GitHub / sony/model_optimization / commits
Model Compression Toolkit (MCT) is an open source project for neural network model optimization under efficient, constrained hardware. This project provides researchers, developers, and engineers advanced quantization and compression tools for deploying state-of-the-art neural networks.
| SHA | Message | Author | Date | Stats |
|---|---|---|---|---|
| b4ff01ec | Add IMX500 Converter 3.18 test badges in supported version table |
YingYang-YY <7****Y@u****m>
Committed by: GitHub <n****y@g****m> |
4 months ago | |
| 08979ce7 | Add TPC version test of conv3.18 |
YingYang-YY <7****Y@u****m>
Committed by: GitHub <n****y@g****m> |
4 months ago | |
| cefa4023 | Feature doc explane quant algorithms 2 (#1677) (#1678) |
KazunoriSumiya <S****i@j****m>
Committed by: GitHub <n****y@g****m> |
5 months ago | |
| 09dfd0e2 | [RedSuns Review]Midify progress bar visualizaion |
kenta kawasaki <k****a@m****m>
Committed by: GitHub <n****y@g****m> |
6 months ago | |
| 95cc888f | [SSS Review] Add progress visualization feature (#1671) |
kenta kawasaki <k****a@m****m>
Committed by: GitHub <n****y@g****m> |
6 months ago | |
| bd6aeb47 | Replace requirement mct-quantizers from git source to official 1.7.0(for main... |
ueno-hiroshi002 <1****2@u****m>
Committed by: GitHub <n****y@g****m> |
6 months ago | |
| c863e521 | [SSS Review] Add tutorial notebook for quantization troubleshooting using the... |
KazunoriSumiya <S****i@j****m>
Committed by: GitHub <n****y@g****m> |
6 months ago | |
| b8699d62 | Support for scalar and 1d-tensor in statistics collector (#1662) |
kawakami-masaki0 <k****i@j****m>
Committed by: GitHub <n****y@g****m> |
6 months ago | |
| edd005d5 | Update Contributors and Typo Correction (#1658) |
yt0705 <t****0@j****m>
Committed by: GitHub <n****y@g****m> |
7 months ago | |
| 4327386f | Version Update v.2.6.0 at main branch |
ueno-hiroshi002 <1****2@u****m>
Committed by: GitHub <n****y@g****m> |
7 months ago | |
| a7c50094 | resolve conflict of searchindex.js (#1653) |
ueno-hiroshi002 <1****2@u****m>
Committed by: GitHub <n****y@g****m> |
7 months ago | |
| a0e73a01 | Refactoring XQuant_Extension_Tool Documents (#1648) |
ueno-hiroshi002 <1****2@u****m>
Committed by: GitHub <n****y@g****m> |
7 months ago | |
| 6579f583 | TPC Improvement (Additional of #1635) (#1650) |
ueno-hiroshi002 <1****2@u****m>
Committed by: GitHub <n****y@g****m> |
7 months ago | |
| eacba530 | Disable weight quantization for Sin, Cos, and Exp (#1647) |
kawakami-masaki0 <k****i@j****m>
Committed by: GitHub <n****y@g****m> |
7 months ago | |
| ef857251 | TPC Improvement (Exclude tutorials and wrapper API documents) (#1635) |
ueno-hiroshi002 <1****2@u****m>
Committed by: GitHub <n****y@g****m> |
8 months ago | |
| d7966e0f | Maintenance Tutorials (#1632) |
ueno-hiroshi002 <1****2@u****m>
Committed by: GitHub <n****y@g****m> |
8 months ago | |
| b983a310 | Change the activation quantization error tolerance when exporting (#1626) |
kawakami-masaki0 <k****i@j****m>
Committed by: GitHub <n****y@g****m> |
8 months ago | |
| 419ff207 | Update version 2.5.1 (#1617) |
ueno-hiroshi002 <1****2@u****m>
Committed by: GitHub <n****y@g****m> |
8 months ago | |
| b1dfa576 | [SSS Review]Feature log onoff (#1610) |
KazunoriSumiya <S****i@j****m>
Committed by: GitHub <n****y@g****m> |
8 months ago | |
| d606657f | Support NumPy 2.4.0 (#1609) |
ueno-hiroshi002 <1****2@u****m>
Committed by: GitHub <n****y@g****m> |
8 months ago | |
| 61130653 | Colab Support for task_tutorials (#1603) |
ueno-hiroshi002 <1****2@u****m>
Committed by: GitHub <n****y@g****m> |
8 months ago | |
| b57da4e0 | limit version of numpy(<2.4.0) (#1602) |
ueno-hiroshi002 <1****2@u****m>
Committed by: GitHub <n****y@g****m> |
8 months ago | |
| 26bc320c | Add EfficientDet(Keras) Tutorial. (#1596) |
ueno-hiroshi002 <1****2@u****m>
Committed by: GitHub <n****y@g****m> |
9 months ago | |
| f631999d | Support torch.index_select (#1591) |
ueno-hiroshi002 <1****2@u****m>
Committed by: GitHub <n****y@g****m> |
9 months ago | |
| 8c20bce7 | [SSS Review] Feature develop mctwrapper (#1578) |
KazunoriSumiya <S****i@j****m>
Committed by: GitHub <n****y@g****m> |
9 months ago | |
| e2e18d4f | Change edgemdt-cl>=1.1 (to main branch) (#1577) |
kawakami-masaki0 <k****i@j****m>
Committed by: GitHub <n****y@g****m> |
9 months ago | |
| 5b2de2c4 | Add Deeplabv3+(Pytorch) Tutorial. (#1563) |
ueno-hiroshi002 <1****2@u****m>
Committed by: GitHub <n****y@g****m> |
9 months ago | |
| 874ccb56 | Add new TPCv6.0 (torch.take) (#1570) |
kawakami-masaki0 <k****i@j****m>
Committed by: GitHub <n****y@g****m> |
9 months ago | |
| 55d1c4bc | Limit networkx version (#1566) |
ueno-hiroshi002 <1****2@u****m>
Committed by: GitHub <n****y@g****m> |
9 months ago | |
| 92673273 | Add YOLOX(PyTorch) Tutorial (#1558) |
ueno-hiroshi002 <1****2@u****m>
Committed by: GitHub <n****y@g****m> |
10 months ago | |
| 420e1609 | Exclude Python 3.9 from support(#1551) |
yt0705 <t****0@j****m>
Committed by: GitHub <n****y@g****m> |
10 months ago | |
| 68ddf7ba | Update Version 2.5.0 and Rebuild Docs (#1545) |
ueno-hiroshi002 <1****2@u****m>
Committed by: GitHub <n****y@g****m> |
11 months ago | |
| 1bff60c6 | Version and Docs update for v2.4.4 (#1540) |
ueno-hiroshi002 <1****2@u****m>
Committed by: GitHub <n****y@g****m> |
11 months ago | |
| e28edbfb | limit pydantic version(<2.12.0) (#1526) |
ueno-hiroshi002 <1****2@u****m>
Committed by: GitHub <n****y@g****m> |
11 months ago | |
| 28ccd789 | Add posenet tutorial. (#1524) |
ueno-hiroshi002 <1****2@u****m>
Committed by: GitHub <n****y@g****m> |
11 months ago | |
| 27e42fbf | [SSS Review] Stack bugfix, Delete nightly release, and Update version (#1523) |
kawakami-masaki0 <k****i@j****m>
Committed by: GitHub <n****y@g****m> |
11 months ago | |
| f04e64a6 | Update CI/CD, Dependabot, Copyrights, setup.py, README.md. (#1517) |
ueno-hiroshi002 <1****2@u****m>
Committed by: GitHub <n****y@g****m> |
11 months ago | |
| abd8f71e | Attach MulticlassNMSOBB (#1508) |
ueno-hiroshi002 <1****2@u****m>
Committed by: GitHub <n****y@g****m> |
11 months ago | |
| 7a5fe9b0 | Develop XQuant Extension Tool (#1502) |
ueno-hiroshi002 <1****2@u****m>
Committed by: GitHub <n****y@g****m> |
11 months ago | |
| 3c1ff38f | Generating docs | ariell <a****d@a****m> | about 1 year ago | |
| fa607e3f | Update example in exporter docsrc for output_names usage | ariell <a****d@a****m> | about 1 year ago | |
| 282de225 | API change: Add option to choose ouput names when exporting using ONNX | ariell <a****d@a****m> | about 1 year ago | |
| bd0cf9fd | Update version to 2.4.2 | ariell <a****d@a****m> | about 1 year ago | |
| e8a17071 | Update version to 2.4.1 | reuvenp <r****p@a****m> | about 1 year ago | |
| 77c5d153 | Replace activation quantizers with identity only when nb>8 and using fakely q... |
Ariel Lapid <5****2@u****m>
Committed by: reuvenp <r****p@a****m> |
about 1 year ago | |
| a6593bda | Fix bug in export using FQ ONNX in replacing activation holder with more than... |
Ariel Lapid <5****2@u****m>
Committed by: GitHub <n****y@g****m> |
about 1 year ago | |
| 35697d28 | Fix check in model collector for quantized FLN node (#1462) |
Ofir Gordon <o****6@g****m>
Committed by: GitHub <n****y@g****m> |
about 1 year ago | |
| ab5da23f |
Fix FQ exporter for 16bit act quant and constant quantization (#1463)
Co-authored-by: Ofir Gordon <o****o@s****m> |
Elad Cohen <7****c@u****m>
Committed by: GitHub <n****y@g****m> |
about 1 year ago | |
| de182fb8 | Add ability to export Positional weights in fakely quant onnx (#1461) |
Ariel Lapid <5****2@u****m>
Committed by: GitHub <n****y@g****m> |
about 1 year ago | |
| a3af1c7f | Update __init__.py with version 2.4.0 (#1458) |
Reuven Peretz <4****z@u****m>
Committed by: GitHub <n****y@g****m> |
about 1 year ago | |
| 9d456aa9 |
Update docsrc (#1459)
Co-authored-by: reuvenp <r****p@a****m> |
Reuven Peretz <4****z@u****m>
Committed by: GitHub <n****y@g****m> |
about 1 year ago | |
| 8b505bb4 |
Enable activation status collection on FLN node. (#1447)
Co-authored-by: KazunoriSumiya <S****i@j****m>, kawasaki.kenta <k****a@m****m> |
gouda-youichi <1****i@u****m>
Committed by: GitHub <n****y@g****m> |
about 1 year ago | |
| 7e4f105b | Update EPTQ reference in readme (#1457) |
Ofir Gordon <o****6@g****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| 0a75cc2a | Set more informative log in bias correction computation function. (#1456) |
Elad Cohen <7****c@u****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| 981568d3 | Add exponential distance weighting (#1453) |
Irena Byzalov <1****y@u****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| a1f6ec7a | Update custom layers version in requirements to >=1.0 (#1455) |
Ofir Gordon <o****6@g****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| eea3af95 |
Add e2e tests for fusing (#1454)
Co-authored-by: reuvenp <r****p@a****m> |
Reuven Peretz <4****z@u****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| 5539da2f | Extract metrics computation into separate classes in sensitivity evaluation (... |
Irena Byzalov <1****y@u****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| 8621d479 | Fix activation bias correction with preserving node (#1449) |
Ariel Lapid <5****2@u****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| cc8189dd | Search for metadata in model submodules during PyTorch export and assign to m... |
Ariel Lapid <5****2@u****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| df79d745 |
Fix issue with computing max cut when SNC is used (#1435)
Co-authored-by: reuvenp <r****p@a****m> |
Reuven Peretz <4****z@u****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| 0c476158 | Update MCTQ package version in requirements and revert usage of nightly in wo... |
Ofir Gordon <o****6@g****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| 21d1def7 | Use tpc preservig flag (#1444) |
Ofir Gordon <o****6@g****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| 8f2b451b | Schema V1 to V2 (#1443) |
Yarden Yagil <y****l@s****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| 056da80e | Add functionality to FusingInfo class to preserve quantization config (#1426) |
kenta kawasaki <k****a@m****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| 09ed05cc | Use float mp model baseline instead of maxbit configuration (#1441) |
Irena Byzalov <1****y@u****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| 77e3d113 | Add an option to dynamically disable configurable quantizers, separate config... |
Irena Byzalov <1****y@u****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| 5bfd07da | Updated imports with the new custom layer package name (#1440) |
Idan Ben Ami <1****i@u****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| d073525f | Remove self.model.node_sort from export (#1442) |
Yarden Yagil <y****l@s****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| 788e74ae | Add holder for quantization preserving (#1406) |
kawakami-masaki0 <k****i@j****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| 51d2aafd | Multiple Inputs Support | Onnx Opset V20 | Add torch2.6 & Remove torch2.2 (#1... |
Yarden Yagil <y****l@s****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| 10e70d6e | Refactor virtual config reconstruction helper, support separate configuration... |
Irena Byzalov <1****y@u****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| da3ef973 | Update setup.py (#1437) |
Reuven Peretz <4****z@u****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| a3b65d8b | fixed readme banner for Tensorflow versions (#1436) |
Amir Servi <1****M@u****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| 8b6fb22b | Add a flag for TPC schema v2 for enabling quantization preserving quantizers ... |
Ofir Gordon <o****6@g****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| 7a6da34e | Export onnx output names list (#1430) |
Ariel Lapid <5****2@u****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| 8a6fa11e |
Update README.md to remove TensorFlow versions 2.12 and 2.13 from the list of...
Co-authored-by: liord <l****n@a****m> |
Lior Dikstein <7****n@u****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| 357b42b2 | add support for custom metric function for mixed precision (#1420) |
Itai Berman <i****n@s****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| 6345f8d6 |
Integrate max cut on FLNs in mixed-precision (#1427)
Co-authored-by: reuvenp <r****p@a****m> |
Reuven Peretz <4****z@u****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| e70c3598 | Weights/Constant Reuse Bugfix (#1418) |
Yarden Yagil <y****l@s****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| 532f6ee6 | Node Supported Bitwidth Filtering Bugfix (#1422) |
Yarden Yagil <y****l@s****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| 8597082d | fix pypi page images (#1419) |
Chizkiyahu Raful <3****u@u****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| 549c6d99 | Add tests for manual bit width config for postional weights (#1424) |
Ariel Lapid <5****2@u****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| 0dbb7607 | Handle quantization preserving with 2 inputs (#1421) |
Elad Cohen <7****c@u****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| 3e78fce8 | Add tests for quantization preserving handling in resource_utilization_calcul... |
Elad Cohen <7****c@u****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| 8d99193c | Add OpQuantizationConfig to schema v2 of Fusing class (#1414) |
kenta kawasaki <k****a@m****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| 018b3520 |
Integrate manual weights bitwidth selection into the MCT core (#1395)
Co-authored-by: gouda-youichi <g****i@j****m> |
kawakami-masaki0 <k****i@j****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| 9f8df99e | Simplify MP manager - LP solver candidates representation, add MP, LP unit te... |
Irena Byzalov <1****y@u****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| 0282c79f | Add exp to schema v2 (#1416) |
Ariel Lapid <5****2@u****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| 660ba89e |
Remove `pydantic<2.0` pin in requirements.txt (#1398)
Co-authored-by: liord <l****n@a****m> |
Glenn Jocher <g****r@u****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| 19134e7e |
Fix error message formatting in configuration validation (#1412)
Signed-off-by: Emmanuel Ferdman <e****n@g****m> |
Emmanuel Ferdman <e****n@g****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| 2dbd8307 | Schema Compatibility (#1404) |
Yarden Yagil <y****l@s****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| 81e9217e | Add a debug option to bypass MCT facade (#1410) |
Ofir Gordon <o****6@g****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| cc23e5fa | add stack to MP interest points (#1411) |
Irena Byzalov <1****y@u****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| 0420d070 |
Add Pytorch end-to-end test for different quantization error methods. (#1397)
Co-authored-by: liord <l****n@a****m>, Ofir Gordon <o****o@s****m> |
Lior Dikstein <7****n@u****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| 23af7f09 | Support Custom Layers in MCT & Move to FX Tracer (#1385) |
Yarden Yagil <y****l@s****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| be82158b | Use "quant_mode" enum in NodeQuantizationConfig, instead of different flags (... |
Elad Cohen <7****c@u****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| 2d5b07f1 | Update requirements to use mctq nightly (#1409) |
Reuven Peretz <4****z@u****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| f5215260 | Fix bug in positional weights quantization config suuport (#1408) |
Ariel Lapid <5****2@u****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago | |
| 219f0014 | Add Positional weights quantization config support (#1399) |
Ariel Lapid <5****2@u****m>
Committed by: GitHub <n****y@g****m> |
over 1 year ago |