Closed alexanderlewis99 closed 10 months ago
Options for making it so that each core gets exactly the same number of samples, such as follows:
First determine how many cores to use based on n_samples and batch_size: n_cores = round_up(n_samples/batch_size)
Then readjust the batch_size so it's (almost) exactly equal across the cores: batch_size = n_samples/n_cores
Example: 10100 cells, 10K sample size, 2 cores ==> batch_size = 10100/2 = 5050 cells.
Solved by commit 9f7ddd8
Options for making it so that each core gets exactly the same number of samples, such as follows:
First determine how many cores to use based on n_samples and batch_size: n_cores = round_up(n_samples/batch_size)
Then readjust the batch_size so it's (almost) exactly equal across the cores: batch_size = n_samples/n_cores
Example: 10100 cells, 10K sample size, 2 cores ==> batch_size = 10100/2 = 5050 cells.