26 lines
720 B
Markdown
26 lines
720 B
Markdown
|
---
|
||
|
|
base_model: unsloth/qwen2.5-1.5b-instruct-unsloth-bnb-4bit
|
||
|
|
tags:
|
||
|
|
- text-generation-inference
|
||
|
|
- transformers
|
||
|
|
- unsloth
|
||
|
|
- qwen2
|
||
|
|
license: apache-2.0
|
||
|
|
language:
|
||
|
|
- en
|
||
|
|
---
|
||
|
|
|
||
|
|
# Reasoning Qwen2.5 1.5B
|
||
|
|
|
||
|
|
Reasoning Qwen2.5 1.5B model to solve grade-level math with explicit structure: a short scratchpad in `<reasoning>…</reasoning>` and a single final number in `<answer>…</answer>`.
|
||
|
|
|
||
|
|
Training: https://github.com/KickItLikeShika/llm-reasoning
|
||
|
|
|
||
|
|
I split the training in two stages:
|
||
|
|
1. Short LoRA SFT on 100 random GSM8K training examples, to teach format + roughly sensible traces, not to maximize benchmark score.
|
||
|
|
2. GRPO on top of that adapter for 2,000 steps.
|
||
|
|
|
||
|
|
W&B Report: https://api.wandb.ai/links/egyttsteam/uja0job9
|
||
|
|
|
||
|
|
|