Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Branch 194768567 #18983

Merged
merged 33 commits into from
Apr 30, 2018
Merged

Branch 194768567 #18983

merged 33 commits into from
Apr 30, 2018

Conversation

caisq
Copy link
Contributor

@caisq caisq commented Apr 30, 2018

No description provided.

tensorflower-gardener and others added 30 commits April 27, 2018 11:44
@tensorflower-gardener
Internally rewrite @recompute_grad to use @custom_gradient
d1e0a73 
PiperOrigin-RevId: 194571125
@tensorflower-gardener
Replace GrapplerFunctionItem input with a constant.
6d793e1 
PiperOrigin-RevId: 194579253
@asimshankar @tensorflower-gardener
eager: Improve error message when GradientTape is used incorrectly.
de26d3e 
PiperOrigin-RevId: 194580654
@tensorflower-gardener
Remove scope name from bfloat16
69465b0 
PiperOrigin-RevId: 194580957
@tensorflower-gardener
[TF:XLA] Bump open source llvm revision to r330950
ec580e6 
PiperOrigin-RevId: 194588403
@tensorflower-gardener
Remove whitespaces from tags in saved_model_cli.
6da711a 
This currently causes tags mismatch because a leading whitespace is added
within the saved_model_cli when doing ', '.join(tag_set).

PiperOrigin-RevId: 194590154
@tensorflower-gardener
If two identical functions are given different grad func,
a4dbc33 
they should be named differently. Otherwise, tf.gradients
gets confused.

PiperOrigin-RevId: 194593519
@tensorflower-gardener
[XLA:CPU] Implement fusion for the Gather HLO
8477e7c 
PiperOrigin-RevId: 194594759
@tensorflower-gardener
Minor eager service proto clarification.
d84f982 
PiperOrigin-RevId: 194596337
@tensorflower-gardener
Fix broken ElementWiseFusionTest.
fbd9ecd 
PiperOrigin-RevId: 194602336
@nickdesaulniers @tensorflower-gardener
[TF:XLA:INTERPRETER] implement bfloat16 comparisons
a52f64d 
PiperOrigin-RevId: 194608854
@petrosmol @tensorflower-gardener
Minor fix to SDCAOptimizer documentation.
95e297c 
PiperOrigin-RevId: 194609850
@saeta @tensorflower-gardener
Make RetryingFileSystem a template.
864e056 
PiperOrigin-RevId: 194614877
@tensorflower-gardener
[XLA] Fix bug in ShapeUtil::StripDegenerateDimensions
b2b8dca 
PiperOrigin-RevId: 194621163
@tensorflower-gardener
Split up ElementaIrEmitter::MakeElementGenerator into smaller functio…
68efa50 
…ns; NFC

PiperOrigin-RevId: 194622198
@tensorflower-gardener
Add internal uint b stats to TfOpStats.
ce8e19a 
PiperOrigin-RevId: 194625155
@tensorflower-gardener
HLO profiling for tfcompile.
7474743 
This CL extends the --xla_hlo_profile knob to tfcompile.  tf_library rules can
now set enable_xla_hlo_profiling to True to:

 - Have the generated code update per-HLO profile counters as it executes.
 - Have tfcompile generate and serialize an instance HloProfilePrinterData with
   a compiled model that can be used to pretty-print the collected profile
   counters.

PiperOrigin-RevId: 194627272
@tensorflower-gardener
Properly export recurrent in contrib.
1aef48e 
The following symbols are available:
- tf.contrib.recurrent.bidirectional_functional_rnn
- tf.contrib.recurrent.functional_rnn
- tf.contrib.recurrent.Recurrent

PiperOrigin-RevId: 194632138
@tensorflower-gardener
Add test case on compiling dense layer node with XLA.
d047a36 
PiperOrigin-RevId: 194634563
@tensorflower-gardener
Fix kernel creation bug, due to constant folding always use CPU.
04e17da 
PiperOrigin-RevId: 194636076
@annarev @tensorflower-gardener
Removing hidden_ops.txt file.
102b0c8 
PiperOrigin-RevId: 194637892
@tensorflower-gardener
This is Part 1 of Swift<->TF sends/recvs: support sending tensors fro…
4c56995 
…m TF to

Swift via direct session.

The changes are:

1. Added an experimental TF C API TF_DequeueNamedTensor() to consume the queued
tensors from a dequeue op. One use case is for the Swift host program to consume
tensors sent by TF, where the queue is a Fifo queue managed by TF.

Enqueuing tensors are done by running an enqueue op in a graph. The queued
tensors are not persisted, and will be lost if the process/machine dies. The
queue has a bounded capacity, to prevent producer from being unboundedly ahead
of consumer.

while caller of TF_DequeueNamedTensor() could have run the Fifo dequeue op
directly, the extra level of indirection provided by this API allows us to more
easily switch the queuing impl to another mechanism. If and once we stabilize on
the Fifo queue based impl, we can remove this API.

2. Added a new S4TF runtime API _TFCReceiveTensorHandle() that receives a tensor
  via TF_DequeueNamedTensor().

3. To support tensor receives in host program, taught PartitionCloner in
  TFPartition to insert SIL code to call _TFCReceiveTensorHandle().

4. To support tensor sends in accelerator program, taught TFGraphLowering in
  generate QueueEnqueueV2 nodes in the TF graphs, with appropriate control
  dependence to make sure these nodes get executed.

a) The enqueue produces no output tensor, and is executed only for its side
effect. To ensure it is executed properly, control dependence is wired up. The
general design is: before a TF_Function (can be a top level function or the body
function of a while op) produces an output tensor OT, make OT control dependent
on the enqueue op, so that enqueue gets run before the function returns.

b) If tensor send occurs in a while loop body, the body logic currently gets
lowered in 3 places: the while op cond function, the while op body function, and
the ops at the same level as the while op itself (for running the last loop
iteration). In this case, the correct TFGraph lowering is to run the enqueue in
the last 2 out of the 3 places above.

After this CL, the dual versions of the above (dequeuing via an op, and
enqueuing via C API) will be added.

PiperOrigin-RevId: 194658511
@tensorflower-gardener
Allow not specifying eval_spec when evaluation is not necessarily run.
c018583 
PiperOrigin-RevId: 194661814
@saeta @tensorflower-gardener
[tf.data] Use core::ScopedUnref to avoid resource leakage.
fb10697 
If for whatever reason iterator_resource->set_iterator did not return Status::OK(), we would leak a reference on the iterator_resource. With this change, we won't leak the resource.

PiperOrigin-RevId: 194662412
@asimshankar @tensorflower-gardener
Java: Release 1.8.0
d07a8d4 
PiperOrigin-RevId: 194663800
@tensorflower-gardener
Internally rewrite RevBlock to use @custom_gradient
3a9c513 
PiperOrigin-RevId: 194679657
@dimvar @tensorflower-gardener
[TF:XLA]
d02745e 
- Require a module config when creating an HloModule.
- All tests using HloTestBase create a module using CreateNewModule.

PiperOrigin-RevId: 194684585
@rxwei @tensorflower-gardener
Update the Swift for TensorFlow community page.
2e1f3ef 
PiperOrigin-RevId: 194687897
@sherrym @tensorflower-gardener
Added del_hparam(), the counter part of add_hparam.
45529aa 
PiperOrigin-RevId: 194711291
@rjpower @tensorflower-gardener
Add support for a clean checkpoint and shutdown in response to a term…
c41b546 
…ination notice.

PiperOrigin-RevId: 194722985
@caisq caisq requested a review from protoget April 30, 2018 14:52
@caisq caisq merged commit 6f532f3 into tensorflow:master Apr 30, 2018
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@av8ramit av8ramit av8ramit approved these changes

@protoget protoget Awaiting requested review from protoget

Assignees
No one assigned
Labels
cla: yes
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.

None yet