top_p/check_hybrid (#43)

oracle-samples · Nov 15, 2024 · a86b4d3 · a86b4d3
1 parent 162f087
commit a86b4d3
Show file tree

Hide file tree

Showing 2 changed files with 10 additions and 12 deletions.
diff --git a/app/src/modules/metadata.py b/app/src/modules/metadata.py
@@ -95,7 +95,7 @@ def ll_models():
             "openai_compat": False,
             "context_length": 127072,
             "temperature": [0.3, 0.3, 0.0, 2.0],
-            "top_p": [0.75, 0.75, 0.0, 1.0],
+            "top_p": [1.0, 1.0, 0.0, 1.0],
             "max_tokens": [100, 100, 1, 4096],
             "frequency_penalty": [0.0, 0.0, -1.0, 1.0],
             "presence_penalty": [0.0, 0.0, -2.0, 2.0],
@@ -108,7 +108,7 @@ def ll_models():
             "openai_compat": True,
             "context_length": 4191,
             "temperature": [1.0, 1.0, 0.0, 2.0],
-            "top_p": [0.9, 0.9, 0.0, 1.0],
+            "top_p": [1.0, 1.0, 0.0, 1.0],
             "max_tokens": [256, 256, 1, 4096],
             "frequency_penalty": [0.0, 0.0, -1.0, 1.0],
             "presence_penalty": [0.0, 0.0, -2.0, 2.0],
@@ -121,7 +121,7 @@ def ll_models():
             "openai_compat": True,
             "context_length": 127072,
             "temperature": [1.0, 1.0, 0.0, 2.0],
-            "top_p": [0.9, 0.9, 0.0, 1.0],
+            "top_p": [1.0, 1.0, 0.0, 1.0],
             "max_tokens": [256, 256, 1, 4096],
             "frequency_penalty": [0.0, 0.0, -1.0, 1.0],
             "presence_penalty": [0.0, 0.0, -2.0, 2.0],
@@ -134,7 +134,7 @@ def ll_models():
             "openai_compat": True,
             "context_length": 127072,
             "temperature": [1.0, 1.0, 0.0, 2.0],
-            "top_p": [0.9, 0.9, 0.0, 1.0],
+            "top_p": [1.0, 1.0, 0.0, 1.0],
             "max_tokens": [256, 256, 1, 8191],
             "frequency_penalty": [0.0, 0.0, -1.0, 1.0],
             "presence_penalty": [0.0, 0.0, -2.0, 2.0],
@@ -147,7 +147,7 @@ def ll_models():
             "openai_compat": True,
             "context_length": 127072,
             "temperature": [1.0, 1.0, 0.0, 2.0],
-            "top_p": [0.9, 0.9, 0.0, 1.0],
+            "top_p": [1.0, 1.0, 0.0, 1.0],
             "max_tokens": [256, 256, 1, 4095],
             "frequency_penalty": [0.0, 0.0, -1.0, 1.0],
             "presence_penalty": [0.0, 0.0, -2.0, 2.0],
@@ -160,7 +160,7 @@ def ll_models():
             "openai_compat": False,
             "context_length": 127072,
             "temperature": [0.2, 0.2, 0.0, 2.0],
-            "top_p": [0.9, 0.9, 0.0, 1.0],
+            "top_p": [1.0, 1.0, 0.0, 1.0],
             "max_tokens": [256, 256, 1, 28000],
             "frequency_penalty": [0.0, 0.0, -1.0, 1.0],
             "presence_penalty": [0.0, 0.0, -2.0, 2.0],
@@ -173,7 +173,7 @@ def ll_models():
             "openai_compat": False,
             "context_length": 127072,
             "temperature": [0.2, 0.2, 0.0, 2.0],
-            "top_p": [0.9, 0.9, 0.0, 1.0],
+            "top_p": [1.0, 1.0, 0.0, 1.0],
             "max_tokens": [256, 256, 1, 28000],
             "frequency_penalty": [0.0, 0.0, -1.0, 1.0],
             "presence_penalty": [0.0, 0.0, -2.0, 2.0],

diff --git a/app/src/modules/st_common.py b/app/src/modules/st_common.py
@@ -436,11 +436,9 @@ def create_zip(state_dict_filt, provider):
 # Check if the conf is full ollama or openai, currently supported for springai export
 def check_hybrid_conf(session_state_json):
 
-    embedding_models = meta.embedding_models()
-    chat_models = meta.ll_models()
-
-    embModel =  embedding_models.get(session_state_json["rag_params"].get("model"))
-    chatModel = chat_models.get(session_state_json["ll_model"])
+    chatModel = state.ll_model_config.get(session_state_json["ll_model"])
+    embModel = state.embed_model_config.get(state.rag_params["model"])
+
     logger.info("Model: %s",session_state_json["ll_model"])
     logger.info("Embedding Model embModel: %s",embModel)
     logger.info("Chat Model: %s",chatModel)