torch-mlir/include
Jiawei Wu b75d0e3f8b
[stablehlo] fix: enhance torch's index-like op lowering to stablehlo's gather/scatter (#3829)
In torch.index_put like ops, `values` is only required to be
broadcastable to `input[indices]`, rather than exact dimension match.
This patch fixes the problem by add additional
stablehlo.dynamic_broadcast_in_dim before creating stablehlo.scatter op.
BTW, this patch also enhance the `getBroadcastResultShape` utility in
hlo namespace.
2024-11-05 19:15:11 +08:00
..
torch-mlir [stablehlo] fix: enhance torch's index-like op lowering to stablehlo's gather/scatter (#3829) 2024-11-05 19:15:11 +08:00
torch-mlir-c [ONNX] add int16 quantization support (#3446) 2024-06-12 10:37:22 +05:30
torch-mlir-dialects [Torch] [TMTensor] Added mask and is_causal support for torch.aten.scaled_dot_product_attention (#3690) 2024-09-09 15:51:41 -07:00
CMakeLists.txt [NFC reformat] Run pre-commit on all files and format misc. 2024-04-27 14:08:09 -07:00