more changes to support no gpu

app.py

@@ -1,7 +1,11 @@
+import os
 import gradio as gr
 from unsloth import FastLanguageModel
 import torch
 
+# Disable CUDA explicitly by setting the environment variable
+os.environ["CUDA_VISIBLE_DEVICES"] = ""  # Disabling CUDA
+
 # Set device to CPU
 device = torch.device("cpu")
 
@@ -15,7 +19,7 @@ model, tokenizer = FastLanguageModel.from_pretrained(
     max_seq_length=max_seq_length,
     dtype=dtype,
     load_in_4bit=True,
-).to(device)  # Make sure the model is on CPU
+).to(device)  # Ensure the model is on CPU
 
 # Enable native faster inference if possible
 FastLanguageModel.for_inference(model)

requirements.txt

@@ -2,4 +2,4 @@ unsloth==2024.12.4
 transformers==4.47.0
 gradio==5.8.0
 bitsandbytes==0.45.0
-torch==2.5.1
+torch==2.1.0+cpu