Spaces:
Sleeping
Sleeping
File size: 21,157 Bytes
29a316b |
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 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 |
from saged import Pipeline
from tqdm import tqdm
from pathlib import Path
from saged import SAGEDData as dt
import streamlit as st
import json
import http.client
from openai import AzureOpenAI
import ollama
import time # Use time.sleep to simulate processing steps
import logging
from io import StringIO
import sys
# Create a custom logging handler to capture log messages
class StreamlitLogHandler(logging.Handler):
def __init__(self):
super().__init__()
self.log_capture_string = StringIO()
def emit(self, record):
# Write each log message to the StringIO buffer
message = self.format(record)
self.log_capture_string.write(message + "\n")
def get_logs(self):
# Return the log contents
return self.log_capture_string.getvalue()
def clear_logs(self):
# Clear the log buffer
self.log_capture_string.truncate(0)
self.log_capture_string.seek(0)
# Define ContentFormatter class
class ContentFormatter:
@staticmethod
def chat_completions(text, settings_params):
message = [
{"role": "system", "content": "You are a helpful assistant."},
{"role": "user", "content": text}
]
data = {"messages": message, **settings_params}
return json.dumps(data)
# Define OllamaModel (For local Ollama interaction)
class OllamaModel:
def __init__(self, base_model='llama3', system_prompt='You are a helpful assistant', model_name='llama3o',
**kwargs):
self.base_model = base_model
self.model_name = model_name
self.model_create(model_name, system_prompt, base_model, **kwargs)
def model_create(self, model_name, system_prompt, base_model, **kwargs):
modelfile = f'FROM {base_model}\nSYSTEM {system_prompt}\n'
if kwargs:
for key, value in kwargs.items():
modelfile += f'PARAMETER {key.lower()} {value}\n'
ollama.create(model=model_name, modelfile=modelfile)
def invoke(self, prompt):
answer = ollama.generate(model=self.model_name, prompt=prompt)
return answer['response']
# Define GPTAgent (For OpenAI GPT models)
class GPTAgent:
def __init__(self, model_name, azure_key, azure_version, azure_endpoint, deployment_name):
self.client = AzureOpenAI(
api_key=azure_key,
api_version=azure_version,
azure_endpoint=azure_endpoint
)
self.deployment_name = deployment_name
def invoke(self, prompt, settings_params=None):
if not settings_params:
settings_params = {}
formatted_input = ContentFormatter.chat_completions(prompt, settings_params)
response = self.client.chat.completions.create(
model=self.deployment_name,
messages=json.loads(formatted_input)['messages'],
**settings_params
)
return response.choices[0].message.content
# Define AzureAgent (For Azure OpenAI models)
class AzureAgent:
def __init__(self, model_name, azure_uri, azure_api_key):
self.azure_uri = azure_uri
self.headers = {
'Authorization': f"Bearer {azure_api_key}",
'Content-Type': 'application/json'
}
self.chat_formatter = ContentFormatter
def invoke(self, prompt, settings_params=None):
if not settings_params:
settings_params = {}
body = self.chat_formatter.chat_completions(prompt, {**settings_params})
conn = http.client.HTTPSConnection(self.azure_uri)
conn.request("POST", '/v1/chat/completions', body=body, headers=self.headers)
response = conn.getresponse()
data = response.read()
conn.close()
decoded_data = data.decode("utf-8")
parsed_data = json.loads(decoded_data)
content = parsed_data["choices"][0]["message"]["content"]
return content
# Renew Source Finder Button
def renew_source_finder(domain, concept_list):
if 'generated_synthetic_files' in st.session_state:
del st.session_state['generated_synthetic_files']
if not domain or not concept_list:
st.error("Please fill in all the required fields before proceeding.")
else:
with st.spinner("Renewing source info files..."):
base_path = Path('data/customized/source_finder/')
for concept in concept_list:
file_path = base_path / f'{domain}_{concept}_source_finder.json'
if file_path.exists():
try:
file_path.unlink() # Delete the file
st.info(f"Deleted source info file: {file_path}")
except Exception as e:
st.error(f"An error occurred while deleting the file {file_path}: {e}")
st.success("Source info files renewal completed!")
def create_source_finder(domain, concept):
source_specification_item = f"data/customized/local_files/{domain}/{concept}.txt"
if not Path(source_specification_item).exists():
st.warning(f"Local file does not exist: {source_specification_item}")
instance = dt.create_data(domain, concept, 'source_finder')
instance.data[0]['keywords'] = {concept: dt.default_keyword_metadata.copy()}
category_shared_source_item = dt.default_source_item.copy()
category_shared_source_item['source_type'] = "local_paths"
category_shared_source_item['source_specification'] = [source_specification_item]
instance.data[0]['category_shared_source'] = [category_shared_source_item]
return instance.data.copy()
def check_and_create_source_files(domain, concept_list):
"""
Checks if the required source finder files exist for each concept in the domain.
If a file does not exist or is invalid, it creates an empty JSON file for that concept.
"""
base_path = Path('data/customized/source_finder/')
base_path.mkdir(parents=True, exist_ok=True)
for concept in concept_list:
file_path = base_path / f'{domain}_{concept}_source_finder.json'
if not file_path.exists():
# Create a new source finder file using create_source_finder
data = create_source_finder(domain, concept)
with open(file_path, 'w', encoding='utf-8') as f:
json.dump(data, f, indent=4)
st.info(f"Created missing source finder file: {file_path}")
else:
# Attempt to load the file to verify its validity
instance = dt.load_file(domain, concept, 'source_finder', file_path)
if instance is None:
# If loading fails, create a new valid file
data = create_source_finder(domain, concept)
with open(file_path, 'w', encoding='utf-8') as f:
json.dump(data, f, indent=4)
st.info(f"Recreated invalid source finder file: {file_path}")
def clean_spaces(data):
"""
Removes trailing or leading spaces from a string or from each element in a list.
"""
if isinstance(data, str):
return data.strip()
elif isinstance(data, list):
return [item.strip() if isinstance(item, str) else item for item in data]
else:
raise TypeError("Input should be either a string or a list of strings")
def create_replacement_dict(concept_list, replacer):
replacement = {}
for concept in concept_list:
replacement[concept] = {}
for company in replacer:
replacement[concept][company] = {concept: company}
return replacement
# Title of the app
st.title("SAGED-bias Benchmark-Building Demo")
# Initialize session state variables
if 'domain' not in st.session_state:
st.session_state['domain'] = None
if 'concept_list' not in st.session_state:
st.session_state['concept_list'] = None
if 'gpt_model' not in st.session_state:
st.session_state['gpt_model'] = None
if 'azure_model' not in st.session_state:
st.session_state['azure_model'] = None
if 'ollama_model' not in st.session_state:
st.session_state['ollama_model'] = None
# Sidebar: Model Selection
with st.sidebar:
st.header("Model Configuration")
# Selection of which model to use
model_selection = st.radio("Select Model Type", ['GPT-Azure', 'Azure', 'Ollama'])
# Collapsible Additional Configuration Section
with st.expander("Model Configuration"):
if model_selection == 'Ollama':
# Ollama Configuration
ollama_deployment_name = st.text_input("Enter Ollama Model Deployment Name", placeholder="e.g., llama3")
ollama_system_prompt = st.text_input("Enter System Prompt for Ollama",
placeholder="e.g., You are a helpful assistant.")
if ollama_deployment_name and ollama_system_prompt:
confirm_ollama = st.button("Confirm Ollama Configuration")
if confirm_ollama:
st.session_state['ollama_model'] = OllamaModel(
model_name=ollama_deployment_name,
system_prompt=ollama_system_prompt
)
st.success("Ollama model configured successfully.")
else:
st.warning("Please provide both Ollama deployment name and system prompt.")
elif model_selection == 'GPT-Azure' or model_selection == 'Azure':
# GPT / Azure Configuration
gpt_azure_endpoint = st.text_input("Enter Azure Endpoint URL",
placeholder="e.g., https://your-resource-name.openai.azure.com/")
gpt_azure_api_key = st.text_input("Enter Azure API Key", type="password")
gpt_azure_model_name = st.text_input("Enter Azure Model Name", placeholder="e.g., GPT-3.5-turbo")
gpt_azure_deployment_name = st.text_input("Enter Azure Deployment Name",
placeholder="e.g., gpt-3-5-deployment")
if gpt_azure_endpoint and gpt_azure_api_key and gpt_azure_model_name and gpt_azure_deployment_name:
confirm_gpt_azure = st.button("Confirm GPT/Azure Configuration")
if confirm_gpt_azure:
if model_selection == 'GPT-Azure':
st.session_state['gpt_model'] = GPTAgent(
model_name=gpt_azure_model_name,
azure_key=gpt_azure_api_key,
azure_version='2023-05-15', # Update if necessary
azure_endpoint=gpt_azure_endpoint,
deployment_name=gpt_azure_deployment_name
)
st.success("GPT model configured successfully.")
elif model_selection == 'Azure':
st.session_state['azure_model'] = AzureAgent(
model_name=gpt_azure_model_name,
azure_uri=gpt_azure_endpoint,
azure_api_key=gpt_azure_api_key
)
st.success("Azure model configured successfully.")
else:
st.warning("Please provide all fields for GPT/Azure configuration.")
# Main interaction based on configured model
if st.session_state.get('ollama_model'):
model = st.session_state['ollama_model']
elif st.session_state.get('gpt_model'):
model = st.session_state['gpt_model']
elif st.session_state.get('azure_model'):
model = st.session_state['azure_model']
else:
model = None
# User input: Domain and Concepts
with st.form(key='domain_concept_form'):
domain = clean_spaces(
st.text_input("Enter the domain: (e.g., Stocks, Education)", placeholder="Enter domain here..."))
# User input: Concepts
concept_text = st.text_area("Enter the concepts (separated by commas):",
placeholder="e.g., excel-stock, ok-stock, bad-stock")
concept_list = clean_spaces(concept_text.split(','))
submit_button = st.form_submit_button(label='Confirm Domain and Concepts')
if submit_button:
if not domain:
st.warning("Please enter a domain.")
elif not concept_list or concept_text.strip() == "":
st.warning("Please enter at least one concept.")
else:
st.session_state['domain'] = domain
st.session_state['concept_list'] = concept_list
st.success("Domain and concepts confirmed.")
# Display further options only after domain and concepts are confirmed
if st.session_state['domain'] and st.session_state['concept_list']:
with st.expander("Additional Options"):
# User input: Method
scraper_method = st.radio("Select the scraper method:", (('wiki', 'local_files', 'synthetic_files')))
# Initiate the source_finder_requirement and keyword_finder_requirement if 'wiki' is selected
if scraper_method == 'wiki':
st.session_state['keyword_finder_requirement'] = True
st.session_state['source_finder_requirement'] = True
st.session_state['check_source_finder'] = False
# File upload for each concept if 'local_files' is selected
if scraper_method == 'local_files':
uploaded_files = {}
st.session_state['keyword_finder_requirement'] = False
st.session_state['source_finder_requirement'] = False
st.session_state['check_source_finder'] = True
for concept in st.session_state['concept_list']:
uploaded_file = st.file_uploader(f"Upload file for concept '{concept}':", type=['txt'],
key=f"file_{concept}")
if uploaded_file:
uploaded_files[concept] = uploaded_file
# Save uploaded file
save_path = Path(f"data/customized/local_files/{st.session_state['domain']}/{concept}.txt")
save_path.parent.mkdir(parents=True, exist_ok=True)
with open(save_path, 'wb') as f:
f.write(uploaded_file.getbuffer())
st.success(f"File for concept '{concept}' saved successfully.")
# Generate synthetic files if 'synthetic_files' is selected
if scraper_method == 'synthetic_files':
scraper_method = 'local_files'
st.session_state['keyword_finder_requirement'] = False
st.session_state['source_finder_requirement'] = False
st.session_state['check_source_finder'] = True
if 'generated_synthetic_files' not in st.session_state:
st.session_state['generated_synthetic_files'] = set()
prompt_inputs = {}
for concept in st.session_state['concept_list']:
if concept not in st.session_state['generated_synthetic_files']:
prompt_inputs[concept] = st.text_input(
f"Enter the prompt for concept '{concept}':",
value=f"Write a long article introducing the {concept} in the {st.session_state['domain']}. Use the {concept} as much as possible.",
key=f"prompt_{concept}"
)
if st.button("Generate Synthetic Files for All Concepts"):
if model:
for concept, prompt in prompt_inputs.items():
if prompt:
with st.spinner(f"Generating content for concept '{concept}'..."):
synthetic_content = model.invoke(prompt)
save_path = Path(
f"data/customized/local_files/{st.session_state['domain']}/{concept}.txt")
save_path.parent.mkdir(parents=True, exist_ok=True)
with open(save_path, 'w', encoding='utf-8') as f:
f.write(synthetic_content)
st.session_state['generated_synthetic_files'].add(concept)
st.success(f"Synthetic file for concept '{concept}' created successfully.")
else:
st.warning("Please configure a model to generate synthetic files.")
# User input: Prompt Method
prompt_method = st.radio("Select the prompt method:", ('split_sentences', 'questions'), index = 0)
# User input: Max Benchmark Length
max_benchmark_length = st.slider("Select the maximum prompts per concepts:", 1, 199, 10)
# User input: Branching
branching = st.radio("Enable branching:", ('Yes', 'No'), index=1)
branching_enabled = True if branching == 'Yes' else False
# User input: Replacer (only if branching is enabled)
replacer = []
replacement = {}
if branching_enabled:
replacer_text = st.text_area("Enter the replacer list (list of strings, separated by commas):",
placeholder="e.g., Company A, Company B")
replacer = clean_spaces(replacer_text.split(','))
replacement = create_replacement_dict(st.session_state['concept_list'], replacer)
# Configuration
concept_specified_config = {
x: {'keyword_finder': {'manual_keywords': [x]}} for x in st.session_state['concept_list']
}
concept_configuration = {
'keyword_finder': {
'require': st.session_state['keyword_finder_requirement'],
'keyword_number': 1,
},
'source_finder': {
'require': st.session_state['source_finder_requirement'],
'scrap_number': 10,
'method': scraper_method,
},
'scraper': {
'require': True,
'method': scraper_method,
},
'prompt_maker': {
'method': prompt_method,
'generation_function': model.invoke if model else None,
'max_benchmark_length': max_benchmark_length,
},
}
domain_configuration = {
'categories': st.session_state['concept_list'],
'branching': branching_enabled,
'branching_config': {
'generation_function': model.invoke if model else None,
'keyword_reference': st.session_state['concept_list'],
'replacement_descriptor_require': False,
'replacement_description': replacement,
'branching_pairs': 'not all',
'direction': 'not both',
},
'shared_config': concept_configuration,
'category_specified_config': concept_specified_config
}
# Renew Source Finder Button
if st.button('Renew Source info'):
renew_source_finder(st.session_state['domain'], st.session_state['concept_list'])
# Save the original stdout to print to the terminal if needed later
original_stdout = sys.stdout
# Define StreamToText to capture and display logs in real-time within Streamlit only
class StreamToText:
def __init__(self):
self.output = StringIO()
def write(self, message):
if message.strip(): # Avoid adding empty messages
# Only append to Streamlit display, not the terminal
st.session_state.log_messages.append(message.strip())
log_placeholder.text("\n".join(st.session_state.log_messages)) # Flush updated logs
def flush(self):
pass # Required for compatibility with sys.stdout
# Initialize session state for log messages
if 'log_messages' not in st.session_state:
st.session_state.log_messages = []
# Replace sys.stdout with our custom StreamToText instance
stream_to_text = StreamToText()
sys.stdout = stream_to_text
# Placeholder for displaying logs within a collapsible expander
with st.expander("Show Logs", expanded=False):
log_placeholder = st.empty() # Placeholder for dynamic log display
# Define the Create Benchmark button
if st.button("Create a Benchmark"):
st.session_state.log_messages = [] # Clear previous logs
with st.spinner("Creating benchmark..."):
if st.session_state['check_source_finder']:
# Check for relevant materials
check_and_create_source_files(st.session_state['domain'], st.session_state['concept_list'])
try:
# Display progress bar and log messages
progress_bar = st.progress(0)
for i in tqdm(range(1, 101)):
progress_bar.progress(i)
time.sleep(0.05) # Short delay to simulate processing time
# Run the benchmark creation function
benchmark = Pipeline.domain_benchmark_building(st.session_state['domain'], domain_configuration)
st.success("Benchmark creation completed!")
st.dataframe(benchmark.data)
except Exception as e:
st.error(f"An error occurred during benchmark creation: {e}")
|