subbuc/qwen3-8b-sft-lmsys

Qwen3-8B SFT LMSYS (Baseline)

This is the SFT baseline model for comparison with the DPO version.

Training Details

• Base Model: unsloth/Qwen3-8B-4bit
• Training Method: Supervised Fine-Tuning (SFT)
• Dataset: LMSYS Arena Human Preference 55k (chosen responses only)
• Training Steps: 60

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("subbuc/qwen3-8b-sft-lmsys")
tokenizer = AutoTokenizer.from_pretrained("subbuc/qwen3-8b-sft-lmsys")

# Your inference code here