【Hackathon 5th No.14】Add combinations API to Paddle #57792
Conversation
|
你的PR提交成功，感谢你对开源项目的贡献! |
Patrick-Star125
changed the title
|
有中文RFC文档吗 |
|
|
Sorry to inform you that a02bc01's CIs have passed for more than 7 days. To prevent PR conflicts, you need to re-run all CIs manually. |
|
codestyle-check流水线失败。安装pre-commit后，重新提交。 |
| def modify_setting(self): | ||
| self.dtype_np = 'int64' | ||
| self.x_shape = [10] | ||
| self.r = 0 |
| class TestIndexFillAPI2(TestCombinationsAPIBase): | ||
| def modify_setting(self): | ||
| self.dtype_np = 'int64' | ||
| self.x_shape = [10] |
|
增加了输入shape为[0]的测试同时r>x_shape的测试 |
python/paddle/tensor/math.py
Outdated
|
|
||
| """ | ||
| if len(x.shape) != 1: | ||
| raise TypeError(f"Expect a 1-D vector, but got x shape {x.shape}") | ||
| if not isinstance(r, int) or r < 0: | ||
| raise ValueError(f"Expect a non-negative int, but got r={r}") | ||
|
|
||
| if r == 0: | ||
| if r == 0 or r > x.shape[0]: | ||
| return paddle.empty([0], dtype=x.dtype) |
There was a problem hiding this comment.
empty的形状设计好像有些问题，pytorch在不同情况下的empty的size是不同的。相关设计后续补充到rfc文档中。
There was a problem hiding this comment.
我之前错误理解empty表示输入形状为空的tensor了，已添加empty相关测试
|
Sorry to inform you that 7b45322's CIs have passed for more than 7 days. To prevent PR conflicts, you need to re-run all CIs manually. |
| expected = paddle.empty([0, 4]) | ||
| np.testing.assert_allclose(c, expected) | ||
|
|
||
| # test empty imput |
| paddle.device.set_device(place) | ||
| a = paddle.to_tensor([1, 2, 3]) | ||
| c = paddle.combinations(a, r=4) | ||
| expected = paddle.empty([0, 4]) |
There was a problem hiding this comment.
一般是不推荐人为给定期望输入用来测试的，是否有其他类似库函数用于测试。
| a = paddle.empty([0]) | ||
| c1 = paddle.combinations(a) | ||
| c2 = paddle.combinations(a, with_replacement=True) | ||
| expected = paddle.empty([0, 2]) |
|
1.更改了测试的输入和输出 |
确认下coverage流水线没有通过的原因 |
|
LGTM |
| """ | ||
| Compute combinations of length r of the given tensor. The behavior is similar to python’s itertools.combinations | ||
| when with_replacement is set to False, and itertools.combinations_with_replacement when with_replacement is set to True. | ||
| Args: |
There was a problem hiding this comment.
| Args: | |
| Args: |
python/paddle/tensor/math.py
Outdated
| Returns: | ||
| out (Tensor): tensor concatenated by combinations, same dtype with x |
There was a problem hiding this comment.
| Returns: | |
| out (Tensor): tensor concatenated by combinations, same dtype with x | |
| Returns: | |
| out (Tensor). Tensor concatenated by combinations, same dtype with x. |
| >>> import paddle | ||
| >>> x = paddle.to_tensor([1, 2, 3], dtype='int32') | ||
| >>> res = paddle.combinations(x) | ||
| >>> print(res) | ||
| Tensor(shape=[3, 2], dtype=int32, place=Place(gpu:0), stop_gradient=True, | ||
| [[1, 2], | ||
| [1, 3], | ||
| [2, 3]]) |
There was a problem hiding this comment.
| >>> import paddle | |
| >>> x = paddle.to_tensor([1, 2, 3], dtype='int32') | |
| >>> res = paddle.combinations(x) | |
| >>> print(res) | |
| Tensor(shape=[3, 2], dtype=int32, place=Place(gpu:0), stop_gradient=True, | |
| [[1, 2], | |
| [1, 3], | |
| [2, 3]]) | |
| >>> import paddle | |
| >>> x = paddle.to_tensor([1, 2, 3], dtype='int32') | |
| >>> res = paddle.combinations(x) | |
| >>> print(res) | |
| Tensor(shape=[3, 2], dtype=int32, place=Place(gpu:0), stop_gradient=True, | |
| [[1, 2], | |
| [1, 3], | |
| [2, 3]]) |
|
Sorry to inform you that d96cb13's CIs have passed for more than 7 days. To prevent PR conflicts, you need to re-run all CIs manually. |
|
@Patrick-Star125 请处理下冲突 |
|
Done |
PR types
New features
PR changes
APIs
Description
Add combinations API to Paddle
Link
Rfc PR: PaddlePaddle/community#654
docs PR: PaddlePaddle/docs#6287
待完成