 177 |
+
{
|
| 178 |
+
"epoch": 0.7453416149068323,
|
| 179 |
+
"grad_norm": 1.6138923168182373,
|
| 180 |
+
"learning_rate": 3.984810761939578e-06,
|
| 181 |
+
"logits/chosen": -1.2249107360839844,
|
| 182 |
+
"logits/rejected": -1.1240571737289429,
|
| 183 |
+
"logps/chosen": -563.8658447265625,
|
| 184 |
+
"logps/rejected": -429.28021240234375,
|
| 185 |
+
"loss": 0.14866907596588136,
|
| 186 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 187 |
+
"rewards/chosen": 0.36940866708755493,
|
| 188 |
+
"rewards/margins": 2.9657340049743652,
|
| 189 |
+
"rewards/rejected": -2.596325397491455,
|
| 190 |
+
"step": 120
|
| 191 |
+
},
|
| 192 |
+
{
|
| 193 |
+
"epoch": 0.8074534161490683,
|
| 194 |
+
"grad_norm": 1.6024304628372192,
|
| 195 |
+
"learning_rate": 3.757836929610803e-06,
|
| 196 |
+
"logits/chosen": -1.2749364376068115,
|
| 197 |
+
"logits/rejected": -1.2431113719940186,
|
| 198 |
+
"logps/chosen": -653.6280517578125,
|
| 199 |
+
"logps/rejected": -399.4922790527344,
|
| 200 |
+
"loss": 0.1276548385620117,
|
| 201 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 202 |
+
"rewards/chosen": 0.43943363428115845,
|
| 203 |
+
"rewards/margins": 3.6862716674804688,
|
| 204 |
+
"rewards/rejected": -3.246838331222534,
|
| 205 |
+
"step": 130
|
| 206 |
+
},
|
| 207 |
+
{
|
| 208 |
+
"epoch": 0.8695652173913043,
|
| 209 |
+
"grad_norm": 1.489258885383606,
|
| 210 |
+
"learning_rate": 3.5160139781789615e-06,
|
| 211 |
+
"logits/chosen": -1.1981886625289917,
|
| 212 |
+
"logits/rejected": -1.1743550300598145,
|
| 213 |
+
"logps/chosen": -630.0133056640625,
|
| 214 |
+
"logps/rejected": -453.70233154296875,
|
| 215 |
+
"loss": 0.13757349252700807,
|
| 216 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 217 |
+
"rewards/chosen": 1.116315484046936,
|
| 218 |
+
"rewards/margins": 3.7938003540039062,
|
| 219 |
+
"rewards/rejected": -2.6774849891662598,
|
| 220 |
+
"step": 140
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"epoch": 0.9316770186335404,
|
| 224 |
+
"grad_norm": 1.529561996459961,
|
| 225 |
+
"learning_rate": 3.2621966956719265e-06,
|
| 226 |
+
"logits/chosen": -1.1383417844772339,
|
| 227 |
+
"logits/rejected": -1.1288671493530273,
|
| 228 |
+
"logps/chosen": -605.6137084960938,
|
| 229 |
+
"logps/rejected": -464.63616943359375,
|
| 230 |
+
"loss": 0.1510193705558777,
|
| 231 |
+
"rewards/accuracies": 0.9375,
|
| 232 |
+
"rewards/chosen": 1.3723418712615967,
|
| 233 |
+
"rewards/margins": 3.875206708908081,
|
| 234 |
+
"rewards/rejected": -2.5028648376464844,
|
| 235 |
+
"step": 150
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 0.9937888198757764,
|
| 239 |
+
"grad_norm": 1.2801965475082397,
|
| 240 |
+
"learning_rate": 2.9993814665729605e-06,
|
| 241 |
+
"logits/chosen": -1.1155999898910522,
|
| 242 |
+
"logits/rejected": -1.1486127376556396,
|
| 243 |
+
"logps/chosen": -607.6810913085938,
|
| 244 |
+
"logps/rejected": -428.43902587890625,
|
| 245 |
+
"loss": 0.14707612991333008,
|
| 246 |
+
"rewards/accuracies": 0.949999988079071,
|
| 247 |
+
"rewards/chosen": 1.2416237592697144,
|
| 248 |
+
"rewards/margins": 4.007672309875488,
|
| 249 |
+
"rewards/rejected": -2.7660484313964844,
|
| 250 |
+
"step": 160
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"epoch": 1.0559006211180124,
|
| 254 |
+
"grad_norm": 0.6223127245903015,
|
| 255 |
+
"learning_rate": 2.730670898658255e-06,
|
| 256 |
+
"logits/chosen": -1.2678742408752441,
|
| 257 |
+
"logits/rejected": -1.2089016437530518,
|
| 258 |
+
"logps/chosen": -546.5106811523438,
|
| 259 |
+
"logps/rejected": -435.56201171875,
|
| 260 |
+
"loss": 0.07639291286468505,
|
| 261 |
+
"rewards/accuracies": 0.987500011920929,
|
| 262 |
+
"rewards/chosen": 1.2402414083480835,
|
| 263 |
+
"rewards/margins": 4.630614280700684,
|
| 264 |
+
"rewards/rejected": -3.3903732299804688,
|
| 265 |
+
"step": 170
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 1.1180124223602483,
|
| 269 |
+
"grad_norm": 0.27513387799263,
|
| 270 |
+
"learning_rate": 2.459237195838577e-06,
|
| 271 |
+
"logits/chosen": -1.1183346509933472,
|
| 272 |
+
"logits/rejected": -1.2728922367095947,
|
| 273 |
+
"logps/chosen": -560.7864990234375,
|
| 274 |
+
"logps/rejected": -461.84332275390625,
|
| 275 |
+
"loss": 0.06406531929969787,
|
| 276 |
+
"rewards/accuracies": 0.987500011920929,
|
| 277 |
+
"rewards/chosen": 1.1345335245132446,
|
| 278 |
+
"rewards/margins": 5.346749305725098,
|
| 279 |
+
"rewards/rejected": -4.212216377258301,
|
| 280 |
+
"step": 180
|
| 281 |
+
},
|
| 282 |
+
{
|
| 283 |
+
"epoch": 1.1801242236024845,
|
| 284 |
+
"grad_norm": 0.5560200810432434,
|
| 285 |
+
"learning_rate": 2.1882847093989544e-06,
|
| 286 |
+
"logits/chosen": -1.2576462030410767,
|
| 287 |
+
"logits/rejected": -1.2605979442596436,
|
| 288 |
+
"logps/chosen": -598.9000854492188,
|
| 289 |
+
"logps/rejected": -403.6759033203125,
|
| 290 |
+
"loss": 0.07317939400672913,
|
| 291 |
+
"rewards/accuracies": 0.987500011920929,
|
| 292 |
+
"rewards/chosen": 1.0486507415771484,
|
| 293 |
+
"rewards/margins": 4.786526203155518,
|
| 294 |
+
"rewards/rejected": -3.737874984741211,
|
| 295 |
+
"step": 190
|
| 296 |
+
},
|
| 297 |
+
{
|
| 298 |
+
"epoch": 1.2422360248447206,
|
| 299 |
+
"grad_norm": 3.307779312133789,
|
| 300 |
+
"learning_rate": 1.921012109729562e-06,
|
| 301 |
+
"logits/chosen": -1.1867707967758179,
|
| 302 |
+
"logits/rejected": -1.1512377262115479,
|
| 303 |
+
"logps/chosen": -564.1589965820312,
|
| 304 |
+
"logps/rejected": -499.6703186035156,
|
| 305 |
+
"loss": 0.07128837704658508,
|
| 306 |
+
"rewards/accuracies": 0.987500011920929,
|
| 307 |
+
"rewards/chosen": 1.1387567520141602,
|
| 308 |
+
"rewards/margins": 4.9200215339660645,
|
| 309 |
+
"rewards/rejected": -3.781264543533325,
|
| 310 |
+
"step": 200
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"epoch": 1.3043478260869565,
|
| 314 |
+
"grad_norm": 0.951438307762146,
|
| 315 |
+
"learning_rate": 1.6605746251211642e-06,
|
| 316 |
+
"logits/chosen": -1.3087455034255981,
|
| 317 |
+
"logits/rejected": -1.4264456033706665,
|
| 318 |
+
"logps/chosen": -574.0267333984375,
|
| 319 |
+
"logps/rejected": -467.0670471191406,
|
| 320 |
+
"loss": 0.05501532554626465,
|
| 321 |
+
"rewards/accuracies": 0.987500011920929,
|
| 322 |
+
"rewards/chosen": 1.2620090246200562,
|
| 323 |
+
"rewards/margins": 5.137342929840088,
|
| 324 |
+
"rewards/rejected": -3.875333309173584,
|
| 325 |
+
"step": 210
|
| 326 |
+
},
|
| 327 |
+
{
|
| 328 |
+
"epoch": 1.3664596273291925,
|
| 329 |
+
"grad_norm": 0.4313580095767975,
|
| 330 |
+
"learning_rate": 1.4100467934067775e-06,
|
| 331 |
+
"logits/chosen": -1.2540130615234375,
|
| 332 |
+
"logits/rejected": -1.2541478872299194,
|
| 333 |
+
"logps/chosen": -557.1004028320312,
|
| 334 |
+
"logps/rejected": -483.9710388183594,
|
| 335 |
+
"loss": 0.08154834508895874,
|
| 336 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 337 |
+
"rewards/chosen": 1.3047633171081543,
|
| 338 |
+
"rewards/margins": 5.482510566711426,
|
| 339 |
+
"rewards/rejected": -4.1777472496032715,
|
| 340 |
+
"step": 220
|
| 341 |
+
},
|
| 342 |
+
{
|
| 343 |
+
"epoch": 1.4285714285714286,
|
| 344 |
+
"grad_norm": 0.3101217746734619,
|
| 345 |
+
"learning_rate": 1.1723861661769e-06,
|
| 346 |
+
"logits/chosen": -1.071692705154419,
|
| 347 |
+
"logits/rejected": -1.244997262954712,
|
| 348 |
+
"logps/chosen": -586.4631958007812,
|
| 349 |
+
"logps/rejected": -464.3282165527344,
|
| 350 |
+
"loss": 0.05381497144699097,
|
| 351 |
+
"rewards/accuracies": 1.0,
|
| 352 |
+
"rewards/chosen": 1.0372705459594727,
|
| 353 |
+
"rewards/margins": 4.841385841369629,
|
| 354 |
+
"rewards/rejected": -3.8041152954101562,
|
| 355 |
+
"step": 230
|
| 356 |
+
},
|
| 357 |
+
{
|
| 358 |
+
"epoch": 1.4906832298136645,
|
| 359 |
+
"grad_norm": 0.4230075180530548,
|
| 360 |
+
"learning_rate": 9.503983940502795e-07,
|
| 361 |
+
"logits/chosen": -1.3364367485046387,
|
| 362 |
+
"logits/rejected": -1.3070945739746094,
|
| 363 |
+
"logps/chosen": -579.2850341796875,
|
| 364 |
+
"logps/rejected": -420.6127014160156,
|
| 365 |
+
"loss": 0.08604136109352112,
|
| 366 |
+
"rewards/accuracies": 0.987500011920929,
|
| 367 |
+
"rewards/chosen": 0.7024589776992798,
|
| 368 |
+
"rewards/margins": 4.885396480560303,
|
| 369 |
+
"rewards/rejected": -4.1829376220703125,
|
| 370 |
+
"step": 240
|
| 371 |
+
},
|
| 372 |
+
{
|
| 373 |
+
"epoch": 1.5527950310559007,
|
| 374 |
+
"grad_norm": 0.31521838903427124,
|
| 375 |
+
"learning_rate": 7.467041051784155e-07,
|
| 376 |
+
"logits/chosen": -1.3780875205993652,
|
| 377 |
+
"logits/rejected": -1.4410358667373657,
|
| 378 |
+
"logps/chosen": -569.69189453125,
|
| 379 |
+
"logps/rejected": -491.8089904785156,
|
| 380 |
+
"loss": 0.09177066087722778,
|
| 381 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 382 |
+
"rewards/chosen": 0.8250600099563599,
|
| 383 |
+
"rewards/margins": 6.094225883483887,
|
| 384 |
+
"rewards/rejected": -5.269165992736816,
|
| 385 |
+
"step": 250
|
| 386 |
+
},
|
| 387 |
+
{
|
| 388 |
+
"epoch": 1.6149068322981366,
|
| 389 |
+
"grad_norm": 0.15115521848201752,
|
| 390 |
+
"learning_rate": 5.637079679923796e-07,
|
| 391 |
+
"logits/chosen": -1.3521003723144531,
|
| 392 |
+
"logits/rejected": -1.344463586807251,
|
| 393 |
+
"logps/chosen": -596.4379272460938,
|
| 394 |
+
"logps/rejected": -428.6333923339844,
|
| 395 |
+
"loss": 0.048569518327713015,
|
| 396 |
+
"rewards/accuracies": 1.0,
|
| 397 |
+
"rewards/chosen": 0.9481250047683716,
|
| 398 |
+
"rewards/margins": 6.150216102600098,
|
| 399 |
+
"rewards/rejected": -5.202090263366699,
|
| 400 |
+
"step": 260
|
| 401 |
+
},
|
| 402 |
+
{
|
| 403 |
+
"epoch": 1.6770186335403725,
|
| 404 |
+
"grad_norm": 0.23105435073375702,
|
| 405 |
+
"learning_rate": 4.035703034149463e-07,
|
| 406 |
+
"logits/chosen": -1.19241464138031,
|
| 407 |
+
"logits/rejected": -1.4326884746551514,
|
| 408 |
+
"logps/chosen": -554.9596557617188,
|
| 409 |
+
"logps/rejected": -461.3987731933594,
|
| 410 |
+
"loss": 0.05657889246940613,
|
| 411 |
+
"rewards/accuracies": 1.0,
|
| 412 |
+
"rewards/chosen": 0.8795555233955383,
|
| 413 |
+
"rewards/margins": 5.471189498901367,
|
| 414 |
+
"rewards/rejected": -4.5916337966918945,
|
| 415 |
+
"step": 270
|
| 416 |
+
},
|
| 417 |
+
{
|
| 418 |
+
"epoch": 1.7391304347826086,
|
| 419 |
+
"grad_norm": 0.14941351115703583,
|
| 420 |
+
"learning_rate": 2.681815816638503e-07,
|
| 421 |
+
"logits/chosen": -1.240888237953186,
|
| 422 |
+
"logits/rejected": -1.4535605907440186,
|
| 423 |
+
"logps/chosen": -582.7335205078125,
|
| 424 |
+
"logps/rejected": -481.8042907714844,
|
| 425 |
+
"loss": 0.043333661556243894,
|
| 426 |
+
"rewards/accuracies": 1.0,
|
| 427 |
+
"rewards/chosen": 0.9755797386169434,
|
| 428 |
+
"rewards/margins": 6.308679580688477,
|
| 429 |
+
"rewards/rejected": -5.333099365234375,
|
| 430 |
+
"step": 280
|
| 431 |
+
},
|
| 432 |
+
{
|
| 433 |
+
"epoch": 1.8012422360248448,
|
| 434 |
+
"grad_norm": 0.7390910983085632,
|
| 435 |
+
"learning_rate": 1.5914010471859947e-07,
|
| 436 |
+
"logits/chosen": -1.3577600717544556,
|
| 437 |
+
"logits/rejected": -1.4665708541870117,
|
| 438 |
+
"logps/chosen": -542.1375732421875,
|
| 439 |
+
"logps/rejected": -501.33306884765625,
|
| 440 |
+
"loss": 0.051105821132659913,
|
| 441 |
+
"rewards/accuracies": 1.0,
|
| 442 |
+
"rewards/chosen": 0.9414765238761902,
|
| 443 |
+
"rewards/margins": 6.377541542053223,
|
| 444 |
+
"rewards/rejected": -5.436064720153809,
|
| 445 |
+
"step": 290
|
| 446 |
+
},
|
| 447 |
+
{
|
| 448 |
+
"epoch": 1.8633540372670807,
|
| 449 |
+
"grad_norm": 0.29464995861053467,
|
| 450 |
+
"learning_rate": 7.773313791559545e-08,
|
| 451 |
+
"logits/chosen": -1.350490689277649,
|
| 452 |
+
"logits/rejected": -1.069253921508789,
|
| 453 |
+
"logps/chosen": -596.630859375,
|
| 454 |
+
"logps/rejected": -484.7491149902344,
|
| 455 |
+
"loss": 0.07570538520812989,
|
| 456 |
+
"rewards/accuracies": 0.987500011920929,
|
| 457 |
+
"rewards/chosen": 0.8024793863296509,
|
| 458 |
+
"rewards/margins": 5.369864463806152,
|
| 459 |
+
"rewards/rejected": -4.567385196685791,
|
| 460 |
+
"step": 300
|
| 461 |
+
},
|
| 462 |
+
{
|
| 463 |
+
"epoch": 1.9254658385093166,
|
| 464 |
+
"grad_norm": 0.14690952003002167,
|
| 465 |
+
"learning_rate": 2.492171341840405e-08,
|
| 466 |
+
"logits/chosen": -1.299762487411499,
|
| 467 |
+
"logits/rejected": -1.3709434270858765,
|
| 468 |
+
"logps/chosen": -577.23974609375,
|
| 469 |
+
"logps/rejected": -418.92852783203125,
|
| 470 |
+
"loss": 0.026452887058258056,
|
| 471 |
+
"rewards/accuracies": 1.0,
|
| 472 |
+
"rewards/chosen": 0.9688733816146851,
|
| 473 |
+
"rewards/margins": 5.997436046600342,
|
| 474 |
+
"rewards/rejected": -5.028562545776367,
|
| 475 |
+
"step": 310
|
| 476 |
+
},
|
| 477 |
+
{
|
| 478 |
+
"epoch": 1.9875776397515528,
|
| 479 |
+
"grad_norm": 0.13554854691028595,
|
| 480 |
+
"learning_rate": 1.3292849624821357e-09,
|
| 481 |
+
"logits/chosen": -1.3074995279312134,
|
| 482 |
+
"logits/rejected": -1.5039985179901123,
|
| 483 |
+
"logps/chosen": -596.8470458984375,
|
| 484 |
+
"logps/rejected": -474.729248046875,
|
| 485 |
+
"loss": 0.02374868541955948,
|
| 486 |
+
"rewards/accuracies": 1.0,
|
| 487 |
+
"rewards/chosen": 0.9388774037361145,
|
| 488 |
+
"rewards/margins": 6.432289123535156,
|
| 489 |
+
"rewards/rejected": -5.493411064147949,
|
| 490 |
+
"step": 320
|
| 491 |
+
}
|
| 492 |
+
],
|
| 493 |
+
"logging_steps": 10,
|
| 494 |
+
"max_steps": 322,
|
| 495 |
+
"num_input_tokens_seen": 0,
|
| 496 |
+
"num_train_epochs": 2,
|
| 497 |
+
"save_steps": 500,
|
| 498 |
+
"stateful_callbacks": {
|
| 499 |
+
"TrainerControl": {
|
| 500 |
+
"args": {
|
| 501 |
+
"should_epoch_stop": false,
|
| 502 |
+
"should_evaluate": false,
|
| 503 |
+
"should_log": false,
|
| 504 |
+
"should_save": true,
|
| 505 |
+
"should_training_stop": true
|
| 506 |
+
},
|
| 507 |
+
"attributes": {}
|
| 508 |
+
}
|
| 509 |
+
},
|
| 510 |
+
"total_flos": 0.0,
|
| 511 |
+
"train_batch_size": 2,
|
| 512 |
+
"trial_name": null,
|
| 513 |
+
"trial_params": null
|
| 514 |
+
}
|