I meet an exception when I run this code: pytorch_lightning.utilities.exceptions.MisconfigurationException: You requested GPUs: [0] But your machine only has: []

$ ./run_pl.sh Some weights of the model checkpoint at bert-base-cased were not used when initializing BertForTokenClassification: ['cls.predictions.bias', 'cls.predictions.transform.dense.weight', 'cls.predictions.transform.dense.bias', 'cls.predictions.decoder.weight', 'cls.seq_relationship.weight', 'cls.seq_relationship.bias', 'cls.predictions.transform.LayerNorm.weight', 'cls.predictions.transform.LayerNorm.bias']

• This IS expected if you are initializing BertForTokenClassification from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPretraining model).
• This IS NOT expected if you are initializing BertForTokenClassification from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model). Some weights of BertForTokenClassification were not initialized from the model checkpoint at bert-base-cased and are newly initialized: ['classifier.weight', 'classifier.bias'] You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.

I run the run_pl_ner.py and meet an error in： File "run_pl_ner.py", line 34

self.token_classification_task: TokenClassificationTask = token_classification_task_clazz()

SyntaxError: invalid syntax

How should I deal with it？

As described in the paper, the training set and test set are modified in the tag set extension experiment. But what about the development set? If it is masked the same as the training set, then the support set cannot be sampled when testing. If it is masked the same as the test set, then how to pick up checkpoints during training? I am confused with this setting, could you please give me some advice and I would be very grateful.

Hi! Thank you for this repository.

I have a question: is it possible use this code to predict (extract) entities in unlabeled data? I'm building an application and I would like to use few-shot NER to extract entities based on a small support set. However, since it's a real world app, the "test" data won't have any annotations.

If you could give me some guidance and examples about this possibility I would really appreciate it.

Thank you!

From the original paper

We project token-level representations obtained from the BERT embedders onto a 2-dimentional space using t-SNE.

And the paper claims that Figure 3 shows the usefulness of pretraining on OntoNotes by showing more compact clusters. However, as the word embeddings returned by transformer model are contextualized, I am wondering how you get the embeddings of individual tokens in the test set and then apply the t-SNE technique. Do you obtain all of the embeddings and then do the average?

Additionally, I could not find the associated code for visualizing embeddings. Would it be possible the code to obtain Figure 3 provided?

Hi,

Thank you for publishing the source code. I have access to OntoNotes and try to reproduce the results in your paper.

In the case of 1-shot learning with NNshot, I could obtain the same average score, i.e., 33.2, although the scores of each dataset are different from yours. However, I'm struggling with Structshot. My scores on CoNLL is much lower than yours, resulting in a lower average score on the three datasets.

In the case of 5-shot learning, I couldn't obtain as high scores as yours both with NNshot and Structshot. In average, Structshot produced worse performance than NNshot.

I wonder if you use a different set of parameters than the one you publish here. I hope you can help me to shed some light here.

Thank you very much for your time!

Best regards, Nhung