Merge pull request #91 from borisdayma/feat-inf

feat: more samples

feat: cleanup

feat: add sample

fix: typo

fix: pmap clip32

feat: cleanup

feat: allow latest version only

refactor: loop over runs

feat: create a table

feat: add scoring

feat: remove cache before creating artifacts

feat: add functions

feat: log to backend

fix: state.step type

Merge branch 'main' of https://github.com/borisdayma/dalle-mini

feat: get rid of global_step + log more metrics

fix(seq2seq): memory issue

feat: update scripts

feat: use optax for gradient accumulation

feat: no need for default values

feat: limit artifacts size

feat: log epoch + check params

feat: update defaults

fix(seq2seq): normalize text

fix(seq2seq): use streaming arg

fix: remove breakpoint

feat: handle streaming

Merge branch 'main' of https://github.com/borisdayma/dalle-mini into fix-opt_state

feat(inference_notebook): dalle-mini is installable

Merge pull request #74 from borisdayma/feat-setup

Replace notebooks with the correct versions.

Notebooks that demonstrate streaming encoding

fix: typo

Merge branch 'main' of https://github.com/borisdayma/dalle-mini into feat-setup

feat: remove symlinks

Normal install of transformers & datasets

fix: actually replace state

fix(seq2seq): opt_state from ckpt + limit cache

Fix requirements.txt to install libtpu from google's page.

fix: use correct branch

feat: remove warning

fix(colab): don't preprocess images twice in CLIP

doc: simplify installation instructions

fix: use correct branch

fix: restore deleted paragraph

feat: link colab to correct branch

feat: remove warnings

feat: add tqdm

chore: reorganize files