gemma-4-E4B
Collection
gemma-4-E4B • 10 items • Updated
How to use NullpoLab/gemma-4-E4B-it-Heretic-ARA-Refusals5 with Transformers:
# Use a pipeline as a high-level helper
from transformers import pipeline
pipe = pipeline("image-text-to-text", model="NullpoLab/gemma-4-E4B-it-Heretic-ARA-Refusals5")
messages = [
{
"role": "user",
"content": [
{"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
{"type": "text", "text": "What animal is on the candy?"}
]
},
]
pipe(text=messages) # Load model directly
from transformers import AutoProcessor, AutoModelForMultimodalLM
processor = AutoProcessor.from_pretrained("NullpoLab/gemma-4-E4B-it-Heretic-ARA-Refusals5")
model = AutoModelForMultimodalLM.from_pretrained("NullpoLab/gemma-4-E4B-it-Heretic-ARA-Refusals5", device_map="auto")
messages = [
{
"role": "user",
"content": [
{"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
{"type": "text", "text": "What animal is on the candy?"}
]
},
]
inputs = processor.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=40)
print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))How to use NullpoLab/gemma-4-E4B-it-Heretic-ARA-Refusals5 with vLLM:
# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "NullpoLab/gemma-4-E4B-it-Heretic-ARA-Refusals5"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
-H "Content-Type: application/json" \
--data '{
"model": "NullpoLab/gemma-4-E4B-it-Heretic-ARA-Refusals5",
"messages": [
{
"role": "user",
"content": [
{
"type": "text",
"text": "Describe this image in one sentence."
},
{
"type": "image_url",
"image_url": {
"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
}
}
]
}
]
}'docker model run hf.co/NullpoLab/gemma-4-E4B-it-Heretic-ARA-Refusals5
How to use NullpoLab/gemma-4-E4B-it-Heretic-ARA-Refusals5 with SGLang:
# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
--model-path "NullpoLab/gemma-4-E4B-it-Heretic-ARA-Refusals5" \
--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": "NullpoLab/gemma-4-E4B-it-Heretic-ARA-Refusals5",
"messages": [
{
"role": "user",
"content": [
{
"type": "text",
"text": "Describe this image in one sentence."
},
{
"type": "image_url",
"image_url": {
"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
}
}
]
}
]
}'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 "NullpoLab/gemma-4-E4B-it-Heretic-ARA-Refusals5" \
--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": "NullpoLab/gemma-4-E4B-it-Heretic-ARA-Refusals5",
"messages": [
{
"role": "user",
"content": [
{
"type": "text",
"text": "Describe this image in one sentence."
},
{
"type": "image_url",
"image_url": {
"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
}
}
]
}
]
}'How to use NullpoLab/gemma-4-E4B-it-Heretic-ARA-Refusals5 with Docker Model Runner:
docker model run hf.co/NullpoLab/gemma-4-E4B-it-Heretic-ARA-Refusals5
google/gemma-4-E4B-it を Heretic v1.2.0 の Arbitrary-Rank Ablation (ARA) 手法を用いて検閲解除したモデルです。
ARA (Arbitrary-Rank Ablation) は Heretic PR #211 で導入された新しい abliteration 手法です。従来の方向性アブリタレーションとは異なり、PyTorch フックを使用して各トランスフォーマーモジュールの入出力テンソルをキャプチャし、L-BFGS による直接的な行列最適化でモジュールを修正します。
最適化は以下の3つの競合する目標のバランスを取ります:
| パラメータ | 値 |
|---|---|
| start_layer_index | 19 |
| end_layer_index | 42 |
| preserve_good_behavior_weight | 0.7662 |
| steer_bad_behavior_weight | 0.0007 |
| overcorrect_relative_weight | 0.9544 |
| neighbor_count | 15 |
| 指標 | このモデル | 元のモデル (google/gemma-4-E4B-it) |
|---|---|---|
| 拒否率 | 5/100 | 99/100 |
| KL発散 | 0.0256 | 0 (定義上) |
評価は mlabonne/harmful_behaviors(テスト 100 プロンプト)で拒否率を、mlabonne/harmless_alpaca(テスト 100 プロンプト)で KL 発散を計測しました。
3b70fe5)