 justheuristic
|
b755f8a8f5
Merge branch 'master' into rfc_optimizer
|
vor 3 Jahren |
|
justheuristic
|
688b514fa6
Fix All-Reduce fault-tolerance: catch Exception instead of BaseException (#424)
|
vor 3 Jahren |
|
justheuristic
|
6da8683975
Improve All-Reduce fault-tolerance (#423)
|
vor 3 Jahren |
|
justheuristic
|
896885a9f0
Minor tweaks learned from the demo experiment (#422)
|
vor 3 Jahren |
 Alexander Borzunov
|
a904cfd58b
Improve hivemind.optim.experimental and averager stability (#421)
|
vor 3 Jahren |
|
Alexander Borzunov
|
249be7ca2b
Do not log caller for INFO messages (#418)
|
vor 3 Jahren |
|
Alexander Borzunov
|
12099ce03c
Remove trailing dots in log messages and errors (#419)
|
vor 3 Jahren |
|
justheuristic
|
c328698b86
Quickstart: typos and references (#420)
|
vor 3 Jahren |
|
justheuristic
|
40726cd56a
Improve Optimizer docs, update quickstart to use Optimizer (#416)
|
vor 3 Jahren |
|
justheuristic
|
318bb7ad48
Hotfix: offload_optimizer in load_state_from_peers (#417)
|
vor 3 Jahren |
 MuXauJl11110
|
a96043804b
DHT Benchmark with asynchronous w/r (#406)
|
vor 3 Jahren |
|
justheuristic
|
5d31c3ba9d
Implement core functionality of hivemind.Optimizer (#403)
|
vor 3 Jahren |
|
justheuristic
|
00538dbef6
Implement state sharing priority (#415)
|
vor 3 Jahren |
 Artem Chumachenko
|
50ab48e240
Replace the invalid link to discord (#414)
|
vor 3 Jahren |
|
justheuristic
|
7a79aea439
Fix StepControl.cancel() in DecentralizedAverager (#411)
|
vor 3 Jahren |
|
justheuristic
|
09e34f8366
Prepare GradScaler for hivemind.Optimizer (#413)
|
vor 3 Jahren |
|
justheuristic
|
22665fdcee
Make target group size optional (#412)
|
vor 3 Jahren |
|
justheuristic
|
1cfc6a3b7b
Fix internal assert in GradientAverager (#410)
|
vor 3 Jahren |
|
justheuristic
|
d883387e37
Add ProgressTracker (#408)
|
vor 3 Jahren |
|
Artem Chumachenko
|
99a0c18ca9
Catch OSError in MPFuture (#409)
|
vor 3 Jahren |
|
justheuristic
|
abbea26c9c
Merge branch 'master' into rfc_optimizer
|
vor 3 Jahren |
|
justheuristic
|
02eee9292d
[hivemind.Optimizer] TrainingStateAverager (#407)
|
vor 3 Jahren |
|
justheuristic
|
8fa0a8e6ae
Add GradientAverager with support for delayed averaging (#404)
|
vor 3 Jahren |
|
justheuristic
|
ed4204009f
move PerformanceEMA to utils, TrainingAverager to optim, update utils (#405)
|
vor 3 Jahren |
|
justheuristic
|
7c4d13f06d
hotfix: replace StepControl.can_modify with began_allreduce
|
vor 3 Jahren |
|
justheuristic
|
4df9665ed2
scheme for rtfd
|
vor 3 Jahren |
|
justheuristic
|
7c7ac805d8
scheme for rtfd
|
vor 3 Jahren |
|
justheuristic
|
793a741950
Move DHT to dht/dht.py, update DHT figure (#399)
|
vor 3 Jahren |
|
justheuristic
|
a09df5492f
Add an option to pre-schedule averaging (#398)
|
vor 3 Jahren |
|
justheuristic
|
025e095d55
backport PerformanceEMA from server_side_averaging (#397)
|
vor 3 Jahren |
