| Commit message (Collapse) | Author | Age |
... | |
|
|
|
|
|
| |
lieu of a new num_cores_per_replica.
PiperOrigin-RevId: 213111326
|
|
|
|
| |
PiperOrigin-RevId: 212986442
|
|
|
|
| |
PiperOrigin-RevId: 212857508
|
|
|
|
|
|
| |
of erring out.
PiperOrigin-RevId: 212385555
|
|
|
|
| |
PiperOrigin-RevId: 212050240
|
|
|
|
| |
PiperOrigin-RevId: 212039241
|
|
|
|
| |
PiperOrigin-RevId: 211984222
|
|
|
|
| |
PiperOrigin-RevId: 211843340
|
|
|
|
| |
PiperOrigin-RevId: 211833891
|
|
|
|
| |
PiperOrigin-RevId: 211724843
|
|
|
|
| |
PiperOrigin-RevId: 211719009
|
|
|
|
|
|
| |
RELNOTES: n/a
PiperOrigin-RevId: 211718248
|
|
|
|
| |
PiperOrigin-RevId: 211707155
|
|
|
|
| |
PiperOrigin-RevId: 211532963
|
|
|
|
| |
PiperOrigin-RevId: 211514287
|
|
|
|
| |
PiperOrigin-RevId: 211501909
|
|
|
|
|
|
| |
With this pipelining change (and a little bit of re-tuning of input pipelines to have _less_ parallelism to avoid thread starvation), we are able to significantly reduce the overheads of supporting dynamic shapes with Keras.
PiperOrigin-RevId: 211157531
|
|
|
|
| |
PiperOrigin-RevId: 211129794
|
|
|
|
| |
PiperOrigin-RevId: 211099558
|
|
|
|
| |
PiperOrigin-RevId: 210855008
|
|
|
|
| |
PiperOrigin-RevId: 210827158
|
|
|
|
|
|
| |
tensor with shape [num_groups, num_replica_per_group].
PiperOrigin-RevId: 210656091
|
|
|
|
|
|
|
|
| |
Also fix the API as num_cores is not a good API. Changed to a debugging oriented argument using_single_core. Ideally we should not need that in future.
The updated lstm example cannot work with TF 1.10 anymore.
PiperOrigin-RevId: 210632592
|
|
|
|
|
|
| |
(Dimension and tensor # were transposed in the error message)
PiperOrigin-RevId: 210598445
|
|
|
|
| |
PiperOrigin-RevId: 210423935
|
|
|
|
| |
PiperOrigin-RevId: 210187980
|
|
|
|
| |
PiperOrigin-RevId: 210185805
|
|
|
|
| |
PiperOrigin-RevId: 209969854
|
|
|
|
|
|
| |
- Removes the use of global variable collections for initialization. This means that variables created externally (such as input feed variables) will have to be initialized explicitly.
PiperOrigin-RevId: 209692235
|
|
|
|
| |
PiperOrigin-RevId: 209682829
|
|
|
|
| |
PiperOrigin-RevId: 209673552
|
|
|
|
| |
PiperOrigin-RevId: 209605597
|
|
|
|
| |
PiperOrigin-RevId: 209507825
|
|
|
|
|
|
| |
Adjust TPUDistributionStrategy API to remove duplicate master address.
PiperOrigin-RevId: 209058959
|
|
|
|
| |
PiperOrigin-RevId: 209010536
|
|
|
|
|
|
|
| |
allowed behavior in tf.Estimator
NB: Calls data/util/nest.flatten() which is slightly different than core/util/nest (already a dependency in tpu_estimator.py) as it maintains structure by treating tuples and lists differently
PiperOrigin-RevId: 208848267
|
|
|
|
| |
PiperOrigin-RevId: 208695032
|
|
|
|
| |
PiperOrigin-RevId: 208491690
|
|
|
|
| |
PiperOrigin-RevId: 208126204
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Back-ticks are now converted to links in the api_docs generator. With the new docs repo we're moving to simplify the docs pipeline, and make everything more readable.
By doing this we no longer get test failures for symbols that don't exist (`tf.does_not_exist` will not get a link).
There is also no way, not to set custom link text. That's okay.
This is the result of the following regex replacement (+ a couple of manual edits.):
re: @\{([^$].*?)(\$.+?)?}
sub: `\1`
Which does the following replacements:
"@{tf.symbol}" --> "`tf.symbol`"
"@{tf.symbol$link_text}" --> "`tf.symbol`"
PiperOrigin-RevId: 208042358
|
|
|
|
|
|
| |
context.
PiperOrigin-RevId: 207996119
|
|
|
|
|
|
| |
Strategy
PiperOrigin-RevId: 207802033
|
|
|
|
| |
PiperOrigin-RevId: 207780309
|
|
|
|
| |
PiperOrigin-RevId: 207569055
|
|
|
|
| |
PiperOrigin-RevId: 207326276
|
|
|
|
| |
PiperOrigin-RevId: 206997688
|
|
|
|
| |
PiperOrigin-RevId: 206972475
|
|
|
|
|
|
| |
be executed as part of the program.
PiperOrigin-RevId: 206956131
|
|
|
|
|
|
| |
RELNOTES: n/a
PiperOrigin-RevId: 206865154
|
|
|
|
| |
PiperOrigin-RevId: 206844376
|