dalle-mini / dev /seq2seq /run_seq2seq_flax.py

Commit History

feat: use_auth_token + seed for dataset and model
eac6890

boris commited on

feat: use custom TrainingArguments
85748ef

boris commited on

fix: comments
bab75aa

boris commited on

fix: fixes training script
0df810d

boris commited on

fix: log train_metric only if defined
9bf9397

boris commited on

feat: add metrics + cleanup
6523a6d

boris commited on

feat: simplify parameters
87fac28

boris commited on

feat: use model definition
803c7df

boris commited on

fix: OOM
86c6c90

boris commited on

fix: OOM with checkpoints
e2400cc

boris commited on

fix: comment
36cb737

boris commited on

feat: cleanup training script
3cd6d41

boris commited on

feat: remove cache before creating artifacts
5f6b691

boris commited on

fix: state.step type
47e006f

boris commited on

Merge branch 'main' of https://github.com/borisdayma/dalle-mini
5faf0fd

boris commited on

feat: get rid of global_step + log more metrics
4a4820f

boris commited on

fix(seq2seq): memory issue
708a42c

boris commited on

feat: use optax for gradient accumulation
69cf636

boris commited on

feat: no need for default values
a37cd75

boris commited on

feat: limit artifacts size
7253e56

boris commited on

feat: log epoch + check params
074c5e1

boris commited on

feat: update defaults
9ed6378

boris commited on

fix(seq2seq): normalize text
061c06b

boris commited on

fix(seq2seq): use streaming arg
0c992bd

boris commited on

fix: remove breakpoint
b75e0e9

boris commited on

feat: handle streaming
a96f44d

boris commited on

fix: actually replace state
1d04ab3

boris commited on

fix(seq2seq): opt_state from ckpt + limit cache
0c9ff65

boris commited on

feat: remove unused metrics
0d94b71

boris commited on

doc: note about model definition
6c5fc6a

boris commited on

feat: remove hardcoded values
93c5ac8

boris commited on

chore: move files around
31da1e5

boris commited on