Mingrui Wu

修改script.py中default_evaluation_params()的参数GT_SAMPLE_NAME_2_ID和DET_SAMPLE_NAME_2_ID

And this is my result and result from your pretrained model. My result seems to non-smooth, is there any possible solution?

I guess that, after the `accelerator.is_main_process` is commented, the multiple processes will execute `save_model_hook` repeatedly. So excute `weights.pop()` under the main process should work.

我遇到一样的问题，升级gcc了也没用

你解决这个问题了吗

我发现这应该只是在mac上的问题，mac对c++内存分配严格，所以c++代码要改，但是我对c++不熟

the same issue.

> @essunny310 similar situation. batch size 2 * accum 16 * cards 8, 15 epoch, lr=1e-3 miou~12.8 batch size 2 * accum 16 * cards 8, 15 epoch, lr=1e-4 miou~32...

Were you train Total-Text successfully？Could you please tell me what changes to code need I do? I want to train ICDAR2019Art dataset, but I find that the coordinates of the...