Merge branch 'ko3n1g/fix/gelu' into 'main'

fix: GeLUFunction

See merge request ADLR/megatron-lm!1993

megatron/core/fusions/fused_bias_gelu.py

@@ -49,7 +49,7 @@ def backward(ctx, grad_output):
     # This is required to make Sphinx happy :-(
     @classmethod
     def apply(cls, *args, **kwargs):
-        super().apply(*args, **kwargs)
+        return super().apply(*args, **kwargs)
 
 
 bias_gelu_impl = GeLUFunction.apply