ViperEkura/AstrAI
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

fix: prevent signal handler test from racing with training completion

Browse Source 
- Set n_epoch=99999 so training runs until parent delivers signal instead of finishing too fast on CPU
- Drop ready-file deadline from 30s to 10s
This commit is contained in:
ViperEkura
2026-08-07 18:32:08 +08:00
parent 81788faef4
commit 6572be4f98
1 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tests/trainer/test_signal_handler.py
+2 -2
View File
@@ -71,7 +71,7 @@ def _inner_run(batch_per_device, ckpt_interval, ckpt_dir, ready_file):
optimizer_fn=optimizer_fn, optimizer_fn=optimizer_fn,
scheduler_fn=scheduler_fn, scheduler_fn=scheduler_fn,
ckpt_dir=ckpt_dir, ckpt_dir=ckpt_dir,
n_epoch=1, n_epoch=99999,
batch_per_device=batch_per_device, batch_per_device=batch_per_device,
ckpt_interval=ckpt_interval, ckpt_interval=ckpt_interval,
grad_accum_steps=1, grad_accum_steps=1,
@@ -94,7 +94,7 @@ def _spawn_train_and_signal(ckpt_dir, sig, timeout=120):
) )
p.start() p.start()
deadline = time.time() + 30 deadline = time.time() + 10
while time.time() < deadline: while time.time() < deadline:
if os.path.exists(ready_file): if os.path.exists(ready_file):
with open(ready_file) as f: with open(ready_file) as f: