 Michael Diskin
|
cc8d39c2ea
Update readthedocs with hivemind.optim (#288)
|
пре 4 година |
 Max Ryabinin
|
5233b6c085
Split hivemind.client into hivemind.averaging and hivemind.moe (#304)
|
пре 4 година |
 justheuristic
|
0a0e290ea3
Add per-tensor compression, make all-reduce faster and more flexible (#272)
|
пре 4 година |
|
Max Ryabinin
|
62652e1717
Add Switch Transformers-like RemoteMixtureOfExperts (#228)
|
пре 4 година |
|
justheuristic
|
053c7c7d13
Disentangle DecentralizedAverager components, add weights (#217)
|
пре 4 година |
|
justheuristic
|
10917b259e
Averager: update group keys after every step, infer nbits dynamically (#141)
|
пре 4 година |
|
justheuristic
|
4edec82bb9
[Suggestions are welcome] Docs v0.8 (#86)
|
пре 5 година |
|
Max Ryabinin
|
35289c7005
tesseract -> hivemind
|
пре 5 година |
|
justheuristic
|
5016002186
remove dependency on run_and_await_k, rename GatingFunction to RemoteMixtureOfExperts
|
пре 5 година |
|
justheuristic
|
795bae239c
add docs in sphinx format: tesseract.runtime
|
пре 5 година |
|
justheuristic
|
e050714f56
add some docs for remoteexpert forward
|
пре 5 година |
|
justheuristic
|
57add150c4
raw html
|
пре 5 година |
|
justheuristic
|
23a1e80de9
indent
|
пре 5 година |
|
justheuristic
|
27b020627d
sphix doc style guide
|
пре 5 година |
|
justheuristic
|
70ed5dcb7f
numpy-style docstr
|
пре 5 година |
