debug

modeling_myolmoe.py

@@ -320,6 +320,7 @@ class MyOLMoESparseMoeBlock(nn.Module):
     def forward(self, hidden_states: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor, torch.Tensor]:
         print(f"DEBUG: MoE forward start - hidden_states shape: {hidden_states.shape}")
         batch_size, seq_len, _ = hidden_states.shape
+        print("absolute precision")
         hidden_states = hidden_states.view(-1, self.hidden_size)
         
         # Get routing weights and selected experts