Nicolas
Nicolas
@jburnim can you please suggest a reviewer? CC @axch
Hi @SiegeLordEx, I have assessed your comments, can you have a look? Thanks
> Hi @nicolaspi Any update on this PR? Please. Thank you! Hi, sorry I am 'out of office' at the moment. I will give an update within two weeks.
Hi, sorry for the delay. I am out of sync with this issue. I am closing it. I will reopen or open a new issue in the future if I...
Unfortunately no, this is old code and the project is not maintained. However, switching into python3 should be fairly easy. Le sam. 15 janv. 2022, 20:46, SteffenEppEB ***@***.***> a écrit...
This is fixed with the new [structured loss feature](https://github.com/keras-team/keras/pull/20358) provided that `y_true` and `y_pred` have the same structure: ```python import numpy as np import tensorflow as tf import keras def...
Wouldn't it be better to wrap `OrderedDict` rather than re-implement `flatten`, which is written in C++ in optree? e.g.: ```python from collections import OrderedDict import optree from keras import tree...
Hi @mehtamansi29 Thanks for the answer. The issue arises from a protection heuristic defined in `tf.data` [here](https://github.com/tensorflow/tensorflow/blame/9286e84d9839e05a36fbf998e72d74f960b86990/tensorflow/python/data/ops/iterator_ops.py#L759C5-L759C5) . The protection is disabled when using eager mode, but this is not...
Potentially related: https://github.com/tensorflow/tensorflow/issues/72388
@tilakrayal Your gist is showing me an error when I run it (even for tf-nightly).