search

Media-Smart / vedastr

A scene text recognition toolbox based on PyTorch
Apache License 2.0
534 stars 100 forks source link

the saved model size #5

Closed haoyuanz13 closed 4 years ago

haoyuanz13 commented 4 years ago

why the trained model size is 2.0G each .pth file?

ChaseMonsterAway commented 4 years ago

The default pth file will save model and optimizer. The size of my trained model is 500M. Maybe your model is too large.

haoyuanz13 commented 4 years ago

hmm i just clone the code and run the demo config...

ChaseMonsterAway commented 4 years ago

which config you used?

haoyuanz13 commented 4 years ago

which config you used?

resnet_fc.py

ChaseMonsterAway commented 4 years ago

which config you used?

resnet_fc.py

The default code will save model and optimizer, the optimizer is very large. You can turn it off, then the model size will be 600M.

haoyuanz13 commented 4 years ago

which config you used?

resnet_fc.py

The default code will save model and optimizer, the optimizer is very large. You can turn it off, then the model size will be 600M.

alright that makes sense now, by the way, what is the usage of saving the opt params in this case?

ChaseMonsterAway commented 4 years ago

which config you used?

resnet_fc.py

The default code will save model and optimizer, the optimizer is very large. You can turn it off, then the model size will be 600M.

alright that makes sense now, by the way, what is the usage of saving the opt params in this case? Just for resuming.