Commit History
feat: use default from_pretrained function
4ac66e4
style: reformat
dca3ada
feat(train): use new HF _do_init api
6b84155
fix: model compatible with do_init
f3a8cbb
fix: apply learning rate offset only when requested
c6263f3
Merge pull request #171 from borisdayma/reduced-requests
209ade7
unverified
Pedro Cuenca
commited on
feat(train): arg to offset lr for resumed runs
89b4c45
Half the number of requests.
d6e9f87
Pedro Cuenca
commited on
feat(demo): use vaild model
23c1ef6
feat(mega): switch to gelu
cdefdd0
feat: layernorm > rmsnorm in long runs
0f2cf98
fix: use correctly cache during inference + allow unscan (#170)
42968cf
unverified
fix: allow non-scanned models (#168)
8ae9176
unverified
feat(demo): do not log search
3500e67
feat: vmap optimizer (#166)
b993d27
unverified
feat(demo): use fixed commit
2f1e5d9
doc: reference dalle playground
648305a
feat: scan layers + gradient checkpointing (#161)
07a6f9a
unverified
Merge pull request #162 from borisdayma/demo-improvements
0199604
unverified
Pedro Cuenca
commited on
Merge branch 'main' of https://github.com/borisdayma/dalle-mini into main
bcd360f
Style: run isort.
58b9afd
Pedro Cuenca
commited on
Style: run black.
2d6dbdd
Pedro Cuenca
commited on
Display info about model and run, below predictions.
13caa60
Pedro Cuenca
commited on
Update year: 2021-2022
f58c732
Pedro Cuenca
commited on
Log dates
687a914
Pedro Cuenca
commited on
Move client api to backend.py
63679e9
Pedro Cuenca
commited on
feat: better multi-node support (#158)
728a3c3
unverified
doc: link to new demo
49be45e
feat(demo): high load
d68264e
feat(demo): update space
20a3626
feat(text): support emojis (#154)
7ef7bd9
unverified
feat: update shampoo
9ecdd3f
fix: smelu
7f2f8ed
fix: sinkformer
2c583b3
fix: support smelu
a2dcee4
fix: einops is required
179282e
feat: allow relative position (#156)
769d20a
unverified
feat: sinkhorn in lse mode (#155)
00d4661
unverified
feat(demo): update model
b9a1a7d
fix: sinkformer gradient
eed4896
feat(model): allow bias (#152)
361a994
unverified
feat(train): google-cloud-storage is optional
02b2308
feat(train): rename logged config
955dc20
feat: add sinkformer + custom final ln + pre-ln (#151)
f139b0b
unverified
feat: placeholders for more config
69bcbeb
feat: add mini_glu config
a7e5050
feat: force final ln in encoder
32f4ba5
feat: allow more configurations
5bd4c20
fix: DeepNet doesn't scale weights of embedding/output layers (#150)
503d6b4
unverified
Shuming Ma
Shuming Ma
commited on