Nelson Liu

Hi @umitkacar , just a heads up that I haven't had a chance to look into building PyTorch 2.0 binaries / if there'd be any major changes for the K40s....

This also occurs with a recurring, all-day events. Things are coerced to naive timezones here https://github.com/irgangla/icalevents/blob/master/icalevents/icalparser.py#L304-L315 , but the `rule.between` here https://github.com/irgangla/icalevents/blob/master/icalevents/icalparser.py#L356 fails because after is time-zone naive, whereas the...

Hi! I wrote a script to reproduce the T5 results here https://gist.github.com/nelson-liu/a4e49549ff14b9f38b657576eba119b6 , but I haven't had a chance to test it yet (and thus haven't merged it). Maybe you...

ah, yes, you're right. you could rewrite that for-block as: ```python for i, generated_sequence in enumerate(outputs): input_ids = inputs[i].ids text = tokenizer.decode(generated_sequence, skip_special_tokens=True, clean_up_tokenization_spaces=True) responses.append(text) ``` sorry about that

fwiw, it seems like Ray provides a distributed multiprocessing.Pool (https://docs.ray.io/en/latest/ray-more-libs/multiprocessing.html), and it's possible that implementing this could be as straightforward as replacing the `multiprocessing.Pool` in the `LocalExecutor` with `ray.util.multiprocessing.Pool`.

@guipenedo I added an implementation of a Ray executor at #331