taxabind update

requirements.txt

@@ -5,3 +5,4 @@ huggingface_hub>=0.23.1
 einops>=0.8.0
 transformers>=4.41.1
 open_clip_torch>=2.24.0
+torchaudio

rshf/taxabind/model.py

@@ -25,7 +25,7 @@ def get_image_processor(self):
         transforms.Normalize(mean=self.config.normalize_mean, std=self.config.normalize_std)])
 
     def get_audio_encoder(self):
-        return ClapAudioModelWithProjection(self.config.audio_encoder)
+        return ClapAudioModelWithProjection.from_pretrained(self.config.audio_encoder)
 
     def process_audio(self, track, sr):
         processor = ClapProcessor.from_pretrained(self.config.audio_encoder)
@@ -47,7 +47,7 @@ def get_env_processor(self):
         return SINR.preprocess_locs
 
     def get_sat_encoder(self):
-        return CLIPVisionModelWithProjection(self.config.sat_encoder)
+        return CLIPVisionModelWithProjection.from_pretrained(self.config.sat_encoder)
 
     def get_sat_processor(self):
         return transforms.Compose([transforms.Resize((self.config.sat_resize, self.config.sat_resize)),

setup.cfg

@@ -1,6 +1,6 @@
 [metadata]
 name = rshf
-version = 0.1.0
+version = 0.1.1
 author = Srikumar Sastry
 author_email = s.sastry@wustl.edu
 description = RS pretrained models in huggingface style