Markopy
Markopy copied to clipboard
Single Threaded Training is faster than multi threading
Describe the bug
With markopy, -t 1
will complete the training faster than -t 10
To Reproduce Steps to reproduce the behavior:
- Train a base model with a corpus using
-t 1
- Train the same base model with the same corpus using
-t 10
- Compare runtime
- Contemplate the hours i spent input scheduling for training during development
Details:
- Libmarkov
- Linux 5.8.0-59-generic # 66~20.04.1-Ubuntu SMP