Fixes for continuous batching #40828
Fix for CB attn mask and refactor
2c9c94ea
Tests for CB (not all passing)
1d4fd516
Passing tests and a logger fix
f1beeff9
Fixed the KV metrics that were broken when we moved to hybrid alloc
97a4248d
Fix circular import and style
c8f82576
Added tests for FA
e9e04cc6
Unfolded test to have device expectations
d7694e85
Fixes for H100
146f4fc2
more fixes for h100
2acd1d56
H100 are good
f35ae0c4
Style
04893a43
Adding some comments from #40831
f7a44487
remi-or
force pushed
from
ad4ab44a
to
f7a44487
93 days ago
Rename test
0d5cb00b
Avoid 1 letter variables
ed2d6474
Dictonnary is only removed during kwargs
7e2fb212
Merge branch 'main' into cb-fix
880372a7
Test for supported sample
17fb99a7
Fix a unvoluntary slice
1266d4d6
Fixes for non-sliced inputs and small example improvments
0792dce0
Slice inputs is more understandabe
49ae0616
Merge branch 'main' into cb-fix
add2bb8b
Style
06781835
Assignees
No one assigned
Login to write a write a comment.
Login via GitHub