Update TF dependency to 2.12rc0 (#353)

* update to 2.11rc2

* update fixedpoint

* update includes

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* update to 2.11

* set jit_compile=False for all optimizers

* format

* test nightly wheel

* update

* training test update

* update to 2.12rc0

* update

* fix scan ops test

* fix slice op test

* fix and update split op test

* fix and update stack op test

* fix and update concat op test

* remove python 3.7

* fix profiler test

* wip

* fix array ops test

* update array ops test

* add python 3.11

* make gpu name cmd line arg

* typo

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* update image ops test, clipbyvalue, format

* fix includes

* fix condition

* small fixes

* updates for code checks

* remove tf TODO and more linting fixes

* remove platform_types

* keep spaces in gpu name

.github/workflows/github-actions-python-check.yml

@@ -12,10 +12,10 @@ jobs:
     runs-on: ubuntu-latest
     steps:
     - uses: actions/checkout@v2
-    - name: Set Up Python 3.7
+    - name: Set Up Python 3.9
       uses: actions/setup-python@v2
       with:
-        python-version: '3.7'
+        python-version: '3.9'
     - run: pip install -r requirements.txt
     - run: "black . --check --diff --verbose"
     - run: pylint *.py

BUILD.md

@@ -21,7 +21,7 @@ This document contains instructions for producing private builds of tensorflow-d
 Your machine must have a few tools installed before you can get started.
 
 - CMake.
-- A Python 3.7, 3.8, 3.9 or 3.10 environment with the `wheel` package installed.
+- A Python 3.8, 3.9, 3.10 or 3.11 environment with the `wheel` package installed.
 - Build tools appropriate for your platform (MSVC on Windows, Clang on Linux).
 - Linux: a glibc-based distro, like Ubuntu.
 
@@ -65,7 +65,7 @@ This project requires a Python interpreter to produce the TFDML plugin wheel. Th
 
 Feel free to use any version of Python, but we recommend [Miniconda](https://docs.conda.io/en/latest/miniconda.html) to sandbox your project environments. Miniconda is a bundle that includes [Python](https://www.python.org/), a package and environment manager named [Conda](https://docs.conda.io/projects/conda/en/latest/), and a very small set of Python packages. It is a lighter-weight alternative to Anaconda, which contains hundreds of Python packages that aren't necessary for building. See the [user guide](https://docs.conda.io/projects/conda/en/latest/user-guide/getting-started.html) for more details if you're not familiar with its usage.
 
-Once you've installed Miniconda you will want to create and activate a Python 3.7, 3.8, 3.9 or 3.10 environment with the required packages Use this activated environment when building this project.
+Once you've installed Miniconda you will want to create and activate a Python 3.8, 3.9, 3.10, or 3.11 environment with the required packages Use this activated environment when building this project.
 
 **Windows**:
 

README.md

@@ -16,7 +16,7 @@ You can also contact us directly at [email protected].
 
 TensorFlow DirectML Plugin is in early development and is not supported for production yet. For production scenarios, use [TensorFlow 1.15 with DirectML](https://github.com/microsoft/tensorflow-directml) instead.
 
-TensorFlow DirectML Plugin only works with the `tensorflow-cpu==2.10` package, not `tensorflow` or `tensorflow-gpu`. To install the package, run the following commands:
+TensorFlow DirectML Plugin only works with the `tensorflow-cpu>=2.12` package, not `tensorflow` or `tensorflow-gpu`. To install the package, run the following commands:
 
 ```sh
 pip install tensorflow-directml-plugin
@@ -34,7 +34,7 @@ The following resources provide additional background on DirectML and TensorFlow
 ### Windows 10
 
 * Windows 10 Version 1709, 64-bit (Build 16299 or higher)
-* Python x86-64 3.7, 3.8, 3.9 or 3.10<sup>1</sup>
+* Python x86-64 3.8, 3.9, 3.10 or 3.11<sup>1</sup>
 * One of the following supported GPUs:
   * AMD Radeon R5/R7/R9 2xx series or newer
   * Intel HD Graphics 5xx or newer
@@ -43,7 +43,7 @@ The following resources provide additional background on DirectML and TensorFlow
 ### Windows Subsystem for Linux
 
 * Windows 10 Insider Preview, 64-bit (Build 20150 or higher)
-* Python x86-64 3.7, 3.8, 3.9 or 3.10<sup>2</sup>
+* Python x86-64 3.8, 3.9, 3.10 or 3.11<sup>2</sup>
 * One of the following supported GPUs:
   * AMD Radeon R5/R7/R9 2xx series or newer, and [20.20.01.05 WSL driver](https://www.amd.com/en/support/kb/release-notes/rn-rad-win-wsl-support)
   * Intel HD Graphics 6xx or newer, and [28.20.100.8322 WSL driver](https://downloadcenter.intel.com/download/29526)

cmake/dependencies.cmake

@@ -24,8 +24,8 @@ if(WIN32)
     # URLs and SHA256 hashes for the cp37 versions. For stable builds, go to https://pypi.org/project/tensorflow-intel/#files instead.
     FetchContent_Declare(
         tensorflow_whl
-        URL https://files.pythonhosted.org/packages/4a/43/b6c853d9e9532b9dd04191383b76fe5ff7b7dfa40d2be2af98f40c7536ce/tensorflow_intel-2.10.0-cp37-cp37m-win_amd64.whl
-        URL_HASH SHA256=e6096b3a84bbc57eab88cc39ad7d6148e400a0da6a448f94d5e84c8cdbb521b5
+        URL https://files.pythonhosted.org/packages/4c/57/3b37cf30bf1549e617f8328c10b1f3237bf50c9ac7be3de6fcb4c73a19ef/tensorflow_intel-2.12.0rc0-cp39-cp39-win_amd64.whl
+        URL_HASH SHA256=4f6793e0d9238b2fd57f10f39d7b3ab38000e2fca3144c260f581665be35a971
     )
 
     FetchContent_Declare(
@@ -38,8 +38,8 @@ else()
     # URLs and SHA256 hashes for the cp37 versions. For stable builds, go to https://pypi.org/project/tensorflow-cpu/#files instead.
     FetchContent_Declare(
         tensorflow_whl
-        URL https://files.pythonhosted.org/packages/9b/4b/293733cd86f09760b14517f077f79bad16de4af11828273eda1722e9b03a/tensorflow_cpu-2.10.0-cp37-cp37m-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
-        URL_HASH SHA256=34477fdfa97045c854fde55381cc3eb7fa6c63f5958ccf30ae3a21513ee73031
+        URL https://files.pythonhosted.org/packages/61/f7/5888bb138d8ae9c6400996c3f0ffc2d9034c5213d7c6aa6deea79bbfe2c3/tensorflow_cpu-2.12.0rc0-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
+        URL_HASH SHA256=505921b47561f7b26fef8d1e5d781c4afd9aaba8945b02258d36037f70cdfbef
     )
 
     FetchContent_Declare(
@@ -209,7 +209,6 @@ function(tf_proto_cpp proto_path)
 endfunction()
 
 # Generate the necessary .proto files in the TF wheel (performed at build time).
-tf_proto_cpp(tensorflow/core/profiler/protobuf/xplane.proto)
 tf_proto_cpp(tensorflow/core/framework/graph.proto)
 tf_proto_cpp(tensorflow/core/framework/function.proto)
 tf_proto_cpp(tensorflow/core/framework/attr_value.proto)
@@ -224,6 +223,8 @@ tf_proto_cpp(tensorflow/core/framework/versions.proto)
 tf_proto_cpp(tensorflow/core/framework/kernel_def.proto)
 tf_proto_cpp(tensorflow/core/grappler/costs/op_performance_data.proto)
 tf_proto_cpp(tensorflow/core/protobuf/device_properties.proto)
+tf_proto_cpp(tensorflow/core/profiler/protobuf/xplane.proto)
+tf_proto_cpp(tensorflow/tsl/profiler/protobuf/xplane.proto)
 
 # A python interpreter is required to produce the plugin wheel. This python environment
 # must have the 'wheel' package installed.

pipelines/pipeline.yml

@@ -40,16 +40,16 @@ parameters:
   displayName: Build Artifacts
   type: object
   default:
-  - x64-win-release-cp37
-  - x64-win-debug-cp37
+  - x64-win-debug-cp39
   - x64-win-release-cp38
   - x64-win-release-cp39
   - x64-win-release-cp310
-  - x64-linux-release-cp37
-  - x64-linux-debug-cp37
+  - x64-win-release-cp311
+  - x64-linux-debug-cp39
   - x64-linux-release-cp38
   - x64-linux-release-cp39
   - x64-linux-release-cp310
+  - x64-linux-release-cp311
 
 - name: enableTelemetry
   displayName: Enable Telemetry
@@ -59,7 +59,7 @@ parameters:
 - name: testArtifacts
   displayName: Test Artifacts
   type: object
-  default: [x64-win-release-cp37, x64-linux-release-cp37]
+  default: [x64-win-release-cp39, x64-linux-release-cp39]
 
 - name: testGroups
   displayName: Test Groups
@@ -74,12 +74,12 @@ parameters:
 - name: testTensorflowVersion
   displayName: TensorFlow Package Version
   type: string
-  default: 2.10.0
+  default: 2.12.0rc0
 
 - name: testKerasPackage
   displayName: Keras Package Version
   type: string
-  default: keras==2.10.0
+  default: keras==2.12.0rc0
 
 - name: enableTests
   displayName: Enable Tests

pipelines/test.yml

@@ -19,10 +19,10 @@ parameters:
   default: [tensorflow-cpu]
 - name: tensorflowVersion
   type: string
-  default: 2.10.0
+  default: 2.12.0rc0
 - name: kerasPackage
   type: string
-  default: keras==2.10.0
+  default: keras==2.12.0rc0
 - name: pluginBuildPipeline
   type: string
   default: current
@@ -153,10 +153,12 @@ jobs:
             script: |
               $ScriptPath = "$(Build.SourcesDirectory)/test/test.py" -replace '\\','/'
               $ScriptPathWsl = wsl wslpath -a $ScriptPath
+              $GPUInfo = wmic path win32_VideoController get name
+
               $ResultsDirWin = "$(System.ArtifactsDirectory)/results/$(agentName)/${{artifact}}/${{tensorflowPackage}}" -replace '\\','/'
               $ResultsDirWinWsl = wsl wslpath -a $ResultsDirWin
               New-Item -ItemType Directory $ResultsDirWin -Force | Out-Null
-              cmd /c "wsl $(test_env_linux_${{replace(tensorflowPackage, '-', '_')}}.activateCommand); python $ScriptPathWsl --run --parallel --summarize --redirect_output --results_dir $ResultsDirWinWsl --groups ${{join(' ', parameters.testGroups)}}"
+              cmd /c "wsl $(test_env_linux_${{replace(tensorflowPackage, '-', '_')}}.activateCommand); python $ScriptPathWsl --run --parallel --summarize --redirect_output --results_dir $ResultsDirWinWsl --groups ${{join(' ', parameters.testGroups)}} --gpu_name '$GPUInfo'"
 
       - ${{if contains(artifact, '-win-') }}:
         - task: PowerShell@2
@@ -180,7 +182,8 @@ jobs:
             script: |
               Invoke-Expression '$(test_env_win_${{replace(tensorflowPackage, '-', '_')}}.activateCommand)'
               $ResultsDir = "$(System.ArtifactsDirectory)/results/$(agentName)/${{artifact}}/${{tensorflowPackage}}"
-              python $(Build.SourcesDirectory)/test/test.py --run --parallel --summarize --redirect_output --results_dir $ResultsDir --groups ${{join(' ', parameters.testGroups)}}
+              $GPUInfo = wmic path win32_VideoController get name
+              python $(Build.SourcesDirectory)/test/test.py --run --parallel --summarize --redirect_output --results_dir $ResultsDir --groups ${{join(' ', parameters.testGroups)}} --gpu_name "$GPUInfo"
 
     - task: PublishBuildArtifacts@1
       displayName: Publish ${{artifact}}

requirements.txt

@@ -1,6 +1,6 @@
 black==22.6.0
 pylint==2.14.5
 tensorboard_plugin_profile==2.8.0
-tensorflow-cpu==2.10.0
+tensorflow-cpu==2.12.0rc0
 vswhere==1.4.0 ; sys_platform == 'win32'
 portpicker==1.5.2

test/README.md

@@ -1,6 +1,6 @@
 # Testing TensorFlow-DirectML-Plugin
 
-This directory contains tests for validating the DirectML plugin. Testing is done with a combination of python scripts and native test executables. The python tests should be run in a python environment with the `tensorflow-directml-plugin` and `tensorflow-cpu == 2.10.0` packages installed. All testing is driven by the following two files:
+This directory contains tests for validating the DirectML plugin. Testing is done with a combination of python scripts and native test executables. The python tests should be run in a python environment with the `tensorflow-directml-plugin` and `tensorflow-cpu >= 2.12.0` packages installed. All testing is driven by the following two files:
 
 - [test.py](test.py) : main script that will execute all test content (python and native tests) and summarize the results. Uses tests.json to drive testing.
 - [tests.json](tests.json) : describes the test content and how it should be executed.