【PaddlePaddle Hackathon】49. Add test of paddle.nn.functional.conv2d transpose #251
Conversation
…into add-test-of-paddle.nn.functional.conv2d_transpose
OccupyMars2025
changed the title
OccupyMars2025
changed the title
| groups=groups, | ||
| output_size=None, | ||
| data_format="NHWC", | ||
| ) |
There was a problem hiding this comment.
需要增加一些异常输入类的case，比如某个参数的值不在规定范围内
There was a problem hiding this comment.
好的，收到。我增加了两个obj.exception的异常输入类的case
…ttps://github.com/SmirnovKol/PaddleTest into add-test-of-paddle.nn.functional.conv2d_transpose
| """ | ||
| implement | ||
| """ | ||
| self.types = [np.float32] |
There was a problem hiding this comment.
在float16下会报错，如下所示：
E RuntimeError: (NotFound) Operator conv2d_transpose does not have kernel for data_type[::paddle::platform::float16]:data_layout[ANY_LAYOUT]:place[CPUPlace]:library_type[PLAIN].
E [Hint: Expected kernel_iter != kernels.end(), but received kernel_iter == kernels.end().] (at C:\home\workspace\Paddle_release\paddle\fluid\imperative\prepared_operator.cc:135)
E [operator < conv2d_transpose > error]
There was a problem hiding this comment.
在float64下可以测试通过
| weight = np.full(shape=[in_channels, out_channels] + kernel_size, fill_value=3.3) | ||
| bias = np.full(shape=[out_channels], fill_value=-1.7) | ||
| stride = 1 | ||
| padding = 1 |
There was a problem hiding this comment.
padding需要多测一下，可以参照文档中的描述：
padding (int|list|tuple|str，可选) - 填充大小。如果它是一个字符串，可以是"VALID"或者"SAME"，表示填充算法，计算细节可参考上述 padding = "SAME"或 padding = "VALID" 时的计算公式。如果它是一个元组或列表，它可以有3种格式：(1)包含4个二元组：当 data_format 为"NCHW"时为 [[0,0], [0,0], [padding_height_top, padding_height_bottom], [padding_width_left, padding_width_right]]，当 data_format 为"NHWC"时为[[0,0], [padding_height_top, padding_height_bottom], [padding_width_left, padding_width_right], [0,0]]；(2)包含4个整数值：[padding_height_top, padding_height_bottom, padding_width_left, padding_width_right]；(3)包含2个整数值：[padding_height, padding_width]，此时padding_height_top = padding_height_bottom = padding_height， padding_width_left = padding_width_right = padding_width。若为一个整数，padding_height = padding_width = padding。默认值：0。
There was a problem hiding this comment.
已增加更多关于padding的测试
add more test cases of the padding argument
PR types
New features
PR changes
add framework/api/nn/test_functional_conv2d_transpose.py
Describe
Task: PaddlePaddle/Paddle#35956
add pytest of paddle.nn.functional.conv2d_transpose