strisunshinewentingwang/pytorch

Pull Requests Commits

Update to sdp benchmark to take into account pt2.0 stack (#90096)

drisspg committed 1 year ago

58b5a9df

[torchgen] Move Executorch codegen logic into torchgen (#90806)

larryliu0820 committed 1 year ago

909b7ca9

[torchgen] Move Executorch custom ops logic into torchgen (#90099)

larryliu0820 committed 1 year ago

679da8bd

[torchgen] Move Executorch unboxing logic into torchgen (#90098)

larryliu0820 committed 1 year ago

ca52f63f

jacrev : Support chunked computation (#89376)

kshitij12345 committed 1 year ago

f02e93b5

[Vulkan + Profiler] Add Timestamp Adjustment Algorithm (#90672)

salilsdesai committed 1 year ago

e2dc60c6

[Vulkan + Profiler] Use 0 as Vulkan Event Durations During Tree Building (#90671)

salilsdesai committed 1 year ago

0428de06

[Vulkan + Profiler] Report Vulkan Events to Profiler in QueryPool (#90670)

salilsdesai committed 1 year ago

8c80a468

[Vulkan + Profiler] Enable Processing Vulkan Events in Profiler (#90852)

salilsdesai committed 1 year ago

193068cb

[Vulkan] Store entries in a separate queue after resetting query pool (#90668)

salilsdesai committed 1 year ago

7badd0b9

[FSDP] Fix `_mp_shard` `record_stream()` (#91096)

awgu committed 1 year ago

c3457550

[inductor] Add retry after benchmark test fails on CI (#90808)

desertfire committed 1 year ago

2a37ba8e

move inputs to device on root module only (#91078)

zhaojuanmao committed 1 year ago

50ab2b70

functionalization: check for undefined tensors in advanced indexing (#90791)

bdhirsh committed 1 year ago

d6efd25d

fix set_() with functionalization (#90722)

bdhirsh committed 1 year ago

440a3f23

Replace TORCHINDUCTOR_TRACE with TORCH_COMPILE_DEBUG in documentation (#91011)

desertfire committed 1 year ago

548960f6

Allow FSDP to have ignored modules out of wrapped root (#91079)

mrshenli committed 1 year ago

e5a48da6

[Quant] Add fused LinearTanh module for onednn backend (#88923)

Xia-Weiwen committed 1 year ago

6686e9bc

Use scalar implementation to keep the precision in linspace of integral types (#89048)

yanbing-j committed 1 year ago

731f417f

Fix torch.distributed.run init connect timeout by comparing `host` with the current IP list (#90221)

cz-37 committed 1 year ago

f833880b

Dynamo comptime, with public ComptimeContext API (#90983)

ezyang committed 1 year ago

dfe916ca

inductor: separate onednn fx fusion from overriders.py (#90890)

XiaobingSuper committed 1 year ago

ec748cbe

add mixed data type support for GroupNorm (#81852)

mingfeima committed 1 year ago

4bf22fcf

[Quant] Add fused linear-tanh op for onednn backend (#88879)

Xia-Weiwen committed 1 year ago

ea49e769

Type torch._inductor.graph (#90987)

ezyang committed 1 year ago

17d860d0

Apply modernize-use-emplace to aten, c10, torch (#91077)

Skylion007 committed 1 year ago

3916d7a5

Switch use_fake_tensor to True by default (#89663)

ezyang committed 1 year ago

944519a4

[cuDNN][cuDNN V8 API] Fix `benchmark_limit` ignoring failed kernels in FIND (#91032)

eqy committed 1 year ago

ce4900f3

Vectorize exmp1 and log1p (#91074)

EikanWang committed 1 year ago

856651dd

[Dynamo] Support torch.get_default_dtype (#89790)

yanboliang committed 1 year ago

490c1cf6

Newer Older