File size: 2,073 Bytes
028f06a
 
 
 
 
8087bbe
028f06a
 
23cc740
 
 
 
 
 
 
 
 
56421a8
028f06a
 
 
 
 
b4957d9
 
028f06a
8087bbe
028f06a
 
 
8087bbe
b4957d9
8087bbe
b4957d9
 
 
 
 
 
 
 
 
8087bbe
b4957d9
8087bbe
028f06a
 
56421a8
028f06a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
7860df6
56421a8
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
import os
import torch
from fastapi import FastAPI
from pydantic import BaseModel
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
from peft import PeftModel, PeftConfig
import uvicorn

from huggingface_hub import login

# Authenticate with Hugging Face Hub
HF_TOKEN = os.getenv("HF_TOKEN")
if HF_TOKEN:
    login(token=HF_TOKEN)
else:
    raise ValueError("Hugging Face token not found. Please set the HF_TOKEN environment variable.")

# Define a Pydantic model for request validation
class Query(BaseModel):
    text: str

app = FastAPI(title="Financial Chatbot API")

# Load the base model
base_model_name = "meta-llama/Llama-3.2-3B"  # Update if using a different base model
model = AutoModelForCausalLM.from_pretrained(
    base_model_name,
    device_map="auto",
    trust_remote_code=True
)

# Load adapter from your checkpoint with a workaround for 'eva_config'
peft_model_id = "Phoenix21/llama-3-2-3b-finetuned-finance_checkpoint2"
# Load the PEFT configuration first
peft_config = PeftConfig.from_pretrained(peft_model_id)
# Remove 'eva_config' if it exists in the configuration
peft_config_dict = peft_config.to_dict()
if "eva_config" in peft_config_dict:
    peft_config_dict.pop("eva_config")
    peft_config = PeftConfig.from_dict(peft_config_dict)
# Load the adapter using the filtered configuration
model = PeftModel.from_pretrained(model, peft_model_id, config=peft_config)

# Load tokenizer from the base model
tokenizer = AutoTokenizer.from_pretrained(base_model_name, trust_remote_code=True)
tokenizer.pad_token = tokenizer.eos_token

# Create a text-generation pipeline
chat_pipe = pipeline(
    "text-generation",
    model=model,
    tokenizer=tokenizer,
    max_new_tokens=256,
    temperature=0.7,
    top_p=0.95,
)

@app.post("/generate")
def generate(query: Query):
    prompt = f"Question: {query.text}\nAnswer: "
    response = chat_pipe(prompt)[0]["generated_text"]
    return {"response": response}

if __name__ == "__main__":
    port = int(os.environ.get("PORT", 7860))
    uvicorn.run(app, host="0.0.0.0", port=port)