MKD2World-9B

LoRA adapter on Qwen/Qwen3.5-9B.

Adapter

Item	Value
Type	PEFT LoRA
Rank / alpha / dropout	32 / 64 / 0.05
Task	CAUSAL_LM
Base model	Qwen/Qwen3.5-9B
Weight file	adapter_model.safetensors

Load with PEFT

import torch
from peft import PeftModel
from transformers import AutoProcessor, Qwen3_5ForConditionalGeneration

base_id = "Qwen/Qwen3.5-9B"
adapter_id = "ZhihaoNan/MKD2World-9B"

model = Qwen3_5ForConditionalGeneration.from_pretrained(
    base_id,
    torch_dtype=torch.bfloat16,
    device_map="auto",
)
model = PeftModel.from_pretrained(model, adapter_id)
processor = AutoProcessor.from_pretrained(base_id)

Serve with vLLM

vllm serve Qwen/Qwen3.5-9B \
  --enable-lora \
  --lora-modules MKD2World-9B=ZhihaoNan/MKD2World-9B \
  --max-lora-rank 64 \
  --max-model-len 60000 \
  --limit-mm-per-prompt '{"image": 1}'