strisunshinewentingwang/pytorch

Pull Requests Commits

fix some comments in accelerator_partitioner.py (#49104)

scottxu0730 committed 3 years ago

33bc7918

Decouple direct access to native::scalar_tensor from TensorIndexing.h (#48761)

iseeyuan committed 3 years ago

c7b8f3e2

Revert D25433268: [PyTorch Mobile] Preserve bundled input related methods when calling optimize_for_mobile

lw committed 3 years ago

2255e68d

Cache the DataPtrs in CUDAFuture (#48788)

lw committed 3 years ago

b5a7e250

Split out reusable CUDAFuture from FutureNCCL (#48506)

lw committed 3 years ago

030fa6cf

Merge common parts of FutureNCCL into at::ivalue::Future (#48505)

lw committed 3 years ago

4c425e8d

Split FutureNCCL's CUDA-specific parts from generic future logic (#48504)

lw committed 3 years ago

9078088e

Support wider range of types in FutureNCCL (#48502)

lw committed 3 years ago

a6778989

Don't store device indices separately on FutureNCCL (#48501)

lw committed 3 years ago

9fe3ac36

Add multi-GPU support to FutureNCCL (#48500)

lw committed 3 years ago

e294c2d8

Fix FutureNCCL not recording dataptrs with caching alloc in wait() (#48563)

lw committed 3 years ago

91ad3ed8

Fix FutureNCCL's completed() disagreeing with wait() (#48503)

lw committed 3 years ago

003c30ba

Record CUDA events for "follow-up" FutureNCCL inside markCompleted (#48499)

lw committed 3 years ago

b91b0872

Use fresh stream from pool for each FutureNCCL callback (#48498)

lw committed 3 years ago

6157f8ae

Make FutureNCCL record events in current stream (#48497)

lw committed 3 years ago

8fb52e7f

Have FutureNCCL record streams w/ allocator in addCallback (#48496)

lw committed 3 years ago

e4267eb4

Add some safeguards to FutureNCCL (#48562)

lw committed 3 years ago

868a1a48

Remove NCCL dependency from PythonFutureWrapper (#48495)

lw committed 3 years ago

b7f5aa98

Avoid using FutureNCCL before it's ready (#48561)

lw committed 3 years ago

7f7f0fa3

[numpy] `torch.exp{2, m1}`: promote integer inputs to float (#48926)

kshitij12345 committed 3 years ago

eb9516ea

Port `eig` CPU from TH to ATen (#43215)

kurtamohler committed 3 years ago

27f7d1c2

[PyTorch Mobile] Preserve bundled input related methods when calling optimize_for_mobile

bearzx committed 3 years ago

95233870

op to gen quant params from min-max thresholds

Xiaohan Wei committed 3 years ago

9417e927

[NNC] Dead Store Elimination (#49030)

nickgg committed 3 years ago

c5bc6b40

Revert D25416620: [pytorch][PR] Add version_info tuple

ezyang committed 3 years ago

7a2abbd8

[PyTorch] Avoid storage refcount bump in copy_tensor_metadata (#48877)

swolchok committed 3 years ago

3123f878

Add version_info tuple (#48414)

jonashaag committed 3 years ago

e69c2f85

Add type annotations to conv-relu (#47680)

guilhermeleobas committed 3 years ago

5375a479

[PyTorch Mobile] Add continuous build config for xplat/caffe2

dhruvbird committed 3 years ago

e9ef1fe3

Class-based structured kernels, with migration of add to framework (#48718)

ezyang committed 3 years ago

16b8e6ab

Newer Older