When I train my own dataset, it works fine, but console doesn't print the information.

WXinlong / SOLO

SOLO and SOLOv2 for instance segmentation, ECCV 2020 & NeurIPS 2020.

Other

1.69k stars 307 forks source link

When I train my own dataset, it works fine, but console doesn't print the information. #199

Closed zhmou closed 2 years ago

zhmou commented 2 years ago

It just print two lines of mmdet info as the image shown and cursor stuck there until training is over, doesn't presenting any training information(expected behavior like: mmdet - INFO - Epoch [1][xx/xx] lr: xxx , eta: xxx, time: xxx, data_time: xxx, memory: xxx, loss_ins: xxx, loss_cate: xxx, loss: xxx), but actually .pth file is writing to the work_dir properly, I don't know why. please show me how to fix that or give me some helpful instructions. Many thanks~

zhmou commented 2 years ago

well, I got the solution. I forgot to change the interval variable in config file:

log_config = dict(
    interval = 20,
    hooks=[
        dict(type='TextLoggerHook'),
    ])

just make it compatible to your own training dataset, and everything goes well.

DanielWuuuuuu commented 2 years ago

But I haven't changed the procedure at all @zhmou