shawntan scattermoe issues - Githubissues

shawntan / scattermoe

Triton-based implementation of Sparse Mixture of Experts.

Apache License 2.0

186 stars 14 forks source link

issues

Newest

Newest Most commented Recently updated Oldest Least commented Least recently updated

Different number of experts for each token

#18 Cy-47 opened 1 day ago
0
torch.autocast errors

#17 JCBrouwer opened 1 month ago
3
module 'torch.library' has no attribute 'custom_op'

#16 YixinSong-e closed 1 month ago
1
Khd prerefactor

#15 shawntan opened 1 month ago
0
Padded indices free.

#14 shawntan opened 1 month ago
0
fix missing parameters in function wrapper

#13 mayank31398 closed 4 months ago
0
Can't use torch.compile

#12 shikhartuli opened 4 months ago
3
Question: Multi-node training

#11 casper-hansen opened 7 months ago
3
Model with balanced load runs slower than the imbalanced

#10 CanyonWind closed 7 months ago
3
No module named 'torch'

#9 winkelstein opened 7 months ago
4
ParallelLinear with bias

#8 CanyonWind closed 7 months ago
2
Megablocks example

#7 ehartford opened 7 months ago
0
Experts with different capacity

#6 CanyonWind closed 7 months ago
4
Accuracy Issues

#5 jeromeku closed 7 months ago
11
Segfault CUDA 12.2

#4 sshleifer opened 8 months ago
1
pytest fail

#3 Eutenacity closed 8 months ago
7
Mixtral inference example

#2 casper-hansen closed 7 months ago
5
Tensor Parallelism

#1 timmytwoteeth opened 8 months ago
3