Update app.py

app.py

@@ -3,7 +3,7 @@ import os
 import gradio as gr
 import torch
 from transformers import AutoTokenizer, AutoModelForCausalLM
-import spaces  # Mandatory library for Hugging Face ZeroGPU 
+import spaces  # Mandatory library for Hugging Face ZeroGPU [1]
 
 MODEL_ID = "DevStudio-AI/Devstudio-Coder-1.5B"
 # Fetch your secure token from the Space Secrets environment
@@ -17,10 +17,10 @@ tokenizer = AutoTokenizer.from_pretrained(
     "Qwen/Qwen2.5-Coder-1.5B-Instruct"
 )
 
-# 2. We load your custom model in 16-bit on CPU first
-# ZeroGPU will automatically move the model to the GPU when the decorated function runs [1]
+# 2. Load the model, pointing directly to the nested directory inside your repo
 model = AutoModelForCausalLM.from_pretrained(
     MODEL_ID,
+    subfolder="models/final_merged",  # Tells the loader to look inside this folder [1]
     torch_dtype=torch.float16,
     device_map="cpu",
     token=HF_TOKEN