popov/gogo2
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix dash actions

Browse Source
This commit is contained in:
Dobromir Popov
2025-07-04 02:24:18 +03:00
parent cf91e090c8
commit d0cf04536c
7 changed files with 309 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
utils/checkpoint_manager.py
View File

@ -124,7 +124,7 @@ class CheckpointManager:
self._rotate_checkpoints(model_name)
self._save_metadata()
logger.info(f"Saved checkpoint: {checkpoint_id} (score: {performance_score:.4f})")
logger.debug(f"Saved checkpoint: {checkpoint_id} (score: {performance_score:.4f})")
return metadata
except Exception as e:
@ -232,7 +232,7 @@ class CheckpointManager:
# Save more frequently during active training (every 5th attempt instead of 10th)
if random.random() < 0.2: # 20% chance to save anyway
logger.info(f"Saving checkpoint for {model_name} - periodic save during active training")
logger.debug(f"Saving checkpoint for {model_name} - periodic save during active training")
return True
return False
@ -268,7 +268,7 @@ class CheckpointManager:
file_path = Path(checkpoint.file_path)
if file_path.exists():
file_path.unlink()
logger.info(f"Rotated out checkpoint: {checkpoint.checkpoint_id}")
logger.debug(f"Rotated out checkpoint: {checkpoint.checkpoint_id}")
except Exception as e:
logger.error(f"Error removing rotated checkpoint {checkpoint.checkpoint_id}: {e}")