Commits

Branch selector

 master

User selector

benoitsteiner

Datepicker

All time

Commit History

Commits on Oct 13, 2017

• Determine peak memory usage from the result of a simulation.

  Show description for 33fc95f
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  33fc95f

  Copy full SHA for 33fc95f

Commits on Oct 12, 2017

• Register a dummy estimation function for _Send ops. This ensures that we don't get negative cost estimates for _Send ops.

  Show description for 4efb8cf
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  4efb8cf

  Copy full SHA for 4efb8cf

• Updated the virtual scheduler to use legal names when inserting Send/Recv nodes in the graph.

  Show description for 4241b86
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  4241b86

  Copy full SHA for 4241b86

• Made sure that the virtual placer correctly handles short device names

  Show description for 9148cdd
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  9148cdd

  Copy full SHA for 9148cdd

Commits on Oct 9, 2017

• Rewrote the clip_by_norm op to avoid generating infinite intermediate results

  Show description for 1ba562a
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  1ba562a

  Copy full SHA for 1ba562a

Commits on Oct 4, 2017

• Place all the nodes created by the trivial_test_graph_input_yielder

  Show description for cc521eb
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  cc521eb

  Copy full SHA for cc521eb

Commits on Oct 3, 2017

• Made sure the save/restore op is preserved during optimization

  Show description for 941a498
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  941a498

  Copy full SHA for 941a498

Commits on Sep 28, 2017

• Added a python API on top of Grappler items and Grappler clusters

  Show description for cf07600
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  cf07600

  Copy full SHA for cf07600

• Fixed outdated comment

  Show description for 6dc4aac
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  6dc4aac

  Copy full SHA for 6dc4aac

Commits on Sep 22, 2017

• Run constant folding iteratively until it converges.

  Show description for d101b3c
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  d101b3c

  Copy full SHA for d101b3c

Commits on Sep 21, 2017

• List the save/restore nodes in the set of nodes to preserve

  Show description for ba2f59f
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  ba2f59f

  Copy full SHA for ba2f59f

• Skip control dependency edges when processing Restore node shapes

  Show description for d2d574b
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  d2d574b

  Copy full SHA for d2d574b

• Don't assert on unsupported input types

  Show description for 0dab6a7
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  0dab6a7

  Copy full SHA for 0dab6a7

Commits on Aug 31, 2017

• Annotate the graph properties with input values if they're known statically.

  Show description for 9acea81
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  9acea81

  Copy full SHA for 9acea81

• Don't use std::pair on GPU since it's not supported by nvcc

  Show description for 5c3977c
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  5c3977c

  Copy full SHA for 5c3977c

Commits on Aug 29, 2017

• The GPU devices have been renamed devices:GPU. Update the code accordingly

  Show description for 1d1fcc0
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  1d1fcc0

  Copy full SHA for 1d1fcc0

• Annotate the performance database with expected distribution of values.

  Show description for ccc8180
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  ccc8180

  Copy full SHA for ccc8180

• Avoid duplicating the functionality in GrapplerItem::NodesToPreserve()

  Show description for 5396e8e
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  5396e8e

  Copy full SHA for 5396e8e

Commits on Aug 28, 2017

• Added a sanity check to make it easier to debug graph corruptions

  Show description for bda0611
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  bda0611

  Copy full SHA for bda0611

• Turn on arithmetic optimization by default

  Show description for 186164b
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  186164b

  Copy full SHA for 186164b

• Don't record persistent memory usage of typed queues since the code crashes when running on GPU.

  Show description for 78da113
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  78da113

  Copy full SHA for 78da113

Commits on Aug 24, 2017

• Error out normally instead of asserting if TensorFlow fails to enqueue a convolution operation.

  Show description for 49dcb6c
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  49dcb6c

  Copy full SHA for 49dcb6c

• Only use hardware traces on a single subgraph since tracing more than that

  Show description for 09fea25
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  09fea25

  Copy full SHA for 09fea25

Commits on Aug 23, 2017

• Error out instead of asserting if TF fails to create a new session

  Show description for 2272987
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  2272987

  Copy full SHA for 2272987

• Converted LOG(FATAL) into regular errors to prevent the process from crashing

  Show description for eb0808f
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  eb0808f

  Copy full SHA for eb0808f

Commits on Aug 22, 2017

• Individual loop identification

  Show description for 0bb9ddd
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  0bb9ddd

  Copy full SHA for 0bb9ddd

Commits on Aug 19, 2017

• Added the ability to load grappler items from a the metagraphdef files contained in a TensorFlow checkpoint directory.

  Show description for 8454e3f
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  8454e3f

  Copy full SHA for 8454e3f

Commits on Aug 18, 2017

• Small improvements to the arithmetic optimizer

  Show description for a271c37
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  a271c37

  Copy full SHA for a271c37

• Fixed BuildOpInfoWithoutDevice

  Show description for 513def0
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  513def0

  Copy full SHA for 513def0

Commits on Aug 17, 2017

• Speed up the graph to graphdef conversion

  Show description for 3099821
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  3099821

  Copy full SHA for 3099821

• Further optimize the encoding of the constants generated by Grappler

  Show description for 909e2a6
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  909e2a6

  Copy full SHA for 909e2a6

• Optimize the encoding of the constants generated by Grappler

  Show description for 055375c
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  055375c

  Copy full SHA for 055375c

• Avoid generating the _output_shape attribute in constant folding since the can

  Show description for 32dca6f
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  32dca6f

  Copy full SHA for 32dca6f

Commits on Aug 16, 2017

• Added preliminary support for arithmetic simplifications

  Show description for c247826
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  c247826

  Copy full SHA for c247826

Commits on Aug 12, 2017

• Don't run constant folding when xla is turned on since it isn't stable enough.

  Show description for 566821c
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  566821c

  Copy full SHA for 566821c

Pagination

Previous

Next