Update app.py

app.py

@@ -18,8 +18,13 @@ whisper_processor = WhisperProcessor.from_pretrained("openai/whisper-base")
 whisper_model = WhisperForConditionalGeneration.from_pretrained("openai/whisper-base")
 
 # Initialize the summarization model and tokenizer
-summarization_model = AutoModelForCausalLM.from_pretrained("meta-llama/Llama-2-7b-hf")
-summarization_tokenizer = AutoTokenizer.from_pretrained("meta-llama/Llama-2-7b-hf")
+# Use a smaller version of the Llama model and load in FP16
+summarization_model = AutoModelForCausalLM.from_pretrained(
+    "meta-llama/Llama-2-3b-hf",
+    torch_dtype=torch.float16,
+    low_cpu_mem_usage=True
+)
+summarization_tokenizer = AutoTokenizer.from_pretrained("meta-llama/Llama-2-3b-hf")
 
 # Function to transcribe audio
 def transcribe_audio(audio_file):