b3eba077b8
DxDispatch: support ONNX models with non-tensor outputs ( #313 )
2022-10-24 14:04:35 -07:00
74c80a9539
Minor fixes for Xbox build for DXD ( #304 )
2022-10-03 08:49:48 -07:00
bf3c23d562
DxDispatch: fix ONNX models with dynamic output shapes ( #300 )
2022-09-26 18:23:12 -07:00
4e48e631bb
Fix implementation of DMLCalcBufferTensorSize for buffers of size >4GB ( #296 )
2022-09-20 15:25:46 -07:00
ec1444c7fc
Update DML version for samples and release notes ( #292 )
...
* Add release note for DML1.9.1
* Updated dml version to 1.9.1
* Refactored change list for 1.9.1
* Updated DML version to 1.9.1 for PyDirectML
* Update directml nuget file hash and verison for dxdispatch
* Fixed Typos
Co-authored-by: Sumit Agarwal <sumitagarwal@microsoft.com>
2022-09-15 19:02:55 -07:00
ef779b6373
DxDispatch: ONNX free dim overrides by denotation ( #289 )
2022-09-06 21:12:39 -07:00
4a144773c8
Build pipeline for DxDispatch ( #288 )
2022-09-02 17:07:24 -07:00
f69be4ff1f
Add `$schema` to `cgmanifest.json` ( #285 )
...
Co-authored-by: Jamie Magee <jamie.magee@microsoft.com>
2022-09-01 17:54:18 -07:00
d1bb24b314
Add ConvolutionInteger & QuantizedLinearConvolution to DirectMLX.h ( #283 )
2022-08-22 17:48:56 -07:00
e601d8be06
Upgrade dxdispatch dependency versions ( #277 )
2022-08-16 20:46:05 -07:00
b3aa57d709
Minor Enhancements to dx-dispatch ( #274 )
2022-08-09 10:35:08 -07:00
9d05178134
Add API for DML_OPERATOR_ELEMENT_WISE_NEGATE to DirectMLX.h ( #275 )
2022-08-08 19:32:59 -07:00
bc01d97e8d
Update samples to use DML 1.9.0, and other small additions ( #268 )
...
*Update DirectML samples to use DML version 1.9.0
*Add metadata to readmes for sample browser automation
*Fix a couple of warnings when building yolov4 sample in VS2022
2022-07-20 15:43:52 -07:00
bb6e396d87
Update Reduce in DirectMLX.h to take in output type for argmin/argmax ( #266 )
...
* update reduce to take in output type for argmin/argmax
* address comments
* output type defaults to unknown
* swap if conditions
2022-07-20 14:27:11 -07:00
91dc865bf5
Update releases.md for DML 1.9 ( #265 )
2022-07-13 14:06:47 -07:00
64cf213c29
Update dxdispatch to DML 1.9 ( #264 )
2022-07-08 19:28:27 -07:00
b9cb7d3989
onnxdispatchable: fix error message formatting ( #263 )
...
when throwing an error for unbound tensor, the '%s' format is not
interpreted by fmt::format
Signed-off-by: Gabriel Ionescu <gabriel.ionescu@intel.com>
2022-07-05 15:12:45 -07:00
3b28a13a51
Fix inconsistent ORT sessions in dxdispatch ( #260 )
2022-06-30 17:56:49 -07:00
8e0587fe01
DxDispatch programmatic capture & ONNX models options ( #252 )
2022-06-30 14:10:26 -07:00
3c5a947e0e
Update TF2 SqueezeNet readme ( #257 )
...
* update readme
* change tf-nightly to tensorflow
2022-06-22 08:56:03 -07:00
7b0f1b6e64
Microsoft mandatory file ( #254 )
...
Co-authored-by: microsoft-github-policy-service[bot] <77245923+microsoft-github-policy-service[bot]@users.noreply.github.com>
2022-06-16 12:21:38 -07:00
df188a34ce
increase default batch size and set optimizer parameters ( #253 )
2022-06-15 15:54:53 -07:00
428a5b523d
Update Releases.md for DML 1.8.2 ( #238 )
...
* Update Releases.md for DML 1.8.2
2022-05-05 18:29:49 -07:00
d16b197c7d
Change ReturnIfFailed to ThrowIfFailed in Python\device.cpp ( #232 )
2022-05-04 09:59:10 -07:00
87c2501f97
Add object detection sample ( #228 )
2022-04-25 16:25:44 -07:00
7735d8ef1a
Use GPU instead of DML to identify DML devices in SqueezeNet model ( #236 )
2022-04-25 12:43:20 -07:00
5f959bad7c
Update Releases.md for 1.8.1 ( #233 )
2022-04-22 13:39:28 -07:00
1e6b2d8723
Support enumerating graphics adapters based on a given GPU preference ( #217 )
...
Thanks!
2022-04-19 13:37:16 -07:00
09f0262813
Add TF2 versions of TF sample models ( #225 )
...
* move tf sample models into tf1 and tf2 folders
* add tf2 versions of sample models
* remove unused file
* update README
* update yolov3
2022-04-07 16:36:37 -07:00
c3f16a701b
gemm example ( #221 )
2022-03-24 16:55:16 -07:00
7aa96917f0
Misc. improvements to DirectMLX.h ( #219 )
2022-03-22 16:16:36 -07:00
aeeef33769
Modify the creation method of input nodes to prevent missing optional inputs. ( #210 )
2022-03-22 12:17:23 -07:00
d4657006a6
Add ability to explicitly specify output sizes for Average/MaxPooling ( #206 )
2022-02-24 10:35:53 -08:00
66d704bdf7
Minor YOLO sample edits to speed up CPU post-processing ( #200 )
2022-02-04 19:10:57 -08:00
57d4cec473
Clean up CMake files for dxdispatch ( #199 )
2022-02-01 11:50:35 -08:00
1cc8ec573e
Add DxDispatch tool ( #198 )
2022-01-27 21:39:53 -08:00
adfa96d350
Update component governance pipeline with updated build pool ( #187 )
2022-01-10 10:44:00 -08:00
13649f718d
Add pytorch yolov3 sample ( #183 )
2021-12-29 09:34:33 -08:00
9068b0ac59
Update sample scripts to train torchvision classification models ( #185 )
2021-12-27 23:29:47 -08:00
ced7b7c8f1
Update Releases.md for DML 1.8.0 ( #178 )
...
* Update Releases.md for DML 1.8.0
* Add row for 1.6
2021-12-02 13:17:49 -08:00
8baf9fd3c1
Enable fast DML resource allocation and residency management. ( #160 )
2021-11-12 15:39:40 -08:00
eb663f7b5b
Fix documentation around how to install pytorch-directml ( #170 )
...
Co-authored-by: Sheil Kumar <sheilk@microsoft.com>
2021-10-26 12:40:37 -07:00
db73b04a61
update link ( #163 )
...
Co-authored-by: Sheil Kumar <sheilk@microsoft.com>
2021-10-21 12:44:17 -07:00
aa904f774e
Introducing PyTorch+DirectML Samples ( #161 )
...
* Main readme updates.
* Fix typo
* Add squeezenet and resnet50 docs
* Add predict
* Minor updates
* install scripts
Co-authored-by: Sheil Kumar <sheilk@microsoft.com>
2021-10-21 11:58:05 -07:00
6681088a94
Fix MSVC PyDML build. ( #158 )
2021-10-20 16:17:25 -07:00
17603b782b
Upgraded DirectML dependency to 1.7.0. Fixed the module export of the overloaded dml::Add function. MeanVarianceNormalization helper has moved to use ONNX MVN-9, remapped the module export accordingly. ( #159 )
2021-10-18 10:01:02 -07:00
6dba666180
Extend DirectMLX to allow different types of optional and also implement NonZeroCoordinates ( #155 )
...
* Changes for linux and add nonzero coordinates
* Add c10 support
* Use optional extensions
* Add optional
* Add option to extend optional
* Remove guids
* remove msc_ver
* Spacing
* PR comment
* PR comments
* PR nit comments
* Add space after for
* Styling
* PR comment
* Small PR comments
2021-10-04 16:09:43 -07:00
33d0d669b8
Remove "Preview" from TensorFlow-DirectML section ( #152 )
...
Remove "Preview" from TensorFlow-DirectML section
2021-09-20 16:35:19 -07:00
9039532f99
Update NOTICES for keras-team/keras-io and tensorflow/text ( #150 )
2021-09-17 13:32:47 -07:00
266aaa69c7
Vision Transformers Sample t-jaidevshah ( #148 )
2021-09-13 16:57:19 -07:00
Justin Stoecker
Adrian Tsai
sumitsays
Jamie Magee
Carson Swope
Ankan Banerjee
Patrice Vignola
maggie1059
Dwayne Robinson
gbi
microsoft-github-policy-service[bot]
mingmingtasd
Xiang Zhang
Bin Miao
Bryan B
Sheil Kumar
Chai Chaoweeraprasit
Ryan Lai
Jaidev Shah