change audiopreprocessing chunk length default to 10 from 20

marqo-ai · Sep 17, 2024 · 357b335 · 357b335
1 parent 1643647
commit 357b335
Show file tree

Hide file tree

Showing 2 changed files with 7 additions and 10 deletions.
diff --git a/src/marqo/s2_inference/model_registry.py b/src/marqo/s2_inference/model_registry.py
@@ -1995,7 +1995,7 @@ def _get_languagebind_properties() -> Dict:
             "model_size": 8,
             "supported_modalities": ["video", "audio", "language", "image"],
             "video_chunk_length": 20,
-            "audio_chunk_length": 20,
+            "audio_chunk_length": 10,
         },
         'LanguageBind/Video_V1.5_FT_Audio_FT': {
             "name": "LanguageBind/Video_V1.5_FT_Audio_FT",
@@ -2005,7 +2005,7 @@ def _get_languagebind_properties() -> Dict:
             "model_size": 5,
             "supported_modalities": ["video", "audio", "language"],
             "video_chunk_length": 20,
-            "audio_chunk_length": 20,
+            "audio_chunk_length": 10,
         },
         'LanguageBind/Video_V1.5_FT_Image': {
             "name": "LanguageBind/Video_V1.5_FT_Image",
@@ -2015,7 +2015,7 @@ def _get_languagebind_properties() -> Dict:
             "model_size": 5,
             "supported_modalities": ["video", "language", "image"],
             "video_chunk_length": 20,
-            "audio_chunk_length": 20,
+            "audio_chunk_length": 10,
         },
         'LanguageBind/Audio_FT_Image': {
             "name": "LanguageBind/Audio_FT_Image",
@@ -2025,7 +2025,7 @@ def _get_languagebind_properties() -> Dict:
             "model_size": 5,
             "supported_modalities": ["audio", "language", "image"],
             "video_chunk_length": 20,
-            "audio_chunk_length": 20,
+            "audio_chunk_length": 10,
         },
         'LanguageBind/Audio_FT': {
             "name": "LanguageBind/Audio_FT",
@@ -2035,7 +2035,7 @@ def _get_languagebind_properties() -> Dict:
             "model_size": 2,
             "supported_modalities": ["video", "language"],
             "video_chunk_length": 20,
-            "audio_chunk_length": 20,
+            "audio_chunk_length": 10,
         },
         'LanguageBind/Video_V1.5_FT': {
             "name": "LanguageBind/Video_V1.5_FT",
@@ -2045,7 +2045,7 @@ def _get_languagebind_properties() -> Dict:
             "model_size": 2,
             "supported_modalities": ["video", "language"],
             "video_chunk_length": 20,
-            "audio_chunk_length": 20,
+            "audio_chunk_length": 10,
         },
 
     }

diff --git a/src/marqo/tensor_search/models/index_settings.py b/src/marqo/tensor_search/models/index_settings.py
@@ -41,7 +41,7 @@ class IndexSettings(StrictBaseModel):
         splitOverlap=3,
     )
     audioPreprocessing: core.AudioPreProcessing = core.AudioPreProcessing(
-        splitLength=20,
+        splitLength=10,
         splitOverlap=3,
     )
     vectorNumericType: core.VectorNumericType = core.VectorNumericType.Float
@@ -57,8 +57,6 @@ class IndexSettings(StrictBaseModel):
     def validate_url_pointer_treatment(cls, values):
         treat_as_images = values.get('treatUrlsAndPointersAsImages')
         treat_as_media = values.get('treatUrlsAndPointersAsMedia')
-        print("treat_as_images", treat_as_images)
-        print("treat_as_media", treat_as_media)
         if treat_as_images is None:
             if treat_as_media is True:
                 treat_as_images = None #False
@@ -153,7 +151,6 @@ def to_marqo_index_request(self, index_name: str) -> MarqoIndexRequest:
                 updated_at=time.time(),
             )
         elif self.type == core.IndexType.Unstructured:
-            print("treatUrlsAndPointersAsImages", self.treatUrlsAndPointersAsImages)
             if self.allFields is not None:
                 raise api_exceptions.InvalidArgError(
                     "allFields is not a valid parameter for unstructured indexes"