grpo_0328_step200

LoRA adapter trained from Qwen/Qwen3-VL-8B-Instruct.

Usage

Load this adapter with PEFT on top of the base model:

from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel

base_model = "Qwen/Qwen3-VL-8B-Instruct"
adapter_repo = "elaine1wan/grpo_0328_step200"

tokenizer = AutoTokenizer.from_pretrained(base_model)
model = AutoModelForCausalLM.from_pretrained(base_model)
model = PeftModel.from_pretrained(model, adapter_repo)

Training checkpoint

Exported from:

verl/checkpoints/qwen3_vl_8b_grpo_multiturn_new_0328_run3/global_step_200/actor/lora_adapter