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

feat: store metric logs inside each checkpoint dir, remove log_dir config

Browse Source
This commit is contained in:
ViperEkura
2026-07-28 00:22:29 +08:00
parent 5ab18bec48
commit d9a0c72149
7 changed files with 7 additions and 22 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tests/trainer/conftest.py
-1
View File
@@ -39,7 +39,6 @@ def create_train_config(
optimizer_fn=optimizer_fn,
scheduler_fn=scheduler_fn,
ckpt_dir=test_dir,
log_dir=os.path.join(test_dir, "logs"),
n_epoch=n_epoch,
batch_per_device=batch_per_device,
ckpt_interval=ckpt_interval,
tests/trainer/test_online_e2e.py
-1
View File
@@ -101,7 +101,6 @@ def test_online_dpo_end_to_end(base_test_env):
optimizer_fn=optimizer_fn,
scheduler_fn=scheduler_fn,
ckpt_dir=os.path.join(test_dir, "ckpt"),
log_dir=os.path.join(test_dir, "logs"),
n_epoch=1,
batch_per_device=2,
ckpt_interval=100,
tests/trainer/test_signal_handler.py
+2 -4
View File
@@ -50,7 +50,7 @@ class _ReadyCallback:
os.fsync(f.fileno())
def _inner_run(batch_per_device, ckpt_interval, ckpt_dir, log_dir, ready_file):
def _inner_run(batch_per_device, ckpt_interval, ckpt_dir, ready_file):
dataset = PicklableDataset()
def model_fn():
@@ -71,7 +71,6 @@ def _inner_run(batch_per_device, ckpt_interval, ckpt_dir, log_dir, ready_file):
optimizer_fn=optimizer_fn,
scheduler_fn=scheduler_fn,
ckpt_dir=ckpt_dir,
log_dir=log_dir,
n_epoch=1,
batch_per_device=batch_per_device,
ckpt_interval=ckpt_interval,
@@ -86,13 +85,12 @@ def _inner_run(batch_per_device, ckpt_interval, ckpt_dir, log_dir, ready_file):
def _spawn_train_and_signal(ckpt_dir, sig, timeout=120):
log_dir = os.path.join(ckpt_dir, "logs")
ready_file = os.path.join(ckpt_dir, "ready.txt")
ctx = mp.get_context("spawn")
p = ctx.Process(
target=_inner_run,
args=(2, 1000, ckpt_dir, log_dir, ready_file),
args=(2, 1000, ckpt_dir, ready_file),
)
p.start()