* init docs

* update

* Update loading_methods.mdx

---------

Co-authored-by: Quentin Lhoest <42851186+lhoestq@users.noreply.github.com>

* feat: avoid some copies in torch formatter

* fix: handle kwargs

* fix: run ruff

* fix: handle dtype

* fix: handle non writable np arrays

* fix: remove comment map_nested

* fix: adjust import for lint

* Let's test like this

* code quality

* add back requests

* install transformers from source

* will it work?

* to remove later: don't fail fast

* don't fail fast

* fix test fixture

* fix OfflineModeIsEnabled test

* huggingface_hub 1.0.0 even if deps latest

* will be broken but better

* pip list in CI

* revert branch

* install latest only in latest tests

* offline

* get back to normal

* better

* ofc

* why not

* as before

* this time is good

* fix yaml format

* system

* fix import in o.x

* :/

* Bump minimal version to 0.25.0

* x-compatible offline helper

* code quality

* fix utils tests

* fixing last bits

* x-version compat

* final commit

* define CI future

* 3.12

* fix tests

* Sample without replacement option

* Exit early for non arrow iterable.

* Add new stopping strategy

* Remove sample_with_replacement argument

* fix CyclingMultiSourcesExamplesIterable.shard_data_sources

* Add sampling without replacement logic for map style datasets.

* Update process.mdx

* Update stream.mdx

---------

Co-authored-by: Petru Radulescu <petradul@amazon.com>
Co-authored-by: Quentin Lhoest <lhoest.q@gmail.com>
Co-authored-by: Quentin Lhoest <42851186+lhoestq@users.noreply.github.com>

parquet scan options and docs

* more parquet stream arg docs

* minor

* minor

less api calls when resolving data_files

add on_bad_file