core/validate.py#L176 checks the current TensorFlow Version on your local machine. It then asserts that the value is in the list from `get_cloud_tpu_supported_tf_version` which contains the list `[“2.1”]`. This makes it currently impossible to perform TPU training (from what I can tell, you can’t get a TF version where `tf.__version__` is exactly “2.1” with no postpended subversion.

tensorflow / cloud

The TensorFlow Cloud repository provides APIs that will allow to easily go from debugging and training your Keras and TensorFlow code in a local environment to distributed training in the cloud.

https://github.com/tensorflow/cloud

Apache License 2.0

370 stars 84 forks source link

core/validate.py#L176 checks the current TensorFlow Version on your local machine. It then asserts that the value is in the list from `get_cloud_tpu_supported_tf_version` which contains the list `[“2.1”]`. This makes it currently impossible to perform TPU training (from what I can tell, you can’t get a TF version where `tf.version` is exactly “2.1” with no postpended subversion. #363

Closed copybara-service[bot] closed 2 years ago

copybara-service[bot] commented 2 years ago

core/validate.py#L176 checks the current TensorFlow Version on your local machine. It then asserts that the value is in the list from get_cloud_tpu_supported_tf_version which contains the list [“2.1”]. This makes it currently impossible to perform TPU training (from what I can tell, you can’t get a TF version where tf.__version__ is exactly “2.1” with no postpended subversion.

Removes this check so TPU training will work again