Open Pale-Blue-Dot-97 opened 5 months ago
What appears to be happening here is that each process of a distributed process group sees that the dataset requested does not exist, hence they all try to independently create and they cache the dataset. As this will be in the same location, a conflict arises when the slower processes try caching to a now extant dataset
The solution is to ensure that only process 0
attempts to create the dataset. All other processes should then wait until 0
is finished, then they can load the dataset from the new cache.
Describe the bug A clear and concise description of what the bug is.
To Reproduce Steps to reproduce the behavior:
cache==True
formake_dataset
Expected behavior No error. If the dataset is not already cached, it should be created then cached under the unique hash. If it exists, the hash should be recognised and the dataset loaded.
Environment (please complete the following information):