Fix: Load CPU-safe model for HF Space

app.py

@@ -1,4 +1,4 @@
-import gradio as gr
+import streamlit as st
 import torch
 from transformers import AutoTokenizer, AutoModelForCausalLM
 import re
@@ -9,31 +9,14 @@ import io
 
 # Load model and tokenizer from local fine-tuned directory
 # Define base and adapter model paths
-#BASE_MODEL = "stanford-crfm/BioMedLM"  # or the path you used originally
-#ADAPTER_PATH = "Tufan1/BioMedLM-Cardio-Fold4-CPU"
-#tokenizer = AutoTokenizer.from_pretrained(BASE_MODEL)
-MODEL_PATH = "Tufan1/BioMedLM-Cardio-Fold4-CPU"
+BASE_MODEL = "stanford-crfm/BioMedLM"  # or the path you used originally
+ADAPTER_PATH = "Tufan1/BioMedLM-Cardio-Fold4-CPU"
+tokenizer = AutoTokenizer.from_pretrained(BASE_MODEL)
+
 # Force CPU-safe model loading
-#base_model = AutoModelForCausalLM.from_pretrained(BASE_MODEL, torch_dtype=torch.float32)
-#model = PeftModel.from_pretrained(base_model, ADAPTER_PATH, device_map=None).to("cpu")
-
-tokenizer = AutoTokenizer.from_pretrained(MODEL_PATH)
-model = AutoModelForCausalLM.from_pretrained(MODEL_PATH, torch_dtype=torch.float16).to("cuda")
-
-def predict_disease(prompt):
-    inputs = tokenizer(prompt, return_tensors="pt").to("cuda")
-    outputs = model.generate(**inputs, max_new_tokens=100)
-    return tokenizer.decode(outputs[0], skip_special_tokens=True)
-
-iface = gr.Interface(
-    fn=predict_disease,
-    inputs=gr.Textbox(label="Enter Symptoms"),
-    outputs=gr.Textbox(label="Predicted Diagnosis"),
-    title="Cardiovascular Disease Predictor",
-    description="Enter patient symptoms to receive diagnosis based on BioMedLM."
-)
-
-iface.launch()
+base_model = AutoModelForCausalLM.from_pretrained(BASE_MODEL, torch_dtype=torch.float32)
+model = PeftModel.from_pretrained(base_model, ADAPTER_PATH, device_map=None).to("cpu")
+
 
 # Dictionaries to decode user inputs
 gender_map = {1: "Female", 2: "Male"}