Spaces:
Sleeping
Sleeping
File size: 1,885 Bytes
5fa76ab 824a4d9 e6fa3d8 473963a 5fa76ab 824a4d9 5fa76ab 2d5e645 5fa76ab 824a4d9 5fa76ab 2d5e645 5fa76ab 2d5e645 5fa76ab 824a4d9 5fa76ab 824a4d9 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 |
from fastapi import FastAPI
from pydantic import BaseModel
from huggingface_hub import InferenceClient
import uvicorn
app = FastAPI()
client = InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1")
class Item(BaseModel):
prompt: str
history: list
system_prompt: str
temperature: float = 0.1 # Updated minimum value for stability
max_new_tokens: int = 1048
top_p: float = 0.15
repetition_penalty: float = 1.0
def format_prompt(message, history):
prompt = "<s>"
for user_prompt, bot_response in history:
prompt += f"[INST] {user_prompt} [/INST] {bot_response}</s> "
prompt += f"[INST] {message} [/INST]"
return prompt
def generate(item: Item):
temperature = max(0.1, item.temperature)
top_p = float(item.top_p)
generate_kwargs = dict(
temperature=temperature,
max_new_tokens=item.max_new_tokens,
top_p=top_p,
repetition_penalty=item.repetition_penalty,
do_sample=True,
seed=42,
)
formatted_prompt = format_prompt(f"{item.system_prompt}, {item.prompt}", item.history)
response_stream = client.text_generation(
formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False
)
output = ""
try:
for response in response_stream:
if hasattr(response, "token") and response.token is not None:
output += response.token.text
else:
output += str(response) # Obsługa przypadków, gdy `response` nie ma atrybutu `token`
except Exception as e:
output = f"Błąd: {str(e)}"
return output
@app.post("/generate/")
async def generate_text(item: Item):
try:
return {"response": generate(item)}
except Exception as e:
return {"error": str(e)}
if __name__ == "__main__":
uvicorn.run(app, host="0.0.0.0", port=8000)
|