LamiaYT's picture
Fix
9f29ca9
raw
history blame
13.5 kB
import os
import gradio as gr
import requests
import json
import re
import numexpr
import pandas as pd
import time
import torch
import math
import pdfminer
from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
from duckduckgo_search import DDGS
from pdfminer.high_level import extract_text
from bs4 import BeautifulSoup
import html2text
from typing import Dict, Any, List, Tuple, Callable
from dotenv import load_dotenv
# --- Load Environment Variables ---
load_dotenv()
SERPER_API_KEY = os.getenv("SERPER_API_KEY")
# --- Constants ---
DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
MAX_STEPS = 6
MAX_TOKENS = 256
MODEL_NAME = "microsoft/Phi-3-mini-4k-instruct"
# --- Configure Environment for Hugging Face Spaces ---
os.environ["PIP_BREAK_SYSTEM_PACKAGES"] = "1"
os.environ["HF_HUB_DISABLE_SYMLINKS_WARNING"] = "1"
os.environ["BITSANDBYTES_NOWELCOME"] = "1"
# --- Load Quantized Model ---
print("Loading quantized model...")
start_time = time.time()
# Configure 4-bit quantization
quant_config = BitsAndBytesConfig(
load_in_4bit=True,
bnb_4bit_quant_type="nf4",
bnb_4bit_use_double_quant=True,
bnb_4bit_compute_dtype=torch.bfloat16
)
# Load model and tokenizer
model = AutoModelForCausalLM.from_pretrained(
MODEL_NAME,
device_map="auto",
quantization_config=quant_config,
trust_remote_code=True
)
tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
load_time = time.time() - start_time
print(f"Model loaded in {load_time:.2f} seconds")
# --- Tools for GAIA Agent ---
def web_search(query: str) -> str:
"""Search the web using DuckDuckGo or Serper API"""
try:
if SERPER_API_KEY:
# Use Serper API if key is available
params = {
'q': query,
'num': 3,
'hl': 'en',
'gl': 'us'
}
headers = {
'X-API-KEY': SERPER_API_KEY,
'Content-Type': 'application/json'
}
response = requests.post(
'https://google.serper.dev/search',
headers=headers,
json=params
)
results = response.json()
if 'organic' in results:
return json.dumps([r['title'] + ": " + r['snippet'] for r in results['organic'][:3]])
return "No results found"
else:
# Fallback to DuckDuckGo
with DDGS() as ddgs:
results = [r for r in ddgs.text(query, max_results=3)]
return json.dumps([r['title'] + ": " + r['body'] for r in results])
except Exception as e:
return f"Search error: {str(e)}"
def calculator(expression: str) -> str:
"""Evaluate mathematical expressions safely"""
try:
return str(numexpr.evaluate(expression))
except Exception as e:
return f"Calculation error: {str(e)}"
def read_pdf(file_path: str) -> str:
"""Extract text from PDF files"""
try:
return extract_text(file_path)[:2000] # Limit to first 2000 characters
except Exception as e:
return f"PDF read error: {str(e)}"
def read_webpage(url: str) -> str:
"""Fetch and extract text from web pages"""
try:
response = requests.get(url, timeout=10)
soup = BeautifulSoup(response.text, 'html.parser')
return soup.get_text(separator=' ', strip=True)[:2000] # Limit text
except Exception as e:
return f"Webpage read error: {str(e)}"
TOOLS = {
"web_search": web_search,
"calculator": calculator,
"read_pdf": read_pdf,
"read_webpage": read_webpage
}
# --- GAIA Agent Implementation ---
class GAIA_Agent:
def __init__(self):
self.tools = TOOLS
self.history = []
self.system_prompt = (
"You are an expert GAIA problem solver. Use these tools: {web_search, calculator, read_pdf, read_webpage}.\n"
"Guidelines:\n"
"1. Think step-by-step. Explain reasoning\n"
"2. Use tools for calculations, searches, or file operations\n"
"3. Tools must be called as: ```json\n{'tool': 'tool_name', 'args': {'arg1': value}}```\n"
"4. Final Answer must be exact and standalone\n\n"
"Example:\n"
"Question: \"What's the population density of France? (File: france_data.pdf)\"\n"
"Thought: Need population and area. Read PDF first.\n"
"Action: ```json\n{'tool': 'read_pdf', 'args': {'file_path': 'france_data.pdf'}}```\n"
"Observation: Population: 67.8M, Area: 643,801 km²\n"
"Thought: Now calculate density: 67,800,000 / 643,801\n"
"Action: ```json\n{'tool': 'calculator', 'args': {'expression': '67800000 / 643801'}}```\n"
"Observation: 105.32\n"
"Final Answer: 105.32 people/km²"
)
def __call__(self, question: str) -> str:
print(f"\nProcessing: {question[:80]}...")
self.history = [f"Question: {question}"]
for step in range(MAX_STEPS):
prompt = self._build_prompt()
response = self._call_model(prompt)
if "Final Answer" in response:
answer = response.split("Final Answer:")[-1].strip()
print(f"Final Answer: {answer}")
return answer
tool_call = self._parse_tool_call(response)
if tool_call:
tool_name, args = tool_call
observation = self._use_tool(tool_name, args)
self.history.append(f"Observation: {observation}")
else:
self.history.append(f"Thought: {response}")
return "Agent couldn't find solution within step limit"
def _build_prompt(self) -> str:
prompt = "<|system|>\n" + self.system_prompt + "<|end|>\n"
prompt += "<|user|>\n" + "\n".join(self.history) + "<|end|>\n"
prompt += "<|assistant|>"
return prompt
def _call_model(self, prompt: str) -> str:
start_time = time.time()
inputs = tokenizer(prompt, return_tensors="pt", return_attention_mask=True).to(model.device)
outputs = model.generate(
**inputs,
max_new_tokens=MAX_TOKENS,
temperature=0.01,
do_sample=True,
pad_token_id=tokenizer.eos_token_id
)
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
response = response.split("<|assistant|>")[-1].strip()
gen_time = time.time() - start_time
print(f"Generated {len(response)} tokens in {gen_time:.2f}s: {response[:60]}...")
return response
def _parse_tool_call(self, text: str) -> Tuple[str, Dict] or None:
try:
json_match = re.search(r'```json\s*({.*?})\s*```', text, re.DOTALL)
if json_match:
tool_call = json.loads(json_match.group(1))
return tool_call["tool"], tool_call["args"]
except Exception as e:
print(f"Tool parse error: {str(e)}")
return None
def _use_tool(self, tool_name: str, args: Dict) -> str:
if tool_name not in self.tools:
return f"Error: Unknown tool {tool_name}"
print(f"Using tool: {tool_name}({args})")
try:
start_time = time.time()
result = self.tools[tool_name](**args)
exec_time = time.time() - start_time
print(f"Tool executed in {exec_time:.2f}s")
return str(result)[:500] # Truncate long outputs
except Exception as e:
return f"Tool error: {str(e)}"
# --- Evaluation Runner ---
def run_and_submit_all(profile: gr.OAuthProfile | None):
"""Fetches questions, runs agent, submits answers, and displays results"""
space_id = os.getenv("SPACE_ID")
if profile:
username = f"{profile.username}"
print(f"User logged in: {username}")
else:
print("User not logged in.")
return "Please Login to Hugging Face with the button.", None
api_url = DEFAULT_API_URL
questions_url = f"{api_url}/questions"
submit_url = f"{api_url}/submit"
try:
agent = GAIA_Agent()
except Exception as e:
print(f"Error instantiating agent: {e}")
return f"Error initializing agent: {e}", None
agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main"
print(agent_code)
# Fetch Questions
print(f"Fetching questions from: {questions_url}")
try:
response = requests.get(questions_url, timeout=15)
response.raise_for_status()
questions_data = response.json()
if not questions_data:
print("Fetched questions list is empty.")
return "Fetched questions list is empty or invalid format.", None
print(f"Fetched {len(questions_data)} questions.")
except requests.exceptions.RequestException as e:
print(f"Error fetching questions: {e}")
return f"Error fetching questions: {e}", None
except Exception as e:
print(f"An unexpected error occurred fetching questions: {e}")
return f"An unexpected error occurred fetching questions: {e}", None
# Run Agent
results_log = []
answers_payload = []
print(f"Running agent on {len(questions_data)} questions...")
for item in questions_data:
task_id = item.get("task_id")
question_text = item.get("question")
if not task_id or question_text is None:
print(f"Skipping item with missing task_id or question: {item}")
continue
try:
submitted_answer = agent(question_text)
answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": submitted_answer})
except Exception as e:
print(f"Error running agent on task {task_id}: {e}")
results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": f"AGENT ERROR: {e}"})
if not answers_payload:
print("Agent did not produce any answers to submit.")
return "Agent did not produce any answers to submit.", pd.DataFrame(results_log)
# Prepare Submission
submission_data = {
"username": username.strip(),
"agent_code": agent_code,
"answers": answers_payload
}
status_update = f"Agent finished. Submitting {len(answers_payload)} answers for user '{username}'..."
print(status_update)
# Submit
print(f"Submitting {len(answers_payload)} answers to: {submit_url}")
try:
response = requests.post(submit_url, json=submission_data, timeout=60)
response.raise_for_status()
result_data = response.json()
final_status = (
f"Submission Successful!\n"
f"User: {result_data.get('username')}\n"
f"Overall Score: {result_data.get('score', 'N/A')}% "
f"({result_data.get('correct_count', '?')}/{result_data.get('total_attempted', '?')} correct)\n"
f"Message: {result_data.get('message', 'No message received.')}"
)
print("Submission successful.")
results_df = pd.DataFrame(results_log)
return final_status, results_df
except requests.exceptions.HTTPError as e:
error_detail = f"Server responded with status {e.response.status_code}."
try:
error_json = e.response.json()
error_detail += f" Detail: {error_json.get('detail', e.response.text)}"
except requests.exceptions.JSONDecodeError:
error_detail += f" Response: {e.response.text[:500]}"
status_message = f"Submission Failed: {error_detail}"
print(status_message)
results_df = pd.DataFrame(results_log)
return status_message, results_df
except Exception as e:
status_message = f"An unexpected error occurred during submission: {e}"
print(status_message)
results_df = pd.DataFrame(results_log)
return status_message, results_df
# --- Gradio Interface ---
with gr.Blocks() as demo:
gr.Markdown("# GAIA Agent Evaluation Runner")
gr.Markdown(
"""
**Instructions:**
1. Log in to your Hugging Face account
2. Click 'Run Evaluation & Submit All Answers'
3. View results and score
**Agent Info:**
- Model: Phi-3-mini-4k-instruct (4-bit quantized)
- Tools: Web Search, Calculator, PDF Reader, Webpage Reader
- Max Steps: 6
"""
)
gr.LoginButton()
run_button = gr.Button("Run Evaluation & Submit All Answers", variant="primary")
status_output = gr.Textbox(label="Run Status / Submission Result", lines=5, interactive=False)
results_table = gr.DataFrame(label="Questions and Agent Answers", wrap=True)
run_button.click(
fn=run_and_submit_all,
outputs=[status_output, results_table]
)
if __name__ == "__main__":
print("\n" + "-"*30 + " App Starting " + "-"*30)
space_host = os.getenv("SPACE_HOST")
space_id = os.getenv("SPACE_ID")
if space_host:
print(f"✅ SPACE_HOST found: {space_host}")
if space_id:
print(f"✅ SPACE_ID found: {space_id}")
print("-"*(60 + len(" App Starting ")) + "\n")
print("Launching Gradio Interface...")
demo.launch(debug=True, share=False)