4th commit - change position of spaces.GPU

app.py

@@ -172,7 +172,6 @@ def query_vector_store(vector_store: FAISS, query, k=4, relevance_threshold=0.8)
     context = [doc.page_content for doc in similar_docs]
     return context
 
-@spaces.GPU
 class ModelWorker:
     def __init__(self, model_path, device='cuda'):
         self.device = device
@@ -212,6 +211,7 @@ class ModelWorker:
         for token_id in streamer:
             yield token_id
 
+    @spaces.GPU
     def generate_stream_gate(self, params):
         try:
             for x in self.generate_stream(params):
@@ -263,7 +263,7 @@ if __name__ == "__main__":
     parser.add_argument("--flow-path", type=str, default="THUDM/glm-4-voice-decoder")
     parser.add_argument("--model-path", type=str, default="THUDM/glm-4-voice-9b")
     parser.add_argument("--tokenizer-path", type=str, default="THUDM/glm-4-voice-tokenizer")
-    parser.add_argument("--whisper_model", type=str, default="base")
+    # parser.add_argument("--whisper_model", type=str, default="base")
     parser.add_argument("--share", action='store_true')
     args = parser.parse_args()
 
@@ -285,9 +285,9 @@ if __name__ == "__main__":
     model_worker = None
 
     # RAG configuration
-    Embedding_Model = '/root/autodl-tmp/rag/multilingual-e5-large-instruct'
-    file_paths = ['/root/autodl-tmp/rag/me.txt', "/root/autodl-tmp/rag/2024-Wealth-Outlook-MidYear-Edition.pdf"]
-    store_path = '/root/autodl-tmp/rag/me.faiss'
+    Embedding_Model = 'intfloat/multilingual-e5-large-instruct'
+    file_paths = []
+    store_path = './data.faiss'
 
     def initialize_fn():
         global audio_decoder, feature_extractor, whisper_model, glm_model, glm_tokenizer
@@ -312,7 +312,7 @@ if __name__ == "__main__":
         embedding_model = create_embedding_model(Embedding_Model)
         vector_store = load_or_create_store(store_path, file_paths, embedding_model)
     
-        whisper_transcribe_model = whisper.load_model("/root/autodl-tmp/whisper/base/base.pt")
+        whisper_transcribe_model = whisper.load_model("base")
 
     def clear_fn():
         return [], [], '', '', '', None, None