Jack Kwok

Hi, I tried (the only config change I made was pointing to my own custom coco JSON files): ``` ./tools/dist_train.sh configs/mask2former/my_custom_mask2former_swin-s-p4-w7-224_lsj_8x2_50e_coco.py 8 ``` For the last 30 min, it is...

@roboserg : No actually, do you have the same issue?

@roboserg : Thanks for letting me know. It's good to know I am not the only one facing the issue. Do you mind posting your environment info?

Any solution? I am running in all NaN predictions as well. I trained on the kaggle toxic training set for 6 epochs. I used 1 GPU, no apex. ``` learner...

Not sure if this will help solve the original issue but my issue was solved by lower the learning rate to 1e-4.

Hi, I am new to this repo and I face the same issue. It appears new_vocab_list is not supported int he fine-tuning step. See: https://github.com/neuspell/neuspell/blob/master/neuspell/corrector_subwordbert.py#L81 Seems like the only option...

Good job on this. I would love to see iOS 7 look and feel.

I am also eagerly waiting for Tensorflow Make3D model.

I am happy to hear you've found my library useful! I didn't find recursive nesting to be a popular use case so I didn't support it.

Hi Bruno, Thanks! You may need to subclass both of them. What kind of customizations do you want to do? Jack