HIGGS Quantization Support #34997
higgs init
08b347cf
working with crunches
14a0c82c
per-model workspaces
1c5b9e7d
style
9f2ef77e
style 2
0ff58c37
Merge branch 'huggingface:main' into main
2e9adc61
tests and style
b6bad710
higgs tests passing
c2bcf39b
protecting torch import
a1e7b35c
removed torch.Tensor type annotations
8f1a0a6a
torch.nn.Module inheritance fix maybe
120f3600
hide inputs inside quantizer calls
fdb71a5d
style structure something
127c5f0b
Merge branch 'main' into main
947a53dd
Merge branch 'main' into main
e6ddc415
Update src/transformers/quantizers/quantizer_higgs.py
0de97f14
reworked num_sms
1f08cb0b
Merge branch 'main' of github.com:BlackSamorez/transformers
ed369be9
Update src/transformers/integrations/higgs.py
96023ab6
revamped device checks
60ce44be
docstring upd
81424439
Update src/transformers/quantizers/quantizer_higgs.py
1d636ac0
edited tests and device map assertions
66ece1d1
Merge branch 'main' of github.com:BlackSamorez/transformers
5a68cd6e
minor edits
1cb9f0c1
updated flute cuda version in docker
257c39b0
Added p=1 and 2,3bit HIGGS
f82d1a38
flute version check update
b7479804
incorporated `modules_to_not_convert`
398d5b18
less hardcoding
0ede69c2
Merge branch 'main' into main
ebe6766e
Fixed comment
4e730a9d
Added docs
c3848f37
Fixed gemma support
c91182a3
example in docs
8e1da3de
MekkCyber
approved these changes
on 2024-12-02
fixed torch_dtype for HIGGS
15f37898
SunMarc
approved these changes
on 2024-12-05
Update docs/source/en/quantization/higgs.md
b10375f2
Collection link
2abdc849
dequantize interface
5074b608
newer flute version, torch.compile support
2fab798c
unittest message fix
db27c598
SunMarc
approved these changes
on 2024-12-13
docs update compile
6917eceb
Merge branch 'main' into main
7c003c28
isort
e317fc5a
ValueError instead of assert
53e6827f
SunMarc
approved these changes
on 2024-12-23
SunMarc
merged
64c05eec
into main 1 year ago
Assignees
No one assigned
Login to write a write a comment.
Login via GitHub