gpu update cast type

mindspore/ops/_op_impl/akg/gpu/cast.py

@@ -23,6 +23,7 @@ cast_op_info = AkgGpuRegOp("Cast") \
     .dtype_format(DataType.F16_Default, DataType.F32_Default) \
     .dtype_format(DataType.F16_Default, DataType.I32_Default) \
     .dtype_format(DataType.F16_Default, DataType.F64_Default) \
+    .dtype_format(DataType.F16_Default, DataType.BOOL_Default) \
     .dtype_format(DataType.I32_Default, DataType.F16_Default) \
     .dtype_format(DataType.I32_Default, DataType.F32_Default) \
     .dtype_format(DataType.I32_Default, DataType.I8_Default) \
@@ -34,6 +35,7 @@ cast_op_info = AkgGpuRegOp("Cast") \
     .dtype_format(DataType.I8_Default, DataType.I32_Default) \
     .dtype_format(DataType.I8_Default, DataType.I16_Default) \
     .dtype_format(DataType.I8_Default, DataType.I64_Default) \
+    .dtype_format(DataType.I8_Default, DataType.BOOL_Default) \
     .dtype_format(DataType.BOOL_Default, DataType.F32_Default) \
     .dtype_format(DataType.BOOL_Default, DataType.F16_Default) \
     .dtype_format(DataType.BOOL_Default, DataType.F64_Default) \
@@ -44,19 +46,29 @@ cast_op_info = AkgGpuRegOp("Cast") \
     .dtype_format(DataType.U8_Default, DataType.F32_Default) \
     .dtype_format(DataType.U8_Default, DataType.F16_Default) \
     .dtype_format(DataType.U8_Default, DataType.I32_Default) \
+    .dtype_format(DataType.U8_Default, DataType.BOOL_Default) \
     .dtype_format(DataType.I16_Default, DataType.F64_Default) \
     .dtype_format(DataType.I16_Default, DataType.F32_Default) \
     .dtype_format(DataType.I16_Default, DataType.F16_Default) \
     .dtype_format(DataType.I16_Default, DataType.I32_Default) \
     .dtype_format(DataType.I16_Default, DataType.I64_Default) \
+    .dtype_format(DataType.I16_Default, DataType.F32_Default) \
+    .dtype_format(DataType.I16_Default, DataType.BOOL_Default) \
     .dtype_format(DataType.I64_Default, DataType.F64_Default) \
     .dtype_format(DataType.I64_Default, DataType.F32_Default) \
     .dtype_format(DataType.I64_Default, DataType.F16_Default) \
     .dtype_format(DataType.I64_Default, DataType.I32_Default) \
-    .dtype_format(DataType.I16_Default, DataType.F32_Default) \
-    .dtype_format(DataType.I16_Default, DataType.F16_Default) \
+    .dtype_format(DataType.I64_Default, DataType.BOOL_Default) \
     .dtype_format(DataType.F32_Default, DataType.I32_Default) \
     .dtype_format(DataType.F32_Default, DataType.F16_Default) \
+    .dtype_format(DataType.F32_Default, DataType.BOOL_Default) \
+    .dtype_format(DataType.F64_Default, DataType.BOOL_Default) \
+    .dtype_format(DataType.F64_Default, DataType.F32_Default) \
+    .dtype_format(DataType.F64_Default, DataType.F16_Default) \
+    .dtype_format(DataType.F64_Default, DataType.I64_Default) \
+    .dtype_format(DataType.F64_Default, DataType.I32_Default) \
+    .dtype_format(DataType.F64_Default, DataType.I16_Default) \
+    .dtype_format(DataType.F64_Default, DataType.I8_Default) \
     .get_op_info()
 
 

tests/st/ops/gpu/test_cast_op.py

@@ -359,3 +359,122 @@ def test_cast18():
     assert type0 == 'float32'
     type1 = output[1].asnumpy().dtype
     assert type1 == 'float16'
+
+@pytest.mark.level0
+@pytest.mark.platform_x86_gpu_training
+@pytest.mark.env_onecard
+def test_cast19():
+    x0 = Tensor(np.arange(24).reshape((4, 3, 2)).astype(np.int8))
+    t0 = mstype.bool_
+    x1 = Tensor(np.arange(24).reshape((4, 3, 2)).astype(np.int16))
+    t1 = mstype.bool_
+
+    context.set_context(mode=context.GRAPH_MODE, device_target='GPU')
+    net = Net(t0, t1)
+    output = net(x0, x1)
+    type0 = output[0].asnumpy().dtype
+    assert type0 == 'bool'
+    type1 = output[1].asnumpy().dtype
+    assert type1 == 'bool'
+
+@pytest.mark.level0
+@pytest.mark.platform_x86_gpu_training
+@pytest.mark.env_onecard
+def test_cast20():
+    x0 = Tensor(np.arange(24).reshape((4, 3, 2)).astype(np.int64))
+    t0 = mstype.bool_
+    x1 = Tensor(np.arange(24).reshape((4, 3, 2)).astype(np.float16))
+    t1 = mstype.bool_
+
+    context.set_context(mode=context.GRAPH_MODE, device_target='GPU')
+    net = Net(t0, t1)
+    output = net(x0, x1)
+    type0 = output[0].asnumpy().dtype
+    assert type0 == 'bool'
+    type1 = output[1].asnumpy().dtype
+    assert type1 == 'bool'
+
+@pytest.mark.level0
+@pytest.mark.platform_x86_gpu_training
+@pytest.mark.env_onecard
+def test_cast21():
+    x0 = Tensor(np.arange(24).reshape((4, 3, 2)).astype(np.float32))
+    t0 = mstype.bool_
+    x1 = Tensor(np.arange(24).reshape((4, 3, 2)).astype(np.float64))
+    t1 = mstype.bool_
+
+    context.set_context(mode=context.GRAPH_MODE, device_target='GPU')
+    net = Net(t0, t1)
+    output = net(x0, x1)
+    type0 = output[0].asnumpy().dtype
+    assert type0 == 'bool'
+    type1 = output[1].asnumpy().dtype
+    assert type1 == 'bool'
+
+@pytest.mark.level0
+@pytest.mark.platform_x86_gpu_training
+@pytest.mark.env_onecard
+def test_cast22():
+    x0 = Tensor(np.arange(24).reshape((4, 3, 2)).astype(np.uint8))
+    t0 = mstype.bool_
+    x1 = Tensor(np.arange(24).reshape((4, 3, 2)).astype(np.int32))
+    t1 = mstype.bool_
+
+    context.set_context(mode=context.GRAPH_MODE, device_target='GPU')
+    net = Net(t0, t1)
+    output = net(x0, x1)
+    type0 = output[0].asnumpy().dtype
+    assert type0 == 'bool'
+    type1 = output[1].asnumpy().dtype
+    assert type1 == 'bool'
+
+@pytest.mark.level0
+@pytest.mark.platform_x86_gpu_training
+@pytest.mark.env_onecard
+def test_cast23():
+    x0 = Tensor(np.arange(24).reshape((4, 3, 2)).astype(np.float64))
+    t0 = mstype.float32
+    x1 = Tensor(np.arange(24).reshape((4, 3, 2)).astype(np.float64))
+    t1 = mstype.float16
+
+    context.set_context(mode=context.GRAPH_MODE, device_target='GPU')
+    net = Net(t0, t1)
+    output = net(x0, x1)
+    type0 = output[0].asnumpy().dtype
+    assert type0 == 'float32'
+    type1 = output[1].asnumpy().dtype
+    assert type1 == 'float16'
+
+@pytest.mark.level0
+@pytest.mark.platform_x86_gpu_training
+@pytest.mark.env_onecard
+def test_cast24():
+    x0 = Tensor(np.arange(24).reshape((4, 3, 2)).astype(np.float64))
+    t0 = mstype.int64
+    x1 = Tensor(np.arange(24).reshape((4, 3, 2)).astype(np.float64))
+    t1 = mstype.int32
+
+    context.set_context(mode=context.GRAPH_MODE, device_target='GPU')
+    net = Net(t0, t1)
+    output = net(x0, x1)
+    type0 = output[0].asnumpy().dtype
+    assert type0 == 'int64'
+    type1 = output[1].asnumpy().dtype
+    assert type1 == 'int32'
+
+@pytest.mark.level0
+@pytest.mark.platform_x86_gpu_training
+@pytest.mark.env_onecard
+def test_cast25():
+    x0 = Tensor(np.arange(24).reshape((4, 3, 2)).astype(np.float64))
+    t0 = mstype.int16
+    x1 = Tensor(np.arange(24).reshape((4, 3, 2)).astype(np.float64))
+    t1 = mstype.int8
+
+    context.set_context(mode=context.GRAPH_MODE, device_target='GPU')
+    net = Net(t0, t1)
+    output = net(x0, x1)
+    type0 = output[0].asnumpy().dtype
+    assert type0 == 'int16'
+    type1 = output[1].asnumpy().dtype
+    assert type1 == 'int8'