[ci] Try to support paddle

.github/workflows/scripts/unix_test.sh

@@ -22,11 +22,17 @@ python3 -m pip install dist/*.whl
 if [ -z "$GPU_TEST" ]; then
     python3 -m pip install -r requirements_test.txt
     python3 -m pip install "torch; python_version < '3.10'"
+    if [[ $PLATFORM == *"linux"* ]]; then
+        python3 -m pip install "paddlepaddle==0.0.0; python_version < '3.10'" -f https://www.paddlepaddle.org.cn/whl/linux/cpu-mkl/develop.html
+    fi
 else
     ## Only GPU machine uses system python.
     export PATH=$PATH:$HOME/.local/bin
     # pip will skip packages if already installed
     python3 -m pip install -r requirements_test.txt
+    if [[ $PLATFORM == *"linux"* ]]; then
+        python3 -m pip install "paddlepaddle-gpu==0.0.0.post112; python_version < '3.10'" -f https://www.paddlepaddle.org.cn/whl/linux/gpu/develop.html
+    fi
 fi
 ti diagnose
 ti changelog
@@ -38,27 +44,35 @@ TI_LIB_DIR="$TI_PATH/_lib/runtime" ./build/taichi_cpp_tests
 if [ -z "$GPU_TEST" ]; then
     if [[ $PLATFORM == *"m1"* ]]; then
 	# Split per arch to avoid flaky test
-        python3 tests/run_tests.py -vr2 -t4 -k "not torch" -a cpu
+        python3 tests/run_tests.py -vr2 -t4 -k "not torch and not paddle" -a cpu
         # Run metal and vulkan separately so that they don't use M1 chip simultaneously.
-        python3 tests/run_tests.py -vr2 -t4 -k "not torch" -a vulkan
-        python3 tests/run_tests.py -vr2 -t2 -k "not torch" -a metal
+        python3 tests/run_tests.py -vr2 -t4 -k "not torch and not paddle" -a vulkan
+        python3 tests/run_tests.py -vr2 -t2 -k "not torch and not paddle" -a metal
         python3 tests/run_tests.py -vr2 -t1 -k "torch" -a "$TI_WANTED_ARCHS"
+        # Paddle's paddle.fluid.core.Tensor._ptr() is only available on develop branch
+        # Not support Apple M1 and no GPU version on MAC
     else
-        python3 tests/run_tests.py -vr2 -t4 -a "$TI_WANTED_ARCHS"
+        python3 tests/run_tests.py -vr2 -t4 -k "not paddle" -a "$TI_WANTED_ARCHS"
+        if [[ $PLATFORM == *"linux"* ]]; then
+            python3 tests/run_tests.py -vr2 -t4 -k "paddle" -a "$TI_WANTED_ARCHS"
+        fi
     fi
 else
     # Split per arch to increase parallelism for linux GPU tests
     if [[ $TI_WANTED_ARCHS == *"cuda"* ]]; then
-        python3 tests/run_tests.py -vr2 -t4 -k "not torch" -a cuda
+        python3 tests/run_tests.py -vr2 -t4 -k "not torch and not paddle" -a cuda
     fi
     if [[ $TI_WANTED_ARCHS == *"cpu"* ]]; then
-        python3 tests/run_tests.py -vr2 -t8 -k "not torch" -a cpu
+        python3 tests/run_tests.py -vr2 -t8 -k "not torch and not paddle" -a cpu
     fi
     if [[ $TI_WANTED_ARCHS == *"vulkan"* ]]; then
-        python3 tests/run_tests.py -vr2 -t8 -k "not torch" -a vulkan
+        python3 tests/run_tests.py -vr2 -t8 -k "not torch and not paddle" -a vulkan
     fi
     if [[ $TI_WANTED_ARCHS == *"opengl"* ]]; then
-        python3 tests/run_tests.py -vr2 -t4 -k "not torch" -a opengl
+        python3 tests/run_tests.py -vr2 -t4 -k "not torch and not paddle" -a opengl
     fi
     python3 tests/run_tests.py -vr2 -t1 -k "torch" -a "$TI_WANTED_ARCHS"
+    if [[ $PLATFORM == *"linux"* ]]; then
+        python3 tests/run_tests.py -vr2 -t1 -k "paddle" -a "$TI_WANTED_ARCHS"
+    fi
 fi

.github/workflows/scripts/win_test.ps1

@@ -9,20 +9,23 @@ pip install -r requirements_test.txt
 # TODO relax this when torch supports 3.10
 if ("$env:TI_WANTED_ARCHS".Contains("cuda")) {
     pip install "torch==1.10.1+cu113; python_version < '3.10'" -f https://download.pytorch.org/whl/cu113/torch_stable.html
+    pip install paddlepaddle-gpu==0.0.0.post112 -f https://www.paddlepaddle.org.cn/whl/windows/gpu/develop.html
 } else {
     pip install "torch; python_version < '3.10'"
+    pip install paddlepaddle==0.0.0 -f https://www.paddlepaddle.org.cn/whl/windows/cpu-mkl-avx/develop.html
 }
 if ("$env:TI_WANTED_ARCHS".Contains("cuda")) {
-  python tests/run_tests.py -vr2 -t4 -k "not torch" -a cuda
+  python tests/run_tests.py -vr2 -t4 -k "not torch and not paddle" -a cuda
   if (-not $?) { exit 1 }
 }
 if ("$env:TI_WANTED_ARCHS".Contains("cpu")) {
-  python tests/run_tests.py -vr2 -t6 -k "not torch" -a cpu
+  python tests/run_tests.py -vr2 -t6 -k "not torch and not paddle" -a cpu
   if (-not $?) { exit 1 }
 }
 if ("$env:TI_WANTED_ARCHS".Contains("opengl")) {
-  python tests/run_tests.py -vr2 -t4 -k "not torch" -a opengl
+  python tests/run_tests.py -vr2 -t4 -k "not torch and not paddle" -a opengl
   if (-not $?) { exit 1 }
 }
 python tests/run_tests.py -vr2 -t2 -k "torch" -a "$env:TI_WANTED_ARCHS"
+python tests/run_tests.py -vr2 -t2 -k "paddle" -a "$env:TI_WANTED_ARCHS"
 if (-not $?) { exit 1 }

ci/scripts/ubuntu_build_test.sh

@@ -24,12 +24,14 @@ git checkout $SHA
 python3 -m pip install -r requirements_dev.txt -i http://repo.taichigraphics.com/repository/pypi/simple --trusted-host repo.taichigraphics.com
 # Update Torch version, otherwise cuda tests fail. See #2969.
 python3 -m pip install torch==1.9.0+cu111 -f https://download.pytorch.org/whl/torch_stable.html -i http://repo.taichigraphics.com/repository/pypi/simple --trusted-host repo.taichigraphics.com
+# Paddle's paddle.fluid.core.Tensor._ptr() is only available on PaddlePaddle's develop branch
+python3 -m pip install paddlepaddle-gpu==0.0.0.post112 -f https://www.paddlepaddle.org.cn/whl/linux/gpu/develop.html
 TAICHI_CMAKE_ARGS="-DTI_WITH_VULKAN:BOOL=ON -DTI_WITH_CUDA:BOOL=ON -DTI_WITH_OPENGL:BOOL=ON" python3 setup.py install
 
 # Add Docker specific ENV
 export TI_IN_DOCKER=true
 
 # Run tests
 ti diagnose
-python tests/run_tests.py -vr2 -t2 -k "not ndarray and not torch"
-python tests/run_tests.py -vr2 -t1 -k "ndarray or torch"
+python tests/run_tests.py -vr2 -t2 -k "not ndarray and not torch and not paddle"
+python tests/run_tests.py -vr2 -t1 -k "ndarray or torch or paddle"

ci/scripts/ubuntu_build_test_cpu.sh

@@ -22,12 +22,14 @@ git clone --recursive https://github.com/taichi-dev/taichi --branch=master
 cd taichi
 git checkout $SHA
 python3 -m pip install -r requirements_dev.txt -i http://repo.taichigraphics.com/repository/pypi/simple --trusted-host repo.taichigraphics.com
+# Paddle's paddle.fluid.core.Tensor._ptr() is only available on PaddlePaddle's develop branch
+python3 -m pip install paddlepaddle==0.0.0 -f https://www.paddlepaddle.org.cn/whl/linux/cpu-mkl/develop.html
 TAICHI_CMAKE_ARGS="-DTI_WITH_VULKAN:BOOL=OFF -DTI_WITH_CUDA:BOOL=OFF -DTI_WITH_OPENGL:BOOL=OFF" python3 setup.py install
 
 # Add Docker specific ENV
 export TI_IN_DOCKER=true
 
 # Run tests
 ti diagnose
-python tests/run_tests.py -vr2 -t2 -k "not ndarray and not torch"
-python tests/run_tests.py -vr2 -t1 -k "ndarray or torch"
+python tests/run_tests.py -vr2 -t2 -k "not ndarray and not torch and not paddle"
+python tests/run_tests.py -vr2 -t1 -k "ndarray or torch or paddle"

ci/windows/win_build_test.ps1

@@ -59,5 +59,5 @@ python setup.py develop
 WriteInfo("Build finished")
 
 WriteInfo("Testing Taichi")
-python tests/run_tests.py -vr2 -t2 -k "not torch" -a cpu
+python tests/run_tests.py -vr2 -t2 -k "not torch and not paddle" -a cpu
 WriteInfo("Test finished")

python/taichi/lang/field.py

@@ -298,9 +298,8 @@ def to_paddle(self, device=None):
         import paddle  # pylint: disable=C0415
 
         # pylint: disable=E1101
-        arr = paddle.zeros(size=self.shape,
-                           dtype=to_paddle_type(self.dtype),
-                           device=device)
+        arr = paddle.zeros(shape=self.shape,
+                           dtype=to_paddle_type(self.dtype))
         from taichi._kernels import tensor_to_ext_arr  # pylint: disable=C0415
         tensor_to_ext_arr(self, arr)
         taichi.lang.runtime_ops.sync()

python/taichi/lang/kernel_impl.py

@@ -676,7 +676,7 @@ def func__(*args):
                             actual_argument_slot, int(tmp.data_ptr()),
                             tmp.element_size() * tmp.nelement(), v.shape)
                     else:
-                        # For now, paddle.fluid.core.Tensor._ptr() is only available on PaddlePaddle's develop branch
+                        # For now, paddle.fluid.core.Tensor._ptr() is only available on develop branch
                         tmp, paddle_callbacks = self.get_paddle_callbacks(
                             v, has_pp)
                         callbacks += paddle_callbacks

tests/python/test_get_external_tensor_shape.py

@@ -83,7 +83,7 @@ def test_get_external_tensor_shape_access_paddle(size):
     def func(x: ti.types.ndarray(), index: ti.template()) -> ti.i32:
         return x.shape[index]
 
-    x_hat = paddle.ones(size, dtype=paddle.int32, device=paddle.CPUPlace())
+    x_hat = paddle.ones([size], dtype=paddle.int32)
     for idx, y_ref in enumerate(size):
         y_hat = func(x_hat, idx)
         assert y_ref == y_hat, "Size of axis {} should equal {} and not {}.".format(