Spaces:
Building
on
T4
Building
on
T4
Update app.py
Browse files
app.py
CHANGED
@@ -215,6 +215,7 @@ def main(query: str, client: QdrantClient, collection_name: str, llm, dense_mode
|
|
215 |
|
216 |
regex = build_regex_from_schema(schema, r"[\n ]?")
|
217 |
gen_text = outlines.generate.regex(llm, regex)
|
|
|
218 |
|
219 |
gen_choice = outlines.generate.choice(llm, choices=['Yes', 'No'])
|
220 |
prompt = route_llm(context, query)
|
@@ -231,7 +232,7 @@ def main(query: str, client: QdrantClient, collection_name: str, llm, dense_mode
|
|
231 |
result_metadatas = "\n\n".join(f'{value}' for value in filtered_metadatas)
|
232 |
|
233 |
prompt = answer_with_context(context, query)
|
234 |
-
answer = gen_text(prompt, max_tokens=300, sampling_params=SamplingParams(temperature=0.6, top_p=0.9))
|
235 |
answer = f"{answer}\n\n\nSource(s) :\n\n{result_metadatas}"
|
236 |
|
237 |
if not st.session_state.documents_only:
|
@@ -243,14 +244,15 @@ def main(query: str, client: QdrantClient, collection_name: str, llm, dense_mode
|
|
243 |
print(f'Choice 2: {action}')
|
244 |
if action == 'General Question':
|
245 |
prompt = open_query_prompt(past_messages, query)
|
246 |
-
answer = gen_text(prompt, max_tokens=300, sampling_params=SamplingParams(temperature=0.6, top_p=0.9))
|
247 |
else:
|
248 |
if st.session_state.documents_only:
|
249 |
prompt = idk(query)
|
250 |
-
answer = gen_text(prompt, max_tokens=128, sampling_params=SamplingParams(temperature=0.6, top_p=0.9))
|
|
|
251 |
else:
|
252 |
prompt = self_knowledge(query)
|
253 |
-
answer = gen_text(prompt, max_tokens=300, sampling_params=SamplingParams(temperature=0.6, top_p=0.9))
|
254 |
answer = f'Internal Knowledge :\n\n{answer}'
|
255 |
|
256 |
torch.cuda.empty_cache()
|
|
|
215 |
|
216 |
regex = build_regex_from_schema(schema, r"[\n ]?")
|
217 |
gen_text = outlines.generate.regex(llm, regex)
|
218 |
+
gen_text.format_sequence = lambda x: schema_object.parse_raw(x)
|
219 |
|
220 |
gen_choice = outlines.generate.choice(llm, choices=['Yes', 'No'])
|
221 |
prompt = route_llm(context, query)
|
|
|
232 |
result_metadatas = "\n\n".join(f'{value}' for value in filtered_metadatas)
|
233 |
|
234 |
prompt = answer_with_context(context, query)
|
235 |
+
answer = gen_text(prompt, max_tokens=300, sampling_params=SamplingParams(temperature=0.6, top_p=0.9))
|
236 |
answer = f"{answer}\n\n\nSource(s) :\n\n{result_metadatas}"
|
237 |
|
238 |
if not st.session_state.documents_only:
|
|
|
244 |
print(f'Choice 2: {action}')
|
245 |
if action == 'General Question':
|
246 |
prompt = open_query_prompt(past_messages, query)
|
247 |
+
answer = gen_text(prompt, max_tokens=300, sampling_params=SamplingParams(temperature=0.6, top_p=0.9))
|
248 |
else:
|
249 |
if st.session_state.documents_only:
|
250 |
prompt = idk(query)
|
251 |
+
answer = gen_text(prompt, max_tokens=128, sampling_params=SamplingParams(temperature=0.6, top_p=0.9))
|
252 |
+
print(f'TYPE: {type(answer)})
|
253 |
else:
|
254 |
prompt = self_knowledge(query)
|
255 |
+
answer = gen_text(prompt, max_tokens=300, sampling_params=SamplingParams(temperature=0.6, top_p=0.9))
|
256 |
answer = f'Internal Knowledge :\n\n{answer}'
|
257 |
|
258 |
torch.cuda.empty_cache()
|