verbose about libfewshot HOT 6 CLOSED

Thanks for your issue.
In fact, we have already shown train/val/test acc, data time, cal time and loss during training. Do you mean the graph display like Tensorboard?
If you are referring to the tensorboard, we also have saved the tensorboard files in the results directory, which can be viewed using the tensorboard command.

from libfewshot.

You are welcome.
I mean how can I see train/val/test acc, data time, cal time and loss during training or how can I active verbose.
During training I can not see these parameters.and I have to open related "log" files of them in result folder to see them. is there any option or argument to activate it?

from libfewshot.

Hello, we have not met this problem you mentioned yet.
We are wondering if you could check or tell us the following points:

1. Have any modifications been made to the source code or configuration file?
2. Check the information of OS and Python version.
3. Check the environment variables.

from libfewshot.

Thank you.
I am running the code on Google Colab.

from libfewshot.

This problem may caused by some limitation of the Colab(or jupyter), could you try to add a magic %%python before run your cell?

%% python
import sys
sys.dont_write_bytecode = True
from core.config import Config
from core import Trainer

if __name__ == "__main__":
    config = Config("./config/proto.yaml").get_config_dict()
    trainer = Trainer(config)
    trainer.train_loop()

Generally, the LibFewShot should be run via shell commands, could you try run the run_trainer.py by following command in the Colab?

!python3 run_trainer.py

from libfewshot.

Thank you so much.
It works good via shell.

from libfewshot.