main

Ggymgit1feat(conv1d) : add conv1d op

5b911f30创建于 29 天前历史提交

文件	最后提交记录	最后更新时间
cann	!1764 merge conv1d_main into main feat(conv1d) : add conv1d op Created-by: gymgit1 Commit-by: gymgit1 Merged-by: ascend-robot Description: 添加conv1d op 位置： third_party/ascend/language/cann/extension 参考torch的conv1d接口，ta侧做透传，进行参数解析，将tl.conv1d转换为hfusion.conv1d，torch接口如图（ta侧删去padding_mode）：  具体更改： core.py 注册&参数解析 semantic.py 语义逻辑 triton_ascend.cc pybind11绑定 TritonAscendOps.td 定义Conv1dOp及其属性 TritonAscendOps.cpp 校验&形状推导 TritonToHFusion.cpp 注册Conversion Patten&提取参数&创建hfusion::Conv1DOp 参数解释： :param input: Input tensor of shape (N, C_in, L_in) or (C_in, L_in). N is a batch size, C denotes a number of channels, L is a length of signal sequence. :type input: tensor :param weight: Weight tensor of shape (C_out, C_in // groups, kernel_size). :type weight: tensor :param bias: Bias tensor of shape (C_out) or None. Default: None. :type bias: tensor or None :param stride: The stride of the convolution kernel. Can be an int or a 1-element tuple. :type stride: int or Tuple[int] :param padding_size: Padding added to both sides of the input. Can be an int, a 1-element tuple, or a string. Can be a string {'valid', 'same'}, single number or a one-element tuple. padding_size='valid' is the same as no padding. padding_size='same' pads the input so the output has the same shape as the input. However, this mode doesn't support any stride values other than 1. :type padding_size: int, Tuple[int], or str :param dilation: The spacing between kernel elements. Can be an int or a 1-element tuple. :type dilation: int or Tuple[int] :param groups: Number of blocked connections from input to output channels. :type groups: int See merge request: Ascend/triton-ascend!1764	23 天前
kernels	feat(gather_simd): add gather_2d_simd kernel impl Co-authored-by: KanuaK<zhouyihan1@huawei.com> # message auto-generated for no-merge-commit merge: !1031 merge feat/gather-simd-main into main feat(gather_simd): add gather_2d_simd kernel impl Created-by: KanuaK Commit-by: KanuaK Merged-by: ascend-robot Description: ## 描述 <!--在这里详细描述你的改动，包括改动的原因和所采取的方法。--> ## checklist <!-- [x] 表示选中 --> - [ ] 是否通过本地IDE对代码进行静态检查 - [ ] 是否通过本地IDE对代码进行格式化处理 - [ ] 是否进行空指针校验 - [ ] 是否进行返回值校验 - [ ] 是否正确释放new/malloc申请的内存 - [ ] 是否充分考虑接口的异常场景 - [ ] 是否正确记录错误日志 See merge request: Ascend/triton-ascend!1031	4 个月前