Instructions to use mlabonne/phixtral-2x2_8 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use mlabonne/phixtral-2x2_8 with Transformers:

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

pipe = pipeline("text-generation", model="mlabonne/phixtral-2x2_8", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("mlabonne/phixtral-2x2_8", trust_remote_code=True, dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use mlabonne/phixtral-2x2_8 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "mlabonne/phixtral-2x2_8"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "mlabonne/phixtral-2x2_8",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/mlabonne/phixtral-2x2_8

SGLang

How to use mlabonne/phixtral-2x2_8 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 "mlabonne/phixtral-2x2_8" \
    --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": "mlabonne/phixtral-2x2_8",
		"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 "mlabonne/phixtral-2x2_8" \
        --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": "mlabonne/phixtral-2x2_8",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use mlabonne/phixtral-2x2_8 with Docker Model Runner:
```
docker model run hf.co/mlabonne/phixtral-2x2_8
```

Unable to replicate using LazyMergeKit Colab

by sudhir2016 - opened Jan 17, 2024

Discussion

sudhir2016

Jan 17, 2024

I tried to replicate using your LazyMergeKit notebook as a learning exercise. This yaml_config gives an error that positive_prompts are same for both experts.
base_model: cognitivecomputations/dolphin-2_6-phi-2
gate_mode: cheap_embed
experts:

source_model: cognitivecomputations/dolphin-2_6-phi-2
positive_prompts: [""]
source_model: lxuechen/phi-2-dpo
positive_prompts: [""]

Then I changed yaml_config like this
MODEL_NAME = "Phixtral-Merge"
yaml_config = """
base_model: cognitivecomputations/dolphin-2_6-phi-2
gate_mode: cheap_embed
experts:

source_model: cognitivecomputations/dolphin-2_6-phi-2
positive_prompts: ["code"]
source_model: lxuechen/phi-2-dpo
positive_prompts: ["math"]
"""
This time I got this error.
File "/content/mergekit/mergekit/io/lazy_tensor_loader.py", line 127, in get_tensor
raise KeyError(key)
KeyError: 'model.embed_tokens.weight'

Please help.

mlabonne

Owner Jan 18, 2024

Sorry that's normal, I modified mergekit's code to produce phixtral. This branch hasn't been released yet (still need to work on it).

sudhir2016

Jan 18, 2024

OK thanks.

sudhir2016 changed discussion status to closed Jan 18, 2024

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment