Spaces:
Running
Running
File size: 2,192 Bytes
6c0215b 8b03c54 812d4f6 6c0215b 042861c 6c0215b 9f70441 6c0215b 8b03c54 6c0215b 8b03c54 6c0215b 8b03c54 6c0215b 8b03c54 6c0215b |
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 67 68 69 70 |
from fastapi import FastAPI, Query
from pydantic import BaseModel
import cloudscraper
from bs4 import BeautifulSoup
from transformers import pipeline
import torch
import re
import os
os.environ["HF_HOME"] = "/home/user/huggingface"
os.environ["TRANSFORMERS_CACHE"] = "/home/user/huggingface"
app = FastAPI()
class ThreadResponse(BaseModel):
question: str
replies: list[str]
def clean_text(text: str) -> str:
text = text.strip()
text = re.sub(r"\b\d+\s*likes?,?\s*\d*\s*replies?$", "", text, flags=re.IGNORECASE).strip()
return text
@app.get("/scrape", response_model=ThreadResponse)
def scrape(url: str = Query(...)):
scraper = cloudscraper.create_scraper()
response = scraper.get(url)
if response.status_code == 200:
soup = BeautifulSoup(response.content, 'html.parser')
comment_containers = soup.find_all('div', class_='post__content')
if comment_containers:
question = clean_text(comment_containers[0].get_text(strip=True, separator="\n"))
replies = [clean_text(comment.get_text(strip=True, separator="\n")) for comment in comment_containers[1:]]
return ThreadResponse(question=question, replies=replies)
return ThreadResponse(question="", replies=[])
MODEL_NAME = "google/flan-t5-small"
# Load the pipeline once at startup with device auto-mapping
text_generator = pipeline(
"text2text-generation",
model=MODEL_NAME,
device=0 if torch.cuda.is_available() else -1,
max_new_tokens=512,
temperature=0.5
)
class PromptRequest(BaseModel):
prompt: str
@app.post("/generate")
async def generate_text(request: PromptRequest):
# Use the pipeline to generate text directly
output = text_generator(request.prompt)[0]['generated_text']
# Extract reasoning and content parts if thinking tags are present
if "</think>" in output:
reasoning_content = output.split("</think>")[0].strip()
content = output.split("</think>")[1].strip().rstrip("</s>")
else:
reasoning_content = ""
content = output.strip().rstrip("</s>")
return {
"reasoning_content": reasoning_content,
"generated_text": content
}
|