Instructions to use eyepyon/rcgemma2_9b_it-finetuned with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use eyepyon/rcgemma2_9b_it-finetuned with Transformers:

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

pipe = pipeline("text-generation", model="eyepyon/rcgemma2_9b_it-finetuned")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("eyepyon/rcgemma2_9b_it-finetuned", dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use eyepyon/rcgemma2_9b_it-finetuned with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "eyepyon/rcgemma2_9b_it-finetuned"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "eyepyon/rcgemma2_9b_it-finetuned",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/eyepyon/rcgemma2_9b_it-finetuned

SGLang

How to use eyepyon/rcgemma2_9b_it-finetuned 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 "eyepyon/rcgemma2_9b_it-finetuned" \
    --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": "eyepyon/rcgemma2_9b_it-finetuned",
		"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 "eyepyon/rcgemma2_9b_it-finetuned" \
        --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": "eyepyon/rcgemma2_9b_it-finetuned",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use eyepyon/rcgemma2_9b_it-finetuned with Docker Model Runner:
```
docker model run hf.co/eyepyon/rcgemma2_9b_it-finetuned
```

eyepyon/rcgemma2_9b_it-finetuned

このモデルは、google/gemma-2-9b-it をベースにLoRAでファインチューニングされたモデルです。

🔧 モデル情報

ベースモデル: google/gemma-2-9b-it
ファインチューニング手法: LoRA (Low-Rank Adaptation)
アテンション実装: eager (Gemma推奨)
量子化: 4ビット (QLoRA)
対応言語: 日本語
タスク: 質問応答

🚀 使用方法

from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel

# ベースモデルとトークナイザーを読み込み
base_model = AutoModelForCausalLM.from_pretrained(
    "google/gemma-2-9b-it",
    torch_dtype="auto",
    device_map="auto",
    attn_implementation="eager",  # Gemma推奨
    trust_remote_code=True
)
tokenizer = AutoTokenizer.from_pretrained("google/gemma-2-9b-it")

# LoRAアダプターを適用
model = PeftModel.from_pretrained(base_model, "eyepyon/rcgemma2_9b_it-finetuned")

# 推論
def generate_response(context, question):
    input_text = f"### コンテキスト:\n{context}\n\n### 質問:\n{question}\n\n### 回答:\n"
    inputs = tokenizer(input_text, return_tensors="pt")
    
    # GPUに移動
    if torch.cuda.is_available():
        inputs = {k: v.to(model.device) for k, v in inputs.items()}
    
    with torch.no_grad():
        outputs = model.generate(
            **inputs,
            max_length=512,
            do_sample=True,
            temperature=0.7,
            top_p=0.9,
            pad_token_id=tokenizer.eos_token_id
        )
    
    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
    if "### 回答:" in response:
        response = response.split("### 回答:")[-1].strip()
    return response

# 使用例
context = "人工知能は機械学習技術を使用してデータから学習します。"
question = "機械学習の特徴は何ですか？"
answer = generate_response(context, question)
print(answer)

📊 入力フォーマット

### コンテキスト:
[背景情報やコンテキスト]

### 質問:
[ユーザーの質問]

### 回答:
[期待される回答]

⚙️ トレーニング設定

LoRAランク: 8
LoRA Alpha: 16
ターゲットモジュール: q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj
学習率: 2e-4
バッチサイズ: 1 × 4 (gradient accumulation)
グラディエントチェックポイント: use_reentrant=False

⚠️ 重要な注意事項

Gemma特有の設定

eager attention必須: attn_implementation="eager"を使用してください
use_cache=False: グラディエントチェックポイントとの互換性のため
use_reentrant=False: 新しい推奨設定

制限事項

主に日本語での質問応答に最適化
生成される回答の事実確認が必要
特定のドメイン知識でのファインチューニング

📄 ライセンス

Apache 2.0 License

🔗 関連リンク

最終更新: 2025年05月25日

Downloads last month: -; Downloads are not tracked for this model. How to track

Model tree for eyepyon/rcgemma2_9b_it-finetuned

Base model

google/gemma-2-9b

Finetuned

google/gemma-2-9b-it

Adapter

(457)

this model