 Jay Shah
|
a7cce59d25
adjust tolerances in test script for kv cache
|
2 månader sedan |
|
Jay Shah
|
d437d3dd5c
remove smem usage for when rmem -> gmem epilogue is used
|
2 månader sedan |
 Ganesh Bikshandi
|
e49cb5f77c
passes except for hdim=256.
|
2 månader sedan |
|
Ganesh Bikshandi
|
dc2c952f37
compiles and builes. Not validates.
|
2 månader sedan |
|
Ganesh Bikshandi
|
a075e769fb
handle gqa_parallel with rmem-to-gmem. Not validating yet.
|
2 månader sedan |
|
Ganesh Bikshandi
|
f0b49460ec
changes to use tiledcopy (still not passing).
|
2 månader sedan |
|
Ganesh Bikshandi
|
0085f04b6a
add fp8 test case.
|
2 månader sedan |
|
Ganesh Bikshandi
|
2b840ef32d
fix the test case and re-factor too.
|
2 månader sedan |
|
Jay Shah
|
2472e5e0b4
add 'in principle' fp8 kv cache support
|
2 månader sedan |
|
Ganesh Bikshandi
|
cd55fb3b5d
set correct tolerance limit
|
2 månader sedan |
|
Ganesh Bikshandi
|
70ff847363
all cases passed.
|
2 månader sedan |
|
Ganesh Bikshandi
|
9a4941cb35
add variable seqlen case.
|
2 månader sedan |
|
Ganesh Bikshandi
|
c516d6349d
Adding another test case.
|
2 månader sedan |
|
Ganesh Bikshandi
|
24b4b4f7b5
add reference from python.
|
2 månader sedan |
|
Jay Shah
|
3615696245
Add some more cases to test script and raise thresholds a bit for max error
|
2 månader sedan |
|
Jay Shah
|
4a77193890
update test script to use heuristic
|
2 månader sedan |
|
Ganesh Bikshandi
|
ffa48eb8cc
more refactoring.
|
2 månader sedan |
|
Ganesh Bikshandi
|
a30863fe27
recent version.
|
2 månader sedan |
