 Alexander Borzunov
|
c6e1b5a8e5
Add various server timeouts, lower --max_batch_size and --inference_max_length defaults (#97)
|
vor 3 Jahren |
|
Alexander Borzunov
|
d8ef09146e
Improve server's logging (#96)
|
vor 3 Jahren |
|
Alexander Borzunov
|
fef7257fe0
Try to fix protobuf versions once again (#95)
|
vor 3 Jahren |
 Aleksandr Borzunov
|
1b51703444
Revert protobuf version change
|
vor 3 Jahren |
|
Alexander Borzunov
|
b26b0b7121
Require hivemind with fixed compression and protobuf working on Colab (#94)
|
vor 3 Jahren |
|
Alexander Borzunov
|
11d6ba683c
Make inference, forward, and backward fully fault-tolerant (#91)
|
vor 3 Jahren |
|
Alexander Borzunov
|
57e8d2e721
Implement exponential backoff for forward & backward (#85)
|
vor 3 Jahren |
|
Alexander Borzunov
|
f64eb3a665
Update hivemind to 1.1.2, mark `model` argument as required (#81)
|
vor 3 Jahren |
 justheuristic
|
fef48d7d99
Use bitsandbytes==0.34.0, update readme (#76)
|
vor 3 Jahren |
|
justheuristic
|
8caf1145a8
Quality of life changes: update readme, simplify run_server interface (#75)
|
vor 3 Jahren |
|
justheuristic
|
3fdcc55a56
fix protobuf version (#74)
|
vor 3 Jahren |
|
justheuristic
|
e92487e5d2
Update dependency versions (#71)
|
vor 3 Jahren |
|
justheuristic
|
d271b75dd4
Let users specify sequence length instead of assuming 2048 (#52)
|
vor 3 Jahren |
 Dmitry Baranchuk
|
11a424837f
integrate mixed-8bit model (#39)
|
vor 3 Jahren |
|
Dmitry Baranchuk
|
04a2b6f5e3
Support various backend dtypes & async serialization (#38)
|
vor 3 Jahren |
|
justheuristic
|
e2711a033b
Add automated tests (#23)
|
vor 3 Jahren |
|
justheuristic
|
99059ae667
install script
|
vor 3 Jahren |
|
justheuristic
|
b370b43110
freeze hivemind and transformers versions
|
vor 3 Jahren |
