Spaces:
Runtime error
Runtime error
File size: 2,435 Bytes
06696b5 |
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 71 72 73 74 75 76 77 78 79 80 81 |
import gradio as gr
import spaces
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
# from retriever.vectordb_rerank import search_documents # π§ RAG κ²μκΈ° λΆλ¬μ€κΈ°
from services.rag_pipeline import rag_pipeline
model_name = "dasomaru/gemma-3-4bit-it-demo"
# 1. λͺ¨λΈ/ν ν¬λμ΄μ 1ν λ‘λ©
tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
# π modelμ CPUλ‘λ§ λ¨Όμ μ¬λ¦Ό (GPU μμ§ μμ)
model = AutoModelForCausalLM.from_pretrained(
model_name,
torch_dtype=torch.float16, # 4bit modelμ΄λκΉ
device_map="auto", # β
μ€μ: μλμΌλ‘ GPU ν λΉ
trust_remote_code=True,
)
# 2. μΊμ κ΄λ¦¬
search_cache = {}
@spaces.GPU(duration=300)
def generate_response(query: str):
tokenizer = AutoTokenizer.from_pretrained(
"dasomaru/gemma-3-4bit-it-demo",
trust_remote_code=True,
)
model = AutoModelForCausalLM.from_pretrained(
"dasomaru/gemma-3-4bit-it-demo",
torch_dtype=torch.float16, # 4bit modelμ΄λκΉ
device_map="auto", # β
μ€μ: μλμΌλ‘ GPU ν λΉ
trust_remote_code=True,
)
model.to("cuda")
if query in search_cache:
print(f"β‘ μΊμ μ¬μ©: '{query}'")
return search_cache[query]
# π₯ rag_pipelineμ νΈμΆν΄μ κ²μ + μμ±
# κ²μ
top_k = 5
results = rag_pipeline(query, top_k=top_k)
# κ²°κ³Όκ° listμΌ κ²½μ° ν©μΉκΈ°
if isinstance(results, list):
results = "\n\n".join(results)
search_cache[query] = results
# return results
inputs = tokenizer(results, return_tensors="pt").to(model.device) # β
model.device
outputs = model.generate(
**inputs,
max_new_tokens=512,
temperature=0.7,
top_p=0.9,
top_k=50,
do_sample=True,
)
return tokenizer.decode(outputs[0], skip_special_tokens=True)
# 3. Gradio μΈν°νμ΄μ€
demo = gr.Interface(
fn=generate_response,
# inputs=gr.Textbox(lines=2, placeholder="μ§λ¬Έμ μ
λ ₯νμΈμ"),
inputs="text",
outputs="text",
title="Law RAG Assistant",
description="λ²λ Ή κΈ°λ° RAG νμ΄νλΌμΈ ν
μ€νΈ",
)
# demo.launch(server_name="0.0.0.0", server_port=7860) # π API λ°°ν¬ μ€λΉ κ°λ₯
# demo.launch()
demo.launch(debug=True)
|