Issue resuming training on tansformer based NER

I'm using the nightly version, I have successfully trained a transformer based NER model and saved it; now I'm trying to resume training on it.

Firstly, I'm not sure if I have set up the config file correctly, the relevant part looks like this:

```
[components]

[components.ner]
# This is the path to my trained model
source='best-model'

[components.ner.model]
@architectures = "spacy.TransitionBasedParser.v1"
state_type = "ner"
extra_state_tokens = false
hidden_width = 64
maxout_pieces = 2
use_upper = false
nO = null

[components.ner.model.tok2vec]
@architectures = "spacy-transformers.TransformerListener.v1"
grad_factor = 1.0
pooling = {"@layers":"reduce_mean.v1"}

[components.transformer]
# This is the path to my trained model
source='best-model'

[components.transformer.model]
@architectures = "spacy-transformers.TransformerModel.v1"
name = "dccuchile/bert-base-spanish-wwm-uncased"

[components.transformer.model.get_spans]
@span_getters = "spacy-transformers.strided_spans.v1"
window = 128
stride = 96

[components.transformer.model.tokenizer_config]
use_fast = true
```

Now, after trying to train like this:
```!python -m spacy train 'config.cfg' --output='model_t' --gpu-id=0 --paths.train train.spacy --paths.dev test.spacy ```

I'm getting this error message:

```
2020-10-29 14:36:11.541313: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library libcudart.so.10.1
ℹ Using GPU: 0

=========================== Initializing pipeline ===========================
Set up nlp object from config
Pipeline: ['transformer', 'ner']
Resuming training for: ['ner', 'transformer']
Created vocabulary
Finished initializing nlp object
Initialized pipeline components: []
✔ Initialized pipeline

============================= Training pipeline =============================
ℹ Pipeline: ['transformer', 'ner']
ℹ Initial learn rate: 0.0
E # LOSS TRANS... LOSS NER ENTS_F ENTS_P ENTS_R SCORE
--- ------ ------------- -------- ------ ------ ------ ------
⚠ Aborting and saving the final best model. Encountered exception: CUDA
out of memory. Tried to allocate 94.00 MiB (GPU 0; 15.75 GiB total capacity;
13.81 GiB already allocated; 78.88 MiB free; 14.34 GiB reserved in total by
PyTorch)
✔ Saved pipeline to output directory
model_t2/model-last
Traceback (most recent call last):
File "/usr/lib/python3.6/runpy.py", line 193, in _run_module_as_main
"__main__", mod_spec)
File "/usr/lib/python3.6/runpy.py", line 85, in _run_code
exec(code, run_globals)
File "/usr/local/lib/python3.6/dist-packages/spacy/__main__.py", line 4, in
setup_cli()
File "/usr/local/lib/python3.6/dist-packages/spacy/cli/_util.py", line 65, in setup_cli
command(prog_name=COMMAND)
File "/usr/local/lib/python3.6/dist-packages/click/core.py", line 829, in __call__
return self.main(*args, **kwargs)
File "/usr/local/lib/python3.6/dist-packages/click/core.py", line 782, in main
rv = self.invoke(ctx)
File "/usr/local/lib/python3.6/dist-packages/click/core.py", line 1259, in invoke
return _process_result(sub_ctx.command.invoke(sub_ctx))
File "/usr/local/lib/python3.6/dist-packages/click/core.py", line 1066, in invoke
return ctx.invoke(self.callback, **ctx.params)
File "/usr/local/lib/python3.6/dist-packages/click/core.py", line 610, in invoke
return callback(*args, **kwargs)
File "/usr/local/lib/python3.6/dist-packages/typer/main.py", line 497, in wrapper
return callback(**use_params) # type: ignore
File "/usr/local/lib/python3.6/dist-packages/spacy/cli/train.py", line 59, in train_cli
train(nlp, output_path, use_gpu=use_gpu, stdout=sys.stdout, stderr=sys.stderr)
File "/usr/local/lib/python3.6/dist-packages/spacy/training/loop.py", line 105, in train
raise e
File "/usr/local/lib/python3.6/dist-packages/spacy/training/loop.py", line 85, in train
for batch, info, is_best_checkpoint in training_step_iterator:
File "/usr/local/lib/python3.6/dist-packages/spacy/training/loop.py", line 201, in train_while_improving
score, other_scores = evaluate()
File "/usr/local/lib/python3.6/dist-packages/spacy/training/loop.py", line 253, in evaluate
scores = nlp.evaluate(dev_examples)
File "/usr/local/lib/python3.6/dist-packages/spacy/language.py", line 1312, in evaluate
docs = list(docs)
File "/usr/local/lib/python3.6/dist-packages/spacy/util.py", line 1363, in _pipe
yield from proc.pipe(docs, **kwargs)
File "spacy/pipeline/transition_parser.pyx", line 170, in pipe
File "/usr/local/lib/python3.6/dist-packages/spacy/util.py", line 1322, in minibatch
batch = list(itertools.islice(items, int(batch_size)))
File "/usr/local/lib/python3.6/dist-packages/spacy/util.py", line 1363, in _pipe
yield from proc.pipe(docs, **kwargs)
File "/usr/local/lib/python3.6/dist-packages/spacy_transformers/pipeline_component.py", line 173, in pipe
self.set_annotations(subbatch, self.predict(subbatch))
File "/usr/local/lib/python3.6/dist-packages/spacy_transformers/pipeline_component.py", line 189, in predict
activations = self.model.predict(docs)
File "/usr/local/lib/python3.6/dist-packages/thinc/model.py", line 312, in predict
return self._func(self, X, is_train=False)[0]
File "/usr/local/lib/python3.6/dist-packages/spacy_transformers/layers/transformer_model.py", line 111, in forward
tensors, bp_tensors = transformer(token_data, is_train)
File "/usr/local/lib/python3.6/dist-packages/thinc/model.py", line 288, in __call__
return self._func(self, X, is_train=is_train)
File "/usr/local/lib/python3.6/dist-packages/thinc/layers/pytorchwrapper.py", line 79, in forward
Ytorch, torch_backprop = model.shims[0](Xtorch, is_train)
File "/usr/local/lib/python3.6/dist-packages/thinc/shims/pytorch.py", line 29, in __call__
return self.predict(inputs), lambda a: ...
File "/usr/local/lib/python3.6/dist-packages/thinc/shims/pytorch.py", line 38, in predict
outputs = self._model(*inputs.args, **inputs.kwargs)
File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py", line 722, in _call_impl
result = self.forward(*input, **kwargs)
File "/usr/local/lib/python3.6/dist-packages/transformers/modeling_bert.py", line 762, in forward
output_hidden_states=output_hidden_states,
File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py", line 722, in _call_impl
result = self.forward(*input, **kwargs)
File "/usr/local/lib/python3.6/dist-packages/transformers/modeling_bert.py", line 439, in forward
output_attentions,
File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py", line 722, in _call_impl
result = self.forward(*input, **kwargs)
File "/usr/local/lib/python3.6/dist-packages/transformers/modeling_bert.py", line 388, in forward
intermediate_output = self.intermediate(attention_output)
File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py", line 722, in _call_impl
result = self.forward(*input, **kwargs)
File "/usr/local/lib/python3.6/dist-packages/transformers/modeling_bert.py", line 333, in forward
hidden_states = self.intermediate_act_fn(hidden_states)
File "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py", line 1369, in gelu
return torch._C._nn.gelu(input)
**RuntimeError: CUDA out of memory. Tried to allocate 94.00 MiB (GPU 0; 15.75 GiB total capacity; 13.81 GiB already allocated; 78.88 MiB free; 14.34 GiB reserved in total by PyTorch)**
```

I understand the message is telling me I'm out of memory, but it seems weird that I'm able to train from scratch with no issues but getting this error when trying to resume training on the saved model. Any help is appreciated.

## Your Environment
- **spaCy version:** 3.0.0rc2
- **Platform:** Linux-4.19.112+-x86_64-with-Ubuntu-18.04-bionic
- **Python version:** 3.6.9
- **Pipelines:** es_core_news_md (3.0.0a0), es_dep_news_trf (3.0.0a0)