Modelpublic

qisein/Qwen3-8B-OPSD-ep64

sourceHugging Faceapache-2.0updated 29d agoView on Hugging Face

0likes23downloads

Model Card

Qwen3-8B OPSD adapter

A PEFT LoRA adapter produced by the unprojected privileged-teacher self-distillation branch (OPSD). The adapter must be loaded on the pinned Qwen3-8B base model; it is not a merged model.

Load

python

import torch
from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer


tokenizer = AutoTokenizer.from_pretrained(base_id, revision=base_revision)
base = AutoModelForCausalLM.from_pretrained(
    base_id,
    revision=base_revision,
    torch_dtype=torch.bfloat16,
    device_map="auto",
)
model = PeftModel.from_pretrained(base, adapter_id, is_trainable=False)
model.eval()

This research artifact inherits the base model's license.