sft_quant_d8

SFT checkpoints of Qwen3-4B trained on the sft_quant_d8 dataset.

Branches:

• step1000: global_step_1000 checkpoint
• step2000: global_step_2000 checkpoint
• final (global_step_2292): final checkpoint at end of training

Load a specific checkpoint with:

from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("wtl666wtl/sft_quant_d8", revision="step1000")