NielsRogge

Hi, Thanks for your interest in using LayoutLMv2. You should probably change this line: `model = nn.DataParallel(model).cuda()` to just this: `model = nn.DataParallel(model)`. You can perhaps check out the [official...

As explained in the [docs](https://huggingface.co/transformers/model_doc/layoutlmv2.html) of LayoutLMv2: > If you want to train the model in a distributed environment, make sure to call `synchronize_batch_norm()` on the model in order to...

Yes, you should call it on the base model, i.e. `model.layoutlmv2.synchronize_batch_norm()`. The method itself is defined for the base model only.

You probably first need to call `torch.distributed.init_process_group()` before starting training.

Pinging @lewtun here, he's an expert regarding ONNX.

Hi, I just tried this out in Google Colab, it runs fine. Perhaps you need to re-install transformers using pip.

When cloning the DETR repo, make sure to do % to change the current working directory for all cells:

Hi, > My question was if there’s any good approaches to joining all these labeled words into only one labeled box? This is a very relevant question indeed! It turns...

It can be found here: https://github.com/NielsRogge/Transformers-Tutorials/tree/master/LayoutLMv3

Hi, Thanks for your interest in deformable DETR! I have a separate branch for this (add_deformable_detr). I'll open a PR soon on the Transformers library.