strisunshinewentingwang/pytorch

Pull Requests Commits

[android] turn on USE_VULKAN for android builds by default

IvanKobzarev committed 3 years ago

6a153980

torch.cuda.memory_allocated to return `{}` if not initialized (#51179)

malfet committed 3 years ago

43f0ccd1

[quant][fx] Update name of packed weight attributes (#51259)

supriyar committed 3 years ago

916af892

memory efficient per-channel fq: use it everywhere, delete old version (#51265)

vkuzo committed 3 years ago

05c8cd74

fake_quant: more memory efficient per-channel backward (#51255)

vkuzo committed 3 years ago

267e2430

Back out "Revert D26077905: Back out "Revert D25850783: Add torch::deploy, an embedded torch-python interpreter"" (#51267)

wconstab committed 3 years ago

f2e41257

[Gradient Compression] Refactor default_hooks.py and powerSGD_hook.py by creating a util function that make a vanilla allreduce future (#51094)

Yi Wang committed 3 years ago

e7b34962

[Gradient Compression] Explicitly specify the dtype of the error tensor (#50985)

Yi Wang committed 3 years ago

9d731e87

[Gradient Compression] Simplify the implementation of error feedback and warm-start (#50981)

Yi Wang committed 3 years ago

b619d37b

clone pytorch.github.io with depth 1 (#48115)

Randall Hunt committed 3 years ago

00d4ec84

Remove debug-only assertion from vulkan::api::Command::Command as the buffer can legitimately be null. (#51160)

Ashkan Aliabadi committed 3 years ago

8a8fac66

Define static constexpr variable in at::native::vulkan:::api::Handle. (#51006)

Ashkan Aliabadi committed 3 years ago

592a8ad1

DataPipe naming convension update (#51262)

lixinyu committed 3 years ago

5ed0ad4b

Add serialization logic for complex numbers (#51287)

anjali411 committed 3 years ago

f9f22c8b

Port cholesky_inverse to ATen (#50269)

IvanYashchuk committed 3 years ago

6e4746c1

Update third_party/build_bundled.py (#51161)

mattip committed 3 years ago

9f6e0de5

[quant][graphmode][fx] Add support for functional conv1d and conv3d (#51155) (#51254)

jerryzh168 committed 3 years ago

7097c0d4

.github: Remove title from stale alert (#51306)

seemethere committed 3 years ago

35990b5f

Add private mechanism to toggle vmap fallback warnings (#51218)

zou3519 committed 3 years ago

1379842f

.github: Update stale messaging add newlines (#51298)

seemethere committed 3 years ago

f68e5f1d

Add missing -inf order for linalg.norm OpInfo (#51233)

kurtamohler committed 3 years ago

b0286536

add mising VSX dispatches (#51217)

quickwritereader committed 3 years ago

8b27c2cc

[Pipe] Refactor convert_to_balance under non-test package. (#50860)

pritamdamania committed 3 years ago

96cedefd

Make DeviceCachingAllocator's error handling more defensive and a bit easier to read (#51158)

mcarilli committed 3 years ago

cedfa4cc

[fx] improve args mutation error (#51175)

zdevito committed 3 years ago

33d51806

Enable TensorPipe's CUDA GDR channel (#50763)

lw committed 3 years ago

4288f08d

.github: Add workflow to stale pull requests (#51237)

seemethere committed 3 years ago

cc211bb4

Enable TensorPipe's InfiniBand transport (#50761)

lw committed 3 years ago

c9cebaf9

[quant][fx] Make scale, zero_point buffers in the model, use FQN (for quantize_per_tensor ops) (#51171)

supriyar committed 3 years ago

288b94a8

[quant][fx] Make scale, zero_point buffers in the model and use FQN (for quantized ops) (#51166)

supriyar committed 3 years ago

4c3f59b7

Older