❌ 文档门禁未通过！ | 检查项 | 检查结果 | | --- | --- | | markdownlint | ❌ 未通过 | 💡 本次检查出 106 项错误，详情内容如下： | 序号 | 错误详情 | | --- | --- | | 1 | [文件路径]：docs/en/0x_new_versions/catlass_v1.6.md
[错误位置]：第 169 行
[检查类型]：markdownlint
[错误信息]：MD001：标题等级只能每次扩大一个。期望：h4；当前：h5
[错误内容]：##### GMToL1 | | 2 | [文件路径]：docs/en/0x_new_versions/catlass_v1.6.md
[错误位置]：第 64 行
[检查类型]：markdownlint
[错误信息]：MD029：有序列表项前的数字请按顺序递增。期望：1；当前：2; Style: 1/2/3
[错误内容]：2. `TileCopyTla`: Specialized for GMToL1 and L1ToL0 MxScale-related types. | | 3 | [文件路径]：docs/en/0x_new_versions/catlass_v1.6.md
[错误位置]：第 110 行
[检查类型]：markdownlint
[错误信息]：MD029：有序列表项前的数字请按顺序递增。期望：1；当前：3; Style: 1/2/3
[错误内容]：3. `PackedMxTileCopyTla`: TileCopy encapsulation for MxScale movement | | 4 | [文件路径]：docs/en/0x_new_versions/catlass_v1.6.md
[错误位置]：第 245 行
[检查类型]：markdownlint
[错误信息]：MD029：有序列表项前的数字请按顺序递增。期望：1；当前：2; Style: 1/2/3
[错误内容]：2. Specialized TileCopyTla for MxScale movement (see the MxScale section above) | | 5 | [文件路径]：docs/en/0x_new_versions/catlass_v1.6.md
[错误位置]：第 314 行
[检查类型]：markdownlint
[错误信息]：MD029：有序列表项前的数字请按顺序递增。期望：1；当前：2; Style: 1/2/3
[错误内容]：2. Specialized TileCopyTla for MxScale movement (see the MxScale section above) | | 6 | [文件路径]：docs/en/0x_new_versions/catlass_v1.6.md
[错误位置]：第 346 行
[检查类型]：markdownlint
[错误信息]：MD029：有序列表项前的数字请按顺序递增。期望：1；当前：2; Style: 1/2/3
[错误内容]：2. Added CopyL0CToDst (optional): On the 950 platform, L0C can be moved to either GM or UB. To keep ... | | 7 | [文件路径]：docs/en/0x_new_versions/catlass_v1.6.md
[错误位置]：第 110 行
[检查类型]：markdownlint
[错误信息]：MD030：列表前缀符号后面请保留一个空格。期望：1；当前：2
[错误内容]：3. `PackedMxTileCopyTla`: TileCopy encapsulation for MxScale movement | | 8 | [文件路径]：docs/en/0x_new_versions/catlass_v1.6.md
[错误位置]：第 156 行
[检查类型]：markdownlint
[错误信息]：MD031：代码块前后应该换行
[错误内容]：```c++ | | 9 | [文件路径]：docs/en/0x_new_versions/catlass_v1.6.md
[错误位置]：第 155 行
[检查类型]：markdownlint
[错误信息]：MD032：列表前后应该换行
[错误内容]：![Left: RowMajor, where the K axis is automatically padded with zero. Right: ColumnMajor, where the ... | | 10 | [文件路径]：docs/en/1_Practice/evaluation/bottleneck_analysis_and_optimization.md
[错误位置]：第 120 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 11 | [文件路径]：docs/en/1_Practice/evaluation/bottleneck_analysis_and_optimization.md
[错误位置]：第 130 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 12 | [文件路径]：docs/en/1_Practice/evaluation/precision_debug.md
[错误位置]：第 161 行
[检查类型]：markdownlint
[错误信息]：MD026：标题的末尾存在标点符号，请删除多余的标点符号
[错误内容]：#### Step 1 (only for FA/MLA/dynamic samples): Distinguish tiling vs. kernel issues. | | 13 | [文件路径]：docs/en/1_Practice/evaluation/precision_debug.md
[错误位置]：第 170 行
[检查类型]：markdownlint
[错误信息]：MD026：标题的末尾存在标点符号，请删除多余的标点符号
[错误内容]：#### Step 2: Perform block-level binary search (BlockMmad vs. BlockEpilogue). | | 14 | [文件路径]：docs/en/1_Practice/evaluation/precision_debug.md
[错误位置]：第 198 行
[检查类型]：markdownlint
[错误信息]：MD026：标题的末尾存在标点符号，请删除多余的标点符号
[错误内容]：#### Step 3: Perform tile-level binary search (inside the faulty block). | | 15 | [文件路径]：docs/en/1_Practice/evaluation/precision_debug.md
[错误位置]：第 18 行
[检查类型]：markdownlint
[错误信息]：MD032：列表前后应该换行
[错误内容]：- Computation starts before data movement completes (missing pipeline synchronization). | | 16 | [文件路径]：docs/en/1_Practice/evaluation/precision_debug.md
[错误位置]：第 30 行
[检查类型]：markdownlint
[错误信息]：MD032：列表前后应该换行
[错误内容]：- Accumulated rounding errors from low-precision data types (FP16/BF16) | | 17 | [文件路径]：docs/en/1_Practice/evaluation/precision_debug.md
[错误位置]：第 118 行
[检查类型]：markdownlint
[错误信息]：MD032：列表前后应该换行
[错误内容]：> - **For samples without tiling** (the majority), precision investigation goes directly to componen... | | 18 | [文件路径]：docs/en/1_Practice/evaluation/precision_debug.md
[错误位置]：第 375 行
[检查类型]：markdownlint
[错误信息]：MD032：列表前后应该换行
[错误内容]：- Check whether the corresponding `SetFlag` operation (such as `SetFlag`) exists after Da... | | 19 | [文件路径]：docs/en/1_Practice/evaluation/precision_debug.md
[错误位置]：第 388 行
[检查类型]：markdownlint
[错误信息]：MD032：列表前后应该换行
[错误内容]：- Check whether DataCopyPad (with padding) is used for non-aligned scenarios. | | 20 | [文件路径]：docs/en/1_Practice/evaluation/precision_debug.md
[错误位置]：第 396 行
[检查类型]：markdownlint
[错误信息]：MD032：列表前后应该换行
[错误内容]：- FP16 maximum is approximately 65504, and BF16 maximum is approximately 3.39e38. | | 21 | [文件路径]：docs/en/1_Practice/evaluation/precision_debug.md
[错误位置]：第 401 行
[检查类型]：markdownlint
[错误信息]：MD032：列表前后应该换行
[错误内容]：- Estimate the maximum possible values of intermediate results. | | 22 | [文件路径]：docs/en/1_Practice/evaluation/precision_debug.md
[错误位置]：第 412 行
[检查类型]：markdownlint
[错误信息]：MD032：列表前后应该换行
[错误内容]：- Ensure that the `ElementC` template parameter of BlockMmad is `float` (FP32 accumulation). | | 23 | [文件路径]：docs/en/1_Practice/evaluation/precision_debug.md
[错误位置]：第 422 行
[检查类型]：markdownlint
[错误信息]：MD032：列表前后应该换行
[错误内容]：- Manually derive the complete computation formula for the epilogue. | | 24 | [文件路径]：docs/en/1_Practice/evaluation/precision_debug.md
[错误位置]：第 433 行
[检查类型]：markdownlint
[错误信息]：MD032：列表前后应该换行
[错误内容]：- Ensure that LayoutA, LayoutB, and LayoutC are consistent in the golden function and the sample. | | 25 | [文件路径]：docs/en/1_Practice/evaluation/precision_debug.md
[错误位置]：第 491 行
[检查类型]：markdownlint
[错误信息]：MD032：列表前后应该换行
[错误内容]：- Change only one variable at a time to keep a clear track of the change outcomes. | | 26 | [文件路径]：docs/en/1_Practice/evaluation/precision_debug.md
[错误位置]：第 500 行
[检查类型]：markdownlint
[错误信息]：MD032：列表前后应该换行
[错误内容]：1. Find a reference sample with similar functionality and normal precision (e.g., `00_basic_matmul`)... | | 27 | [文件路径]：docs/en/1_Practice/evaluation/precision_debug.md
[错误位置]：第 555 行
[检查类型]：markdownlint
[错误信息]：MD032：列表前后应该换行
[错误内容]：- [ ] Read [Precision Analysis Basics](./precision_analysis_basics.md) to understand precision compa... | | 28 | [文件路径]：docs/en/1_Practice/evaluation/precision_debug.md
[错误位置]：第 561 行
[检查类型]：markdownlint
[错误信息]：MD032：列表前后应该换行
[错误内容]：- [ ] Determine whether it's a complete miscalculation or a precision error. | | 29 | [文件路径]：docs/en/1_Practice/evaluation/precision_debug.md
[错误位置]：第 565 行
[检查类型]：markdownlint
[错误信息]：MD032：列表前后应该换行
[错误内容]：- [ ] The correct golden function is selected (ComputeMatmul vs. ComputeGemm, etc.). | | 30 | [文件路径]：docs/en/1_Practice/evaluation/precision_debug.md
[错误位置]：第 571 行
[检查类型]：markdownlint
[错误信息]：MD032：列表前后应该换行
[错误内容]：- [ ] Pipeline synchronization issues (EnQue/DeQue after DataCopy?) | | 31 | [文件路径]：docs/en/1_Practice/evaluation/precision_debug.md
[错误位置]：第 577 行
[检查类型]：markdownlint
[错误信息]：MD032：列表前后应该换行
[错误内容]：- [ ] Try replacing tiling parameters (for samples with tiling). | | 32 | [文件路径]：docs/en/1_Practice/evaluation/precision_debug.md
[错误位置]：第 127 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 33 | [文件路径]：docs/en/1_Practice/evaluation/precision_debug.md
[错误位置]：第 204 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 34 | [文件路径]：docs/en/1_Practice/evaluation/precision_debug.md
[错误位置]：第 220 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 35 | [文件路径]：docs/en/1_Practice/evaluation/precision_debug.md
[错误位置]：第 261 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 36 | [文件路径]：docs/en/1_Practice/evaluation/precision_debug.md
[错误位置]：第 292 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 37 | [文件路径]：docs/en/1_Practice/evaluation/precision_debug.md
[错误位置]：第 324 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 38 | [文件路径]：docs/en/1_Practice/evaluation/precision_debug.md
[错误位置]：第 468 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 39 | [文件路径]：docs/en/1_Practice/evaluation/precision_debug.md
[错误位置]：第 513 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 40 | [文件路径]：docs/en/1_Practice/evaluation/print.md
[错误位置]：第 54 行
[检查类型]：markdownlint
[错误信息]：MD029：有序列表项前的数字请按顺序递增。期望：1；当前：2; Style: 1/2/3
[错误内容]：2. Switch to the `output/bin` directory where the executable file is built and execute the operator ... | | 41 | [文件路径]：docs/en/3_API/include/catlass/epilogue/block/block_epilogue.md
[错误位置]：第 15 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 42 | [文件路径]：docs/en/3_API/include/catlass/epilogue/block/block_epilogue_per_token_dequant.md
[错误位置]：第 49 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 43 | [文件路径]：docs/en/3_API/include/catlass/epilogue/block/block_epilogue_per_token_dequant.md
[错误位置]：第 64 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 44 | [文件路径]：docs/en/3_API/include/catlass/epilogue/block/block_epilogue_per_token_dequant.md
[错误位置]：第 85 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 45 | [文件路径]：docs/en/3_API/include/catlass/epilogue/block/block_epilogue_per_token_dequant.md
[错误位置]：第 104 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 46 | [文件路径]：docs/en/3_API/include/catlass/epilogue/block/block_epilogue_per_token_dequant.md
[错误位置]：第 112 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 47 | [文件路径]：docs/en/3_API/include/catlass/epilogue/block/block_epilogue_per_token_dequant.md
[错误位置]：第 129 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 48 | [文件路径]：docs/en/3_API/include/catlass/epilogue/tile/copy_gm_to_ub/copy_gm_to_ub_aligned.md
[错误位置]：第 49 行
[检查类型]：markdownlint
[错误信息]：MD032：列表前后应该换行
[错误内容]：- No stride and no stride on the destination: directly `DataCopy(dst, src, rows * cols)` | | 49 | [文件路径]：docs/en/3_API/include/catlass/gemm/block/block_mmad.md
[错误位置]：第 15 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 50 | [文件路径]：docs/en/3_API/include/catlass/gemm/block/block_mmad_pingpong.md
[错误位置]：第 11 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 51 | [文件路径]：docs/en/3_API/include/catlass/gemm/block/block_mmad_pingpong.md
[错误位置]：第 28 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 52 | [文件路径]：docs/en/3_API/include/catlass/gemm/block/block_mmad_pingpong.md
[错误位置]：第 38 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 53 | [文件路径]：docs/en/3_API/include/catlass/gemm/block/block_mmad_pingpong.md
[错误位置]：第 46 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 54 | [文件路径]：docs/en/3_API/include/catlass/gemm/block/block_mmad_pingpong.md
[错误位置]：第 55 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 55 | [文件路径]：docs/en/3_API/include/catlass/gemm/block/block_swizzle/GemmIdentityBlockSwizzle.md
[错误位置]：第 32 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 56 | [文件路径]：docs/en/3_API/include/catlass/gemm/block/block_swizzle/GemmIdentityBlockSwizzle.md
[错误位置]：第 46 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 57 | [文件路径]：docs/en/3_API/include/catlass/gemm/block/block_swizzle/GemmIdentityBlockSwizzle.md
[错误位置]：第 52 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 58 | [文件路径]：docs/en/3_API/include/catlass/gemm/block/block_swizzle/GemmIdentityBlockSwizzle.md
[错误位置]：第 58 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 59 | [文件路径]：docs/en/3_API/include/catlass/gemm/block/block_swizzle/GemmIdentityBlockSwizzle.md
[错误位置]：第 73 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 60 | [文件路径]：docs/en/3_API/include/catlass/gemm/block/block_swizzle/GemmIdentityBlockSwizzle.md
[错误位置]：第 79 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 61 | [文件路径]：docs/en/3_API/include/catlass/gemm/block/block_swizzle/GemmIdentityBlockSwizzle.md
[错误位置]：第 90 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 62 | [文件路径]：docs/en/3_API/include/catlass/gemm/kernel/basic_matmul.md
[错误位置]：第 17 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 63 | [文件路径]：docs/en/3_API/include/catlass/gemm/kernel/basic_matmul.md
[错误位置]：第 32 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 64 | [文件路径]：docs/en/3_API/include/catlass/gemm/kernel/basic_matmul.md
[错误位置]：第 45 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 65 | [文件路径]：docs/en/3_API/include/catlass/gemm/tile/cast_fp8_to_fp16.md
[错误位置]：第 46 行
[检查类型]：markdownlint
[错误信息]：MD032：列表前后应该换行
[错误内容]：- `inputBuffer[2]` × `COMPUTE_LENGTH` × 1 byte (FP8 input) | | 66 | [文件路径]：docs/en/3_API/include/catlass/gemm/tile/cast_int4_to_int8.md
[错误位置]：第 48 行
[检查类型]：markdownlint
[错误信息]：MD032：列表前后应该换行
[错误内容]：- `ubInTensor[2]` × `COMPUTE_LEN / 2` bytes (INT4 compact storage input) | | 67 | [文件路径]：docs/en/3_API/include/catlass/gemm/tile/cast_int8_to_fp16.md
[错误位置]：第 47 行
[检查类型]：markdownlint
[错误信息]：MD032：列表前后应该换行
[错误内容]：- `ubInTensor[2]` × `COMPUTE_LEN` bytes (INT8 input) | | 68 | [文件路径]：docs/en/3_API/include/catlass/gemm/tile/copy_gm_to_l1/copy_gm_to_l1.md
[错误位置]：第 12 行
[检查类型]：markdownlint
[错误信息]：MD032：列表前后应该换行
[错误内容]：- `Arch::AtlasA2`(ARCH 2201):[atlasa2/copy_gm_to_l1.hpp](../../../../../../../../include/catlass/gem... | | 69 | [文件路径]：docs/en/3_API/include/catlass/gemm/tile/copy_gm_to_l1.md
[错误位置]：第 15 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 70 | [文件路径]：docs/en/3_API/include/catlass/gemm/tile/copy_gm_to_l1.md
[错误位置]：第 49 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 71 | [文件路径]：docs/en/3_API/include/catlass/gemm/tile/copy_gm_to_l1.md
[错误位置]：第 66 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 72 | [文件路径]：docs/en/3_API/include/catlass/gemm/tile/copy_gm_to_l1.md
[错误位置]：第 92 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 73 | [文件路径]：docs/en/3_API/include/catlass/gemm/tile/copy_gm_to_l1.md
[错误位置]：第 115 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 74 | [文件路径]：docs/en/3_API/include/catlass/gemm/tile/copy_gm_to_l1.md
[错误位置]：第 143 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 75 | [文件路径]：docs/en/3_API/include/catlass/gemm/tile/copy_gm_to_l1.md
[错误位置]：第 183 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 76 | [文件路径]：docs/en/3_API/include/catlass/gemm/tile/copy_gm_to_ub/tile_copy_tla.md
[错误位置]：第 43 行
[检查类型]：markdownlint
[错误信息]：MD032：列表前后应该换行
[错误内容]：- `TensorSrc::position == GM`; `TensorSrc::Layout` is RowMajor. | | 77 | [文件路径]：docs/en/3_API/include/catlass/gemm/tile/copy_l0c_to_dst/README.md
[错误位置]：第 21 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 78 | [文件路径]：docs/en/3_API/include/catlass/gemm/tile/copy_l0c_to_ub.md
[错误位置]：第 57 行
[检查类型]：markdownlint
[错误信息]：MD032：列表前后应该换行
[错误内容]：- `TensorDst::Layout` is `RowMajor`. | | 79 | [文件路径]：docs/en/3_API/include/catlass/gemm/tile/copy_l1_to_bt/copy_l1_to_bt.md
[错误位置]：第 39 行
[检查类型]：markdownlint
[错误信息]：MD012：文档中有多余的空行，请删除多余的空行。期望：1；当前：2 | | 80 | [文件路径]：docs/en/3_API/include/catlass/gemm/tile/copy_l1_to_bt/copy_l1_to_bt.md
[错误位置]：第 40 行
[检查类型]：markdownlint
[错误信息]：MD012：文档中有多余的空行，请删除多余的空行。期望：1；当前：3 | | 81 | [文件路径]：docs/en/3_API/include/catlass/gemm/tile/copy_l1_to_l0a/copy_l1_to_l0a.md
[错误位置]：第 12 行
[检查类型]：markdownlint
[错误信息]：MD032：列表前后应该换行
[错误内容]：- **zN → zZ**: Nd transposed copy (`ifTranspose = false`) | | 82 | [文件路径]：docs/en/3_API/include/catlass/gemm/tile/copy_l1_to_l0b/copy_l1_to_l0b.md
[错误位置]：第 12 行
[检查类型]：markdownlint
[错误信息]：MD032：列表前后应该换行
[错误内容]：- **zZ → nZ**: transposed copy (Transpose B) with `ifTranspose = true`. For float data, the `LoadDat... | | 83 | [文件路径]：docs/en/3_API/include/catlass/gemm/tile/copy_ub_to_gm/copy_ub_to_gm.md
[错误位置]：第 34 行
[检查类型]：markdownlint
[错误信息]：MD032：列表前后应该换行
[错误内容]：- `srcStride = (layoutSrc.stride(0) - shape(1)) / ELE_NUM_PER_C0` | | 84 | [文件路径]：docs/en/3_API/include/catlass/gemm/tile/copy_ub_to_gm/tile_copy_tla.md
[错误位置]：第 43 行
[检查类型]：markdownlint
[错误信息]：MD032：列表前后应该换行
[错误内容]：- `TensorSrc::position == VECCALC`, and `TensorSrc::Layout` is RowMajor. | | 85 | [文件路径]：docs/en/3_API/include/catlass/gemm/tile/tile_copy/README.md
[错误位置]：第 40 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 86 | [文件路径]：docs/en/3_API/include/catlass/gemm/tile/tile_copy/packed_mx_a8w4_tile_copy_tla.md
[错误位置]：第 12 行
[检查类型]：markdownlint
[错误信息]：MD032：列表前后应该换行
[错误内容]：- Matrix A is in FP8 format (`ElementA_`) and is transferred with MX Scale. | | 87 | [文件路径]：docs/en/3_API/include/catlass/gemm/tile/tile_copy/packed_tile_copy_tla.md
[错误位置]：第 57 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 88 | [文件路径]：docs/en/3_API/include/catlass/gemm/tile/tile_copy/padding_packed_tile_copy_tla.md
[错误位置]：第 48 行
[检查类型]：markdownlint
[错误信息]：MD031：代码块前后应该换行
[错误内容]：```cpp | | 89 | [文件路径]：docs/en/3_API/include/catlass/gemm/tile/tile_copy/tile_copy_prologue.md
[错误位置]：第 45 行
[检查类型]：markdownlint
[错误信息]：MD031：代码块前后应该换行
[错误内容]：```cpp | | 90 | [文件路径]：docs/en/3_API/include/catlass/gemm/tile/tile_copy_tla/README.md
[错误位置]：第 23 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 91 | [文件路径]：docs/en/3_API/include/catlass/gemm/tile/tile_mmad/tile_mmad.md
[错误位置]：第 12 行
[检查类型]：markdownlint
[错误信息]：MD032：列表前后应该换行
[错误内容]：- Without Bias: standard matrix multiply-accumulate | | 92 | [文件路径]：docs/en/3_API/include/catlass/gemm/tile/tile_mmad/tile_mmad.md
[错误位置]：第 9 行
[检查类型]：markdownlint
[错误信息]：MD034：请勿直接书写链接，请使用 markdown 语法 或 [xxx](xxx)
[错误内容]：`TileMmad` utilizes basic API [AscendC::Mmad] (https://www.hiascend.com/document/detail/en/CANNCommu... | | 93 | [文件路径]：docs/en/3_API/include/catlass/gemm/tile/tile_mmad/tile_mmad_tla.md
[错误位置]：第 12 行
[检查类型]：markdownlint
[错误信息]：MD032：列表前后应该换行
[错误内容]：1. Standard matrix multiply-accumulate (without bias) | | 94 | [文件路径]：docs/en/3_API/include/catlass/gemm/tile/tile_mmad/tile_mmad_tla.md
[错误位置]：第 97 行
[检查类型]：markdownlint
[错误信息]：MD032：列表前后应该换行
[错误内容]：- `m = tla::get<0>(l0CTensor.layout().originShape())` | | 95 | [文件路径]：docs/en/3_API/include/catlass/gemm/tile/tile_mmad.md
[错误位置]：第 17 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 96 | [文件路径]：docs/en/3_API/include/catlass/gemm/tile/tile_mmad.md
[错误位置]：第 29 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 97 | [文件路径]：docs/en/3_API/include/catlass/gemm/tile/tile_mmad.md
[错误位置]：第 44 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 98 | [文件路径]：docs/en/3_API/include/catlass/gemm/tile/tile_mmad.md
[错误位置]：第 72 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 99 | [文件路径]：docs/en/3_API/include/catlass/gemm/tile/tile_mmad.md
[错误位置]：第 83 行
[检查类型]：markdownlint
[错误信息]：MD040：代码块应该指定一种编程语言
[错误内容]：``` | | 100 | [文件路径]：docs/en/3_API/include/catlass/gemm/tile/tile_muls.md
[错误位置]：第 43 行
[检查类型]：markdownlint
[错误信息]：MD032：列表前后应该换行
[错误内容]：1. `SetMaskCount()` → `SetVectorMask(len)` Set mask. | | 101 | [文件路径]：docs/en/3_API/include/catlass/gemm/tile/tile_traits.md
[错误位置]：第 12 行
[检查类型]：markdownlint
[错误信息]：MD032：列表前后应该换行
[错误内容]：- Deduces `TensorSrc`/`TensorDst` as fully-qualified `AscendC::GlobalTensor` types based on `Prol... | | 102 | [文件路径]：docs/en/3_API/include/catlass/gemv/tile/tile_vmad.md
[错误位置]：第 37 行
[检查类型]：markdownlint
[错误信息]：MD032：列表前后应该换行
[错误内容]：1. `Duplicate` initializes the accumulator buffer temp to zero. | | 103 | [文件路径]：docs/en/3_API/include/catlass/gemv/tile/tile_vmad.md
[错误位置]：第 44 行
[检查类型]：markdownlint
[错误信息]：MD032：列表前后应该换行
[错误内容]：1. `Duplicate` initializes temp to zero. | | 104 | [文件路径]：docs/zh/1_Practice/evaluation/precision_debug.md
[错误位置]：第 74 行
[检查类型]：markdownlint
[错误信息]：MD029：有序列表项前的数字请按顺序递增。期望：1；当前：3; Style: 1/2/3
[错误内容]：3. **Layout参数是否正确**：确认`layoutA`、`layoutB`、`layoutC`与样例实际使用的内存排布一致。行优先（RowMajor）和列优先（ColumnMajor）搞反是最... | | 105 | [文件路径]：docs/zh/1_Practice/evaluation/precision_debug.md
[错误位置]：第 76 行
[检查类型]：markdownlint
[错误信息]：MD029：有序列表项前的数字请按顺序递增。期望：2；当前：4; Style: 1/2/3
[错误内容]：4. **标杆输入数据与NPU输入一致**：确认传入标杆的`hostA`/`hostB`与拷贝到Device的数据完全相同。如果在拷贝后修改了host数据再传标杆，会导致标杆与NPU计算不同的数据。 | | 106 | [文件路径]：docs/zh/1_Practice/evaluation/precision_debug.md
[错误位置]：第 78 行
[检查类型]：markdownlint
[错误信息]：MD029：有序列表项前的数字请按顺序递增。期望：3；当前：5; Style: 1/2/3
[错误内容]：5. **简单用例交叉验证**：用一个极简单的用例（如M=N=K=2，数据全为1.0）手动计算预期结果，与标杆输出对比，确认标杆逻辑正确。 |