Datasets

Name	🤗	Tested
BERT	`bert-large-uncased-whole-word-masking-finetuned-squad`
DistilBERT	`distilbert-base-uncased-distilled-squad`

Name	🤗	Tested
ALBERT	`mfeb/albert-xxlarge-v2-squad2`
BERT	`deepset/bert-large-uncased-whole-word-masking-squad2`
ELECTRA	`deepset/electra-base-squad2`
MiniLM	`deepset/minilm-uncased-squad2`
RoBERTa	`deepset/roberta-base-squad2`
XLNet	`jkgrad/xlnet-base-squadv2`

QQP is a paraphrase identification dataset of two classes, contains 390965 examples, and is part of the glue benchmark.

Name	🤗	`lgxa`	`lig`	`lime`	`occ`	`svs`
ALBERT	`textattack/albert-base-v2-QQP`
BERT	`textattack/bert-base-uncased-QQP`
ELECTRA	`howey/electra-base-qqp`
XLNet	`textattack/xlnet-base-cased-QQP`

trec is a question classification dataset with 6 classes.

Name	🤗	Tested
BERT	`aychang/bert-base-cased-trec-coarse`
DistilBERT	`aychang/distilbert-base-cased-trec-coarse`

SST-2 is a sentiment analysis dataset with 2 classes and part of the glue benchmark.
There are no labels available for the test set.

Name	🤗	`lgxa`	`lig`	`lime`	`occ`	`svs`
ALBERT (`albert`)	`textattack/albert-base-v2-SST-2`
BERT (`bert`)	`textattack/bert-base-uncased-SST-2`
ELECTRA (`electra`)	`howey/electra-base-sst2`
RoBERTa (`roberta`)	`textattack/roberta-base-SST-2`
XLNet (`xlnet`)	`textattack/xlnet-base-cased-SST-2`

Apply MNLI-trained models (already in-place)

Apply QQP-trained models (see table above)

Provide feedback