| Commit message (Collapse) | Author | Age |
... | |
|
|
|
| |
PiperOrigin-RevId: 172118528
|
|
|
|
| |
PiperOrigin-RevId: 172117003
|
|
|
|
| |
PiperOrigin-RevId: 172114960
|
|
|
|
|
|
| |
redundant now that bazel-toolchains repo is live
PiperOrigin-RevId: 172113861
|
|
|
|
|
|
| |
from the core Estimator API.
PiperOrigin-RevId: 172108321
|
|
|
|
| |
PiperOrigin-RevId: 172107872
|
|
|
|
| |
PiperOrigin-RevId: 172105420
|
|
|
|
|
|
| |
Change the logic that identifies topK choices with a cache friendly alternative.
PiperOrigin-RevId: 172101068
|
|
|
|
| |
PiperOrigin-RevId: 172091595
|
|
|
|
| |
PiperOrigin-RevId: 172091245
|
|
|
|
| |
PiperOrigin-RevId: 172073518
|
|
|
|
|
|
| |
computation.
PiperOrigin-RevId: 172071664
|
|
|
|
| |
PiperOrigin-RevId: 172065800
|
|
|
|
|
|
|
|
|
| |
tools/graph_transforms.
By setting this option to false, the transformer will not strip off the shape
information stored as attributes.
PiperOrigin-RevId: 172057283
|
|
|
|
|
|
| |
from the graph def that were not present in the consumer (server).
PiperOrigin-RevId: 172051437
|
|
|
|
| |
PiperOrigin-RevId: 172051036
|
|
|
|
| |
PiperOrigin-RevId: 172050536
|
|
|
|
|
|
| |
streaming_false_{negative,positive}_rate_at_thresholds.
PiperOrigin-RevId: 172048554
|
|
|
|
| |
PiperOrigin-RevId: 172045110
|
|
|
|
|
|
| |
bool)
PiperOrigin-RevId: 172044654
|
|
|
|
| |
PiperOrigin-RevId: 172043591
|
|
|
|
| |
PiperOrigin-RevId: 172041381
|
|
|
|
| |
PiperOrigin-RevId: 172041133
|
|
|
|
| |
PiperOrigin-RevId: 172040631
|
|
|
|
|
|
| |
arbitrary nesting in layer inputs & outputs.
PiperOrigin-RevId: 172040243
|
|
|
|
|
|
| |
underlying resource. This makes the lifetime of the underlying resource match that of its corresponding Python object.
PiperOrigin-RevId: 172039259
|
|
|
|
| |
PiperOrigin-RevId: 172038787
|
|
|
|
| |
PiperOrigin-RevId: 172037998
|
|
|
|
| |
PiperOrigin-RevId: 172034428
|
|
|
|
|
|
|
|
|
| |
1. Use a id_to_string map to reduce the profile size (2/3 in xception)
2. dedup code view's function name with extra file base name.
3. remove code view display heuristic that doesn't work in some cases.
4. make the profile_context thread-safe.
PiperOrigin-RevId: 172031528
|
|
|
|
|
|
| |
clean/organized graph displays in TensorBoard.
PiperOrigin-RevId: 172028555
|
|
|
|
|
|
| |
with tf.linalg.adjoint(foo) or tf.transpose(foo, conjugate=True), and clean up a few places that can avoid explicit adjoints as inputs to matmul.
PiperOrigin-RevId: 172027859
|
|
|
|
|
|
| |
synchronization, which can have a significant performance impact.
PiperOrigin-RevId: 172025744
|
|
|
|
| |
PiperOrigin-RevId: 172023756
|
|
|
|
|
|
| |
than trying to launch kernels with an invalid config and leaving the stream in an error state. The latter is much harder to debug.
PiperOrigin-RevId: 172019169
|
|
|
|
| |
PiperOrigin-RevId: 172018709
|
|
|
|
|
|
|
| |
TF_RET_CHECK logs the call stack and HloEvaluator is often called with
parameters and non-constant ops which results in a logging output in cases where the caller expects it could fail.
PiperOrigin-RevId: 172018263
|
|
|
|
| |
PiperOrigin-RevId: 172014544
|
|
|
|
|
|
|
|
|
|
|
| |
unknown.
Right now, TPUEstimator will raise an error message complaining that sharding
failed because (None % integer value) is not a well-defined operation. I hope
the new error message will make it easier for people to figure out what's going
on.
PiperOrigin-RevId: 172013663
|
|
|
|
| |
PiperOrigin-RevId: 172013619
|
|
|
|
| |
PiperOrigin-RevId: 172013289
|
|
|
|
|
|
|
|
|
|
|
|
| |
to virtual_batch_size
Key changes:
1) adding support for multi-axis batch norm by allowing the axis to be either an int or a list of ints
2) multi-axis batch norm is handled entirely by TensorFlow ops at the moment (no special kernel) and the performance is heavily dependent on the Tensor format (see reduce_* kernels for reduction rules)
3) Fix ghost batch norm by sharing the same gamma/beta/mean/var parameter across all virtual batches
4) Change ghost batch norm API to virtual_batch_size to be more consistent with its intended use case.
PiperOrigin-RevId: 172012360
|
|
|
|
| |
PiperOrigin-RevId: 172009823
|
|
|
|
| |
PiperOrigin-RevId: 172005195
|
|
|
|
| |
PiperOrigin-RevId: 172001904
|
|
|
|
|
|
| |
don't get negative cost estimates for _Send ops.
PiperOrigin-RevId: 171999586
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
learning losses,
starting with:
* [Contrastive loss](http://yann.lecun.com/exdb/publis/pdf/hadsell-chopra-lecun-06.pdf)
* [Triplet Loss with semihard negative mining](https://arxiv.org/abs/1503.03832)
* [Npairs loss](http://www.nec-labs.com/uploads/images/Department-Images/MediaAnalytics/papers/nips16_npairmetriclearning.pdf)
* Npairs loss w/ multilabel support
* [Lifted structured loss](http://www.cv-foundation.org/openaccess/content_cvpr_2016/papers/Song_Deep_Metric_Learning_CVPR_2016_paper.pdf)
* [Structured clustering embedding](https://arxiv.org/abs/1612.01213)
PiperOrigin-RevId: 171997156
|
|
|
|
| |
PiperOrigin-RevId: 171987329
|
|
|
|
|
|
| |
nodes in the graph.
PiperOrigin-RevId: 171986401
|
|
|
|
| |
PiperOrigin-RevId: 171983725
|