zw

It works for me! Thank you!

By the way, this issue seems to be only related to BEVDet training. I've also tried to train CenterPoint with this repo using distributed training (no other modification needed), and...

That is not the reason why the program cannot run under distributed training setting. Please read my issue description carefully. Simply delete the registered swin transformer in mmdet (maybe it...

Also , I see the source code of InceptionV3. I found this function > convert_all_kernels_in_model And I use it before load weight `model=InceptionV3(include_top=False, weights=None, input_tensor=None, input_shape=None, pooling=None) convert_all_kernels_in_model(model) model.load_weights(weight_path+'/inception_v3_weights_tf_dim_ordering_tf_kernels_notop.h5')` However,...

Now I can download weight from[ https://github.com/fchollet/deep-learning-models/releases/download/v0.5/inception_v3_weights_tf_dim_ordering_tf_kernels_notop.h5:](url) But I still get: > TypeError: Indexing elements must be in increasing order

I too send a e-mail to [email protected] via my student email. But I havn't got the answer so far. Could you please share leftImg8bit_trainvaltest_foggy.zip through google drive or baiduyun?

And by the way, I think this issue is more severe when lr scheduler steps in each iteration instead of each epoch. I know that timm's default behaviour is to...

Do you mean this? ``` import torch_xla.debug.metrics as met print(met.metrics_report()) ``` It seems for now there is not a plausible solution for this bug now...

Yes, I do see some differences. But I have never used this feature before and don't know how to read this file. Could you please help me on that? Here...

Hi, It has been 11days. Have you got any chance to look at these files?