Instructions to use Hyukkyu/Qwen3-8B-Base-RAQUEL-TOFU-Unlearn-GA-GD-LoRA-v1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Hyukkyu/Qwen3-8B-Base-RAQUEL-TOFU-Unlearn-GA-GD-LoRA-v1 with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="Hyukkyu/Qwen3-8B-Base-RAQUEL-TOFU-Unlearn-GA-GD-LoRA-v1")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Hyukkyu/Qwen3-8B-Base-RAQUEL-TOFU-Unlearn-GA-GD-LoRA-v1")
model = AutoModelForCausalLM.from_pretrained("Hyukkyu/Qwen3-8B-Base-RAQUEL-TOFU-Unlearn-GA-GD-LoRA-v1", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=256)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Hyukkyu/Qwen3-8B-Base-RAQUEL-TOFU-Unlearn-GA-GD-LoRA-v1 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Hyukkyu/Qwen3-8B-Base-RAQUEL-TOFU-Unlearn-GA-GD-LoRA-v1"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Hyukkyu/Qwen3-8B-Base-RAQUEL-TOFU-Unlearn-GA-GD-LoRA-v1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Hyukkyu/Qwen3-8B-Base-RAQUEL-TOFU-Unlearn-GA-GD-LoRA-v1

SGLang

How to use Hyukkyu/Qwen3-8B-Base-RAQUEL-TOFU-Unlearn-GA-GD-LoRA-v1 with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "Hyukkyu/Qwen3-8B-Base-RAQUEL-TOFU-Unlearn-GA-GD-LoRA-v1" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Hyukkyu/Qwen3-8B-Base-RAQUEL-TOFU-Unlearn-GA-GD-LoRA-v1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "Hyukkyu/Qwen3-8B-Base-RAQUEL-TOFU-Unlearn-GA-GD-LoRA-v1" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Hyukkyu/Qwen3-8B-Base-RAQUEL-TOFU-Unlearn-GA-GD-LoRA-v1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Hyukkyu/Qwen3-8B-Base-RAQUEL-TOFU-Unlearn-GA-GD-LoRA-v1 with Docker Model Runner:
```
docker model run hf.co/Hyukkyu/Qwen3-8B-Base-RAQUEL-TOFU-Unlearn-GA-GD-LoRA-v1
```

Qwen3-8B-Base-RAQUEL-TOFU-Unlearn-GA-GD-LoRA-v1 / training_recipe.json

Hyukkyu

RAQUEL2 LoRA release (tofu_qwen3_ga, data ac827565)

b739e59 verified 10 days ago

Raw History Blame Contribute Delete

2.32 kB

	{
	"adapter_weights_sha256": "cdf3f3b5a77fc78fa418932bcfb2245687923202678e058879e8c2d52dc45cf8",
	"base_model": "Hyukkyu/Qwen3-8B-Base-RAQUEL-TOFU-M-orig-LoRA-v1",
	"base_revision": "4f55767275e481122f80a29a53fc0bacf07b859a",
	"data": {
	"forget_rows": 3600,
	"max_length": 512,
	"pairing": "all forget and all retain source QA; the smaller side cycled in a seeded order, paired 1:1",
	"qa_serialization": "paper_v1",
	"repository": "Hyukkyu/RAQUEL2-ICLR",
	"retain_rows": 3600,
	"revision": "ac82756570fcce84441fb413ab523de8da679efd"
	},
	"early_stopping": {
	"config": {
	"cached_decode_agreement": 1.0,
	"continue_after_stop": true,
	"every_steps": 4,
	"max_new_tokens": 64,
	"rule": "max retain ROUGE-L s.t. forget ROUGE-L <= M_ret (margin 0; user 2026-09-24)",
	"seed": 0,
	"subset_size": 100,
	"target_forget_rouge": 0.3903,
	"target_source": "early_stop/tofu_qwen3_M_ret.json"
	},
	"released_adapter": "step 352",
	"rule": "max retain ROUGE-L s.t. forget ROUGE-L <= M_ret (user 2026-09-24)",
	"selected_forget_rougeL_recall": 0.0017251051893408134,
	"selected_retain_rougeL_recall": 0.9983076923076923,
	"selected_step": 352
	},
	"export_merge_algorithm": "fp32_sum_then_cast_once",
	"export_precision": "bfloat16",
	"learning_rate_note": "10x the full-parameter protocol's learning rates (LoRA convention), fixed before any LoRA result",
	"lora": {
	"adapter_dtype": "float32",
	"alpha": 128.0,
	"base_dtype": "bfloat16",
	"dropout": 0.05,
	"initialisation": "peft default (B=0: start == M_orig)",
	"rank": 64,
	"targets": [
	"q_proj",
	"k_proj",
	"v_proj",
	"o_proj",
	"gate_proj",
	"up_proj",
	"down_proj"
	]
	},
	"merged_weights_sha256": "39452776724d67b76e5562bf90d61f8f564883460341f42fcf96e21b94a0a2f5",
	"method": {
	"idk_target": "I don't know.",
	"method": "ga",
	"preference_beta": 0.1,
	"retain_weight": 4.0
	},
	"optimizer": {
	"global_batch_size": 32,
	"learning_rate": 0.0001,
	"scheduler": "constant",
	"seed": 0,
	"weight_decay": 0.0
	},
	"parameter_counts": {
	"total": 8365323264,
	"trainable": 174587904
	},
	"schedule": {
	"epochs": 5,
	"total_steps": 565
	},
	"training_complete": true
	}