Commits

Breadcrumbs

History for

onmaster

User selector

All users

All time

Commit History

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

• Fold fetch nodes.

  Show description for 57498a8
  

  Yao Zhang

  authored and

  tensorflower-gardener

  committed
  57498a8

  Copy full SHA for 57498a8

Commits on Sep 8, 2017

• Pass in the CPU device to grappler, instead of making a new one, when possible.

  Show description for 2b50116
  tensorflower-gardener

  committed
  2b50116

  Copy full SHA for 2b50116

Commits on Aug 29, 2017

• Support loop for layout optimizer.

  Show description for a0cb8c5
  

  Yao Zhang

  authored and

  tensorflower-gardener

  committed
  a0cb8c5

  Copy full SHA for a0cb8c5

Commits on Aug 28, 2017

• Do not keep unused, folded nodes in the graph to reduce the graph size.

  Show description for ffe641b
  

  Yao Zhang

  authored and

  tensorflower-gardener

  committed
  ffe641b

  Copy full SHA for ffe641b

Commits on Aug 19, 2017

• Produce valid node names for transpose.

  Show description for 181267c
  

  Jingyue Wu

  authored and

  tensorflower-gardener

  committed
  181267c

  Copy full SHA for 181267c

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

Commits on Aug 15, 2017

• Avoid creating new nodes and updating the input names of output nodes, if a folded node has only 1 output; instead, simply replace the existing folded node with a const node. This is to speed up th…

  Show description for 39b5ffc
  

  Yao Zhang

  authored and

  tensorflower-gardener

  committed
  39b5ffc

  Copy full SHA for 39b5ffc

Commits on Aug 2, 2017

• Prune nodes that won't be run. This makes subsequent optimizers faster since they don't have to process as many nodes.

  Show description for c44250d
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  c44250d

  Copy full SHA for c44250d

Commits on Jul 31, 2017

• Grappler memory optimization: allow inputs to gradients with non-standard names to be recomputed

  Show description for 6263539
  

  allenlavoie

  authored and

  tensorflower-gardener

  committed
  6263539

  Copy full SHA for 6263539

Commits on Jul 29, 2017

• Replace constants with variables in memory optimizer test, so that they are not removed by the constant folding pass.

  Show description for 5270d48
  

  Yao Zhang

  authored and

  tensorflower-gardener

  committed
  5270d48

  Copy full SHA for 5270d48

Commits on Jun 29, 2017

• Prepare to remove a bunch of proto.h includes from tensorflow/core headers

  Show description for e85d3df
  

  girving

  authored and

  tensorflower-gardener

  committed
  e85d3df

  Copy full SHA for e85d3df

Commits on Jun 26, 2017

• Added a tool to report the static information that can be extracted from a TF model.

  Show description for bdc928c
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  bdc928c

  Copy full SHA for bdc928c

Commits on Jun 23, 2017

• Don't crash if a metagraph fails to load.

  Show description for 0eff699
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  0eff699

  Copy full SHA for 0eff699

Commits on Jun 16, 2017

• Support optimizers in cost analyzer tool.

  Show description for 8bb8099
  

  Yao Zhang

  authored and

  tensorflower-gardener

  committed
  8bb8099

  Copy full SHA for 8bb8099

Commits on Jun 15, 2017

• Use gfile for reading a file.

  Show description for f564625
  

  Yao Zhang

  authored and

  tensorflower-gardener

  committed
  f564625

  Copy full SHA for f564625

Commits on Jun 14, 2017

• Add a heuristic to Grappler's memory optimizer to recompute elementwise ops

  Show description for f0a8bd9
  tensorflower-gardener

  committed
  f0a8bd9

  Copy full SHA for f0a8bd9

• Add a standalone cost analysis tool. Improved logging.

  Show description for 6ffa51f
  

  Yao Zhang

  authored and

  tensorflower-gardener

  committed
  6ffa51f

  Copy full SHA for 6ffa51f

Commits on Jun 13, 2017

• Simplified the virtual placer code

  Show description for 0248500
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  0248500

  Copy full SHA for 0248500

Commits on Jun 9, 2017

• Add statically inferred output shapes to the graph. Update a

  Show description for 9f8ca0c
  

  Yao Zhang

  authored and

  tensorflower-gardener

  committed
  9f8ca0c

  Copy full SHA for 9f8ca0c

Commits on Jun 3, 2017

• Don't use the _output_shape attribute in the op_level_cost_estimator since

  Show description for bb7a8d8
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  bb7a8d8

  Copy full SHA for bb7a8d8

Commits on Jun 1, 2017

• Add flag to determine whether to do L1 optimizations and inline functions. Default is to do them. In tf_optimizer don't inline or do l1 optimizations.

  Show description for d9620ca
  tensorflower-gardener

  committed
  d9620ca

  Copy full SHA for d9620ca

• Fix discrepancy between measured and analytical cost graph. Use tf_cuda_library for utils.

  Show description for e5088cb
  

  Yao Zhang

  authored and

  tensorflower-gardener

  committed
  e5088cb

  Copy full SHA for e5088cb

Commits on May 26, 2017

• Open sourced the cost analyzer

  Show description for 2251633
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  2251633

  Copy full SHA for 2251633

Commits on May 23, 2017

• Add a python test for LayoutOptimizer.

  Show description for 60957e8
  

  Yao Zhang

  authored and

  tensorflower-gardener

  committed
  60957e8

  Copy full SHA for 60957e8

Commits on May 15, 2017

• Add control dependencies to prevent nodes from being swapped back in too soon.

  Show description for e91712d
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  e91712d

  Copy full SHA for e91712d

Commits on May 10, 2017

• Propagate the cluster to the grappler optimizers: this will enable the development of smarter optimizers that can take the available hardware into account when optimizing a graph.

  Show description for 1937df8
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  1937df8

  Copy full SHA for 1937df8

Commits on Apr 25, 2017

• Added a python API to the meta graph optimizer

  Show description for 58fe576
  

  benoitsteiner

  authored and

  tensorflower-gardener

  committed
  58fe576

  Copy full SHA for 58fe576

Loading