Delete app.py

app.py

@@ -1,64 +0,0 @@
-import gradio as gr
-import torch
-from transformers import AutoTokenizer, AutoModelForCausalLM
-import os
-
-# Set the device for model inference
-# This will automatically use the GPU if one is available and configured
-device = "cuda" if torch.cuda.is_available() else "cpu"
-
-# Load the tokenizer
-tokenizer = AutoTokenizer.from_pretrained("google/medgemma-4b-it")
-
-# Load the model and ensure it's on the correct device
-# We've added `load_in_8bit=True` to reduce the memory footprint.
-# We've also added `offload_folder` to explicitly enable disk offloading
-# for the model when it can't fit into VRAM or system RAM.
-model = AutoModelForCausalLM.from_pretrained(
-    "google/medgemma-4b-it",
-    torch_dtype=torch.bfloat16,
-    device_map="auto",
-    load_in_8bit=True,
-    offload_folder="./offload_dir"
-)
-
-def chat_interface(message, history):
-    """
-    Main chat function to interact with the model.
-    """
-    # The Gradio history is a list of lists.
-    # We need to format the chat history for the tokenizer's chat template.
-    chat_history = []
-    for user_msg, bot_msg in history:
-        chat_history.append({"role": "user", "content": user_msg})
-        chat_history.append({"role": "assistant", "content": bot_msg})
-    
-    # Add the current user message to the chat history
-    chat_history.append({"role": "user", "content": message})
-    
-    # Apply the tokenizer's chat template
-    prompt = tokenizer.apply_chat_template(chat_history, tokenize=False, add_generation_prompt=True)
-    
-    # Generate the response
-    input_ids = tokenizer.encode(prompt, add_special_tokens=False, return_tensors="pt")
-    outputs = model.generate(
-        input_ids.to(device),
-        max_new_tokens=256,
-        do_sample=True,
-        temperature=0.7,
-        top_k=50,
-        top_p=0.95
-    )
-    
-    # Decode the response and extract the model's part
-    response = tokenizer.decode(outputs[0])
-    response = response.split("<end_of_turn>")[1].strip()
-    
-    return response
-
-# Create the Gradio interface
-gr.ChatInterface(
-    fn=chat_interface,
-    title="MedGemma-4B-IT Medical Assistant",
-    description="A fine-tuned model for medical-related questions."
-).launch()