 Aleksandr Borzunov
|
6190a5909e
Move inference/fwd/bwd outputs to the same devices and dtypes as inputs
|
2 anni fa |
|
Aleksandr Borzunov
|
e37b2f526a
Show tracebacks in case of empty error messages
|
2 anni fa |
 Alexander Borzunov
|
11d6ba683c
Make inference, forward, and backward fully fault-tolerant (#91)
|
2 anni fa |
 Pavel Samygin
|
50535a8435
Priority tasks (#47)
|
2 anni fa |
 Artem Chumachenko
|
ada98a1b37
Add deep prompt inference (#66)
|
3 anni fa |
 justheuristic
|
d271b75dd4
Let users specify sequence length instead of assuming 2048 (#52)
|
3 anni fa |
 Dmitry Baranchuk
|
11a424837f
integrate mixed-8bit model (#39)
|
3 anni fa |
|
justheuristic
|
f0c7383181
Implement RemoteSequential slicing and extra repr, add tests (#30)
|
3 anni fa |
