Update app.py

app.py

@@ -54,7 +54,7 @@ config = 'configs/audio/ePALM_audio_caption.yaml'
 config = yaml.load(open(config, 'r'))
 
 text_model = 'facebook/opt-2.7b' 
-vision_model_name = 'vit_base_patch16_224'
+vision_model_name = 'ast'
 
 
 start_layer_idx = 19
@@ -72,7 +72,7 @@ MODEL = ePALM(opt_model_name=text_model,
 print("Model Built")
 MODEL.to(device)
 
-checkpoint_path = 'checkpoints/float32/ePALM_caption/checkpoint_best.pth'
+checkpoint_path = 'checkpoints/float32/ePALM_audio_caption/checkpoint_best.pth'
 checkpoint = torch.load(checkpoint_path, map_location='cpu') 
 state_dict = checkpoint['model']
 msg = MODEL.load_state_dict(state_dict,strict=False)  
@@ -82,13 +82,6 @@ MODEL.bfloat16()
 
 
 
-# Audio Captioning
-checkpoint_path = 'checkpoints/float32/ePALM_audio_caption/checkpoint_best.pth'
-checkpoint = torch.load(checkpoint_path, map_location='cpu') 
-state_dict_audio_caption = checkpoint['model']
-
-
-
 
 ## Load tokenizer
 tokenizer = AutoTokenizer.from_pretrained(text_model, use_fast=False)