projects / platform / upstream / tensorflow.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 753cc5b) | patch
Ensure that saving/restoring iterator in CheckpointInputPipelineHook is performed...
authorSaurabh Saxena <srbs@google.com>
Mon, 21 May 2018 23:43:53 +0000 (16:43 -0700)
committerTensorFlower Gardener <gardener@tensorflow.org>
Mon, 21 May 2018 23:45:55 +0000 (16:45 -0700)
commit0ad7f20ed6876809a2b804365293a5c21dbcd374
treeaed9a68a8db537a11759321b5c7e09bc77366cedtree | snapshot
parent753cc5b3f7461b0b3f59605cba10b965aca0e3adcommit | diff
Ensure that saving/restoring iterator in CheckpointInputPipelineHook is performed *after* the _DatasetInitializerHook has been run.
In the TPUEstimator the _DatasetInitializerHook is present in the
EstimatorSpec.training_hooks. Since these are executed after the `hooks`
passed to Estimator.train the input pipeline checkpointing hook fails
since it finds an uninitialized iterator.

PiperOrigin-RevId: 197482609
tensorflow/contrib/data/python/ops/iterator_ops.py diff | blob | history
Domain: Machine Learning / ML Framework;