Update app.py

app.py

@@ -68,7 +68,7 @@ qdrant_client = QdrantClient(
 
 print(qdrant_client.get_collections())
 
-class ChatGroq:
+class CustomChatGroq:
     def __init__(self, temperature, model_name, api_key):
         self.temperature = temperature
         self.model_name = model_name
@@ -135,13 +135,13 @@ bnb_config = BitsAndBytesConfig(
     load_in_8bit=True,
 )
 
-mistral_tokenizer = AutoTokenizer.from_pretrained(MODEL_PATH,use_auth_token=True)
+mistral_tokenizer = AutoTokenizer.from_pretrained(MODEL_PATH,token=hf_api_key)
 
 judge_llm = AutoModelForCausalLM.from_pretrained(
     MODEL_PATH,
     quantization_config=bnb_config,torch_dtype=torch.float16,
     device_map="auto",
-    use_auth_token=True
+    token=hf_api_key
 )
 
 print(judge_llm.hf_device_map)
@@ -601,7 +601,7 @@ def eval_question_quality(
         }
 
     except Exception as e:
-        logging.error(f"Error in eval_question_quality: {type(e)._name_}: {e}", exc_info=True)
+        logging.error(f"Error in eval_question_quality: {type(e).__name__}: {e}", exc_info=True)
         return {
             "Score": "Poor",
             "Reasoning": f"Critical error occurred: {str(e)}. Defaulted to 'Poor'.",
@@ -1459,13 +1459,15 @@ import json
 from moviepy.editor import VideoFileClip
 import shutil
 from transformers import BarkModel, AutoProcessor
-import torch
+import torch, gc
 import whisper
 from transformers import Wav2Vec2Processor, Wav2Vec2ForSequenceClassification
 import librosa
 
 import torch
 print(torch.cuda.is_available())  # ✅ Tells you if GPU is available
+torch.cuda.empty_cache()
+gc.collect()
 
 
 # Bark TTS