Spaces:
Running
Running
Commit History
feat: output_length considers bos and eos
8bb2236
fix: missing arg
bc01f78
[WIP] Encoding YFC100M dataset.
b4dfea0
Pedro Cuenca
commited on
Merge pull request #15 from borisdayma/feat-fix-lr
dc5ae57
unverified
feat: requirements for tb logging
8884d40
feat: shared cache folder
42ce7dd
feat: update lr range
dbbd01a
fix: accumulation vs lr
4d55db6
Move generate nb by @ghosh-r to demo
dcbf091
Pedro Cuenca
commited on
Merge pull request #14 from khalidsaifullaah/main
66bfb99
unverified
Merge pull request #13 from borisdayma/model-generate-notebook
894a546
unverified
Updated with train_file flag to resolve the error
8f058ae
unverified
notebook example for model.generate
67221fc
Ritobrata Ghosh
commited on
text-heneration-notebook
1c2552a
Ritobrata Ghosh
commited on
feat: update sweep parameters
2f69241
feat: update script
650ecb1
fix: wandb logging with sync_tensorboard
8ba598c
Merge pull request #12 from borisdayma/feat-sweeps
06f1345
unverified
feat: add sweep for parameter search
dad6d93
feat: lower default source length
48c07ca
Merge pull request #11 from borisdayma/feat-cumul
ba73e00
unverified
fix: typos
5960e87
feat: gradient accumulation
c9e9575
Merge pull request #10 from borisdayma/feat-loss
cbeacb9
unverified
feat: simplify loss function
9db361a
feat: padding mask not required
d61405b
fix: typo
833a2d5
feat: default to 1000 warmup steps
b66b951
feat: add adafactor
600ad79
feat: log metrics more frequently
498559f
* Make padding mask optional.
ecafe5e
Pedro Cuenca
commited on
* Only perform validation if requested
32dc2d8
Pedro Cuenca
commited on
Preprocessing: return "labels", "decoder_input_ids" and
df3c7bd
Pedro Cuenca
commited on
Decoder: set eos to an unreachable value, set min_length=max_length to
a841a4c
Pedro Cuenca
commited on
accept tsv files as inputs.
a104edb
Pedro Cuenca
commited on
val_max_target_length set to OUTPUT_LENGTH
6c27b0d
Pedro Cuenca
commited on
use pylogging to refer to logging.
d9f5a35
Pedro Cuenca
commited on
fix typos and update requirements
de74f11
Pedro Cuenca
commited on
feat: adjust seq2seq script for dalle
3f0364c
feat: add seq2seq requirements
fad333f
feat: add run_seq2seq_flax
46cb01f
* Prepend [bos] to image encodings, rename to "labels".
86ba774
Pedro Cuenca
commited on
* Data preprocessing pipeline proof of concept.
95d2faf
Pedro Cuenca
commited on
* Notebook that processes CC12M and creates a version with encodings.
16f038a
Pedro Cuenca
commited on
* dalle_mini package with models and utilities:
150ed18
Pedro Cuenca
commited on
* Ignore __pycache__
550b472
Pedro Cuenca
commited on
* JIT outside the loop.
4b8c3a8
Pedro Cuenca
commited on