Closed faysou closed 5 years ago
When using tensorflow 2 and tensorflow.compat.v1 some functions in "tensor_train_batch.py" and "decompositions.py" have .value after some numbers used for ranks, it leads to the code bugging. When removing these .value the code works again.
.value
For example in tensor_train_batch.py
self._batch_size = tt_cores[0].get_shape()[0].value (line 58)
TF 2 is not fully supported, you have to use
tf.disable_v2_behavior() tf.enable_eager_execution()
to disable some of v2 features (like in the tutorial).
Ok thank you.
When using tensorflow 2 and tensorflow.compat.v1 some functions in "tensor_train_batch.py" and "decompositions.py" have
.value
after some numbers used for ranks, it leads to the code bugging. When removing these.value
the code works again.For example in tensor_train_batch.py