zhuchen03 / FreeLB

Adversarial Training for Natural Language Understanding

zhuchen03/FreeLB Issues

NaN encounted if FreeLB is used at the beginning of finetune stage
Updated 2 years ago
Some confusion about the detach operation and embeds_init
Updated 3 years ago
Errors generated during data preprocessing
Closed 3 years ago2
词向量空间的不变性
Closed 3 years ago
API Key
Closed 4 years ago1
A few questions about FreeLB and dropout
Closed 4 years ago2
the "dp_mask" problem
Closed 4 years ago4
Having issues with training RoBERTa. Loss not decreasing
Closed 4 years ago2
FreeLB-RoBERTa within HuggingFace's transformers?
Closed 4 years ago1
ImportError: cannot import name 'glue_criterion_metrics' from 'transformers'
Closed 4 years ago6
Reproducing results from the paper with roberta using fairseq
Closed 4 years ago5
Is it still working with update_freq > 1?
Closed 4 years ago2
FreeLB didn't use the original training samples?
Closed 4 years ago2
Would you please release the hyper-parameters for FreeLB based on ALBERT(hugging-face)
Closed 4 years ago6
Does anyone meet the Nan error during the end epochs of training?
Closed 4 years ago5
Regarding the release of FreeLB ^_^
Updated 4 years ago
'AlbertForSequenceClassification' object has no attribute 'encoder'
Closed 4 years ago2
Could you add some comments in the code？
Updated 4 years ago1