add mlx support

[davidberenstein1957]

Use it individually

from distilabel.models.llms import MlxLLM

llm = MlxLLM(model="mlx-community/Meta-Llama-3.1-8B-Instruct-4bit")

llm.load()

# Call the model
output = llm.generate_outputs(inputs=[[{"role": "user", "content": "Hello world!"}]])

Use it with magpie

from distilabel.models.llms.mlx import MlxLLM
from distilabel.steps.tasks import Magpie

model = MlxLLM(
    path_or_hf_repo="mlx-community/Meta-Llama-3.1-8B-Instruct-8bit",
    use_magpie_template=True,
    magpie_pre_query_template="llama3",
)
task = Magpie(llm=model)
task.load()

It is relatively easy to spin up an mlx server, but no public Python API clients are available except for LangChain https://python.langchain.com/docs/integrations/chat/mlx/. Currently, the OpenAIAPI does not align with the payloads from both the chat and the text generation API.

[github-actions]

Documentation for this PR has been built. You can view it at: https://distilabel.argilla.io/pr-1089/

[codspeed-hq]

CodSpeed Performance Report

Merging #1089 will not alter performance

_{Comparing feat/add-support-mlx (162f73e) with develop (f1f7d77)}

Summary

✅ 1 untouched benchmarks