Instructions to use patrick-h/Ablit-PhiTest with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use patrick-h/Ablit-PhiTest with Transformers:

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

pipe = pipeline("text-generation", model="patrick-h/Ablit-PhiTest", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("patrick-h/Ablit-PhiTest", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("patrick-h/Ablit-PhiTest", trust_remote_code=True)
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=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use patrick-h/Ablit-PhiTest with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/patrick-h/Ablit-PhiTest

SGLang

How to use patrick-h/Ablit-PhiTest 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 "patrick-h/Ablit-PhiTest" \
    --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": "patrick-h/Ablit-PhiTest",
		"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 "patrick-h/Ablit-PhiTest" \
        --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": "patrick-h/Ablit-PhiTest",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use patrick-h/Ablit-PhiTest with Docker Model Runner:
```
docker model run hf.co/patrick-h/Ablit-PhiTest
```

Model Summary

A modified version of the Phi-3-Mini-128K-Instruct. Ablitered following this very indepth guide https://huggingface.co/blog/mlabonne/abliteration. Results weren't quite perfect and more testing is needed, model does refuse questionable requests less than previous but does still refuse some requests.

Bias, Risks, and Limitations

Use at your own risk. I have no idea what this model's biases or limits are. I wanted to test how this compares after the abliteration process.

Downloads last month: 2

Safetensors

Model size

4B params

Tensor type

BF16