Commits

Branch selector

 master

User selector

er91

Datepicker

All time

Commit History

Commits on Oct 11, 2019

• Fixes for outside compilation.

  Show description for 308a0d4
  

  er91

  authored and

  tensorflower-gardener

  committed
  308a0d4

  Copy full SHA for 308a0d4

• Make sure outside_compilation() handles TF 2.0 functions.

  Show description for 91bd735
  

  er91

  authored and

  tensorflower-gardener

  committed
  91bd735

  Copy full SHA for 91bd735

Commits on Oct 7, 2019

• Only perform outside compilation related logic if there are outside compilation nodes.

  Show description for bb3ac73
  

  er91

  authored and

  tensorflower-gardener

  committed
  bb3ac73

  Copy full SHA for bb3ac73

Commits on Oct 3, 2019

• Use ShapeUtil::Compatible() instead of ShapeUtil::Equal().

  Show description for bbb01d8
  

  er91

  authored and

  tensorflower-gardener

  committed
  bbb01d8

  Copy full SHA for bbb01d8

Commits on Sep 20, 2019

• Remove lookup_library from FunctionalizeControlFlow().

  Show description for a2a51fa
  

  er91

  authored and

  tensorflower-gardener

  committed
  a2a51fa

  Copy full SHA for a2a51fa

Commits on Sep 10, 2019

• In If/Case op, if branch function output is TensorList, call SetTensorListOutput() instead of SetOutput().

  Show description for 794a207
  

  er91

  authored and

  tensorflower-gardener

  committed
  794a207

  Copy full SHA for 794a207

Commits on Sep 7, 2019

• In XlaCompiler, when creating the TUPLE root instruction, set its sharding according to its inputs' shardings.

  Show description for 25e5c91
  

  er91

  authored and

  tensorflower-gardener

  committed
  25e5c91

  Copy full SHA for 25e5c91

Commits on Aug 28, 2019

• Remove tensor input shape from function signature.

  Show description for bb4cacf
  

  er91

  authored and

  tensorflower-gardener

  committed
  bb4cacf

  Copy full SHA for bb4cacf

• Make sure we return proper error message when there is Placeholder node in TPU computation (instead of crashing).

  Show description for c210716
  

  er91

  authored and

  tensorflower-gardener

  committed
  c210716

  Copy full SHA for c210716

• When calculating gradient, we should gather nodes to propagate from y_node_outputs in a backward search.

  Show description for 391eb2d
  

  er91

  authored and

  tensorflower-gardener

  committed
  391eb2d

  Copy full SHA for 391eb2d

Commits on Aug 8, 2019

• Rollback an accidental change that sneaked into commit 8cc20b.

  Show description for 1861639
  

  er91

  authored and

  tensorflower-gardener

  committed
  1861639

  Copy full SHA for 1861639

Commits on Aug 6, 2019

• Reduce stack usage of ExtractOutsideCompilationForFunction().

  Show description for cdfd050
  

  er91

  authored and

  tensorflower-gardener

  committed
  cdfd050

  Copy full SHA for cdfd050

Commits on Aug 2, 2019

• Do not assume in_edges() are sorted by dst_input.

  Show description for 4a9e18e
  

  er91

  authored and

  tensorflower-gardener

  committed
  4a9e18e

  Copy full SHA for 4a9e18e

• Check for stateless If/While nodes as well.

  Show description for 10e4995
  

  er91

  authored and

  tensorflower-gardener

  committed
  10e4995

  Copy full SHA for 10e4995

Commits on Aug 1, 2019

• For outside compilation nodes that are in outside compilation blocks, lower them as V1 function.

  Show description for 8cc20b9
  

  er91

  authored and

  tensorflower-gardener

  committed
  8cc20b9

  Copy full SHA for 8cc20b9

Commits on Jul 29, 2019

• Fix control inputs for while loop's Enter nodes.

  Show description for 9e12642
  

  er91

  authored and

  tensorflower-gardener

  committed
  9e12642

  Copy full SHA for 9e12642

Commits on Jul 26, 2019

• Automated rollback of commit 7042a685613f1d9b307a5fd259de2966c54e09b1

  Show description for a65a4de
  

  er91

  authored and

  tensorflower-gardener

  committed
  a65a4de

  Copy full SHA for a65a4de

• Fix an error in constant_folding.cc.

  Show description for 9ebead9
  

  er91

  authored and

  tensorflower-gardener

  committed
  9ebead9

  Copy full SHA for 9ebead9

Commits on Jul 25, 2019

• Fixes for XlaCompiler shape inference.

  Show description for f085af8
  

  er91

  authored and

  tensorflower-gardener

  committed
  f085af8

  Copy full SHA for f085af8

Commits on Jul 23, 2019

• Correctly handle lifted _Arg nodes in then/else branch: they might be 2 different sets of _Arg nodes.

  Show description for 9dcffc2
  

  er91

  authored and

  tensorflower-gardener

  committed
  9dcffc2

  Copy full SHA for 9dcffc2

• Lift outside compilation only arguments from function call nodes.

  Show description for 90502d1
  

  er91

  authored and

  tensorflower-gardener

  committed
  90502d1

  Copy full SHA for 90502d1

Commits on Jul 19, 2019

• Lower outside compilation functional nodes (If/While/function call).

  Show description for 2b8485c
  

  er91

  authored and

  tensorflower-gardener

  committed
  2b8485c

  Copy full SHA for 2b8485c

Commits on Jul 18, 2019

• Do not remove XlaHostCompute nodes with no input/output edges.

  Show description for a4d0acf
  

  er91

  authored and

  tensorflower-gardener

  committed
  a4d0acf

  Copy full SHA for a4d0acf

Commits on Jul 16, 2019

• Pass attributes when lowering functional If/While.

  Show description for c51eb49
  

  er91

  authored and

  tensorflower-gardener

  committed
  c51eb49

  Copy full SHA for c51eb49

Commits on Jul 15, 2019

• Handle shape inference for "outside compilation to outside compilation data edge" for If/While nodes as well.

  Show description for efafe1b
  

  er91

  authored and

  tensorflower-gardener

  committed
  efafe1b

  Copy full SHA for efafe1b

Commits on Jul 12, 2019

• Return error in IdentityN shape inference function if any input shape is unknown.

  Show description for d2e42b0
  

  er91

  authored and

  tensorflower-gardener

  committed
  d2e42b0

  Copy full SHA for d2e42b0

• Automated rollback of commit 823813768199861be752a06947fcfda340350e25

  Show description for be936b3
  

  er91

  authored and

  tensorflower-gardener

  committed
  be936b3

  Copy full SHA for be936b3

• If an _Arg node of TPU computation is used in outside compilation, use corresponding input in outside compilation host graph directly.

  Show description for 351383f
  

  er91

  authored and

  tensorflower-gardener

  committed
  351383f

  Copy full SHA for 351383f

Commits on Jun 28, 2019

• Support summary in while loop.

  Show description for 7d77191
  

  er91

  authored and

  tensorflower-gardener

  committed
  7d77191

  Copy full SHA for 7d77191

Commits on Jun 24, 2019

• Get representation shape from shape_representation_fn.

  Show description for 99ecd61
  

  er91

  authored and

  tensorflower-gardener

  committed
  99ecd61

  Copy full SHA for 99ecd61

Commits on Jun 21, 2019

• If function is in an implementation selection group, do not inline it in LowerFunctionalOpsPass.

  Show description for 5952a16
  

  er91

  authored and

  tensorflower-gardener

  committed
  5952a16

  Copy full SHA for 5952a16

Commits on Jun 20, 2019

• Add dtypes_and_shapes to ResourceHandle.

  Show description for 512a3c3
  

  er91

  authored and

  tensorflower-gardener

  committed
  512a3c3

  Copy full SHA for 512a3c3

Commits on Jun 18, 2019

• If we are in FuncGraph (e.g. While V2 body func), append function name to TPU cluster name.

  Show description for 962bfb1
  

  er91

  authored and

  tensorflower-gardener

  committed
  962bfb1

  Copy full SHA for 962bfb1

Commits on Jun 12, 2019

• Add uncompilable node info to error message if function is uncompilable.

  Show description for d7a0333
  

  er91

  authored and

  tensorflower-gardener

  committed
  d7a0333

  Copy full SHA for d7a0333

Commits on Jun 10, 2019

• Handle shape inference for loop invariant Merge node.

  Show description for ed01344
  

  er91

  authored and

  tensorflower-gardener

  committed
  ed01344

  Copy full SHA for ed01344

Pagination

Previous

Next