Update app.py

app.py

@@ -52,15 +52,14 @@ text_to_image_pipelines = {}
 text_to_speech_pipelines = {}
 
 # Initialize pipelines for other tasks
-device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
-logger.info(f"Device set to use {device}")
 
-visual_qa_pipeline = pipeline("visual-question-answering", model="dandelin/vilt-b32-finetuned-vqa", device=device)
-document_qa_pipeline = pipeline("question-answering", model="deepset/roberta-base-squad2", device=device)
-image_classification_pipeline = pipeline("image-classification", model="facebook/deit-base-distilled-patch16-224", device=device)
-object_detection_pipeline = pipeline("object-detection", model="facebook/detr-resnet-50", device=device)
-video_classification_pipeline = pipeline("video-classification", model="facebook/timesformer-base-finetuned-k400", device=device)
-summarization_pipeline = pipeline("summarization", model="facebook/bart-large-cnn", device=device)
+
+visual_qa_pipeline = pipeline("visual-question-answering", model="dandelin/vilt-b32-finetuned-vqa")
+document_qa_pipeline = pipeline("question-answering", model="deepset/roberta-base-squad2")
+image_classification_pipeline = pipeline("image-classification", model="facebook/deit-base-distilled-patch16-224")
+object_detection_pipeline = pipeline("object-detection", model="facebook/detr-resnet-50")
+video_classification_pipeline = pipeline("video-classification", model="facebook/timesformer-base-finetuned-k400")
+summarization_pipeline = pipeline("summarization", model="facebook/bart-large-cnn")
 
 # Load speaker embeddings for text-to-audio
 def load_speaker_embeddings(model_name):
@@ -74,14 +73,14 @@ def load_speaker_embeddings(model_name):
 
 # Use a different model for text-to-audio if stabilityai/stable-audio-open-1.0 is not supported
 try:
-    text_to_audio_pipeline = pipeline("text-to-audio", model="stabilityai/stable-audio-open-1.0", device=device)
+    text_to_audio_pipeline = pipeline("text-to-audio", model="stabilityai/stable-audio-open-1.0")
 except ValueError as e:
     logger.error(f"Error loading stabilityai/stable-audio-open-1.0: {e}")
     logger.info("Falling back to a different text-to-audio model.")
-    text_to_audio_pipeline = pipeline("text-to-audio", model="microsoft/speecht5_tts", device=device)
+    text_to_audio_pipeline = pipeline("text-to-audio", model="microsoft/speecht5_tts")
     speaker_embeddings = load_speaker_embeddings("microsoft/speecht5_tts")
 
-audio_classification_pipeline = pipeline("audio-classification", model="facebook/wav2vec2-base", device=device)
+audio_classification_pipeline = pipeline("audio-classification", model="facebook/wav2vec2-base")
 
 def load_conversational_model(model_name):
     if model_name not in conversational_models_loaded:
@@ -115,7 +114,7 @@ def chat(model_name, user_input, history=[]):
     tokenizer, model = load_conversational_model(model_name)
     
     # Encode the input
-    input_ids = tokenizer.encode(user_input + tokenizer.eos_token, return_tensors="pt").to(device)
+    input_ids = tokenizer.encode(user_input + tokenizer.eos_token, return_tensors="pt")
     
     # Generate a response
     with torch.no_grad():