JPBianchi commited on
Commit
fc26027
1 Parent(s): 8d089a5

fixed issue in backend

Browse files
Files changed (2) hide show
  1. app.py +34 -26
  2. backend.py +3 -2
app.py CHANGED
@@ -152,8 +152,6 @@ if 'secrets' in st.secrets:
152
 
153
  hf_token = st.secrets['secrets']['LLAMA2_ENDPOINT_HF_TOKEN_chris']
154
  hf_endpoint = st.secrets['secrets']['LLAMA2_ENDPOINT_UPLIMIT']
155
- model_token_id = st.secrets['secrets']['MODAL_TOKEN_ID']
156
- modal_token_secret = st.secrets['secrets']['MODAL_TOKEN_SECRET']
157
 
158
  else :
159
  # st.write("Loading secrets for Huggingface")
@@ -164,8 +162,6 @@ else :
164
 
165
  hf_token = st.secrets['LLAMA2_ENDPOINT_HF_TOKEN_chris']
166
  hf_endpoint = st.secrets['LLAMA2_ENDPOINT_UPLIMIT']
167
- model_token_id = st.secrets['MODAL_TOKEN_ID']
168
- modal_token_secret = st.secrets['MODAL_TOKEN_SECRET']
169
 
170
  # else:
171
  # # if we want to use env file
@@ -183,7 +179,8 @@ model_default = 'models/finetuned-all-mpnet-base-v2-300' if we_are_not_online \
183
 
184
  available_models = ['sentence-transformers/all-mpnet-base-v2',
185
  'sentence-transformers/all-MiniLM-L6-v2',
186
- 'models/finetuned-all-mpnet-base-v2-300']
 
187
 
188
  #%%
189
  models_urls = {'models/finetuned-all-mpnet-base-v2-300': "https://drive.google.com/drive/folders/1asJ37-AUv5nytLtH6hp6_bVV3_cZOXfj"}
@@ -307,8 +304,10 @@ def main():
307
  st.text("Running OFFLINE")
308
  st.write("----------")
309
 
310
- alpha_input = st.slider(label='Alpha',min_value=0.00, max_value=1.00, value=0.40, step=0.05)
311
- retrieval_limit = st.slider(label='Hybrid Search Results', min_value=10, max_value=300, value=10, step=10)
 
 
312
 
313
  hybrid_filter = st.toggle('Filter Guest', True) # i.e. look only at guests' data
314
 
@@ -322,7 +321,7 @@ def main():
322
 
323
  rag_it = st.toggle('RAG it', True)
324
  if rag_it:
325
- st.sidebar.write(f"Using LLM '{model_nameGPT}'")
326
  llm_temperature = st.slider(label='LLM T˚', min_value=0.0, max_value=2.0, value=0.01, step=0.10 )
327
 
328
  model_name_or_path = st.selectbox(label='Model Name:', options=available_models,
@@ -336,29 +335,35 @@ def main():
336
  from finetune_backend import finetune
337
  if 'finetuned' in model_name_or_path:
338
  st.write("Model already finetuned")
339
- elif model_name_or_path.startswith("models/"):
340
- st.write("Sentence Transformers models only!")
341
  else:
342
- try:
343
- if 'finetuned' in finetune_model:
344
- st.write("Model already finetuned")
345
- else:
346
- model_path = finetune(model_name_or_path, savemodel=True, outpath='models')
347
- if model_path is not None:
348
- if finetune_model.split('/')[-1] not in model_path:
349
- st.write(model_path) # a warning from finetuning in this case
350
- elif model_path not in available_models:
351
- # finetuning generated a model, let's add it
352
- available_models.append(model_path)
353
- st.write("Model saved!")
354
- except Exception:
355
- st.write("Model not found on HF or error")
 
356
  else:
357
  st.write("Finetuning not available on Streamlit online because of space limitations")
358
 
359
  model_name_or_path = check_model(model_name_or_path)
360
- client, available_classes = get_weaviate_client(Wapi_key, url, model_name_or_path, openai_api_key)
361
-
 
 
 
 
 
362
  start_class = 'Impact_theory_all_mpnet_base_v2_finetuned'
363
 
364
  class_name = st.selectbox(
@@ -406,6 +411,9 @@ def main():
406
 
407
  st.write("\u21D0 Open the sidebar to change Search settings \n ") # https://home.unicode.org also 21E0, 21B0 B2 D0
408
 
 
 
 
409
  col1, _ = st.columns([3,7])
410
  with col1:
411
  guest = st.selectbox('Select A Guest',
 
152
 
153
  hf_token = st.secrets['secrets']['LLAMA2_ENDPOINT_HF_TOKEN_chris']
154
  hf_endpoint = st.secrets['secrets']['LLAMA2_ENDPOINT_UPLIMIT']
 
 
155
 
156
  else :
157
  # st.write("Loading secrets for Huggingface")
 
162
 
163
  hf_token = st.secrets['LLAMA2_ENDPOINT_HF_TOKEN_chris']
164
  hf_endpoint = st.secrets['LLAMA2_ENDPOINT_UPLIMIT']
 
 
165
 
166
  # else:
167
  # # if we want to use env file
 
179
 
180
  available_models = ['sentence-transformers/all-mpnet-base-v2',
181
  'sentence-transformers/all-MiniLM-L6-v2',
182
+ 'models/finetuned-all-mpnet-base-v2-300',
183
+ 'sentence-transformers/all-MiniLM-L12-v2']
184
 
185
  #%%
186
  models_urls = {'models/finetuned-all-mpnet-base-v2-300': "https://drive.google.com/drive/folders/1asJ37-AUv5nytLtH6hp6_bVV3_cZOXfj"}
 
304
  st.text("Running OFFLINE")
305
  st.write("----------")
306
 
307
+ hubrid_search = st.toggle('Hybrid Search', True)
308
+ if hubrid_search:
309
+ alpha_input = st.slider(label='Alpha',min_value=0.00, max_value=1.00, value=0.40, step=0.05)
310
+ retrieval_limit = st.slider(label='Hybrid Search Results', min_value=10, max_value=300, value=10, step=10)
311
 
312
  hybrid_filter = st.toggle('Filter Guest', True) # i.e. look only at guests' data
313
 
 
321
 
322
  rag_it = st.toggle('RAG it', True)
323
  if rag_it:
324
+ st.write(f"Using LLM '{model_nameGPT}'")
325
  llm_temperature = st.slider(label='LLM T˚', min_value=0.0, max_value=2.0, value=0.01, step=0.10 )
326
 
327
  model_name_or_path = st.selectbox(label='Model Name:', options=available_models,
 
335
  from finetune_backend import finetune
336
  if 'finetuned' in model_name_or_path:
337
  st.write("Model already finetuned")
338
+ elif "models/" in model_name_or_path:
339
+ st.write("sentence-transformers models only!")
340
  else:
341
+ # try:
342
+ if 'finetuned' in model_name_or_path:
343
+ st.write("Model already finetuned")
344
+ else:
345
+ model_path = finetune(model_name_or_path, savemodel=True, outpath='models')
346
+ st.write(f"model_path returned = {model_path}")
347
+ if model_path is not None:
348
+ if model_name_or_path.split('/')[-1] not in model_path:
349
+ st.write(model_path) # a warning from finetuning in this case
350
+ elif model_path not in available_models:
351
+ # finetuning generated a model, let's add it
352
+ available_models.append(model_path)
353
+ st.write("Model saved!")
354
+ # except Exception:
355
+ # st.write("Model not found on HF or error")
356
  else:
357
  st.write("Finetuning not available on Streamlit online because of space limitations")
358
 
359
  model_name_or_path = check_model(model_name_or_path)
360
+ try:
361
+ client, available_classes = get_weaviate_client(Wapi_key, url, model_name_or_path, openai_api_key)
362
+ except Exception as e:
363
+ # Weaviate doesn't know this model, maybe we're just finetuning a model
364
+ st.sidebar.write(f"Model unknown to Weaviate")
365
+ st.stop()
366
+
367
  start_class = 'Impact_theory_all_mpnet_base_v2_finetuned'
368
 
369
  class_name = st.selectbox(
 
411
 
412
  st.write("\u21D0 Open the sidebar to change Search settings \n ") # https://home.unicode.org also 21E0, 21B0 B2 D0
413
 
414
+ if not hubrid_search:
415
+ st.stop()
416
+
417
  col1, _ = st.columns([3,7])
418
  with col1:
419
  guest = st.selectbox('Select A Guest',
backend.py CHANGED
@@ -100,7 +100,8 @@ def encode_content_splits(content_splits,
100
  )
101
  def finetune(training_path='./data/training_data_300.json',
102
  valid_path='./data/validation_data_100.json',
103
- model_id=None):
 
104
 
105
  import os
106
  print("Data:", os.listdir('/root/data'))
@@ -170,7 +171,7 @@ def finetune(training_path='./data/training_data_300.json',
170
  zippedio = bytes_buffer
171
 
172
  return zippedio
173
- except:
174
  return "Finetuning failed"
175
 
176
 
 
100
  )
101
  def finetune(training_path='./data/training_data_300.json',
102
  valid_path='./data/validation_data_100.json',
103
+ model_id=None,
104
+ ignore_existing=False):
105
 
106
  import os
107
  print("Data:", os.listdir('/root/data'))
 
171
  zippedio = bytes_buffer
172
 
173
  return zippedio
174
+ except Exception:
175
  return "Finetuning failed"
176
 
177