Nathan Fradet

Cache size limit for generation

3

# What does this PR do? Following #20767, it adds a `cache_limit` argument for `generate` for PyTorch and TensorFlow (except xla), limiting the size of the cache (`past_key_values`). `position_ids` is...

GenerationConfig argument for Seq2SeqTrainer / Seq2SeqTrainingArgument

5

### Feature request 👋 The request is for a way to pass a `GenerationConfig` to a `Seq2SeqTrainer` (through `Seq2SeqTrainingArguments`). ### Motivation ATOW, `Seq2SeqTrainer` only supports a few arguments for generation:...

Seq2seq trainer generation config arg

9

# What does this PR do? `Seq2SeqTrainer` can load a `GenerationConfig`, by calling the `from_pretrained` method. This is done with the `generation_config_from_pretrain` argument from `Seq2SeqTrainingArguments` (or in `kwargs` of the...

Refactor of PyTorch data loading classes and methods

1

Following the discussions in #147, this PR implements a new PyTorch `Dataset` class, with the particularity of splitting the MIDI itself instead of splitting the tokens as done previously. ----...

`create_duration_tuples` can create overlapping tuples

1

Tokenization tests with this feature are not passing right now, `_time_ticks_to_tokens` have to be adapted to handle overlapping resolutions ---- 📚 Documentation preview 📚: https://miditok--144.org.readthedocs.build/en/144/

[Metrics] Can't find / acquire lock files in distributed multi-node shared file system

4

Hello, I'm facing an issue when calling `.compute` in distributed multi-node setting. The symptoms are the same as in huggingface/datasets#4420 , however I'm not sure the cause is the same...

### Project to be claimed `mmm`: https://pypi.org/project/mmm ### Your PyPI username `Metacreation`: https://pypi.org/user/Metacreation ### Reasons for the request The project complies to the conditions of what an abandoned project in...