huggingface/transformers

Pull Requests Commits
update cache shape
MekkCyber committed 1 year ago
fbd288fb
update
MekkCyber committed 1 year ago
eb714c4b
chaing v2 default
MekkCyber committed 1 year ago
028e6d6e
add flash sdpa
MekkCyber committed 1 year ago
b6322afd
nits
ArthurZucker committed 1 year ago
04828074
comment latest sync for now
MekkCyber committed 1 year ago
3dc01c46
adding v2
MekkCyber committed 1 year ago
d0be498d
mps support
MekkCyber committed 1 year ago
5e9b251a
seems to work
MekkCyber committed 1 year ago
255bc3c9
updates
MekkCyber committed 1 year ago
1fd77c3d
updates
MekkCyber committed 1 year ago
5189c26e
more fixes
MekkCyber committed 1 year ago
2721261e
kernel either works or not used
MekkCyber committed 1 year ago
571af3e2
figuring out why sdpa without kernel is broken
MekkCyber committed 1 year ago
51e115ac
remove mps device
MekkCyber committed 1 year ago
962f7c16
update flash_paged
ArthurZucker committed 1 year ago
fa3e0db0
I need to test on cuda, this shouldd work
ArthurZucker committed 1 year ago
3cbe8254
the shape expected by the kernel is not the sliced cache I reckon
ArthurZucker committed 1 year ago
6fbe4177
make sure we can compare without reshape kernel
ArthurZucker committed 1 year ago
9f47da14
more cleanup
ArthurZucker committed 1 year ago
da935a68
bunch of cleanup and let's not sample, easier to teset
ArthurZucker committed 1 year ago
f73db2b6
smaller for faster
ArthurZucker committed 1 year ago
c5765160
works with this :wink:
ArthurZucker committed 1 year ago
6eec32cf
where I am at
ArthurZucker committed 1 year ago
cbe3e632
comment out stuff for now!
ArthurZucker committed 1 year ago
c2cf536c
losing my mind, test on cluster
MekkCyber committed 1 year ago
4d86a5bb
change shapes
MekkCyber committed 1 year ago
ccf8a46c
second test
MekkCyber committed 1 year ago
f205a9c9
fisrt integration
MekkCyber committed 1 year ago
458b3522
Update src/transformers/generation/continuous_batching.py
ArthurZucker committed 1 year ago
012d8f47
Terms Of Use Privacy Policy Refund Policy Impressum