yeah we vibecoding

app.py

@@ -54,17 +54,17 @@ def load_model(repo_id, filename, config_override=None):
 
 # Load both models once at the start
 try:
-    # This is the baseline model from your portfolio
+    # FIX #1: Add the correct subdirectory for the baseline model
     BASELINE_MODEL, BASELINE_CFG = load_model(
         repo_id="SQCU/pgptlformer-tinystories", 
-        filename="state_step040500.pt"
+        filename="re-pqt-rmsXrmsx2-70b91221-a39c-4824-a69c-48a034963529/state_step040500.pt"
     )
 
-    # This is the shift-attn model. Note the config_override.
+    # FIX #2: The shift-attn model already had the directory, but ensure it's correct
     SHIFT_ATTN_MODEL, SHIFT_ATTN_CFG = load_model(
         repo_id="SQCU/pgptlformer-tinystories",
         filename="re-pqt-rmsXrmsx2x2-ATTNII-791967c5-5c59-4a5f-a2c5-07772bcf65ab/state_step040500.pt",
-        config_override={"attention_deux": True} # Crucial: This enables the shift-attn mechanism in your code
+        config_override={"attention_deux": True} 
     )
 except Exception as e:
     # If loading fails, show an error in the Gradio app instead of crashing