Rewrite tfl.batch_matmul with batch_1 constants of ones into reduce_sum, to...
Rewrite tfl.batch_matmul with batch_1 constants of ones into reduce_sum, to avoid materializing the constant. PiperOrigin-RevId: 613852834
Showing
- tensorflow/compiler/mlir/lite/tests/optimize_batch_matmul.mlir 20 additions, 0 deletions...rflow/compiler/mlir/lite/tests/optimize_batch_matmul.mlir
- tensorflow/compiler/mlir/lite/transforms/optimize_batch_matmul.cc 71 additions, 1 deletion...ow/compiler/mlir/lite/transforms/optimize_batch_matmul.cc
Please register or sign in to comment