 Aleksandr Borzunov
|
c61c61b20d
Use t5-small tokenizer
|
%!s(int64=3) %!d(string=hai) anos |
|
Aleksandr Borzunov
|
144d35ebce
Set share_input_output_emb=True
|
%!s(int64=3) %!d(string=hai) anos |
|
Aleksandr Borzunov
|
380a5f3231
Set num threads to <= 4
|
%!s(int64=3) %!d(string=hai) anos |
|
Aleksandr Borzunov
|
df54ab6da5
Log number of params
|
%!s(int64=3) %!d(string=hai) anos |
|
Aleksandr Borzunov
|
c5493b9baf
Update per_device_batch_sizes and shuffle_buffer_size
|
%!s(int64=3) %!d(string=hai) anos |
|
Aleksandr Borzunov
|
17da8639c2
Fix and improve preprocessing
|
%!s(int64=3) %!d(string=hai) anos |
|
Aleksandr Borzunov
|
d0985de540
Enable rotary embeddings
|
%!s(int64=3) %!d(string=hai) anos |
|
Aleksandr Borzunov
|
f621362466
Make logging less verbose
|
%!s(int64=3) %!d(string=hai) anos |
|
Aleksandr Borzunov
|
31215c85d3
Handle None captions
|
%!s(int64=3) %!d(string=hai) anos |
|
Aleksandr Borzunov
|
1c30209eba
Increase timeouts
|
%!s(int64=3) %!d(string=hai) anos |
|
Aleksandr Borzunov
|
20e2a3aab2
Use dalle-pytorch instead of LeanAlbert
|
%!s(int64=3) %!d(string=hai) anos |
 Max Ryabinin
|
72fc0bcdb7
Initial commit (ru-max branch without private code)
|
%!s(int64=4) %!d(string=hai) anos |
