Nelson Liu
Nelson Liu
**Is your feature request related to a problem? Please describe.** I'd like the ability to set `requires_grad=True` in the optimizer parameter groups. For instance: ``` ... "text_field_embedder": { "token_embedders": {...
Looks like the train time is pretty long on AWS instances with K80s. Adding MultiGPU data parallelism would be a good way to mitigate this (as done in https://www.tensorflow.org/tutorials/using_gpu#using_multiple_gpus)
right now, the data pipeline will tokenize the input into both words / characters, even if you only want words. This is fine for now since character tokenization isn't that...
Right now, the model can "train" (training on train data / periodically measure validation accuracy / loss) and it can "predict" (given an unlabeled test set, make predictions). It would...
SwitchableDropoutWrapper currently has to run the LSTM cell twice, one to get the dropped out inputs and one to get the un-dropped out inputs (and then use `tf.cond` to output...
NER features seem quite important. For example, how would a model distinguish `Does he live in New York` and `Does he live in Newark`; changing named entities can have drastic...
It's annoying to have to write separate scripts for each model.
Right now, setup involves running a bunch of `make` commands and doing some things in between at times (e.g. unzipping the GloVe download). It'd be nice to have a unified...
Hi! I noticed that the `basevocab` SciBERT models aren't in the HuggingFace hub ( https://huggingface.co/allenai )---would it be possible to add them? Thanks!