Update app.py
Browse files
app.py
CHANGED
@@ -16,6 +16,8 @@ from anthropic import Anthropic
|
|
16 |
import google.generativeai as genai
|
17 |
import hmac
|
18 |
import hashlib
|
|
|
|
|
19 |
|
20 |
load_dotenv()
|
21 |
|
@@ -32,16 +34,6 @@ if not os.path.exists(DATA_DIR):
|
|
32 |
else:
|
33 |
st.info(f"`{DATA_DIR}` directory already exists.")
|
34 |
|
35 |
-
if os.path.exists(DATA_DIR):
|
36 |
-
files = os.listdir(DATA_DIR)
|
37 |
-
st.write(f"Contents of `{DATA_DIR}` directory:")
|
38 |
-
if files:
|
39 |
-
for file in files:
|
40 |
-
st.write(f"- {file}")
|
41 |
-
else:
|
42 |
-
st.write("The data directory is currently empty.")
|
43 |
-
else:
|
44 |
-
st.error(f"`{DATA_DIR}` directory does not exist.")
|
45 |
|
46 |
def initialize_session_state():
|
47 |
if 'api_configured' not in st.session_state:
|
@@ -62,8 +54,7 @@ def initialize_session_state():
|
|
62 |
st.session_state.last_evaluated_dataset = None
|
63 |
|
64 |
def setup_api_clients():
|
65 |
-
|
66 |
-
|
67 |
with st.sidebar:
|
68 |
st.title("API Configuration")
|
69 |
|
@@ -76,20 +67,24 @@ def setup_api_clients():
|
|
76 |
if st.button("Verify Credentials"):
|
77 |
if (hmac.compare_digest(username, os.environ.get("STREAMLIT_USERNAME", "")) and
|
78 |
hmac.compare_digest(password, os.environ.get("STREAMLIT_PASSWORD", ""))):
|
79 |
-
|
80 |
-
|
81 |
-
|
82 |
-
|
83 |
-
|
84 |
-
|
85 |
-
|
86 |
-
|
87 |
-
|
88 |
-
|
89 |
-
|
90 |
-
|
91 |
-
|
92 |
-
|
|
|
|
|
|
|
|
|
93 |
else:
|
94 |
st.error("Invalid credentials. Please try again or use your own API keys.")
|
95 |
st.session_state.api_configured = False
|
@@ -120,6 +115,7 @@ def setup_api_clients():
|
|
120 |
st.error(f"Error initializing API clients: {str(e)}")
|
121 |
st.session_state.api_configured = False
|
122 |
|
|
|
123 |
MAX_CONCURRENT_CALLS = 5
|
124 |
semaphore = threading.Semaphore(MAX_CONCURRENT_CALLS)
|
125 |
|
@@ -145,7 +141,7 @@ def load_dataset_by_name(dataset_name, split="train"):
|
|
145 |
}
|
146 |
questions.append(question_dict)
|
147 |
|
148 |
-
st.write(f"Loaded {len(questions)} single-select questions from {dataset_name}")
|
149 |
return questions
|
150 |
|
151 |
@retry(
|
@@ -233,14 +229,13 @@ def process_single_evaluation(question, prompt_template, model_name, clients, la
|
|
233 |
'options': ' | '.join(question['options']),
|
234 |
'model_response': answer,
|
235 |
'is_correct': is_correct,
|
236 |
-
'explanation': question['explanation']
|
|
|
237 |
}
|
238 |
-
|
239 |
with WRITE_LOCK:
|
240 |
file_exists = os.path.isfile(RESULTS_FILE)
|
241 |
with open(RESULTS_FILE, 'a', encoding='utf-8', newline='') as f:
|
242 |
-
|
243 |
-
writer.to_csv(f, header=not file_exists, index=False)
|
244 |
|
245 |
return result
|
246 |
|
@@ -264,7 +259,6 @@ def process_evaluations_concurrently(questions, prompt_template, models_to_evalu
|
|
264 |
current_iteration += 1
|
265 |
progress_callback(current_iteration, total_iterations)
|
266 |
continue # Skip already completed evaluations
|
267 |
-
# Pass last_evaluated_dataset as an argument
|
268 |
future = executor.submit(
|
269 |
process_single_evaluation,
|
270 |
question,
|
@@ -283,42 +277,39 @@ def process_evaluations_concurrently(questions, prompt_template, models_to_evalu
|
|
283 |
|
284 |
return results
|
285 |
|
286 |
-
def main():
|
287 |
-
|
288 |
-
initialize_session_state()
|
289 |
-
setup_api_clients()
|
290 |
-
|
291 |
-
if not st.session_state.api_configured:
|
292 |
-
st.warning("Please configure API keys in the sidebar to proceed")
|
293 |
-
st.stop()
|
294 |
|
|
|
295 |
if 'all_results' not in st.session_state:
|
296 |
-
|
297 |
-
|
298 |
-
|
299 |
-
|
300 |
-
|
301 |
-
|
302 |
-
|
303 |
-
|
304 |
-
|
305 |
-
|
306 |
-
|
307 |
-
|
308 |
-
|
309 |
-
|
310 |
-
|
311 |
-
|
312 |
-
st.session_state.detailed_model = None
|
313 |
-
if 'detailed_dataset' not in st.session_state:
|
314 |
-
st.session_state.detailed_dataset = None
|
315 |
-
if 'last_evaluated_dataset' not in st.session_state:
|
316 |
st.session_state.last_evaluated_dataset = None
|
|
|
317 |
|
318 |
with st.sidebar:
|
319 |
if st.button("Reset Results"):
|
320 |
if os.path.exists(RESULTS_FILE):
|
321 |
os.remove(RESULTS_FILE)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
322 |
st.session_state.all_results = {}
|
323 |
st.session_state.last_evaluated_dataset = None
|
324 |
st.success("Results have been reset.")
|
@@ -333,14 +324,15 @@ def main():
|
|
333 |
help="Choose the dataset to evaluate on"
|
334 |
)
|
335 |
with col2:
|
336 |
-
|
337 |
"Select Model(s)",
|
338 |
options=list(MODELS.keys()),
|
339 |
default=[list(MODELS.keys())[0]],
|
340 |
help="Choose one or more models to evaluate."
|
341 |
)
|
342 |
|
343 |
-
models_to_evaluate =
|
|
|
344 |
|
345 |
default_prompt = '''You are a medical AI assistant. Please answer the following multiple choice question.
|
346 |
Question: {question}
|
@@ -365,13 +357,14 @@ Important:
|
|
365 |
- Only the "answer" field will be used for evaluation
|
366 |
- Ensure your response is in valid JSON format'''
|
367 |
|
|
|
368 |
col1, col2 = st.columns([2, 1])
|
369 |
with col1:
|
370 |
prompt_template = st.text_area(
|
371 |
"Customize Prompt Template",
|
372 |
default_prompt,
|
373 |
height=400,
|
374 |
-
help="
|
375 |
)
|
376 |
|
377 |
with col2:
|
@@ -381,28 +374,34 @@ Important:
|
|
381 |
- `{options}`: The multiple choice options
|
382 |
""")
|
383 |
|
|
|
384 |
with st.spinner("Loading dataset..."):
|
385 |
questions = load_dataset_by_name(selected_dataset)
|
|
|
|
|
386 |
subjects = sorted(list(set(q['subject_name'] for q in questions)))
|
387 |
selected_subject = st.selectbox("Filter by subject", ["All"] + subjects)
|
388 |
|
389 |
if selected_subject != "All":
|
390 |
questions = [q for q in questions if q['subject_name'] == selected_subject]
|
391 |
|
392 |
-
|
|
|
|
|
393 |
|
394 |
if st.button("Start Evaluation"):
|
395 |
with st.spinner("Starting evaluation..."):
|
396 |
selected_questions = questions[:num_questions]
|
397 |
|
398 |
-
|
399 |
clients = {
|
400 |
"togetherai": st.session_state["togetherai_client"],
|
401 |
"openai": st.session_state["openai_client"],
|
402 |
"anthropic": st.session_state["anthropic_client"]
|
403 |
}
|
404 |
|
405 |
-
|
|
|
406 |
|
407 |
progress_container = st.container()
|
408 |
progress_bar = progress_container.progress(0)
|
@@ -443,7 +442,6 @@ Important:
|
|
443 |
if st.session_state.all_results:
|
444 |
st.subheader("Evaluation Results")
|
445 |
|
446 |
-
model_metrics = {}
|
447 |
for model_name, results in st.session_state.all_results.items():
|
448 |
df = pd.DataFrame(results)
|
449 |
metrics = {
|
@@ -454,7 +452,6 @@ Important:
|
|
454 |
metrics_df = pd.DataFrame(model_metrics).T
|
455 |
|
456 |
st.subheader("Model Performance Comparison")
|
457 |
-
|
458 |
accuracy_chart = alt.Chart(
|
459 |
metrics_df.reset_index().melt(id_vars=['index'], value_vars=['Accuracy'])
|
460 |
).mark_bar().encode(
|
@@ -473,7 +470,6 @@ Important:
|
|
473 |
)
|
474 |
|
475 |
st.altair_chart(accuracy_chart, use_container_width=True)
|
476 |
-
|
477 |
if st.session_state.all_results:
|
478 |
st.subheader("Detailed Results")
|
479 |
|
@@ -518,11 +514,11 @@ Important:
|
|
518 |
|
519 |
col1, col2 = st.columns(2)
|
520 |
with col1:
|
521 |
-
st.write("**
|
522 |
-
st.code(result.get('
|
523 |
with col2:
|
524 |
-
st.write("**
|
525 |
-
st.code(result.get('
|
526 |
|
527 |
col1, col2 = st.columns(2)
|
528 |
with col1:
|
@@ -534,32 +530,20 @@ Important:
|
|
534 |
else:
|
535 |
st.error("Incorrect")
|
536 |
|
537 |
-
st.write("**
|
538 |
else:
|
539 |
st.info(f"No results available for {selected_model_details} on {selected_dataset_details}. Please run the evaluation first.")
|
540 |
|
541 |
st.markdown("---")
|
|
|
542 |
all_data = []
|
543 |
-
|
544 |
for model_name, results in st.session_state.all_results.items():
|
545 |
-
for
|
546 |
-
row =
|
547 |
-
'dataset': st.session_state.last_evaluated_dataset,
|
548 |
-
'model': model_name,
|
549 |
-
'question': result['question'],
|
550 |
-
'correct_answer': result['correct_answer'],
|
551 |
-
'subject': result['subject'],
|
552 |
-
'options': result['options'],
|
553 |
-
'model_response': result['model_response'],
|
554 |
-
'is_correct': result['is_correct'],
|
555 |
-
'explanation': result['explanation']
|
556 |
-
}
|
557 |
all_data.append(row)
|
558 |
|
559 |
complete_df = pd.DataFrame(all_data)
|
560 |
-
|
561 |
csv = complete_df.to_csv(index=False)
|
562 |
-
|
563 |
st.download_button(
|
564 |
label="Download All Results as CSV",
|
565 |
data=csv,
|
|
|
16 |
import google.generativeai as genai
|
17 |
import hmac
|
18 |
import hashlib
|
19 |
+
from uuid import uuid4
|
20 |
+
from datetime import datetime
|
21 |
|
22 |
load_dotenv()
|
23 |
|
|
|
34 |
else:
|
35 |
st.info(f"`{DATA_DIR}` directory already exists.")
|
36 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
37 |
|
38 |
def initialize_session_state():
|
39 |
if 'api_configured' not in st.session_state:
|
|
|
54 |
st.session_state.last_evaluated_dataset = None
|
55 |
|
56 |
def setup_api_clients():
|
57 |
+
initialize_session_state()
|
|
|
58 |
with st.sidebar:
|
59 |
st.title("API Configuration")
|
60 |
|
|
|
67 |
if st.button("Verify Credentials"):
|
68 |
if (hmac.compare_digest(username, os.environ.get("STREAMLIT_USERNAME", "")) and
|
69 |
hmac.compare_digest(password, os.environ.get("STREAMLIT_PASSWORD", ""))):
|
70 |
+
try:
|
71 |
+
st.session_state.togetherai_client = OpenAI(
|
72 |
+
api_key=os.getenv('TOGETHERAI_API_KEY'),
|
73 |
+
base_url="https://api.together.xyz/v1"
|
74 |
+
)
|
75 |
+
st.session_state.openai_client = OpenAI(
|
76 |
+
api_key=os.getenv('OPENAI_API_KEY')
|
77 |
+
)
|
78 |
+
st.session_state.anthropic_client = Anthropic(
|
79 |
+
api_key=os.getenv('ANTHROPIC_API_KEY')
|
80 |
+
)
|
81 |
+
genai.configure(api_key=os.environ["GEMINI_API_KEY"])
|
82 |
+
|
83 |
+
st.session_state.api_configured = True
|
84 |
+
st.success("Successfully configured the API clients with stored keys!")
|
85 |
+
except Exception as e:
|
86 |
+
st.error(f"Error initializing API clients: {str(e)}")
|
87 |
+
st.session_state.api_configured = False
|
88 |
else:
|
89 |
st.error("Invalid credentials. Please try again or use your own API keys.")
|
90 |
st.session_state.api_configured = False
|
|
|
115 |
st.error(f"Error initializing API clients: {str(e)}")
|
116 |
st.session_state.api_configured = False
|
117 |
|
118 |
+
setup_api_clients()
|
119 |
MAX_CONCURRENT_CALLS = 5
|
120 |
semaphore = threading.Semaphore(MAX_CONCURRENT_CALLS)
|
121 |
|
|
|
141 |
}
|
142 |
questions.append(question_dict)
|
143 |
|
144 |
+
st.write(f"Loaded {len(questions)} single-select questions from `{dataset_name}`")
|
145 |
return questions
|
146 |
|
147 |
@retry(
|
|
|
229 |
'options': ' | '.join(question['options']),
|
230 |
'model_response': answer,
|
231 |
'is_correct': is_correct,
|
232 |
+
'explanation': question['explanation'],
|
233 |
+
'timestamp': datetime.utcnow().isoformat()
|
234 |
}
|
|
|
235 |
with WRITE_LOCK:
|
236 |
file_exists = os.path.isfile(RESULTS_FILE)
|
237 |
with open(RESULTS_FILE, 'a', encoding='utf-8', newline='') as f:
|
238 |
+
pd.DataFrame([result]).to_csv(f, header=not file_exists, index=False)
|
|
|
239 |
|
240 |
return result
|
241 |
|
|
|
259 |
current_iteration += 1
|
260 |
progress_callback(current_iteration, total_iterations)
|
261 |
continue # Skip already completed evaluations
|
|
|
262 |
future = executor.submit(
|
263 |
process_single_evaluation,
|
264 |
question,
|
|
|
277 |
|
278 |
return results
|
279 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
280 |
|
281 |
+
def main():
|
282 |
if 'all_results' not in st.session_state:
|
283 |
+
st.session_state.all_results = {}
|
284 |
+
st.session_state.last_evaluated_dataset = None
|
285 |
+
if os.path.exists(RESULTS_FILE):
|
286 |
+
existing_df = pd.read_csv(RESULTS_FILE)
|
287 |
+
all_results = {}
|
288 |
+
for _, row in existing_df.iterrows():
|
289 |
+
model = row['model']
|
290 |
+
result = row.to_dict()
|
291 |
+
if model not in all_results:
|
292 |
+
all_results[model] = []
|
293 |
+
all_results[model].append(result)
|
294 |
+
st.session_state.all_results = all_results
|
295 |
+
st.session_state.last_evaluated_dataset = existing_df['dataset'].iloc[-1]
|
296 |
+
st.info(f"Loaded existing results from `{RESULTS_FILE}`.")
|
297 |
+
else:
|
298 |
+
st.session_state.all_results = {}
|
|
|
|
|
|
|
|
|
299 |
st.session_state.last_evaluated_dataset = None
|
300 |
+
st.info(f"No existing results found. Ready to start fresh.")
|
301 |
|
302 |
with st.sidebar:
|
303 |
if st.button("Reset Results"):
|
304 |
if os.path.exists(RESULTS_FILE):
|
305 |
os.remove(RESULTS_FILE)
|
306 |
+
for file in os.listdir(DATA_DIR):
|
307 |
+
file_path = os.path.join(DATA_DIR, file)
|
308 |
+
try:
|
309 |
+
if os.path.isfile(file_path):
|
310 |
+
os.unlink(file_path)
|
311 |
+
except Exception as e:
|
312 |
+
st.error(f"Error deleting file {file_path}: {e}")
|
313 |
st.session_state.all_results = {}
|
314 |
st.session_state.last_evaluated_dataset = None
|
315 |
st.success("Results have been reset.")
|
|
|
324 |
help="Choose the dataset to evaluate on"
|
325 |
)
|
326 |
with col2:
|
327 |
+
selected_models = st.multiselect(
|
328 |
"Select Model(s)",
|
329 |
options=list(MODELS.keys()),
|
330 |
default=[list(MODELS.keys())[0]],
|
331 |
help="Choose one or more models to evaluate."
|
332 |
)
|
333 |
|
334 |
+
models_to_evaluate = selected_models
|
335 |
+
|
336 |
|
337 |
default_prompt = '''You are a medical AI assistant. Please answer the following multiple choice question.
|
338 |
Question: {question}
|
|
|
357 |
- Only the "answer" field will be used for evaluation
|
358 |
- Ensure your response is in valid JSON format'''
|
359 |
|
360 |
+
|
361 |
col1, col2 = st.columns([2, 1])
|
362 |
with col1:
|
363 |
prompt_template = st.text_area(
|
364 |
"Customize Prompt Template",
|
365 |
default_prompt,
|
366 |
height=400,
|
367 |
+
help="Edit the prompt template before starting the evaluation."
|
368 |
)
|
369 |
|
370 |
with col2:
|
|
|
374 |
- `{options}`: The multiple choice options
|
375 |
""")
|
376 |
|
377 |
+
|
378 |
with st.spinner("Loading dataset..."):
|
379 |
questions = load_dataset_by_name(selected_dataset)
|
380 |
+
|
381 |
+
|
382 |
subjects = sorted(list(set(q['subject_name'] for q in questions)))
|
383 |
selected_subject = st.selectbox("Filter by subject", ["All"] + subjects)
|
384 |
|
385 |
if selected_subject != "All":
|
386 |
questions = [q for q in questions if q['subject_name'] == selected_subject]
|
387 |
|
388 |
+
|
389 |
+
num_questions = st.number_input("Number of questions to evaluate", min_value=1, max_value=len(questions), value=1, step=1)
|
390 |
+
|
391 |
|
392 |
if st.button("Start Evaluation"):
|
393 |
with st.spinner("Starting evaluation..."):
|
394 |
selected_questions = questions[:num_questions]
|
395 |
|
396 |
+
|
397 |
clients = {
|
398 |
"togetherai": st.session_state["togetherai_client"],
|
399 |
"openai": st.session_state["openai_client"],
|
400 |
"anthropic": st.session_state["anthropic_client"]
|
401 |
}
|
402 |
|
403 |
+
|
404 |
+
last_evaluated_dataset = st.session_state.last_evaluated_dataset if st.session_state.last_evaluated_dataset else selected_dataset
|
405 |
|
406 |
progress_container = st.container()
|
407 |
progress_bar = progress_container.progress(0)
|
|
|
442 |
if st.session_state.all_results:
|
443 |
st.subheader("Evaluation Results")
|
444 |
|
|
|
445 |
for model_name, results in st.session_state.all_results.items():
|
446 |
df = pd.DataFrame(results)
|
447 |
metrics = {
|
|
|
452 |
metrics_df = pd.DataFrame(model_metrics).T
|
453 |
|
454 |
st.subheader("Model Performance Comparison")
|
|
|
455 |
accuracy_chart = alt.Chart(
|
456 |
metrics_df.reset_index().melt(id_vars=['index'], value_vars=['Accuracy'])
|
457 |
).mark_bar().encode(
|
|
|
470 |
)
|
471 |
|
472 |
st.altair_chart(accuracy_chart, use_container_width=True)
|
|
|
473 |
if st.session_state.all_results:
|
474 |
st.subheader("Detailed Results")
|
475 |
|
|
|
514 |
|
515 |
col1, col2 = st.columns(2)
|
516 |
with col1:
|
517 |
+
st.write("**Model Response:**")
|
518 |
+
st.code(result.get('model_response', "N/A"))
|
519 |
with col2:
|
520 |
+
st.write("**Explanation:**")
|
521 |
+
st.code(result.get('explanation', "N/A"))
|
522 |
|
523 |
col1, col2 = st.columns(2)
|
524 |
with col1:
|
|
|
530 |
else:
|
531 |
st.error("Incorrect")
|
532 |
|
533 |
+
st.write("**Timestamp:**", result['timestamp'])
|
534 |
else:
|
535 |
st.info(f"No results available for {selected_model_details} on {selected_dataset_details}. Please run the evaluation first.")
|
536 |
|
537 |
st.markdown("---")
|
538 |
+
st.subheader("Download Results")
|
539 |
all_data = []
|
|
|
540 |
for model_name, results in st.session_state.all_results.items():
|
541 |
+
for result in results:
|
542 |
+
row = result.copy()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
543 |
all_data.append(row)
|
544 |
|
545 |
complete_df = pd.DataFrame(all_data)
|
|
|
546 |
csv = complete_df.to_csv(index=False)
|
|
|
547 |
st.download_button(
|
548 |
label="Download All Results as CSV",
|
549 |
data=csv,
|