COMMITS
/ torchao/quantization/quant_api.py March 13, 2025
A
Add CUTLASS-based row-wise scaled sparse FP8 kernel (#1671)
Aleksandar Samardžić committed
March 12, 2025
V
remove the `set_inductor_config` argument of `quantize_`. (#1865)
Vasiliy Kuznetsov committed
V
enforce AOBaseConfig type in quantize_'s `config` argument (#1861)
Vasiliy Kuznetsov committed
February 14, 2025
V
make quantize_.set_inductor_config None by default (#1716)
Vasiliy Kuznetsov committed
V
config migration: fpx, gemlite, uintx (#1697)
Vasiliy Kuznetsov committed
V
config migration: int* (#1696)
Vasiliy Kuznetsov committed
V
config migration: float8* (#1694)
Vasiliy Kuznetsov committed
V
[bc-breaking] enable direct configuration in quantize_ (#1595)
Vasiliy Kuznetsov committed
February 5, 2025
T
Add CUTLASS-based W4A4 (#1515)
Thien Tran committed
February 4, 2025
D
Fix ruff and make sure pre-commit is at same version (#1658)
Driss Guessous committed
January 30, 2025
N
[Feat]: Add support for kleidiai quantization schemes (#1447)
Nikhil Gupta committed
January 29, 2025
S
Consolidate `ZeroPointDomain.NONE` & `None` zero point domains (#1556)
sanchitintel committed
January 24, 2025
X
[CPU] Fix registration of int4wo linear implementation on CPU (#1578)
Xia Weiwen committed
January 5, 2025
A
W4A8 based on CUTLASS (#880)
Aleksandar Samardžić committed
December 30, 2024
J
int8 dynamic prefill weight only decode (#1436)
Jesse Cai committed
December 20, 2024
J
Add gemlite kernel option to autoquant (#1449)
Jerry Zhang committed
December 17, 2024
M
Expose zero_point_domain as arguments (#1401)
Meng, Hengyu committed
December 16, 2024
H
gemlite integration in torchao (#1034)
HDCharles committed
A
Add new QAT API through quantize_ (#1415)
andrewor14 committed
December 11, 2024
J
Add marlin and semi sparse + quant option to autoquant (#1399)
Jerry Zhang committed
December 1, 2024
A
Update hardware check conditions (#1356)
Apurva Jain committed
November 28, 2024
A
Add support for quantize_() with Float8Linear module (#1344)
Apurva Jain committed
November 27, 2024
Y
Add Int4CPULayout and update int4 woq (#1278)
YanbingJiang committed
November 26, 2024
A
Add hardware check to fp8 quant (#1314)
Apurva Jain committed
November 15, 2024
A
Refactor Affine Quantized Tensor (#1234)
Apurva Jain committed
November 14, 2024
H
support W4A8 Marlin kernel (#1113)
HandH1998 committed
November 13, 2024
J
Update torchao api reference and add contributor guide (#1255)
Jerry Zhang committed
November 12, 2024
A
Ruff Lint fixes (torchao/quantization) (#1244)
Apurva Jain committed
October 29, 2024
D
Add default filtering to remove mis-alinged weights (#1194)
Driss Guessous committed
October 25, 2024
H
is_linear fix for MHA (#1141)
HDCharles committed
A
Fix int8_weight_only group_size (#1165)
andrewor14 committed
October 24, 2024
A
[reland] Move QAT out of prototype (#1152)
andrewor14 committed
October 22, 2024
P
fixed `weights_only=True` load for `float8_dynamic_activation_float8_weight` in quant_api (#1122)
Paarth Bhatnagar committed
J
Add support for asymmetric act quant for int8 dynamic quant (#1131)
Jerry Zhang committed
October 19, 2024
J
Add support for groupwise quantization for int8 weight only quantization (#1121)
Jerry Zhang committed
October 10, 2024
A
Rename AQT#2 LayoutType -> Layout (#1049)
Apurva Jain committed
A
Move and rename GranularityType -> Granularity (#1038)
andrewor14 committed
A
Aqt rename#1 Layout -> TensorImpl (#1046)
Apurva Jain committed
October 8, 2024
M
Revert "Rename Layout -> TensorImpl" (#1040)
Mark Saroufim committed
A
Rename Layout -> TensorImpl (#1028)
Apurva Jain committed
October 2, 2024
A
Add float8_static_activation_float8_weight to user-facing API (#994)
Apurva Jain committed
September 13, 2024
D
[Float8] Add static constructor that will be used in Observer workflow (#869)
Driss Guessous committed
September 12, 2024
J
Renaming fpx to floatx (#877)
Jerry Zhang committed
September 9, 2024
M
Add symmetric quantization with no clipping error in the tensor subclass based API (#845)
Mengtao Yuan committed
T
Add INT8 mixed-precision training (#748)
Thien Tran committed
September 7, 2024
D
[Float8Quant] Add rowwise scaling option to float8 dyanmic quant (#819)
Driss Guessous committed
September 6, 2024
J
Expose hqq through `uintx_weight_only` API (#786)
Jerry Zhang committed
J
Update docs + add deprecation warning (#825)
Jesse Cai committed
D
Add sparse marlin AQT layout (#621)
Diogo Venâncio committed
D
[StaticQuant] add a linear observer class and test (#807)
Driss Guessous committed