test_model.py

from huggingface_hub import InferenceClient
import os

token = os.getenv("HF_TOKEN")
if not token:
    print("Установи HF_TOKEN!")
    exit()

client = InferenceClient(token=token)
messages = [{"role": "user", "content": "Решить 2+2=?"}]
response = client.chat_completion(
    model="deepseek-ai/DeepSeek-R1-Distill-Qwen-32B",
    messages=messages,
    max_tokens=50,
    temperature=0.5
)
print("Ответ:", response.choices[0].message["content"])