Use synthetic data in keras integ test #367
Conversation
Codecov Report
@@ Coverage Diff @@
## master #367 +/- ##
=========================================
+ Coverage 92.85% 92.9% +0.05%
=========================================
Files 51 51
Lines 3568 3568
=========================================
+ Hits 3313 3315 +2
+ Misses 255 253 -2
Continue to review full report at Codecov.
|
|
But customers are expected to download data from s3. |
|
@nadiaya we have other tests that do download data from S3. This change matches what the TF CIFAR test does, which was a change introduced to make our tests more reliable and run faster. |
| def _filenames(mode, data_dir): | ||
| """Returns a list of filenames based on 'mode'.""" | ||
| data_dir = os.path.join(data_dir, 'cifar-10-batches-bin') | ||
| def _generate_synthetic_data(mode, batch_size): |
There was a problem hiding this comment.
can you put this in a different file that's used by both the TF and Keras tests?
There was a problem hiding this comment.
They are actually different. For the keras model created, the label type should match feature which is float32. Since I will talk with Marcio about whether to use real cifar10 data in tests, I think this PR is just a quick fix for current failure. We probably don't need to have a separate file with this function (and with additional parameters for type).
I will update this part after talking with Marcio in the future.
* Use synthetic data in keras integ test * Add new line at end of file * Use PREDICT_INPUTS as prediction input tensor name
Issue #, if available:
Description of changes:
Keras integ tests failed for not downloading cifar10 data. Will use synthetic data instead to fix this problem. The synthetic data is generated in same way as tf_cifar integ test.
Merge Checklist
Put an
xin the boxes that apply. You can also fill these out after creating the PR. If you're unsure about any of them, don't hesitate to ask. We're here to help! This is simply a reminder of what we are going to look for before merging your pull request.By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.