GitLab

We currently save the pretrained_weights of the encoder and decoder in
two separate directories `encoder` and `decoder`. However, for the
`from_pretrained` function to operate with automodels we need to
specify the type of model in the path to the weights.

The path to the encoder/decoder weights is handled by the
`PreTrainedEncoderDecoder` class in the `save_pretrained` function. Sice
there is no easy way to infer the type of model that was initialized for
the encoder and decoder we add a parameter `model_type` to the function.
This is not an ideal solution as it is error prone, and the model type
should be carried by the Model classes somehow.

This is a temporary fix that should be changed before merging.

Since I started my PR the `add_special_token_single_sequence` function
has been deprecated for another; I replaced it with the new function.

Improve global visibility on the run_squad script, remove unused files and fixes related to XLNet.

* license

* changes

* ok

* Update paper link and commands to run

* pointer to uber repo

Create a NER example similar to the Pytorch one. It takes the same options, and can be run the same way.

Co-Authored-By: Piero Molino <w4nderlust@gmail.com>

When evaluating, shouldn't we always use the SequentialSampler instead of DistributedSampler? Evaluation only runs on 1 GPU no matter what, so if you use the DistributedSampler with N GPUs, I think you'll only evaluate on 1/N of the evaluation set. That's at least what I'm finding when I run an older/modified version of this repo.