 Aleksandr Borzunov
|
2fed6ba68e
Use HF username as wandb run name
|
il y a 3 ans |
 Alexander Borzunov
|
c365b2ec9f
Tweak settings for the upcoming demo (#2)
|
il y a 3 ans |
|
Aleksandr Borzunov
|
64dee420da
Upgrade to using hivemind.optim.experimental
|
il y a 3 ans |
|
Aleksandr Borzunov
|
e97e7b8811
Try removing OffloadOptimizer
|
il y a 3 ans |
|
Aleksandr Borzunov
|
3b184c57da
Don't download VQGAN weights
|
il y a 3 ans |
|
Aleksandr Borzunov
|
c61c61b20d
Use t5-small tokenizer
|
il y a 3 ans |
|
Aleksandr Borzunov
|
144d35ebce
Set share_input_output_emb=True
|
il y a 3 ans |
|
Aleksandr Borzunov
|
df54ab6da5
Log number of params
|
il y a 3 ans |
|
Aleksandr Borzunov
|
d0985de540
Enable rotary embeddings
|
il y a 3 ans |
|
Aleksandr Borzunov
|
20e2a3aab2
Use dalle-pytorch instead of LeanAlbert
|
il y a 3 ans |
 Max Ryabinin
|
72fc0bcdb7
Initial commit (ru-max branch without private code)
|
il y a 4 ans |
