amphora/qwen3-30b-dasd-peft

Qwen3-4B Instruct distill LoRA

LoRA adapter from offline distillation (teacher reasoning traces) on Qwen/Qwen3-30B-A3B-Base.

Load

from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel

base = "Qwen/Qwen3-30B-A3B-Base"
adapter = "talzoomanzoo/qwen3-4b-instruct-distill-lora"

tokenizer = AutoTokenizer.from_pretrained(base)
model = AutoModelForCausalLM.from_pretrained(base, torch_dtype="auto", device_map="auto")
model = PeftModel.from_pretrained(model, adapter)