Spaces:
Running
on
T4
Running
on
T4
model changed to Haiku 3.5 for query re=write
Browse files
semantic_search/query_rewrite.py
CHANGED
@@ -235,18 +235,18 @@ def get_new_query_res(query):
|
|
235 |
if(query == ""):
|
236 |
query = st.session_state.input_rekog_label
|
237 |
if(st.session_state.input_is_rewrite_query == 'enabled'):
|
238 |
-
|
239 |
get_prompt = prompt_.format(query=query,schema = schema)
|
240 |
-
print(f"prompt latency: {time.time() - start:.2f} sec")
|
241 |
-
start = time.time()
|
242 |
res = invoke_models.invoke_llm_model( get_prompt,False)
|
243 |
-
print(f"Claude latency: {time.time() - start:.2f} sec")
|
244 |
inter_query = res[7:-3].replace('\\"',"'").replace("\n","")
|
245 |
-
start = time.time()
|
246 |
query_struct = StructuredQueryOutputParser.from_components().parse(inter_query)
|
247 |
opts = OpenSearchTranslator()
|
248 |
result_query_llm = opts.visit_structured_query(query_struct)[1]['filter']
|
249 |
-
print(f"Langchain latency: {time.time() - start:.2f} sec")
|
250 |
draft_new_query = {'bool':{'should':[],'must':[]}}
|
251 |
if('bool' in result_query_llm and ('must' in result_query_llm['bool'] or 'should' in result_query_llm['bool'])):
|
252 |
#draft_new_query['bool']['should'] = []
|
|
|
235 |
if(query == ""):
|
236 |
query = st.session_state.input_rekog_label
|
237 |
if(st.session_state.input_is_rewrite_query == 'enabled'):
|
238 |
+
#tart = time.time()
|
239 |
get_prompt = prompt_.format(query=query,schema = schema)
|
240 |
+
#print(f"prompt latency: {time.time() - start:.2f} sec")
|
241 |
+
#start = time.time()
|
242 |
res = invoke_models.invoke_llm_model( get_prompt,False)
|
243 |
+
#print(f"Claude latency: {time.time() - start:.2f} sec")
|
244 |
inter_query = res[7:-3].replace('\\"',"'").replace("\n","")
|
245 |
+
#start = time.time()
|
246 |
query_struct = StructuredQueryOutputParser.from_components().parse(inter_query)
|
247 |
opts = OpenSearchTranslator()
|
248 |
result_query_llm = opts.visit_structured_query(query_struct)[1]['filter']
|
249 |
+
#print(f"Langchain latency: {time.time() - start:.2f} sec")
|
250 |
draft_new_query = {'bool':{'should':[],'must':[]}}
|
251 |
if('bool' in result_query_llm and ('must' in result_query_llm['bool'] or 'should' in result_query_llm['bool'])):
|
252 |
#draft_new_query['bool']['should'] = []
|
utilities/invoke_models.py
CHANGED
@@ -108,7 +108,7 @@ def read_from_table(file,question):
|
|
108 |
|
109 |
model = BedrockChat(
|
110 |
client=bedrock_runtime_client,
|
111 |
-
model_id=
|
112 |
model_kwargs=bedrock_params,
|
113 |
streaming=False
|
114 |
)
|
@@ -130,7 +130,7 @@ def generate_image_captions_llm(base64_string,question):
|
|
130 |
|
131 |
|
132 |
response = bedrock_runtime_client.invoke_model(
|
133 |
-
modelId= "anthropic.claude-3-haiku-
|
134 |
contentType = "application/json",
|
135 |
accept = "application/json",
|
136 |
|
|
|
108 |
|
109 |
model = BedrockChat(
|
110 |
client=bedrock_runtime_client,
|
111 |
+
model_id="anthropic.claude-3-5-haiku-20241022-v1:0",
|
112 |
model_kwargs=bedrock_params,
|
113 |
streaming=False
|
114 |
)
|
|
|
130 |
|
131 |
|
132 |
response = bedrock_runtime_client.invoke_model(
|
133 |
+
modelId= "anthropic.claude-3-5-haiku-20241022-v1:0",
|
134 |
contentType = "application/json",
|
135 |
accept = "application/json",
|
136 |
|