Dox: add example for defining "gpu_card"

f67c4cf2 · payno · 4a6836b2 · f67c4cf2 · f67c4cf2
Commit f67c4cf2 authored 1 year ago by payno
--- a/doc/tutorials/hello_world.ipynb
+++ b/doc/tutorials/hello_world.ipynb
@@ -32,7 +32,7 @@
    "* `walltime` (str): walltime ( <=> `-t` option in slurm)\n",
    "* `python_venv` (str): path to the python environment to source before executing script lines\n",
    "* `modules` (tuple of str): module names to load before executing script lines\n",
-    "* `sbatch_extra_params`: dict: possible extra parameter to be used when cal the 'sbatch' command. For now handles 'export' only"
+    "* `sbatch_extra_params`: dict: possible extra parameter to be used when calling the 'sbatch' command. For now handles 'export' and 'gpu_card'"
   ]
  },
  {
@@ -80,9 +80,6 @@
  }
 ],
 "metadata": {
-  "nbsphinx": {
-   "allow_errors": true
-  },  
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
@@ -99,6 +96,9 @@
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.3"
+  },
+  "nbsphinx": {
+   "allow_errors": true
  }
 },
 "nbformat": 4,

 %% Cell type:markdown id: tags:

 # How to launch a shell script to slurm using sluurp

 The class to launch a shell script to slurm is `SBatchScriptJob` and with it you must provide:
 * `slurm_config`: configuration to slurl
 * `script`: script to launch
 * `script_path`: where to save the script (if clean_script is set to True) then this script will be removed once processing is done. It can be useful to keep it for debugging
 * `working_directory`: working directory to provide to `SBATCH` when executing the script

 Once the job is created you can submit it to slurm using the `submit` function from `sluurp.executor`

 %% Cell type:markdown id: tags:

 ## slurm_config

 `slurm_config` must be provided as a dictionary and can contain the following keys:

 * `cpu-per-task` (int): define number of cpu to require for each task
 * `n_tasks` (int): define th enumber of tasks
 * `memory` (int): memory in GB ( <=> `--mem` option in slurm)
 * `partition` (str): partition to be used
 * `n_gpus` (int): number of gpu to use
 * `job_name` (str): name of the job ( <=> `-J` option in slurm)
 * `walltime` (str): walltime ( <=> `-t` option in slurm)
 * `python_venv` (str): path to the python environment to source before executing script lines
 * `modules` (tuple of str): module names to load before executing script lines
-* `sbatch_extra_params`: dict: possible extra parameter to be used when cal the 'sbatch' command. For now handles 'export' only
+* `sbatch_extra_params`: dict: possible extra parameter to be used when calling the 'sbatch' command. For now handles 'export' and 'gpu_card'

 %% Cell type:markdown id: tags:

 ## example

 %% Cell type:code id: tags:

 ``` python
 import os
 import tempfile
 from sluurp.executor import submit
 from sluurp.job import SBatchScriptJob

 sleep_time = 10

 with tempfile.TemporaryDirectory() as folder:

    job = SBatchScriptJob(
        script=(f"sleep {sleep_time}s", "echo 'succeed'"),
        slurm_config={
            "partition": "nice",
            "memory": 32,
            "job_name": "hello_world",
        },
        script_path=os.path.join(folder, "script.sh"),
        working_directory=folder,
    )

    submit(job)
 ```

 %% Cell type:code id: tags:

 ``` python
 ```

--- a/example/sleeps.py
+++ b/example/sleeps.py
@@ -15,7 +15,7 @@ slurm_config = {
    # "job_name": "test sluurp",
    # "python_venv": "/scisoft/tomotools/activate dev", # optional python environement to source before script command
    # "modules": ("tomotools", ), # optional module to source before script command
-    # "sbatch_extra_params": {"export": "NONE"}, # extra parameters to provide to sbatch like 'export'
+    # "sbatch_extra_params": {"export": "NONE", "gpu_card": "a40"}, # extra parameters to provide to sbatch like 'export'
 }

 # step 1: launch the jobs