kwrobel.eth

doc_type

1

What kind of `doc_types` are supported? I have tried `html`, but it is not working.

I am using pinned tabs and when they are discarded and I close the browser and open again they are `about:blank`.

ByT5Tokenizer ignores spaces around added tokens

5

### System Info transformers 4.23.1 ### Who can help? @patrickvonplaten @SaulLu ### Information - [ ] The official example scripts - [ ] My own modified scripts ### Tasks -...

Alpaca was released with wrong license BY and corrected to NC (after 2 days). The correct license (NC) was mentioned in theirs blogpost when releasing the repository.

Do you have plans to train other sizes of mDeBERTa model, e.g. large? It is important for non English texts.

``` automaton = ahocorasick.Automaton() input='👨‍👨‍👦' for c in input: automaton.add_word(c, c) automaton.add_word(input, input) automaton.make_automaton() for end_index, value in automaton.iter_long(input): end_index += 1 start_index = end_index - len(value) print('FOUND', start_index, end_index,...

spec.json combining marks

2

I have checked some data from `spec.json`. There are combining marks under `cm` key and "valid" characters from `groups` (primary and secondary). Some combining marks are in these valid characters...

Speed up inference problems

19

I am trying to speed up benchmarking on A100. Below are times of tests on one task in two versions using Mistral.  Unfortunately, using `torch.compile` and flash_attention slows down...

`adduser` tworzy katalog domowy użytkownika oraz od razu nadaje hasło

While testing gpt3.5-instruct I encounter the error for a few tasks: ``` 2024-05-04:13:39:36,070 INFO [_client.py:1026] HTTP Request: POST https://api.openai.com/v1/completions "HTTP/1.1 429 Too Many Requests" 2024-05-04:13:39:36,070 INFO [_base_client.py:952] Retrying request to...