This commit is contained in:
2026-08-14 19:29:52 +09:00
parent 79be9e3ed3
commit 8e8136d562
+25 -121
View File
@@ -1,47 +1,27 @@
import os
import traceback
import multiprocessing as mp
import shutil
import bo
import ptycho
def run_ptycho_worker(
worker_id,
gpu_token,
job_config,
metric,
run_id,
result_queue,
):
# This process, and MATLAB launched from it,
# can see exactly one GPU.
def run_ptycho_worker(worker_id, gpu_token, job_config, metric, run_id, result_queue):
# This process, and MATLAB launched from it, can see exactly one GPU.
os.environ["CUDA_VISIBLE_DEVICES"] = gpu_token
try:
PTYCHOENGINE = ptycho.engines[job_config["ptycho"]["engine"]]
ptycho_engine = PTYCHOENGINE(job_config)
ptycho_engine = ptycho.engines[job_config["ptycho"]["engine"]](job_config)
ptycho_engine.run(run_id=run_id)
y_value = ptycho_engine.metric(metric)
result_queue.put(
(worker_id, y_value, None)
)
result_queue.put((worker_id, y_value, None))
except Exception:
result_queue.put(
(worker_id, None, traceback.format_exc())
)
result_queue.put((worker_id, None, traceback.format_exc()))
def run_batch(
ctx,
gpu_tokens,
job_configs,
metric,
iteration,
):
def run_batch(ctx, gpu_tokens, job_configs, metric, iteration):
result_queue = ctx.Queue()
processes = []
@@ -51,25 +31,12 @@ def run_batch(
p = ctx.Process(
target=run_ptycho_worker,
args=(
i,
gpu_tokens[i],
job_config,
metric,
run_id,
result_queue,
),
args=(i, gpu_tokens[i], job_config, metric, run_id, result_queue),
)
p.start()
processes.append(p)
# Four jobs are now running concurrently.
results = [
result_queue.get()
for _ in processes
]
results = [result_queue.get() for _ in processes]
# Synchronization barrier.
for p in processes:
@@ -80,111 +47,48 @@ def run_batch(
for worker_id, _, error in results:
if error is not None:
raise RuntimeError(
f"Ptycho worker {worker_id} failed:\n{error}"
)
raise RuntimeError(f"Ptycho worker {worker_id} failed:\n{error}")
return [y_value for _, y_value, _ in results]
def sobo_pipeline(config):
result_dir = config["io"]["result_dir"]
if os.path.exists(result_dir):
shutil.rmtree(result_dir)
os.makedirs(result_dir, exist_ok=True)
RANDOM_ITERS = config["job"].get("random_iters", 0)
SOBO_ITERS = config["job"].get("sobo_iters", 0)
METRIC = config["bo"]["metric"]
BO_BATCH = config["bo"]["batch"]
# SLURM should expose the four GPUs allocated to this job.
visible = os.environ.get("CUDA_VISIBLE_DEVICES")
if visible is None:
raise RuntimeError(
"CUDA_VISIBLE_DEVICES is not set"
)
gpu_tokens = [
token.strip()
for token in visible.split(",")
if token.strip()
]
raise RuntimeError("CUDA_VISIBLE_DEVICES is not set")
gpu_tokens = [token.strip() for token in visible.split(",") if token.strip()]
if len(gpu_tokens) < BO_BATCH:
raise RuntimeError(
f"Expected {BO_BATCH} allocated GPUs, got {len(gpu_tokens)}"
)
raise RuntimeError(f"Expected {BO_BATCH} allocated GPUs, got {len(gpu_tokens)}")
# Explicitly use spawn for CUDA / MATLAB isolation.
ctx = mp.get_context("spawn")
# ---------------------------------------------------------
# Random sampling
# ---------------------------------------------------------
############################ RANDOM SAMPLING ###############################
randombo = bo.RandomBOEngine(config)
for j in range(RANDOM_ITERS):
print(
f"RANDOM sampling; iteration {j}",
flush=True,
)
print(f"RANDOM sampling; iteration {j}")
job_configs = randombo.ask(n=BO_BATCH)
y_values = run_batch(ctx=ctx, gpu_tokens=gpu_tokens, job_configs=job_configs, metric=METRIC, iteration=j)
for job_config, y_value in zip(job_configs, y_values):
randombo.tell(job_config, y_value)
y_values = run_batch(
ctx=ctx,
gpu_tokens=gpu_tokens,
job_configs=job_configs,
metric=METRIC,
iteration=j,
)
# Only the parent touches BO state / train_x / train_y.
for job_config, y_value in zip(
job_configs,
y_values,
):
randombo.tell(
job_config,
y_value,
)
# ---------------------------------------------------------
# SOBO
# ---------------------------------------------------------
############################ SOBO SAMPLING ###############################
sobo = bo.SingleObjectiveBOEngine(config)
sobo.train_x = randombo.train_x
sobo.train_y = randombo.train_y
for j in range(SOBO_ITERS):
iteration = RANDOM_ITERS + j
print(
f"SOBO sampling; iteration {iteration}",
flush=True,
)
for j in range(RANDOM_ITERS, SOBO_ITERS+RANDOM_ITERS):
print(f"SOBO sampling iteration {j}")
job_configs = sobo.ask(n=BO_BATCH)
y_values = run_batch(ctx=ctx, gpu_tokens=gpu_tokens, job_configs=job_configs, metric=METRIC, iteration=j)
y_values = run_batch(
ctx=ctx,
gpu_tokens=gpu_tokens,
job_configs=job_configs,
metric=METRIC,
iteration=iteration,
)
for job_config, y_value in zip(
job_configs,
y_values,
):
sobo.tell(
job_config,
y_value,
)
for job_config, y_value in zip(job_configs, y_values):
sobo.tell(job_config, y_value)