row 1.0.0

Row is a command line tool that helps you manage workflows on HPC resources.
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90

# resources

`action.resources`: **table** - Defines the resources that the action will consume.
**Row** passes this information to the scheduler when you submit a job.

Example:
```toml
[action.resources]
processes.per_submission = 10
threads_per_process = 4
walltime.per_submission = "04:00:00"
```
> [!TIP]
> You may omit `[action.resources]` entirely.

## processes

`action.resources.processes`: **table** - Set the number of processes this action
will execute on (launched by `mpi` or similarly capable launcher). The table **must**
have one of two keys: `per_submission` or `per_directory`.

Examples:
```toml
processes.per_submission = 16
```
```toml
processes.per_directory = 8
```

When set to `per_submission`, **row** asks the scheduler to allocate the given
number of processes for each job. When set to `per_directory`, **row** requests the
given value multiplied by the number of directories in the submission group. Use
`per_submission` when your action loops over directories and reuses the same processes
to perform computations. Use `per_directory` when your action parallelizes over the
directories and therefore requires `processes.per_directory * num_directories` total
processes.

When omitted, `processes` defaults to `per_submission = 1`.

## threads_per_process

`action.resources.threads_per_process`: **positive integer** - The number of CPU threads
your action utilizes per process. When omitted, **row** does not make any specific
request for threads from the scheduler. Most schedulers default to 1 thread per process
in this case.

## gpus_per_process

`action.resources.gpus_per_process`: **positive integer** - The number of GPUs your
action utilizes per process. When omitted, **row** does not make any specific request
for GPUs from the scheduler. Most schedulers default to 0 GPUs per process in this case.

## walltime

`action.resources.walltime`: **table** - Set the walltime that this action takes to
execute. The table **must** have one of two keys: `per_submission` or `per_directory`.
Valid walltime strings include `"HH:MM:SS"`, `"D days, HH:MM:SS"`, and all other valid
`Duration` formats parsed by [speedate](https://docs.rs/speedate/latest/speedate/).

Examples:
```toml
walltime.per_submission = "4 days, 12:00:00"
```
```toml
walltime.per_directory = "00:10:00"
```

When set to `per_submission`, **row** asks the scheduler to allocate the given walltime
for each job. When set to `per_directory`, **row** requests the given value multiplied
by the number of directories in the submission group. Use `per_submission` when your
action parallelizes over directories and therefore takes the same amount of time
independent of the submission group size. Use `per_directory` when your action loops
over the directories and therefore the walltime scales with the number of directories.

When omitted, `walltime` defaults to `per_directory = 01:00:00`.

## memory_per_cpu_mb

`action.resources.memory_per_cpu_mb`: **integer** - The number of megabytes of memory to
request per CPU (thread). `memory_per_cpu_mb` is used when `gpus_per_process` is unset.

> [!IMPORTANT]
> Most users should omit `memory_per_cpu_mb`. When omitted **row** will
> automatically select the maximum amount of memory possible without incurring
> extra charges on your HPC resources.

## memory_per_gpu_mb

Similar to `memory_per_cpu_mb`, but for GPUs. `memory_per_gpu_mb` is used when
`gpus_per_process` is set.