mirror of https://github.com/llvm/torch-mlir
bb259f918a
-- This commit adds decompose logic for `aten._softmax` when `half_to_float` is `True`. -- An e2e test case will be added once support for half to float conversion for `aten._softmax` is added upstream. Signed-off-by: Abhishek Varma <abhishek@nod-labs.com> |
||
---|---|---|
.. | ||
Torch | ||
TorchConversion | ||
CMakeLists.txt |