Fix example hpo (#503)

* Use aiaccel-job.

* Add job_config.

* Fix job_config.

* Fix code.

* Fix aiaccel-job command.

* Fix README.

* Remove unnecessary files.

Author: KanaiYuma-aist

examples/hpo/optuna/samplers/coco/README.md

@@ -40,17 +40,21 @@
   - For details, please refer to the following GitHub repository:
     https://github.com/numbbo/coco
 
-- Please replace the virtual environment path in `objective.sh` and the `job_group` in `main_parallel_coco.py` with the appropriate paths and IDs.
+- Please replace the # activate environment and the `job_group` in `job_config.py` with the appropriate commands and IDs.
 - When you run `main_parallel.py`, the validation for each sampler will be executed.
-- The results are saved in `optuna_csv` and `step_csv` under each directory.
 
-cd nelder-mead
-python main_parallel.py
+```
+aiaccel-job pbs --config job_config.yaml gpu --walltime 4:00:00 main_parallel_coco.log -- python3.13 main_parallel_coco.py
+```
+
+- The results are saved in `optuna_csv` and `step_csv` under each directory.
 
 - To run `plot.py`, you need to install pandas and matplotlib.
 
+```
 pip install pandas matplotlib
 python plot.py
+```
 
 ## 3. Checking the Results
 

examples/hpo/optuna/samplers/coco/README_ja.md

@@ -0,0 +1,61 @@
+walltime: "1:0:0"
+
+script_prologue: |
+    echo Job ID: $PBS_JOBID
+    echo Hostname: $(hostname)
+
+    export NVIDIA_VISIBLE_DEVICES=all
+    export JOB_GROUP=job_group
+
+    # activate environment
+
+qsub: "qsub -P $JOB_GROUP -l walltime={args.walltime} -v USE_SSH=1"
+
+cpu:
+    qsub_args: "-q rt_HF -l select=1"
+    job: "{command}"
+
+cpu-array:
+    n_tasks_per_proc: 128
+    n_procs: 24
+    qsub_args: "-q rt_HF -l select=1 -J 1-{args.n_tasks}:$(( {args.n_tasks_per_proc} * {args.n_procs} ))"
+    job: "{command}"
+
+gpu:
+    qsub_args: "-q rt_HF -l select=1"
+    job: "{command}"
+
+gpu-array:
+    n_tasks_per_proc: 1
+    n_procs: 1
+    qsub_args: "-q rt_HF -l select=1 -J 1-{args.n_tasks}:$(( {args.n_tasks_per_proc} * {args.n_procs} ))"
+    job: "CUDA_VISIBLE_DEVICES=$(( LOCAL_PROC_INDEX % 8 )) {command}"
+
+mpi:
+    n_nodes: 1
+    qsub_args: >-
+        -q rt_HF
+        -l select={args.n_nodes}:mpiprocs=$(( {args.n_procs} / {args.n_nodes} )):ompthreads=$(( {args.n_nodes} * 96 / {args.n_procs} ))
+    job: |
+        source /etc/profile.d/modules.sh
+        module load hpcx
+
+        mpirun -np {args.n_procs} -bind-to none -map-by slot \\
+            -mca pml ob1 -mca btl self,tcp -mca btl_tcp_if_include bond0 \\
+            {command}
+
+train:
+    qsub_args: >-
+        -q $( (({args.n_gpus}==1)) && printf rt_HG || printf rt_HF )
+        -l select=$(( ({args.n_gpus} + 7) / 8 )):mpiprocs=$( (({args.n_gpus}==1)) && printf 1 || printf 8 ):ompthreads=$( (({args.n_gpus}==1)) && printf 8 || printf 12 )
+    job: |
+        source /etc/profile.d/modules.sh
+        module load hpcx
+
+        mpirun -np {args.n_gpus} -bind-to none -map-by slot \\
+            -mca pml ob1 -mca btl self,tcp -mca btl_tcp_if_include bond0 \\
+            -x MAIN_ADDR=$(hostname -i) \\
+            -x MAIN_PORT=3000 \\
+            -x COLUMNS=120 \\
+            -x PYTHONUNBUFFERED=true \\
+            {command}

examples/hpo/optuna/samplers/coco/job_config.yaml

@@ -1,13 +1,9 @@
+from concurrent.futures import ThreadPoolExecutor
 from itertools import product
-from pathlib import Path
-
-from aiaccel.job.jobs.abci_job import AbciJob
+import subprocess
 
 
 def main() -> None:
-    job_filename = Path("objective.sh")
-    job_group = "xxx"
-
     sampler_names = ["nelder-mead", "nelder-mead-subTPE", "TPE"]
     func_ids = list(range(1, 25))
     dims = [2, 3, 5, 10, 20, 40]
@@ -19,30 +15,18 @@ def main() -> None:
         sampler_names, func_ids, zip(dims, execute_times, strict=False), zip(instances, optuna_seeds, strict=False)
     )
 
-    for sampler_name, func_id, (dim, execute_time), (instance, optuna_seed) in combinations:
-        execute_time = "0:05:00" if sampler_name == "nelder-mead" else execute_time
-        print(sampler_name, (func_id, execute_time), dim, (instance, optuna_seed))
-        job = AbciJob(
-            job_filename,
-            job_group,
-            qsub_args=[
-                "-l",
-                f"h_rt={execute_time}",
-            ],
-            args=[
-                "--func_id",
-                f"{func_id}",
-                "--dim",
-                f"{dim}",
-                "--instance",
-                f"{instance}",
-                "--optuna_seed",
-                f"{optuna_seed}",
-                "--sampler_name",
-                f"{sampler_name}",
-            ],
-        )
-        job.submit()
+    with ThreadPoolExecutor() as pool:
+        for sampler_name, func_id, (dim, execute_time), (instance, optuna_seed) in combinations:
+            execute_time = "0:05:00" if sampler_name == "nelder-mead" else execute_time
+            print(sampler_name, (func_id, execute_time), dim, (instance, optuna_seed))
+
+            aiaccel_job_command = f"""\
+aiaccel-job pbs --config job_config.yaml cpu --walltime {execute_time} log/job_{func_id}_{dim}_{instance}.log \
+-- python3.13 experiment_coco.py --func_id {func_id} --dim {dim} \
+--instance {instance} --optuna_seed {optuna_seed} --sampler_name {sampler_name}
+"""
+
+            pool.submit(subprocess.run, aiaccel_job_command, shell=True)
 
 
 if __name__ == "__main__":