30 lines
477 B
Markdown
30 lines
477 B
Markdown
---
|
|
license: apache-2.0
|
|
datasets:
|
|
- shibing624/alpaca-zh
|
|
- hiyouga/DPO-En-Zh-20k
|
|
language:
|
|
- zh
|
|
- en
|
|
metrics:
|
|
- accuracy
|
|
base_model: Qwen/Qwen2.5-1.5B
|
|
pipeline_tag: text-generation
|
|
library_name: transformers
|
|
tags:
|
|
- qwen
|
|
- dpo
|
|
- sft
|
|
- alignment
|
|
- ollama
|
|
---
|
|
|
|
A lightweight model fine-tuned from Qwen2.5-1.5B via SFT and DPO alignment. Enjoy!
|
|
|
|
| Phase | Metric | Value |
|
|
| :--- | :--- | :--- |
|
|
| SFT | Final Loss | 1.65 |
|
|
| DPO | Accuracies | 70.4% |
|
|
| DPO | Margins | 1.022 |
|
|
|