Spaces:
Runtime error
Runtime error
fixing ver3
Browse files
app.py
CHANGED
@@ -5,10 +5,9 @@ import json
|
|
5 |
import re
|
6 |
import numexpr
|
7 |
import pandas as pd
|
8 |
-
import math
|
9 |
from pdfminer.high_level import extract_text
|
10 |
from bs4 import BeautifulSoup
|
11 |
-
from typing import
|
12 |
from dotenv import load_dotenv
|
13 |
from transformers import AutoModelForCausalLM, AutoTokenizer, GenerationConfig
|
14 |
import torch
|
@@ -21,14 +20,14 @@ SERPER_API_KEY = os.getenv("SERPER_API_KEY")
|
|
21 |
|
22 |
# --- Constants ---
|
23 |
DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
|
24 |
-
MAX_STEPS = 6
|
25 |
-
MAX_TOKENS = 256
|
26 |
MODEL_NAME = "microsoft/Phi-3-mini-4k-instruct"
|
27 |
-
TIMEOUT_PER_QUESTION = 45
|
28 |
-
MAX_RESULT_LENGTH = 500
|
29 |
|
30 |
-
# --- Model Loading ---
|
31 |
-
print("Loading
|
32 |
start_time = time.time()
|
33 |
|
34 |
model = AutoModelForCausalLM.from_pretrained(
|
@@ -50,12 +49,12 @@ if tokenizer.pad_token is None:
|
|
50 |
|
51 |
print(f"Model loaded in {time.time() - start_time:.2f} seconds")
|
52 |
|
53 |
-
# ---
|
54 |
def web_search(query: str) -> str:
|
55 |
-
"""Enhanced web search with better
|
56 |
try:
|
57 |
if SERPER_API_KEY:
|
58 |
-
params = {'q': query, 'num': 3
|
59 |
headers = {'X-API-KEY': SERPER_API_KEY}
|
60 |
response = requests.post(
|
61 |
'https://google.serper.dev/search',
|
@@ -64,97 +63,64 @@ def web_search(query: str) -> str:
|
|
64 |
timeout=10
|
65 |
)
|
66 |
results = response.json()
|
67 |
-
|
68 |
if 'organic' in results:
|
69 |
-
|
70 |
-
|
71 |
-
if 'title' in r and 'snippet' in r:
|
72 |
-
output.append(f"{r['title']}: {r['snippet']}")
|
73 |
-
return "\n".join(output)[:MAX_RESULT_LENGTH]
|
74 |
-
return "No relevant results found"
|
75 |
else:
|
76 |
-
|
77 |
-
results = [r for r in ddgs.text(query, max_results=3)]
|
78 |
-
return "\n".join([f"{r['title']}: {r['body']}" for r in results])[:MAX_RESULT_LENGTH]
|
79 |
except Exception as e:
|
80 |
return f"Search error: {str(e)}"
|
81 |
|
82 |
def calculator(expression: str) -> str:
|
83 |
-
"""
|
84 |
try:
|
85 |
-
# Clean and validate expression
|
86 |
expression = re.sub(r'[^\d+\-*/().^%,\s]', '', expression)
|
87 |
if not expression:
|
88 |
return "Invalid empty expression"
|
89 |
-
|
90 |
-
# Handle percentages and commas
|
91 |
-
expression = expression.replace('%', '/100').replace(',', '')
|
92 |
-
result = numexpr.evaluate(expression)
|
93 |
-
return str(float(result))
|
94 |
except Exception as e:
|
95 |
return f"Calculation error: {str(e)}"
|
96 |
|
97 |
-
def read_pdf(file_path: str) -> str:
|
98 |
-
"""PDF reader with better text extraction"""
|
99 |
-
try:
|
100 |
-
text = extract_text(file_path)
|
101 |
-
if not text:
|
102 |
-
return "No readable text found in PDF"
|
103 |
-
|
104 |
-
# Clean and condense text
|
105 |
-
text = re.sub(r'\s+', ' ', text).strip()
|
106 |
-
return text[:MAX_RESULT_LENGTH]
|
107 |
-
except Exception as e:
|
108 |
-
return f"PDF read error: {str(e)}"
|
109 |
-
|
110 |
def read_webpage(url: str) -> str:
|
111 |
-
"""
|
112 |
try:
|
113 |
headers = {'User-Agent': 'Mozilla/5.0'}
|
114 |
response = requests.get(url, timeout=10, headers=headers)
|
115 |
-
response.raise_for_status()
|
116 |
-
|
117 |
soup = BeautifulSoup(response.text, 'html.parser')
|
118 |
|
119 |
-
# Remove unwanted elements
|
120 |
for element in soup(['script', 'style', 'nav', 'footer']):
|
121 |
element.decompose()
|
122 |
|
123 |
-
# Get text with better formatting
|
124 |
text = soup.get_text(separator='\n', strip=True)
|
125 |
-
|
126 |
-
|
127 |
-
return text[:MAX_RESULT_LENGTH] if text else "No main content found"
|
128 |
except Exception as e:
|
129 |
-
return f"Webpage
|
130 |
|
131 |
TOOLS = {
|
132 |
"web_search": web_search,
|
133 |
"calculator": calculator,
|
134 |
-
"read_pdf": read_pdf,
|
135 |
"read_webpage": read_webpage
|
136 |
}
|
137 |
|
138 |
-
# ---
|
139 |
class GAIA_Agent:
|
140 |
def __init__(self):
|
141 |
self.tools = TOOLS
|
142 |
self.system_prompt = """You are an advanced GAIA problem solver. Follow these steps:
|
143 |
-
1. Analyze the question
|
144 |
-
2. Choose the
|
145 |
-
3. Process
|
146 |
-
4. Provide
|
147 |
|
148 |
-
|
149 |
-
- web_search: For general knowledge
|
150 |
-
- calculator: For math
|
151 |
-
-
|
152 |
-
- read_webpage: For webpage content extraction
|
153 |
|
154 |
Tool format: ```json
|
155 |
{"tool": "tool_name", "args": {"arg1": value}}```
|
156 |
|
157 |
-
Always end with: Final Answer: [
|
158 |
|
159 |
def __call__(self, question: str) -> str:
|
160 |
start_time = time.time()
|
@@ -169,21 +135,20 @@ Always end with: Final Answer: [your answer]"""
|
|
169 |
response = self._call_model(prompt)
|
170 |
|
171 |
if "Final Answer:" in response:
|
172 |
-
|
173 |
-
return answer[:500] # Limit answer length
|
174 |
|
175 |
tool_call = self._parse_tool_call(response)
|
176 |
if tool_call:
|
177 |
tool_name, args = tool_call
|
178 |
observation = self._use_tool(tool_name, args)
|
179 |
-
history.append(f"Tool
|
180 |
-
history.append(f"
|
181 |
else:
|
182 |
-
history.append(f"
|
183 |
|
184 |
gc.collect()
|
185 |
|
186 |
-
return "Maximum steps reached
|
187 |
except Exception as e:
|
188 |
return f"Error: {str(e)}"
|
189 |
|
@@ -199,21 +164,17 @@ Always end with: Final Answer: [your answer]"""
|
|
199 |
padding=False
|
200 |
)
|
201 |
|
202 |
-
|
|
|
|
|
203 |
max_new_tokens=MAX_TOKENS,
|
204 |
temperature=0.3,
|
205 |
top_p=0.9,
|
206 |
do_sample=True,
|
207 |
-
pad_token_id=tokenizer.pad_token_id
|
|
|
208 |
)
|
209 |
|
210 |
-
with torch.no_grad():
|
211 |
-
outputs = model.generate(
|
212 |
-
inputs.input_ids,
|
213 |
-
generation_config=generation_config,
|
214 |
-
attention_mask=inputs.attention_mask
|
215 |
-
)
|
216 |
-
|
217 |
return tokenizer.decode(outputs[0], skip_special_tokens=True).split("<|assistant|>")[-1].strip()
|
218 |
|
219 |
def _parse_tool_call(self, text: str) -> Optional[Tuple[str, Dict]]:
|
@@ -232,11 +193,9 @@ Always end with: Final Answer: [your answer]"""
|
|
232 |
return f"Unknown tool: {tool_name}"
|
233 |
|
234 |
try:
|
235 |
-
#
|
236 |
if tool_name == "read_webpage" and "url" not in args:
|
237 |
-
if "
|
238 |
-
args = args["args"]
|
239 |
-
elif "http" in str(args):
|
240 |
url = re.search(r'https?://[^\s]+', str(args)).group()
|
241 |
args = {"url": url}
|
242 |
|
@@ -293,14 +252,9 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
|
|
293 |
return f"Submission failed: {str(e)}", pd.DataFrame(results)
|
294 |
|
295 |
# --- Gradio Interface ---
|
296 |
-
with gr.Blocks(title="
|
297 |
-
gr.Markdown("##
|
298 |
-
gr.Markdown(""
|
299 |
-
Improved version with:
|
300 |
-
- Better tool utilization
|
301 |
-
- Increased step/token limits
|
302 |
-
- Enhanced error handling
|
303 |
-
""")
|
304 |
|
305 |
with gr.Row():
|
306 |
gr.LoginButton()
|
|
|
5 |
import re
|
6 |
import numexpr
|
7 |
import pandas as pd
|
|
|
8 |
from pdfminer.high_level import extract_text
|
9 |
from bs4 import BeautifulSoup
|
10 |
+
from typing import List, Dict, Optional, Tuple
|
11 |
from dotenv import load_dotenv
|
12 |
from transformers import AutoModelForCausalLM, AutoTokenizer, GenerationConfig
|
13 |
import torch
|
|
|
20 |
|
21 |
# --- Constants ---
|
22 |
DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
|
23 |
+
MAX_STEPS = 6
|
24 |
+
MAX_TOKENS = 256
|
25 |
MODEL_NAME = "microsoft/Phi-3-mini-4k-instruct"
|
26 |
+
TIMEOUT_PER_QUESTION = 45
|
27 |
+
MAX_RESULT_LENGTH = 500
|
28 |
|
29 |
+
# --- Fixed Model Loading ---
|
30 |
+
print("Loading model with fixed configuration...")
|
31 |
start_time = time.time()
|
32 |
|
33 |
model = AutoModelForCausalLM.from_pretrained(
|
|
|
49 |
|
50 |
print(f"Model loaded in {time.time() - start_time:.2f} seconds")
|
51 |
|
52 |
+
# --- Tools Implementation ---
|
53 |
def web_search(query: str) -> str:
|
54 |
+
"""Enhanced web search with better error handling"""
|
55 |
try:
|
56 |
if SERPER_API_KEY:
|
57 |
+
params = {'q': query, 'num': 3}
|
58 |
headers = {'X-API-KEY': SERPER_API_KEY}
|
59 |
response = requests.post(
|
60 |
'https://google.serper.dev/search',
|
|
|
63 |
timeout=10
|
64 |
)
|
65 |
results = response.json()
|
|
|
66 |
if 'organic' in results:
|
67 |
+
return "\n".join([f"{r['title']}: {r['snippet']}" for r in results['organic'][:3]])[:MAX_RESULT_LENGTH]
|
68 |
+
return "No search results found"
|
|
|
|
|
|
|
|
|
69 |
else:
|
70 |
+
return "Search API key not configured"
|
|
|
|
|
71 |
except Exception as e:
|
72 |
return f"Search error: {str(e)}"
|
73 |
|
74 |
def calculator(expression: str) -> str:
|
75 |
+
"""Safe mathematical evaluation"""
|
76 |
try:
|
|
|
77 |
expression = re.sub(r'[^\d+\-*/().^%,\s]', '', expression)
|
78 |
if not expression:
|
79 |
return "Invalid empty expression"
|
80 |
+
return str(numexpr.evaluate(expression))
|
|
|
|
|
|
|
|
|
81 |
except Exception as e:
|
82 |
return f"Calculation error: {str(e)}"
|
83 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
84 |
def read_webpage(url: str) -> str:
|
85 |
+
"""Robust webpage content extraction"""
|
86 |
try:
|
87 |
headers = {'User-Agent': 'Mozilla/5.0'}
|
88 |
response = requests.get(url, timeout=10, headers=headers)
|
|
|
|
|
89 |
soup = BeautifulSoup(response.text, 'html.parser')
|
90 |
|
|
|
91 |
for element in soup(['script', 'style', 'nav', 'footer']):
|
92 |
element.decompose()
|
93 |
|
|
|
94 |
text = soup.get_text(separator='\n', strip=True)
|
95 |
+
return re.sub(r'\n{3,}', '\n\n', text)[:MAX_RESULT_LENGTH]
|
|
|
|
|
96 |
except Exception as e:
|
97 |
+
return f"Webpage error: {str(e)}"
|
98 |
|
99 |
TOOLS = {
|
100 |
"web_search": web_search,
|
101 |
"calculator": calculator,
|
|
|
102 |
"read_webpage": read_webpage
|
103 |
}
|
104 |
|
105 |
+
# --- Fixed GAIA Agent ---
|
106 |
class GAIA_Agent:
|
107 |
def __init__(self):
|
108 |
self.tools = TOOLS
|
109 |
self.system_prompt = """You are an advanced GAIA problem solver. Follow these steps:
|
110 |
+
1. Analyze the question
|
111 |
+
2. Choose the best tool
|
112 |
+
3. Process results
|
113 |
+
4. Provide final answer
|
114 |
|
115 |
+
Tools:
|
116 |
+
- web_search: For general knowledge
|
117 |
+
- calculator: For math
|
118 |
+
- read_webpage: For web content
|
|
|
119 |
|
120 |
Tool format: ```json
|
121 |
{"tool": "tool_name", "args": {"arg1": value}}```
|
122 |
|
123 |
+
Always end with: Final Answer: [answer]"""
|
124 |
|
125 |
def __call__(self, question: str) -> str:
|
126 |
start_time = time.time()
|
|
|
135 |
response = self._call_model(prompt)
|
136 |
|
137 |
if "Final Answer:" in response:
|
138 |
+
return response.split("Final Answer:")[-1].strip()[:500]
|
|
|
139 |
|
140 |
tool_call = self._parse_tool_call(response)
|
141 |
if tool_call:
|
142 |
tool_name, args = tool_call
|
143 |
observation = self._use_tool(tool_name, args)
|
144 |
+
history.append(f"Tool: {tool_name}")
|
145 |
+
history.append(f"Result: {observation[:300]}...")
|
146 |
else:
|
147 |
+
history.append(f"Thought: {response}")
|
148 |
|
149 |
gc.collect()
|
150 |
|
151 |
+
return "Maximum steps reached"
|
152 |
except Exception as e:
|
153 |
return f"Error: {str(e)}"
|
154 |
|
|
|
164 |
padding=False
|
165 |
)
|
166 |
|
167 |
+
# Fixed generation config without problematic parameters
|
168 |
+
outputs = model.generate(
|
169 |
+
inputs.input_ids,
|
170 |
max_new_tokens=MAX_TOKENS,
|
171 |
temperature=0.3,
|
172 |
top_p=0.9,
|
173 |
do_sample=True,
|
174 |
+
pad_token_id=tokenizer.pad_token_id,
|
175 |
+
attention_mask=inputs.attention_mask
|
176 |
)
|
177 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
178 |
return tokenizer.decode(outputs[0], skip_special_tokens=True).split("<|assistant|>")[-1].strip()
|
179 |
|
180 |
def _parse_tool_call(self, text: str) -> Optional[Tuple[str, Dict]]:
|
|
|
193 |
return f"Unknown tool: {tool_name}"
|
194 |
|
195 |
try:
|
196 |
+
# Handle URL extraction for webpage reading
|
197 |
if tool_name == "read_webpage" and "url" not in args:
|
198 |
+
if "http" in str(args):
|
|
|
|
|
199 |
url = re.search(r'https?://[^\s]+', str(args)).group()
|
200 |
args = {"url": url}
|
201 |
|
|
|
252 |
return f"Submission failed: {str(e)}", pd.DataFrame(results)
|
253 |
|
254 |
# --- Gradio Interface ---
|
255 |
+
with gr.Blocks(title="Fixed GAIA Agent") as demo:
|
256 |
+
gr.Markdown("## 🛠️ Fixed GAIA Agent")
|
257 |
+
gr.Markdown("Resolved the 'DynamicCache' error with improved configuration")
|
|
|
|
|
|
|
|
|
|
|
258 |
|
259 |
with gr.Row():
|
260 |
gr.LoginButton()
|