You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Performance issues in /src/python/tensorflow_cloud/core (by P3) #371
Hello! I've found a performance issue in /tests/testdata/keras_tuner_cifar_example.py: batch() should be called before map(), which could make your program more efficient. Here is the tensorflow document to support it.
Detailed description is listed below:
.batch(BATCH_SIZE)(here) should be called before train_dataset.map(scale)(here).
.batch(BATCH_SIZE)(here) should be called before .map(scale)(here).
Besides, you need to check the function called in map()(e.g., scale called in .map(scale)) whether to be affected or not to make the changed code work properly. For example, if scale needs data with shape (x, y, z) as its input before fix, it would require data with shape (batch_size, x, y, z).
Looking forward to your reply. Btw, I am very glad to create a PR to fix it if you are too busy.
I resolved the performance issue by changing the order of operations in the dataset pipeline, calling .batch(BATCH_SIZE) before .map(scale). This modification improves efficiency by batching the data before applying the map function, reducing the number of individual function calls.
Activity
DLPerf commentedon Nov 4, 2021
Hello, I'm looking forward to your reply~
sharmadhiraj86 commentedon Sep 11, 2022
I would like to contribute in solving this issue. Can you please assign this issue to me? @DLPerf @lamberta @mihaimaruseac @jaeyounkim @martinwicke
mihaimaruseac commentedon Sep 12, 2022
We don't generally assign issue to contributors. Instead, please send a PR when ready.
sahil270803 commentedon Oct 14, 2024
I resolved the performance issue by changing the order of operations in the dataset pipeline, calling .batch(BATCH_SIZE) before .map(scale). This modification improves efficiency by batching the data before applying the map function, reducing the number of individual function calls.