bingshuo/guanghulab
1
0
Fork 0
Code Issues 2 Pull Requests Actions 1 Packages Projects Releases Wiki Activity

fix: fetch_train.py -> read train_coder.log

Browse Source
This commit is contained in:
bingshuo 2026-05-20 01:30:46 +08:00
parent 1a06e738f5
commit 6fa14c4978
1 changed files with 6 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

11
homepage/fetch_train.py
View File

@ -7,18 +7,19 @@ SSH = ['sshpass','-p','HkM43lFVUIsc','ssh','-o','StrictHostKeyChecking=no',
'-o','ConnectTimeout=10','root@connect.westd.seetacloud.com','-p','23647']
try:
r = subprocess.run(SSH + ['tail','-2','/root/autodl-tmp/train_mother.log'],
r = subprocess.run(SSH + ['tail','-5','/root/autodl-tmp/train_coder.log'],
capture_output=True, text=True, timeout=25).stdout
sm = re.search(r'(\d+)/12243', r)
sm = re.search(r'(\d+)/(\d+)', r.split(chr(10))[-1] if chr(10) in r[-5:] else r)
lm = re.search(r"'loss':\s*'([\d.]+)'", r)
step = int(sm.group(1)) if sm else 0
total = int(sm.group(2)) if sm else 11838
loss = lm.group(1) if lm else '--'
except:
step, loss = 0, '--'
step, total, loss = 0, 11838, '--'
data = {'step': step, 'total': 12243, 'loss': loss,
data = {'step': step, 'total': total, 'loss': loss,
'updated': datetime.now(timezone.utc).isoformat()+'Z'}
with open(DATA_FILE, 'w') as f:
json.dump(data, f)
print(f"Step {step} / 12243, Loss: {loss}")
print(f"Step {step} / {total}, Loss: {loss}")