transformers
Fine-grained block-scaled quantization: the shared machinery
#49776
Open
Open
Fine-grained block-scaled quantization: the shared machinery #49776
IlyasMoutawwakil
wants to merge
9 commits into
fg-split/4a-host
from
fg-split/4-machinery
IlyasMoutawwakil
force-pushed the
fg-split/3-experts-tp
branch
from
dc6c12da
to
1b8bd6a5
10 days ago
IlyasMoutawwakil
force-pushed the
fg-split/3-experts-tp
branch
from
1b8bd6a5
to
bf49e7e9
10 days ago
IlyasMoutawwakil
force-pushed the
fg-split/3-experts-tp
branch
from
bf49e7e9
to
caf0319b
10 days ago
IlyasMoutawwakil
changed the title
IlyasMoutawwakil
force-pushed the
fg-split/3-experts-tp
branch
from
a1e72b97
to
2a513c35
10 days ago
IlyasMoutawwakil
force-pushed the
fg-split/3-experts-tp
branch
from
2a513c35
to
602bdd3d
10 days ago
IlyasMoutawwakil
force-pushed the
fg-split/3-experts-tp
branch
from
602bdd3d
to
84d4b701
10 days ago
Add the block-scaled quantizer base and the fine-grained modules
8817bc56
Find a shared op's member through the host's member_for
ceaa9769
Say in the docs that a plan entry can be any method's config
610c2f0b
Let the block-scaled formats extend the standard quantizer hooks
019a5f02
Read skip lists into None entries and drop config_for
2ffdc512
Let a block-scaled op read its own member's config
57fd5839
Test that only a format config lends to a plan
c6b3d110
Document how a passed config merges into a plan, and fused modules
49cbba81
Merge remote-tracking branch 'origin/fg-split/4a-host' into fg-split/…
7a174ca4
Login to write a write a comment.
Login via GitHub