Commits

Breadcrumbs

History for

onmrry-patch-1

User selector

All users

All time

Commit History

Commits on Feb 14, 2017

• Fix code that ignores tensorflow::Status.

  Show description for bc225bf
  

  hawkinsp

  authored and

  tensorflower-gardener

  committed
  bc225bf

  Copy full SHA for bc225bf

Commits on Feb 10, 2017

• Record persistent memory in tracking allocator and cost model.

  Show description for f128935
  

  Yuefeng Zhou

  authored and

  tensorflower-gardener

  committed
  f128935

  Copy full SHA for f128935

• Fixing up GraphTransferer to run shape inference to compute output shapes for each op in the graph.

  Show description for 6177838
  

  rohan100jain

  authored and

  tensorflower-gardener

  committed
  6177838

  Copy full SHA for 6177838

Commits on Feb 9, 2017

• Record kernel's host memory and device memory in cost model.

  Show description for 13a5be6
  

  Yuefeng Zhou

  authored and

  tensorflower-gardener

  committed
  13a5be6

  Copy full SHA for 13a5be6

Commits on Feb 8, 2017

• tfdbg core: add core metadata to debugger data stream + better support of concurrent debugged runs

  Show description for aabc797
  

  caisq

  authored and

  tensorflower-gardener

  committed
  aabc797

  Copy full SHA for aabc797

• Add missing "inline" to PendingCounts::Layout::CreateHandle.

  Show description for b5fa87d
  

  Justin Lebar

  authored and

  tensorflower-gardener

  committed
  b5fa87d

  Copy full SHA for b5fa87d

Commits on Feb 6, 2017

• Streamlined representation of PendingCounts data structure.

  Show description for de94a13
  

  jeffreyadean

  authored and

  tensorflower-gardener

  committed
  de94a13

  Copy full SHA for de94a13

Commits on Feb 4, 2017

• This implemented Jeff's idea to avoid most of the refcount related atomic operations when propagating tensors to the input slots of its consumers.

  Show description for 8cfb9ac
  

  yuanbyu

  authored and

  tensorflower-gardener

  committed
  8cfb9ac

  Copy full SHA for 8cfb9ac

Commits on Feb 2, 2017

• Improve memory system behavior of core graph execution module.

  Show description for 9f827c6
  

  jeffreyadean

  authored and

  tensorflower-gardener

  committed
  9f827c6

  Copy full SHA for 9f827c6

• Add GraphView internal helper class, which is initialized from the Graph

  Show description for 6ced722
  

  jeffreyadean

  authored and

  tensorflower-gardener

  committed
  6ced722

  Copy full SHA for 6ced722

• Update generated C++ API. This is a roll-forward past CLs that were

  Show description for 4957f8a
  tensorflower-gardener

  committed
  4957f8a

  Copy full SHA for 4957f8a

Commits on Feb 1, 2017

• Fix the code that looks up the hardware node stats. If an op is implemented by

  Show description for 760e7a0
  

  Yao Zhang

  authored and

  tensorflower-gardener

  committed
  760e7a0

  Copy full SHA for 760e7a0

• Add a mechanism for hiding, skipping, and modifying the generated op

  Show description for 8fe3202
  tensorflower-gardener

  committed
  8fe3202

  Copy full SHA for 8fe3202

• Don't try to constant-fold operators that do not have a CPU kernel.

  Show description for 7dfbb13
  

  hawkinsp

  authored and

  tensorflower-gardener

  committed
  7dfbb13

  Copy full SHA for 7dfbb13

Commits on Jan 31, 2017

• 1. Always run the cancellation callbacks in the event of any execution error. Without the fix, async ops can deadlock when error occurs. This may fix some of the deadlock problems we have seen with…

  Show description for 26460c8
  

  yuanbyu

  authored and

  tensorflower-gardener

  committed
  26460c8

  Copy full SHA for 26460c8

• Fix the logic for skipping warmup steps in cost model.

  Show description for 1adf9df
  

  Yao Zhang

  authored and

  tensorflower-gardener

  committed
  1adf9df

  Copy full SHA for 1adf9df

Commits on Jan 20, 2017

• Fix the bug that when cpu_allocated is called before session initialization memory cannot be tracked.

  Show description for 448c0dc
  

  Yuefeng Zhou

  authored and

  tensorflower-gardener

  committed
  448c0dc

  Copy full SHA for 448c0dc

Commits on Jan 19, 2017

• Create a fast path for direct session executor lookup to avoid sorting strings.

  Show description for 29e8d98
  

  Suharsh Sivakumar

  authored and

  tensorflower-gardener

  committed
  29e8d98

  Copy full SHA for 29e8d98

• Expose the two constants that control GPU Event polling pacing as

  Show description for 3f45361
  tensorflower-gardener

  committed
  3f45361

  Copy full SHA for 3f45361

• Deal with parallel concat as a registered graph optimization pass.

  Show description for 2b589dc
  tensorflower-gardener

  committed
  2b589dc

  Copy full SHA for 2b589dc

Commits on Jan 18, 2017

• Migrate to new namespace for Input, Output, etc. in the C++ API.

  Show description for 4d69bcf
  tensorflower-gardener

  committed
  4d69bcf

  Copy full SHA for 4d69bcf

• Automated rollback of change 144776821

  Show description for a430a5e
  tensorflower-gardener

  committed
  a430a5e

  Copy full SHA for a430a5e

• Migrate to new namespace for Input, Output, etc. in the C++ API.

  Show description for 59f813b
  tensorflower-gardener

  committed
  59f813b

  Copy full SHA for 59f813b

Commits on Jan 17, 2017

• Fix error message in direct_session.

  Show description for dbf4e01
  

  Suharsh Sivakumar

  authored and

  tensorflower-gardener

  committed
  dbf4e01

  Copy full SHA for dbf4e01

• Simplifications and cleanups in the implementation of ParallelConcat

  Show description for f49cca9
  tensorflower-gardener

  committed
  f49cca9

  Copy full SHA for f49cca9

• Merge changes from github.

  Show description for db24941
  

  Patrick Nguyen

  authored and

  tensorflower-gardener

  committed
  db24941

  Copy full SHA for db24941

Commits on Jan 14, 2017

• Add support for passes that run post-partitioning to OptimizationRegistry.

  Show description for 8eb161e
  

  hawkinsp

  authored and

  tensorflower-gardener

  committed
  8eb161e

  Copy full SHA for 8eb161e

Commits on Jan 13, 2017

• Makes the ops used by ParallelConcat internal.

  Show description for b24a0b2
  tensorflower-gardener

  committed
  b24a0b2

  Copy full SHA for b24a0b2

• Do parallel_stack as a graph rewrite instead of python code.

  Show description for 705cc93
  tensorflower-gardener

  committed
  705cc93

  Copy full SHA for 705cc93

• [TF:XLA] Add support for specifying a device instance for randomized testing.

  Show description for 9f97bba
  

  hawkinsp

  authored and

  tensorflower-gardener

  committed
  9f97bba

  Copy full SHA for 9f97bba

Commits on Jan 12, 2017

• Automated rollback of change 144344623

  Show description for a2f9e99
  

  David G. Andersen

  authored and

  tensorflower-gardener

  committed
  a2f9e99

  Copy full SHA for a2f9e99

• Fail earlier and more clearly if dtype in parsed proto is invalid.

  Show description for 1f409de
  

  David G. Andersen

  authored and

  tensorflower-gardener

  committed
  1f409de

  Copy full SHA for 1f409de

• Switch open-source to use jemalloc for CPU Tensor memory allocation, gRPC, and other places where we call malloc/free.

  Show description for 83c6e0c
  

  Jonathan Hseu

  authored and

  tensorflower-gardener

  committed
  83c6e0c

  Copy full SHA for 83c6e0c

Commits on Jan 9, 2017

• SimplePlacer: apply heuristics only if the candidate device is valid

  Show description for 3f5f958
  

  Vijay Vasudevan

  authored and

  tensorflower-gardener

  committed
  3f5f958

  Copy full SHA for 3f5f958

Commits on Jan 6, 2017

• Switch GetCUDAHostAllocator() return the standard CPU allocator when

  Show description for 699d1e8
  

  Vijay Vasudevan

  authored and

  tensorflower-gardener

  committed
  699d1e8

  Copy full SHA for 699d1e8