SemanticDiff pytorch
17659a7f - FX Partitioner: use "all aten ops that decompose to prims" as supported NVFuser ops

Loading