Upload app.py with huggingface_hub

app.py

@@ -1,6 +1,6 @@
 #!/usr/bin/env python3
 """
-OpenLLM Real Models App - Ultimate fixed version with correct attention bias handling
+OpenLLM Real Models App - Final working version with correct attribute naming
 """
 
 import gradio as gr
@@ -133,7 +133,7 @@ class Block(nn.Module):
         return x
 
 class CausalSelfAttention(nn.Module):
-    """Multi-head self-attention with causal masking - ULTIMATE FIX"""
+    """Multi-head self-attention with causal masking - FINAL WORKING VERSION"""
     def __init__(self, config):
         super().__init__()
         assert config.n_embd % config.n_head == 0
@@ -144,7 +144,7 @@ class CausalSelfAttention(nn.Module):
         self.n_head = config.n_head
         self.n_embd = config.n_embd
         self.dropout = config.dropout
-        self.bias = config.bias
+        self.use_bias = config.bias  # Use different name for the boolean flag
         
         # REGISTER THE ATTENTION BIAS as a buffer (not parameter) to match saved model
         # This is actually an attention mask, not a learnable bias
@@ -152,7 +152,7 @@ class CausalSelfAttention(nn.Module):
             # Create a causal attention mask buffer
             mask = torch.tril(torch.ones(config.block_size, config.block_size))
             mask = mask.view(1, 1, config.block_size, config.block_size)
-            self.register_buffer('bias', mask)
+            self.register_buffer('bias', mask)  # This matches the saved model's 'bias' key
         else:
             self.register_buffer('bias', None)