-
Notifications
You must be signed in to change notification settings - Fork 1k
Distributed training
Tatiana Likhomanenko edited this page Dec 10, 2019
·
1 revision
wav2letter++ supports distributed training on multiple GPUs out of the box. To
run on multiple GPUs set pass the flag -enable_distributed true and run with
MPI:
mpirun -n 8 <train_cpp_binary> [train|continue|fork] \
-enable_distributed true \
<... other flags ..>
The above command will run data parallel training with 8 processes (e.g. on 8 GPUs).