up

chatbot.py

@@ -46,9 +46,7 @@ warnings.filterwarnings('ignore', category=UserWarning)
 os.environ.update({
     'TRANSFORMERS_VERBOSITY': 'error',
     'TOKENIZERS_PARALLELISM': 'false',
-    'BITSANDBYTES_NOWELCOME': '1',
-    'TRANSFORMERS_CACHE': '/tmp/huggingface',  # Set cache directory to /tmp
-    'HF_HOME': '/tmp/huggingface'  # Set Hugging Face home directory to /tmp
+    'BITSANDBYTES_NOWELCOME': '1'
 })
 
 # Define base directory and paths
@@ -290,16 +288,20 @@ Response:"""
                 "text-classification",
                 model="SamLowe/roberta-base-go_emotions",
                 top_k=None,
-                device_map="auto" if self.device == "cuda" else None
+                device_map="auto" if self.device == "cuda" else None,
+                local_files_only=False,  # Force download from Hugging Face Hub
+                use_auth_token=False  # No authentication needed for public models
             )
         except Exception as e:
             logger.error(f"Error loading emotion model: {e}")
-            # Fallback 
+            # Fallback to a simpler model
             return pipeline(
                 "text-classification", 
                 model="j-hartmann/emotion-english-distilroberta-base", 
                 return_all_scores=True,
-                device_map="auto" if self.device == "cuda" else None
+                device_map="auto" if self.device == "cuda" else None,
+                local_files_only=False,
+                use_auth_token=False
             )
     
     def _initialize_llm(self, model_name: str, use_4bit: bool):
@@ -315,23 +317,34 @@ Response:"""
             else:
                 quantization_config = None
 
-            # Load base model
+            # Load base model directly from Hugging Face Hub
             logger.info(f"Loading base model: {model_name}")
             base_model = AutoModelForCausalLM.from_pretrained(
                 model_name,
                 quantization_config=quantization_config,
                 device_map="auto" if self.device == "cuda" else None,
-                trust_remote_code=True
+                trust_remote_code=True,
+                local_files_only=False,
+                use_auth_token=False
             )
 
-            # Load tokenizer
+            # Load tokenizer directly from Hugging Face Hub
             logger.info("Loading tokenizer")
-            tokenizer = AutoTokenizer.from_pretrained(model_name)
+            tokenizer = AutoTokenizer.from_pretrained(
+                model_name,
+                local_files_only=False,
+                use_auth_token=False
+            )
             tokenizer.pad_token = tokenizer.eos_token
 
-            # Load PEFT model from Hugging Face
+            # Load PEFT model directly from Hugging Face Hub
             logger.info(f"Loading PEFT model from {self.peft_model_path}")
-            model = PeftModel.from_pretrained(base_model, self.peft_model_path)
+            model = PeftModel.from_pretrained(
+                base_model, 
+                self.peft_model_path,
+                local_files_only=False,
+                use_auth_token=False
+            )
             logger.info("Successfully loaded PEFT model")
 
             # Create text generation pipeline