drisspg 0f9eea1329 [FlexAttention] Fix multiple calls to flex bug (#140761) ... # Summary Fixes long-standing bug we've had in the backward pass for flex attention. See https://github.com/pytorch/pytorch/issues/135161 for details Pull Request resolved: https://github.com/pytorch/pytorch/pull/140761 Approved by: https://github.com/Chillee, https://github.com/zou3519		2024-11-16 04:57:04 +00:00
..
__init__.py	Add base class for single-subgraph inductor HOPs (#139898)	2024-11-11 16:12:35 +00:00
associative_scan.py	Improvements for associative_scan - slicing of xs (#138858)	2024-11-05 23:38:21 +00:00
auto_functionalize.py	Generate slice.Tensor view operations instead of as_strided when split is used in the original program. (#137225)	2024-10-23 17:42:16 +00:00
cond.py	[hop free symbols] lift free symbols in example_value when create_graph_input (#138363)	2024-11-07 04:44:32 +00:00
effects.py
executorch_call_delegate.py
flex_attention.py	[FlexAttention] Fix multiple calls to flex bug (#140761)	2024-11-16 04:57:04 +00:00
hints_wrap.py
invoke_subgraph.py	[invoke_subgraph] Support symint/int as inputs (#140058)	2024-11-11 22:26:43 +00:00
map.py	[hop free symbols][refactor] make map's save_for_backward to handle int (#138558)	2024-11-04 22:48:07 +00:00
out_dtype.py
prim_hop_base.py	Add base class for single-subgraph inductor HOPs (#139898)	2024-11-11 16:12:35 +00:00
run_const_graph.py
scan.py	[hop free symbols] lift free symbols in example_value when create_graph_input (#138363)	2024-11-07 04:44:32 +00:00
strict_mode.py
torchbind.py
triton_kernel_wrap.py	[RFC] Implement caching for user defined triton kernels (#140326)	2024-11-16 02:37:16 +00:00
utils.py	[hop free symbols] lift free symbols in example_value when create_graph_input (#138363)	2024-11-07 04:44:32 +00:00
while_loop.py	[hop free symbols] lift free symbols in example_value when create_graph_input (#138363)	2024-11-07 04:44:32 +00:00
wrap.py