Tensorboard use

[aparna-nagarajan-GY]

Hi all,

How can I visualize tensorboard plots during training?
I do not have any folders created after training even after enabling use_tensorboard = True in config file and quite= False.
Please let me know how to access plots.

Thanks,
Aparna

[isu873]

Hi,

I was able to use tensorboard during training.

I'm using the git version of this model, not the pip package.

Run the train script as following (on the git version):

python yolo/lazy.py task=train dataset='coco' use_tensorboard=True task.data.batch_size=2 task.epoch=100 device=cuda name=v9-c_test

Then i went into my runs/train/v9-c_test folder, and inside are some files and folders

runs/train/v9-c_test
|_ lazy.log
|_ lightning_logs
|_ output.log

Run the tensorboard command in this folder tensorboard --logdir=lightning_logs/, then you can reach the dashboard on localhost:6006

If the trainer machine is not yours, but you are on the same network, run tensorboard --logdir=lightning_logs/ --bind_all and then you can reach the dashboard on the trainer machine local IP address and same port, example: 192.168.1.111:6006

Hope this helps you out!

[aparna-nagarajan-GY]

Thanks a lot!

[aparna-nagarajan-GY]

@isu873: how can I get plots in epoch? All my plots have x axis as steps and I tried changing on_epoch = True, but it does not work.

[isu873]

As far as I know, you can't, or at least i don't know how :)

You have an epoch plot within tensorboard

There you can see which step number occurs in which epoch.