Update app.py

app.py

@@ -1,9 +1,10 @@
 import gradio as gr
 from peft import AutoPeftModelForCausalLM
 from transformers import AutoTokenizer
-from huggingface_hub import login
+from huggingface_hub import login, snapshot_download
 import torch
-import os  # For env vars
+import os
+import json
 
 # Login using secret (secure, no hardcode)
 login(os.environ["HF_TOKEN"])
@@ -14,7 +15,18 @@ base_model_id = "mistralai/Ministral-8B-Instruct-2410"
 MAX_LENGTH = 32768
 DEVICE = "cuda" if torch.cuda.is_available() else "cpu"  # Use GPU if available (ZeroGPU on HF)
 
-model = AutoPeftModelForCausalLM.from_pretrained(model_id)
+# Workaround: Download model files, edit adapter_config.json to remove 'model_type'
+local_dir = snapshot_download(repo_id=model_id)
+config_path = os.path.join(local_dir, "adapter_config.json")
+with open(config_path, 'r') as f:
+    config = json.load(f)
+if 'model_type' in config:
+    del config['model_type']
+with open(config_path, 'w') as f:
+    json.dump(config, f)
+
+# Now load the model from modified local dir
+model = AutoPeftModelForCausalLM.from_pretrained(local_dir)
 tokenizer = AutoTokenizer.from_pretrained(base_model_id)
 model = model.to(DEVICE)
 model.eval()