[CB] [Major] Upgrade the cache to support different attention types #47809
Consolidate TP
b0bc55dd
rename
1a515c8a
The most breaking commit ever
22b627fa
Rename attributes in cfg
751c59da
Disable compile around attn
4b4ad8d0
Nits
0240a9e0
Cahges to make mask more efficient
08a7a6e8
Rename nits
3e5784a0
Init fixes
4486df35
Further integration
f9951575
Fix cache allocators
c43a8edc
Nit
9445e44f
Sizing
8d2471a2
Add cache pool
feaaf07f
Better and more optimized cache pool
f1de8a4c
Block buggy path
c550e06e
Add forking
caeef1a9
De-duplication and prefix sharing
2e0d6ad2
Offloading is back
50aee742
Restore block table
9cc752ea
Update overall
a5ec4d17
Block table part 2
11713d36
Test
ba7698d7
Fixes
ce5fe45a
Rename
eb0f4f7f
Tests 1/n
032bda56
Test 2/n
f539e4e7
More renaming
acdc0b18
Rename back
8db0c7c7
More test fixes
30eb11aa
remi-or
force pushed
from
6142eefc
to
30eb11aa
5 days ago
Mb
730ba388
Merge branch 'main' into cb-cache-upgrade
16e58590
Assignees
No one assigned
Login to write a write a comment.
Login via GitHub