Added an option to the client to manually set the number of models to spawn on a parameter server. This allows for more parallelism because we will have multiple actors (and thus concurrent threads) per parameter server. The default setting is 1 which is the same behavior as before.
Example usage:
// Construct vector with 10000 keys and 4 models per parameter server (= 4 threads)
client.vector[Long](10000, 4)
// Construct matrix with 1000 rows, 1000 cols and 4 models per parameter server
client.matrix[Long](1000, 1000, 4)
Added an option to the client to manually set the number of models to spawn on a parameter server. This allows for more parallelism because we will have multiple actors (and thus concurrent threads) per parameter server. The default setting is
1which is the same behavior as before.Example usage: