Useful options
Here is a non exhaustive list of option you can use for your jobs:
| Option full name | shortcut | Value | Descritpion |
|---|---|---|---|
Node selection | |||
| –partition | -p | ALL|SCHOOL|<CONTRIBUTOR> | Choose a partition to run the job see Queue and Priority |
| –account | -A | <contributor>_group | Use by contributor to access their priority partition see Queue and Priority |
| –nodelist | -w | watgpu108,watgpu208,... | List of nodes the job can run in |
| –exclude | -x | watgpu108,watgpu208,... | List of job the job will not run in |
Resources selection | |||
| –gres | - | gpu:1 or gpu:<contributor>gpu | Number of GPU for the job |
| –mem | - | ’10GB | Amount of memory for the job |
| –cpus-per-task | -c | 1 | Number of CPU for the job |
| –constraint | -C | `ADA6000 | L40S |
Time constraint | |||
| –time | -t | DD-HH:MM:SS | Time limit for the job |
| –begin | - | YYYY-MM-DDTHH:MM:SS | The job will not run before the specified date |
Logs and notifications | |||
| –signal | - | <sig_num>@<sig_time> | When a job is within sig_time seconds of its end time, send it the signal sig_num |
| –output | -o | <file/path> | File to write the output of the job |
| –error | -e | <file/path> | If specified, the error output will be written in this file |
| –mail-user | - | email@domain.com | User email to receive notification |
| –mail-type | - | NONE, BEGIN, END, FAIL, REQUEUE, ALL also TIME_LIMIT, TIME_LIMIT_90, TIME_LIMIT_80, TIME_LIMIT_50 | Type of notification to receive by email |