mlx-lm : générer du texte

pip install -U mlx-lm

mlx_lm.generate \
  --model mlx-community/Mistral-7B-Instruct-v0.3-4bit \
  --prompt "Explique le machine learning en trois phrases."

from mlx_lm import load, generate

model, tokenizer = load("mlx-community/Mistral-7B-Instruct-v0.3-4bit")

response = generate(
    model,
    tokenizer,
    prompt="Explique l'attention dans un Transformer.",
    max_tokens=300,
)

print(response)

from mlx_lm import load, generate

model, tokenizer = load("mlx-community/Llama-3.2-3B-Instruct-4bit")

messages = [
    {"role": "system", "content": "Tu réponds de manière claire et concise."},
    {"role": "user", "content": "Quand utiliser la quantification 4-bit ?"},
]

prompt = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True,
)

response = generate(model, tokenizer, prompt=prompt, max_tokens=250)
print(response)

from mlx_lm import load, stream_generate

model, tokenizer = load("mlx-community/Mistral-7B-Instruct-v0.3-4bit")

for token in stream_generate(
    model,
    tokenizer,
    prompt="Raconte une courte histoire de science-fiction.",
    max_tokens=200,
):
    print(token, end="", flush=True)

mlx_lm.server \
  --model mlx-community/Mistral-7B-Instruct-v0.3-4bit \
  --port 8080

from openai import OpenAI

client = OpenAI(
    base_url="http://localhost:8080/v1",
    api_key="not-used",
)

response = client.chat.completions.create(
    model="mlx-community/Mistral-7B-Instruct-v0.3-4bit",
    messages=[
        {"role": "user", "content": "Résume MLX en deux points."}
    ],
)

print(response.choices[0].message.content)