diff --git a/dsv4/layers/shared_expert.py b/dsv4/layers/shared_expert.py index 72ca19ba..af3951fa 100644 --- a/dsv4/layers/shared_expert.py +++ b/dsv4/layers/shared_expert.py @@ -433,6 +433,7 @@ class Nvfp4SharedExpert: intermediate = self._run_l1_fused(hidden_states) else: l1_out = self._run_l1(hidden_states) + print(f" SE L1 out: shape={tuple(l1_out.shape)} |max|={l1_out.abs().max().item() if l1_out.numel() > 0 else 'EMPTY'} nan={torch.isnan(l1_out).any().item() if l1_out.numel() > 0 else 'N/A'}", flush=True) if l1_out.shape[1] < 2 * self.intermediate_size: print(f" WARNING: l1_out shape {l1_out.shape} < expected (N, {2*self.intermediate_size})", flush=True)