Hsu Wan Ting

@samanthawyf I just used the command line `ulimit -n 65535` and the error was gone.

@Ineedangoodidea I got a similar zip error. And @WJTUz 's solution does not work for me. I fixed this error by changing 2 lines in `evaluation/totaltext/e2e/prepare_results.py` from https://github.com/MhLiao/MaskTextSpotterV3/blob/9bd595e40c693ced0dc8ce4f33d65048424189e5/evaluation/totaltext/e2e/prepare_results.py#L135-L136 to ```python...

@MhLiao I can't find `synthtext/train_list.txt` either. Or did you just use all SynthText data as training data?

Hi, I just used all images as training data and generated the `train_list.txt` by myself.

Hi, You should use the "evalall" mode for decoding. "eval" mode will do the evaluation during training. So it will always find the latest trained model. "Evalall" mode will load...

Yes, if you don't give a pretrained model path, it will use the latest checkpoint as default.

> The console always outputs recall:0, ratio:1, thres:0 This is normal for **the beginning of the training**. After every training step, I will check the ability of the model. The...

Do you mean the warning (`fail to reach target recall: 0.9`) is gone after 10 training steps? If yes, it's normal. I forget the exact number of training steps when...

Hi, I didn't try the code with a higher TensorFlow version. If you want to run with a higher version, you may need to check the difference between Tensoflow 1.1.0...

Hi @tjruwase , > You can directly pass the torch [implementation](https://pytorch.org/docs/stable/generated/torch.optim.lr_scheduler.CosineAnnealingLR.html) into deepspeed.initialize() as documented [here](https://deepspeed.readthedocs.io/en/latest/initialize.html#training-initialization). For this option, do you mean like the following? ``` model, optimizer, _, lr_scheduler...