Commits

Branch selector

 master

User selector

alextp

Datepicker

All time

Commit History

Commits on Oct 9, 2017

• Removing side outputs from tape code.

  Show description for 8814502
  

  alextp

  authored and

  tensorflower-gardener

  committed
  8814502

  Copy full SHA for 8814502

• Expose tfe.test, tfe.in_eager_mode, tfe.in_graph_mode

  Show description for 7c74d2f
  

  alextp

  authored and

  tensorflower-gardener

  committed
  7c74d2f

  Copy full SHA for 7c74d2f

Commits on Oct 5, 2017

• Start of work towards ordering access to resources in tfe.defun/graph_callable.

  Show description for e4aa9dc
  

  alextp

  authored and

  tensorflower-gardener

  committed
  e4aa9dc

  Copy full SHA for e4aa9dc

• Internal private header file with eager C struct definitions.

  Show description for fca4320
  

  alextp

  authored and

  tensorflower-gardener

  committed
  fca4320

  Copy full SHA for fca4320

• Splits backprop.py in two files, one of which can be converted to C

  Show description for 5f97262
  

  alextp

  authored and

  tensorflower-gardener

  committed
  5f97262

  Copy full SHA for 5f97262

• imperative_gradient doesn't fail if some variables are not connected to the output

  Show description for ae98ba9
  

  alextp

  authored and

  tensorflower-gardener

  committed
  ae98ba9

  Copy full SHA for ae98ba9

• Removes use of _grad_fn_accepts_none_for_indices in magic_gradient_function.

  Show description for 6cf9ffe
  

  alextp

  authored and

  tensorflower-gardener

  committed
  6cf9ffe

  Copy full SHA for 6cf9ffe

Commits on Oct 4, 2017

• Fast path for tf.conj when it should be pass-through.

  Show description for cc8ee6c
  

  alextp

  authored and

  tensorflower-gardener

  committed
  cc8ee6c

  Copy full SHA for cc8ee6c

• Fast path for tf.conj when it should be pass-through.

  Show description for 1515549
  

  alextp

  authored and

  tensorflower-gardener

  committed
  1515549

  Copy full SHA for 1515549

• Removes unnecessary eager-mode call to convert_to_tensor in record_gradient.

  Show description for d4ea993
  

  alextp

  authored and

  tensorflower-gardener

  committed
  d4ea993

  Copy full SHA for d4ea993

• matmul uses shape_tuple internally

  Show description for 0c8dbc1
  

  alextp

  authored and

  tensorflower-gardener

  committed
  0c8dbc1

  Copy full SHA for 0c8dbc1

Commits on Oct 3, 2017

• Fast-path for EagerTensorBase.dtype

  Show description for b925f85
  

  alextp

  authored and

  tensorflower-gardener

  committed
  b925f85

  Copy full SHA for b925f85

• get_variable in graph_callable returns Variable objects and not Tensors.

  Show description for 8fb14b1
  

  alextp

  authored and

  tensorflower-gardener

  committed
  8fb14b1

  Copy full SHA for 8fb14b1

• Makes custom_gradient work in graph mode.

  Show description for 57c5613
  

  alextp

  authored and

  tensorflower-gardener

  committed
  57c5613

  Copy full SHA for 57c5613

• Correct names for contrib summaries.

  Show description for 7020f17
  

  alextp

  authored and

  tensorflower-gardener

  committed
  7020f17

  Copy full SHA for 7020f17

Commits on Sep 15, 2017

• Certain ops don't need eager gradients to keep their inputs / outputs alive.

  Show description for b4224fe
  

  alextp

  authored and

  tensorflower-gardener

  committed
  b4224fe

  Copy full SHA for b4224fe

• Eager gradient tape doesn't keep tensors alive.

  Show description for b51cd8c
  

  alextp

  authored and

  tensorflower-gardener

  committed
  b51cd8c

  Copy full SHA for b51cd8c

Commits on Sep 14, 2017

• Fast path for get_variable if in eager and the variable already exists.

  Show description for 62a59d4
  

  alextp

  authored and

  tensorflower-gardener

  committed
  62a59d4

  Copy full SHA for 62a59d4

• Remove dependency on autograd from tf.

  Show description for 75d2fd5
  

  alextp

  authored and

  tensorflower-gardener

  committed
  75d2fd5

  Copy full SHA for 75d2fd5

Commits on Sep 13, 2017

• Expose ResourceVariable as tfe.Variable

  Show description for acdc1f2
  

  alextp

  authored and

  tensorflower-gardener

  committed
  acdc1f2

  Copy full SHA for acdc1f2

Commits on Sep 12, 2017

• Use the scalar cache in MeanGrad.

  Show description for bf96fcd
  

  alextp

  authored and

  tensorflower-gardener

  committed
  bf96fcd

  Copy full SHA for bf96fcd

• Resurrects autograd-free eager gradients.

  Show description for 655f26f
  

  alextp

  authored and

  tensorflower-gardener

  committed
  655f26f

  Copy full SHA for 655f26f

Commits on Sep 8, 2017

• Scope the scalar cache in the context.

  Show description for 0753b0c
  

  alextp

  authored and

  tensorflower-gardener

  committed
  0753b0c

  Copy full SHA for 0753b0c

• Removes "fast paths" which are not fast in eager mode.

  Show description for f331f52
  

  alextp

  authored and

  tensorflower-gardener

  committed
  f331f52

  Copy full SHA for f331f52

• Disables unnecessary checks in TensorVSpace.

  Show description for 03ebcfd
  

  alextp

  authored and

  tensorflower-gardener

  committed
  03ebcfd

  Copy full SHA for 03ebcfd

Commits on Sep 7, 2017

• Avoids triggering memcpy for empty tensors.

  Show description for bf1c826
  

  alextp

  authored and

  tensorflower-gardener

  committed
  bf1c826

  Copy full SHA for bf1c826

• Cache per-device scalars when creating eager tensors, to reduce number of DMAs needed.

  Show description for 8dbe7dd
  

  alextp

  authored and

  tensorflower-gardener

  committed
  8dbe7dd

  Copy full SHA for 8dbe7dd

Commits on Sep 6, 2017

• contrib summaries work in eager-graph mode (with defun)

  Show description for 86b94d8
  

  alextp

  authored and

  tensorflower-gardener

  committed
  86b94d8

  Copy full SHA for 86b94d8

Commits on Sep 2, 2017

• Typo

  Show description for 1499e91
  

  alextp

  authored and

  tensorflower-gardener

  committed
  1499e91

  Copy full SHA for 1499e91

Commits on Sep 1, 2017

• Contrib ops and kernels for summary ops which write without touching python.

  Show description for e83d8ab
  

  alextp

  authored and

  tensorflower-gardener

  committed
  e83d8ab

  Copy full SHA for e83d8ab

Commits on Aug 31, 2017

• Eager-graph mode should work with gradient computation.

  Show description for 424aa9a
  

  alextp

  authored and

  tensorflower-gardener

  committed
  424aa9a

  Copy full SHA for 424aa9a

Commits on Aug 30, 2017

• Fixes eager higher-order gradients for ops whose gradient function uses their outputs.

  Show description for 8f25fa9
  

  alextp

  authored and

  tensorflower-gardener

  committed
  8f25fa9

  Copy full SHA for 8f25fa9

• Guards references to tensor.op to avoid calling them in eager mode.

  Show description for 822a920
  

  alextp

  authored and

  tensorflower-gardener

  committed
  822a920

  Copy full SHA for 822a920

• Supporting control flow ops in eager mode.

  Show description for b7816d0
  

  alextp

  authored and

  tensorflower-gardener

  committed
  b7816d0

  Copy full SHA for b7816d0

• Fixes stray isinstance(_, Tensor) check in math_ops.

  Show description for 943d606
  

  alextp

  authored and

  tensorflower-gardener

  committed
  943d606

  Copy full SHA for 943d606

Pagination

Previous

Next