Downloads · 30 days
0
Kaito-F/qwen3-4b-grpo-alfworld-v1
qwen3-4b-grpo-alfworld-v1 is a text generation model from Kaito-F. Use it when you need the model to write or continue text. It is set up for peft. The card lists the license as apache-2.0.
This repository provides a LoRA adapter trained with GiGPO (Group-in-Group Policy Optimization) from Kaito-F/qwen3-4b-instruct-lora-v2.
Downloads · 30 days
0
Access
Public
Updated Feb 24, 2026
Parameters
4B
8.1 GB on disk
Likes
0
Public
Click a slice to open those files.
.safetensors8 GB · 100%
From the Hugging Face model README
This repository provides a LoRA adapter trained with GiGPO (Group-in-Group Policy Optimization) from Kaito-F/qwen3-4b-instruct-lora-v2.
This repository contains LoRA adapter weights only. The base model must be loaded separately.
GiGPO (verl-agent, ICML 2025) extends GRPO with two-level advantage:
This addresses the credit assignment problem in multi-turn agent tasks: "which specific step in a 35-step trajectory caused success/failure?"
| Parameter | Value |
|---|---|
| Base model | Kaito-F/qwen3-4b-instruct-lora-v2 |
| Method | GiGPO (LoRA + 4-bit quantization) |
| LoRA r / alpha | 16 / 32 |
| Learning rate | 1e-06 |
| Epochs | 2 |
| Rollout tasks | 100 x K=4 |
| GiGPO lambda | 0.5 |
| Max steps/episode | 35 |
| Rollout success rate | 44.0% |
from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel
import torch
base = "Kaito-F/qwen3-4b-instruct-lora-v2"
adapter = "your-id/your-repo"
tokenizer = AutoTokenizer.from_pretrained(base)
model = AutoModelForCausalLM.from_pretrained(
base, torch_dtype=torch.float16, device_map="auto"
)
model = PeftModel.from_pretrained(model, adapter)