torchrun
Field |
Default |
Description |
Type |
---|---|---|---|
name |
??? |
The referenced name of the object to construct |
str |
nproc_per_node |
??? |
The number of processes per node |
int |
master_addr |
127.0.0.1 |
The address of the master |
str |
master_port |
??? |
The port of the master |
int |
backend |
nccl |
The backend to use for distributed training |
str |
start_method |
spawn |
The method to use to start processes |
str |
torchrun_path |
??? |
The path to the TorchRun script |
str |