Roberta Rocca

following up on this cause I stumbled on the same issue (I think) and want to double-check I understand correctly. I need to do hyperparameter optimization + model comparison for...

thanks for the super quick reply and the pointers! :)

Pointer to new potentially interesting library [tokenizers](https://github.com/huggingface/tokenizers), although some tokenization options area will already be covered by the [transformers](https://github.com/huggingface/transformers) wrapper.

Library for co-reference resolution which works as pipeline extension for SpaCy: [neuralcoref](https://github.com/huggingface/neuralcoref). Finds expressions in text that refer to same entity and groups them in clusters. At token level, can...

New pretty cool object recognition model trained with contrastive learning: https://github.com/google-research/simclr. Paper: https://arxiv.org/abs/2002.05709

Auditory English Lexicon Project (auditory lexical norms): haven't taken a close look yet, but for future reference, paper here: https://doi.org/10.3758/s13428-020-01352-0, website with data here: https://inetapps.nus.edu.sg/aelp/

hey thanks for spotting this and sorry for the delay! Re: the dtype problem. Couldn't we simply convert x to tensor _before_ rescaling and cast to `tf.float32` by default, triggering...

hey @anibalsolon - thanks! Couple of considerations: - let's allow users to pass `dtype` for sure (it also needs to be added to the docstring); - I'm a bit torn...

> @rbroc do you remember if we fixed this? I feel like we did but I'm not seeing a PR for it We didn't. I fixed it within the `AudiosetLabelExtractor`...

okay let's maybe leave this open for now, not super high-priority and we can maybe come back to it later on. For reference, the usage case here was the attempt...