Commits

Branch selector

 master

User selector

mrshenli

Datepicker

All time

Commit History

Commits on Jun 1, 2019

• Fix Python DataParallel RNN in no_grad mode (pytorch#21197)

  Show description for 51ebbe9
  

  mrshenli

  authored and

  facebook-github-bot

  committed
  51ebbe9

  Copy full SHA for 51ebbe9

Commits on May 26, 2019

• Update nccl submodule to v2.4.6 (pytorch#20882)

  Show description for 4c5b1e3
  

  mrshenli

  authored and

  facebook-github-bot

  committed
  4c5b1e3

  Copy full SHA for 4c5b1e3

Commits on May 20, 2019

• Make CUDACachingAllocator::recordStream() a no-op on null ptrs (pytorch#20658)

  Show description for 8acaa28
  

  mrshenli

  authored and

  facebook-github-bot

  committed
  8acaa28

  Copy full SHA for 8acaa28

Commits on May 17, 2019

• Emphasize all DDP forward() outputs must participate in computing loss (pytorch#20586)

  Show description for fa4ca4e
  

  mrshenli

  authored and

  facebook-github-bot

  committed
  fa4ca4e

  Copy full SHA for fa4ca4e

Commits on May 9, 2019

• Only record grad_fn in C++ Scatter and Gather when required so (pytorch#20286)

  Show description for d6815e1
  

  mrshenli

  authored and

  facebook-github-bot

  committed
  d6815e1

  Copy full SHA for d6815e1

Commits on Apr 29, 2019

• Make find_unused_parameters in DDP default to False (pytorch#19895)

  Show description for b695e56
  

  mrshenli

  authored and

  facebook-github-bot

  committed
  b695e56

  Copy full SHA for b695e56

Commits on Apr 22, 2019

• Fix some typos in jit README

  Show description for 1bd5f2c
  

  mrshenli

  authored and

  facebook-github-bot

  committed
  1bd5f2c

  Copy full SHA for 1bd5f2c

Commits on Apr 18, 2019

• Allow DDP to wrap multi-GPU modules (pytorch#19271)

  Show description for 6732358
  

  mrshenli

  authored and

  facebook-github-bot

  committed
  6732358

  Copy full SHA for 6732358

Commits on Apr 16, 2019

• Revert replicate.py to disallow replicating multi-device modules (pytorch#19278)

  Show description for 344acaa
  

  mrshenli

  authored and

  facebook-github-bot

  committed
  344acaa

  Copy full SHA for 344acaa

Commits on Apr 11, 2019

• Fix flaky store timeout test (pytorch#19114)

  Show description for 6b0ca8e
  

  mrshenli

  authored and

  facebook-github-bot

  committed
  6b0ca8e

  Copy full SHA for 6b0ca8e

Commits on Apr 9, 2019

• Remind users to set map_location properly when using DDP

  Show description for 168c079
  

  mrshenli

  authored and

  facebook-github-bot

  committed
  168c079

  Copy full SHA for 168c079

• Avoid calling tensor.data.set_() in DDP

  Show description for 5eb6a2b
  

  mrshenli

  authored and

  facebook-github-bot

  committed
  5eb6a2b

  Copy full SHA for 5eb6a2b

• Propagate ProcessGroup timeout to Store (pytorch#16571)

  Show description for 8f9b11c
  

  mrshenli

  authored and

  facebook-github-bot

  committed
  8f9b11c

  Copy full SHA for 8f9b11c

• Fix BN tests for >= 8 GPU test environments (pytorch#19049)

  Show description for 544783f
  

  mrshenli

  authored and

  facebook-github-bot

  committed
  544783f

  Copy full SHA for 544783f

Commits on Apr 3, 2019

• Support replicating multi-GPU modules (pytorch#18687)

  Show description for 7ae0263
  

  mrshenli

  authored and

  facebook-github-bot

  committed
  7ae0263

  Copy full SHA for 7ae0263

• Highlight NCCL all_reduce and all_gather requirements (pytorch#18741)

  Show description for c0ad674
  

  mrshenli

  authored and

  facebook-github-bot

  committed
  c0ad674

  Copy full SHA for c0ad674

Commits on Mar 28, 2019

• Fix NCCL/Gloo process groups and DDP stream sync bug (pytorch#18465)

  Show description for aea8ee1
  

  mrshenli

  authored and

  facebook-github-bot

  committed
  aea8ee1

  Copy full SHA for aea8ee1

Commits on Feb 27, 2019

• Always synchronize src and dst streams when copying tensors (pytorch#16966)

  Show description for 1154506
  

  mrshenli

  authored and

  facebook-github-bot

  committed
  1154506

  Copy full SHA for 1154506

Commits on Feb 25, 2019

• Restore current streams on dst device after switching streams (pytorch#17439)

  Show description for b527055
  

  mrshenli

  authored and

  facebook-github-bot

  committed
  b527055

  Copy full SHA for b527055

Commits on Feb 15, 2019

• Enforce module device at DataParallel construction time (pytorch#17129)

  Show description for 472cfc0
  

  mrshenli

  authored and

  facebook-github-bot

  committed
  472cfc0

  Copy full SHA for 472cfc0

Commits on Jan 25, 2019

• Relax lower bound for nogil timing test to avoid false alarm (pytorch#16259)

  Show description for 7ce634e
  

  mrshenli

  authored and

  facebook-github-bot

  committed
  7ce634e

  Copy full SHA for 7ce634e

Commits on Jan 22, 2019

• Add default_stream() and enhance current_stream() (pytorch#16200)

  Show description for 2235fb2
  

  mrshenli

  authored and

  facebook-github-bot

  committed
  2235fb2

  Copy full SHA for 2235fb2

• Release GIL when synchronize or wait (pytorch#16182)

  Show description for 1c058de
  

  mrshenli

  authored and

  facebook-github-bot

  committed
  1c058de

  Copy full SHA for 1c058de

Commits on Jan 20, 2019

• Unify device() return type in Stream, Event, and Tensor (pytorch#16150)

  Show description for 898329c
  

  mrshenli

  authored and

  facebook-github-bot

  committed
  898329c

  Copy full SHA for 898329c

Commits on Jan 18, 2019

• Change current device in stream context manager if necessary (pytorch#16128)

  Show description for 292edfb
  

  mrshenli

  authored and

  facebook-github-bot

  committed
  292edfb

  Copy full SHA for 292edfb

• Port legacy any(*) to ATen

  Show description for 1ff8647
  

  mrshenli

  authored and

  facebook-github-bot

  committed
  1ff8647

  Copy full SHA for 1ff8647

Commits on Jan 17, 2019

• Move all Stream and Event Python implementation to C++ (pytorch#15937)

  Show description for 24f4d39
  

  mrshenli

  authored and

  facebook-github-bot

  committed
  24f4d39

  Copy full SHA for 24f4d39

Commits on Jan 16, 2019

• Port legacy all(*) to ATen (pytorch#15540)

  Show description for a2af554
  

  mrshenli

  authored and

  facebook-github-bot

  committed
  a2af554

  Copy full SHA for a2af554

Commits on Jan 11, 2019

• comment out large test cases for tril(u)_indices (pytorch#15959)

  Show description for 03a570c
  

  mrshenli

  authored and

  facebook-github-bot

  committed
  03a570c

  Copy full SHA for 03a570c

Commits on Jan 10, 2019

• Porting legacy reflection_pad2d to ATen

  Show description for 23e28ef
  

  mrshenli

  authored and

  facebook-github-bot

  committed
  23e28ef

  Copy full SHA for 23e28ef

Commits on Jan 9, 2019

• Wrap C10 CUDAStream instead of cudaStream_t in THCPStream

  Show description for 7b9f794
  

  mrshenli

  authored and

  facebook-github-bot

  committed
  7b9f794

  Copy full SHA for 7b9f794

Commits on Jan 8, 2019

• Move Stream.query() implementation down to C++ (pytorch#15737)

  Show description for 99d2743
  

  mrshenli

  authored and

  facebook-github-bot

  committed
  99d2743

  Copy full SHA for 99d2743

Commits on Jan 3, 2019

• A quick fix for Stream operation errors on non-current device (pytorch#15689)

  Show description for 1e9a6d7
  

  mrshenli

  authored and

  facebook-github-bot

  committed
  1e9a6d7

  Copy full SHA for 1e9a6d7

• Port legacy reflection_pad1d to ATen (pytorch#15480)

  Show description for 279ca4a
  

  mrshenli

  authored and

  facebook-github-bot

  committed
  279ca4a

  Copy full SHA for 279ca4a

• Fix vec256 inversion (pytorch#15659)

  Show description for efc3d6b
  

  mrshenli

  authored and

  facebook-github-bot

  committed
  efc3d6b

  Copy full SHA for efc3d6b

Pagination

Previous

Next