fix(setup): gracefully handle misconfiguration

app.py

@@ -31,15 +31,23 @@ MAX_RETRY_ATTEMPTS = 1
 print(f"Is CUDA available: {torch.cuda.is_available()}")
 print(f"CUDA device: {torch.cuda.get_device_name(torch.cuda.current_device())}")
 
-tokenizer = AutoTokenizer.from_pretrained(
-    BASE_MODEL, token=AUTH_TOKEN, device_map="auto"
-)
-model = AutoModelForCausalLM.from_pretrained(BASE_MODEL, token=AUTH_TOKEN)
-model = model.to("cuda")
-model = model.eval()
+model_loaded = False
+
+try:
+    tokenizer = AutoTokenizer.from_pretrained(
+        BASE_MODEL, token=AUTH_TOKEN, device_map="auto"
+    )
+    model = AutoModelForCausalLM.from_pretrained(BASE_MODEL, token=AUTH_TOKEN)
+    model = model.to("cuda")
+    model = model.eval()
+
+    peft_config = PeftConfig.from_pretrained(LORA_ADAPTER, token=AUTH_TOKEN)
+    model = PeftModel.from_pretrained(model, LORA_ADAPTER, token=AUTH_TOKEN)
 
-peft_config = PeftConfig.from_pretrained(LORA_ADAPTER, token=AUTH_TOKEN)
-model = PeftModel.from_pretrained(model, LORA_ADAPTER, token=AUTH_TOKEN)
+    model_loaded = True
+except Exception:
+    print("No HF_TOKEN found. Ensure you follow setup instructions!")
+    # continue on so setup instructions load
 
 
 @spaces.GPU(duration=360)
@@ -242,11 +250,12 @@ with gr.Blocks() as demo:
             graph_out = gr.Image(label="Network Graph Representation", type="pil")
 
     # If files are uploaded, generate prompt and run model
-    files.upload(
-        fn=gen_output_from_files_uploaded,
-        inputs=files,
-        outputs=[json_out, graph_out, sql_out],
-    )
+    if model_loaded:
+        files.upload(
+            fn=gen_output_from_files_uploaded,
+            inputs=files,
+            outputs=[json_out, graph_out, sql_out],
+        )
 
     gr.Markdown("Run out of FreeGPU or having issues? Try the example output!")
     demo_btn = gr.Button("Manually Load Example Output from Previous Run")