Update app.py

app.py

@@ -3,7 +3,6 @@ import torch
 from transformers import pipeline, WhisperProcessor, WhisperForConditionalGeneration, AutoModelForCausalLM, AutoTokenizer
 from huggingface_hub import login
 import os
-from accelerate import init_empty_weights, load_checkpoint_and_dispatch
 
 # Retrieve the token from the environment variable
 hf_api_token = os.getenv("HF_API_TOKEN")
@@ -19,20 +18,8 @@ whisper_processor = WhisperProcessor.from_pretrained("openai/whisper-base")
 whisper_model = WhisperForConditionalGeneration.from_pretrained("openai/whisper-base")
 
 # Initialize the summarization model and tokenizer
-# Load LLAMA 7B model with accelerate
-model_name = "meta-llama/Llama-2-7b-hf"
-with init_empty_weights():
-    summarization_model = AutoModelForCausalLM.from_pretrained(model_name)
-
-# Load checkpoint and dispatch model
-summarization_model = load_checkpoint_and_dispatch(
-    summarization_model,
-    checkpoint=model_name,
-    device_map="auto",
-    dtype=torch.float16
-)
-
-summarization_tokenizer = AutoTokenizer.from_pretrained(model_name)
+summarization_model = AutoModelForCausalLM.from_pretrained("meta-llama/Llama-2-7b-hf")
+summarization_tokenizer = AutoTokenizer.from_pretrained("meta-llama/Llama-2-7b-hf")
 
 # Function to transcribe audio
 def transcribe_audio(audio_file):
@@ -59,7 +46,7 @@ def process_audio(audio_file):
 # Gradio UI
 iface = gr.Interface(
     fn=process_audio,
-    inputs=gr.Audio(type="file"),
+    inputs=gr.Audio(source="upload", type="file"),
     outputs=[
         gr.Textbox(label="Transcription"),
         gr.Textbox(label="Summary")