Skip to content

Commit 3d7253e

Browse files
committed
Deleted unnecessary files
1 parent f536ac6 commit 3d7253e

13 files changed

+320
-963
lines changed

examples/apps/flux-quantization-fp32.py

Lines changed: 0 additions & 168 deletions
This file was deleted.

examples/apps/flux-quantization.py

Lines changed: 0 additions & 203 deletions
This file was deleted.

examples/apps/register_sdpa.py

Lines changed: 5 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -19,9 +19,11 @@
1919

2020
# Remove decompositions for aten.scaled_dot_product_attention, aten._scaled_dot_product_efficient_attention, aten._scaled_dot_product_flash_attention
2121
# This is because we want to have SDPA as a standalone operator in the graph and invoke the custom converter for it.
22-
# TORCH_TRT_DECOMPOSITIONS.pop(torch.ops.aten.scaled_dot_product_attention.default)
23-
# TORCH_TRT_DECOMPOSITIONS.pop(torch.ops.aten._scaled_dot_product_efficient_attention.default)
24-
# TORCH_TRT_DECOMPOSITIONS.pop(torch.ops.aten._scaled_dot_product_flash_attention.default)
22+
TORCH_TRT_DECOMPOSITIONS.pop(torch.ops.aten.scaled_dot_product_attention.default)
23+
TORCH_TRT_DECOMPOSITIONS.pop(
24+
torch.ops.aten._scaled_dot_product_efficient_attention.default
25+
)
26+
TORCH_TRT_DECOMPOSITIONS.pop(torch.ops.aten._scaled_dot_product_flash_attention.default)
2527

2628
REPLACEABLE_ATEN_OPS = {
2729
torch.ops.aten._scaled_dot_product_efficient_attention.default,

0 commit comments

Comments
 (0)