Closed
[Documentation|文档反馈]: ops-nn 本仓 123 处错误码表 rowspan 声明行数多于实际描述行(缺 381 条子项,正式目录 43 处 / 37 个算子) #8190
liuhanwen8Add on  16 days agoClosed at  2 days ago
liuhanwen8
liuhanwen8
16 days ago Create

现象

本条为 #8190 的原位修订:原单把四个仓的内容合并提在 ops-nn 下,且分仓归类有误(image/、gmm/、mc2/ 等目录并不属于本仓)。现改为只覆盖 cann/ops-nn,另三仓各自单独提单。

类型:[Documentation|文档反馈]
算子资料「报错说明」小节的错误码表里,rowspan="N" 声明某错误码有 N 条报错场景,但该错误码下实际只写了 M 条描述(M < N)。渲染后是「说有 N 条、只列 M 条」的残缺表格,调用方无法判断该错误码还会在哪些前置条件下返回。

以本仓缺失最多的 rnn/thnn_fused_lstm_cell_grad(rnn/thnn_fused_lstm_cell_grad/docs/aclnnThnnFusedLstmCellBackward.md:254)为例:

<tr><td rowspan="12">ACLNN_ERR_PARAM_INVALID</td><td rowspan="12">161002</td><td>第 1 条描述</td></tr>
</tbody></table>

rowspan="12" 声明 12 条子项,<tbody> 内实际只有 3 条描述行,另外 9 条缺失。

期望

rowspan 数值应与该错误码下实际列出的描述条数一致:确有 N 种场景就补全描述,若只有 M 种就把 rowspan="N" 改成 rowspan="M"。现状无论按哪种口径都是错的——声明与内容不符,契约不可完整推导。

环境

  • 仓库:cann/ops-nn master ed6471cdb55e90e03eaee377ad60b035aa56df6c(2026-11-05 fetch);纯静态文档扫描,无需编译、无需上板

复现步骤

  1. git clone --depth 1 https://gitcode.com/cann/ops-nn.git && cd ops-nn
  2. 把「复现脚本」存为仓库根目录下的 rowspan_check.py
  3. 执行 python3 rowspan_check.py --repo-root .(rc=1 即复现成功)
  4. 打开任一 [DEFECT] 行给出的 file:line,核对 rowspan 声明值与实际描述条数

复现脚本

# python3 rowspan_check.py --repo-root .   rc=1 表示复现到缺陷
import os, re, sys, argparse
TR = re.compile(r'<tr[^>]*>(.*?)</tr>', re.S | re.I)
TD = re.compile(r'<td([^>]*)>(.*?)</td>', re.S | re.I)
CODE = re.compile(r'^\s*(\d{6})\s*$')
def tds(row):
    out = []
    for m in TD.finditer(row):
        t = re.sub(r'\s+', ' ', re.sub(r'<[^>]+>', ' ', m.group(2))).strip()
        r = re.search(r'rowspan="(\d+)"', m.group(1))
        out.append((int(r.group(1)) if r else 1, t))
    return out
def head(t):
    # 只认前 2 列(返回码列/错误码列)里纯 6 位数字的 td;描述列的 rowspan 折行不算新块
    for sp, tx in t[:2]:
        m = CODE.match(tx)
        if m and sp > 1:
            return sp, m.group(1)
def scan(path):
    s = open(path, encoding='utf-8', errors='ignore').read()
    hits, trs = [], list(TR.finditer(s))
    for i, rm in enumerate(trs):
        t = tds(rm.group(1)); h = head(t)
        if not h:
            continue
        n, code = h
        e = [x for x in (s.find('</tbody>', rm.end()), s.find('</table>', rm.end())) if x > 0]
        lim = min(e) if e else len(s)
        act = 1 if len(t) >= 3 and t[2][1] else 0
        skip = (t[2][0] - 1) if act else 0
        for j in range(i + 1, len(trs)):
            if trs[j].start() >= lim:
                break
            t2 = tds(trs[j].group(1))
            if skip:
                skip -= 1; continue
            if head(t2):
                break
            if not t2:
                continue
            sp, tx = t2[-1] if len(t2) > 1 else t2[0]
            act += 1; skip = sp - 1 if sp > 1 else 0
        if act < n:
            hits.append((s[:rm.start()].count('\n') + 1, code, n, act, n - act))
    return hits
ap = argparse.ArgumentParser()
ap.add_argument('--repo-root', default='.')
root = ap.parse_args().repo_root
tot = mis = 0
for dp, _, fn in os.walk(root):
    if '/.git' in dp.replace(os.sep, '/'):
        continue
    for f in fn:
        if not f.endswith('.md'):
            continue
        p = os.path.join(dp, f)
        for ln, code, n, act, m in scan(p):
            tot += 1; mis += m
            print('[DEFECT] %s:%d code=%s rowspan=%d actual=%d missing=%d'
                  % (p.replace(os.sep, '/'), ln, code, n, act, m))
print('合计 %d 处不符,缺失 %d 条描述' % (tot, mis))
sys.exit(1 if tot else 0)

实测输出(master ed6471cd):

[DEFECT] activation/celu_v2/docs/aclnnCelu&aclnnInplaceCelu.md:344 code=161002 rowspan=8 actual=4 missing=4
[DEFECT] activation/gelu_quant/docs/aclnnGeluQuant.md:272 code=361001 rowspan=3 actual=1 missing=2
[DEFECT] activation/leaky_relu/docs/aclnnLeakyRelu&aclnnInplaceLeakyRelu.md:338 code=161002 rowspan=8 actual=1 missing=7
[DEFECT] activation/leaky_relu/docs/aclnnRReluWithNoise&aclnnInplaceRReluWithNoise.md:462 code=161002 rowspan=8 actual=4 missing=4
合计 123 处不符,缺失 381 条描述

影响

错误码表是唯一说明「这个码在什么场景返回」的载体。缺失子项意味着调用方写 catch 161002 时不知道还有哪些前置条件会触发它,只能靠试错定位;上层框架也无法做前置规避。

建议

  1. 按目录批量修(推荐):多数是「声明 N 条、实际写 1~3 条」,属同一模板复制后忘记补内容,按目录分给对应责任人集中处理。
  2. 接入文档 CI:上面的脚本零依赖,rowspan 与描述条数一致是纯机械可判规则,适合在 PR 阶段自动拦截。短期若补不齐内容,至少先把 rowspan 改成实际条数,避免渲染错位误导读者。

实测统计

范围 不符处数 缺失条目数 涉及算子数
正式目录·中文 43 96 37
正式目录·英文 + experimental 80 285 —
合计 123 381 —
_en.md 与中文是同一处缺陷的两份副本,建议同步修。

同批同类问题(各仓独立单,按责任田分派):cann/ops-math#4535、cann/ops-cv#1560、cann/ops-transformer#7139。本条为 cann/ops-nn 本仓口径,共 123 处 / 381 条。

likedislike
liuhanwen8liuhanwen8
16 days ago Added label: documentation
liuhanwen8liuhanwen8
16 days ago Modify description of issue
liuhanwen8
liuhanwen8
16 days ago Comments:

完整清单(cann/ops-nn 共 123 处 / 381 条缺失)

列说明:file:line / 错误码 / rowspan 声明值 / 实际描述条数 / 缺失条数。_en.md 与中文文档是同一处缺陷的两份副本,修的时候建议同步。

按算子汇总(正式目录·中文,43 处 / 96 条 / 37 个算子)

算子目录 不符处数 缺失条目数
activation/leaky_relu 2 11
rnn/thnn_fused_lstm_cell_grad 1 9
activation/mish 1 6
activation/sigmoid 1 6
pooling/max_pool3d_with_argmax_v2 1 6
activation/swish 1 5
index/apply_top_k_top_p_with_sorted 1 5
activation/celu_v2 1 4
matmul/quant_matmul 2 4
matmul/mv 1 3
pooling/max_pool3d_grad_with_argmax 3 3
activation/gelu_quant 1 2
index/scatter_elements_v2 2 2
matmul/flat_quant 2 2
matmul/mat_mul_v3 1 2
matmul/quant_matmul_dequant 1 2
norm/add_rms_norm_quant_v2 1 2
rnn/bidirection_lstm 1 2
rnn/bidirection_lstmv2 1 2
activation/log_softmax_v2 1 1
conv/convolution_backward 1 1
foreach/foreach_div_scalar 1 1
foreach/foreach_maximum_scalar 1 1
index/index_put_v2 1 1
loss/binary_cross_entropy_grad 1 1
loss/fused_cross_entropy_loss_with_max_sum 1 1
matmul/batch_mat_mul_v3 1 1
matmul/dual_level_quant_batch_matmul 1 1
norm/lp_norm_v2 1 1
pooling/adaptive_avg_pool2d_grad 1 1
pooling/max_pool_v3 1 1
quant/anti_mx_quant 1 1
quant/ascend_quant_v2 1 1
quant/grouped_quant_max 1 1
quant/group_quant 1 1
quant/mx_to_block_mx_quant 1 1
quant/quantize 1 1

正式目录·中文(43 处 / 96 条)

file:line code rowspan actual missing
activation/celu_v2/docs/aclnnCelu&aclnnInplaceCelu.md:344 161002 8 4 4
activation/gelu_quant/docs/aclnnGeluQuant.md:272 361001 3 1 2
activation/leaky_relu/docs/aclnnLeakyRelu&aclnnInplaceLeakyRelu.md:338 161002 8 1 7
activation/leaky_relu/docs/aclnnRReluWithNoise&aclnnInplaceRReluWithNoise.md:462 161002 8 4 4
activation/log_softmax_v2/docs/aclnnLogSoftmax.md:164 161002 3 2 1
activation/mish/docs/aclnnMish&aclnnInplaceMish.md:316 161002 8 2 6
activation/sigmoid/docs/aclnnSigmoid&aclnnInplaceSigmoid.md:165 161002 8 2 6
activation/swish/docs/aclnnSilu.md:151 161002 8 3 5
conv/convolution_backward/docs/aclnnConvolutionBackward.md:373 161001 2 1 1
foreach/foreach_div_scalar/docs/aclnnForeachDivScalarV2.md:162 161002 6 5 1
foreach/foreach_maximum_scalar/docs/aclnnForeachMaximumScalarV2.md:164 161002 6 5 1
index/apply_top_k_top_p_with_sorted/docs/aclnnApplyTopKTopP.md:200 161002 8 3 5
index/index_put_v2/docs/aclnnIndexPutImpl.md:191 161002 6 5 1
index/scatter_elements_v2/docs/aclnnScatterValue&aclnnInplaceScatterValue.md:229 161002 9 8 1
index/scatter_elements_v2/docs/aclnnScatterValue&aclnnInplaceScatterValue.md:430 161002 7 6 1
loss/binary_cross_entropy_grad/docs/aclnnBinaryCrossEntropyBackward.md:214 161002 6 5 1
loss/fused_cross_entropy_loss_with_max_sum/docs/aclnnFusedCrossEntropyLossWithMaxSum.md:225 161002 7 6 1
matmul/batch_mat_mul_v3/docs/aclnnBatchMatMul.md:192 161002 5 4 1
matmul/dual_level_quant_batch_matmul/docs/aclnnDualLevelQuantMatmulWeightNz.md:308 161002 5 4 1
matmul/flat_quant/docs/aclnnFlatQuant.md:246 161002 12 11 1
matmul/flat_quant/docs/aclnnFlatQuantV2.md:305 161002 12 11 1
matmul/mat_mul_v3/docs/aclnnMatmulWeightNz.md:193 161002 5 3 2
matmul/mv/docs/aclnnMv.md:196 161002 7 4 3
matmul/quant_matmul/docs/aclnnQuantMatmul.md:180 161002 7 5 2
matmul/quant_matmul/docs/aclnnQuantMatmulV2.md:201 161002 7 5 2
matmul/quant_matmul_dequant/docs/aclnnQuantMatmulDequant.md:281 561002 3 1 2
norm/add_rms_norm_quant_v2/docs/aclnnAddRmsNormQuantV2.md:342 161002 3 1 2
norm/lp_norm_v2/docs/aclnnLinalgVectorNorm.md:190 161002 7 6 1
pooling/adaptive_avg_pool2d_grad/docs/aclnnAdaptiveAvgPool2dBackward.md:153 161002 6 5 1
pooling/max_pool3d_grad_with_argmax/docs/aclnnMaxPool2dWithIndicesBackward.md:248 161002 11 10 1
pooling/max_pool3d_grad_with_argmax/docs/aclnnMaxPool2dWithMaskBackward.md:214 161002 11 10 1
pooling/max_pool3d_grad_with_argmax/docs/aclnnMaxPool3dWithArgmaxBackward.md:219 161002 12 11 1
pooling/max_pool3d_with_argmax_v2/docs/aclnnMaxPool3dWithArgmax.md:235 161002 13 7 6
pooling/max_pool_v3/docs/aclnnMaxPool.md:234 161002 11 10 1
quant/anti_mx_quant/docs/aclnnAntiMxQuant.md:174 161001 2 1 1
quant/ascend_quant_v2/docs/aclnnAscendQuant.md:222 161002 7 6 1
quant/group_quant/docs/aclnnGroupQuant.md:190 161002 6 5 1
quant/grouped_quant_max/docs/aclnnGroupedQuantMax.md:213 161002 5 4 1
quant/mx_to_block_mx_quant/docs/aclnnMxToBlockMxQuant.md:198 161001 2 1 1
quant/quantize/docs/aclnnQuantize.md:207 161002 7 6 1
rnn/bidirection_lstm/docs/aclnnBidirectionLSTM.md:341 561002 3 1 2
rnn/bidirection_lstmv2/docs/aclnnBidirectionLSTMV2.md:363 561002 3 1 2
rnn/thnn_fused_lstm_cell_grad/docs/aclnnThnnFusedLstmCellBackward.md:254 161002 12 3 9

experimental·中文(22 处 / 110 条)

file:line code rowspan actual missing
experimental/activation/gelu/docs/aclnnGelu.md:132 161002 8 4 4
experimental/activation/gelu_grad/docs/aclnnGeluBackward.md:165 161002 8 3 5
experimental/activation/hard_sigmoid_grad/docs/aclnnHardsigmoidBackward.md:144 161002 8 3 5
experimental/activation/hardtanh_grad/docs/aclnnHardtanhBackward.md:171 161002 8 4 4
experimental/activation/mish_grad/docs/aclnnMishBackward.md:135 161002 8 3 5
experimental/activation/relu_grad/docs/aclnnThresholdBackward.md:153 161002 8 3 5
experimental/activation/relu_v2/docs/aclnnAddRelu&aclnnInplaceAddRelu.md:184 161002 10 5 5
experimental/activation/relu_v2/docs/aclnnAddRelu&aclnnInplaceAddRelu.md:349 161002 10 5 5
experimental/activation/relu_v2/docs/aclnnRelu&aclnnInplaceRelu.md:157 161002 8 3 5
experimental/activation/relu_v2/docs/aclnnRelu&aclnnInplaceRelu.md:298 161002 8 2 6
experimental/activation/sigmoid/docs/aclnnSigmoid&aclnnInplaceSigmoid.md:169 161002 8 2 6
experimental/activation/sigmoid_grad/docs/aclnnSigmoidBackward.md:146 161002 8 2 6
experimental/activation/silu_grad/docs/aclnnSiluBackward.md:158 161002 8 3 5
experimental/activation/silu_mul/docs/aclnnSiluMul.md:153 161002 8 1 7
experimental/activation/softplus_v2/docs/aclnnSoftplus.md:163 161002 8 3 5
experimental/activation/softplus_v2_grad/docs/aclnnSoftplusBackward.md:165 161002 8 2 6
experimental/activation/swish/docs/aclnnSilu.md:134 161002 8 3 5
experimental/activation/swish/docs/aclnnSwish.md:151 161002 8 3 5
experimental/activation/swish_grad/docs/aclnnSwishBackward.md:171 161002 8 3 5
experimental/activation/threshold_grad_v2_d/docs/aclnnThresholdBackward.md:156 161002 8 3 5
experimental/norm/sync_batch_norm_gather_stats/docs/aclnnSyncBatchNormGatherStats.md:230 161002 8 3 5
experimental/quant/ascend_anti_quant_v2/docs/aclnnAscendAntiQuant.md:193 161002 4 3 1

英文文档(58 处 / 175 条)

file:line code rowspan actual missing
activation/celu_v2/docs/aclnnCelu&aclnnInplaceCelu_en.md:346 161002 8 4 4
activation/fatrelu_mul/docs/aclnnFatreluMul_en.md:186 161002 8 1 7
activation/gelu_quant/docs/aclnnGeluQuant_en.md:274 361001 3 1 2
activation/hard_shrink/docs/aclnnHardshrink_en.md:171 161002 8 3 5
activation/leaky_relu/docs/aclnnLeakyRelu&aclnnInplaceLeakyRelu_en.md:338 161002 8 1 7
activation/leaky_relu/docs/aclnnRReluWithNoise&aclnnInplaceRReluWithNoise_en.md:462 161002 8 4 4
activation/log_softmax_v2/docs/aclnnLogSoftmax_en.md:166 161002 3 2 1
activation/mish/docs/aclnnMish&aclnnInplaceMish_en.md:318 161002 8 2 6
activation/sigmoid/docs/aclnnSigmoid&aclnnInplaceSigmoid_en.md:168 161002 8 2 6
activation/swish/docs/aclnnSilu_en.md:153 161002 8 3 5
conv/convolution_backward/docs/aclnnConvolutionBackward_en.md:361 161001 2 1 1
foreach/foreach_add_scalar/docs/aclnnForeachAddScalarV2_en.md:183 161002 6 4 2
foreach/foreach_div_scalar/docs/aclnnForeachDivScalarV2_en.md:164 161002 6 5 1
foreach/foreach_maximum_scalar/docs/aclnnForeachMaximumScalarV2_en.md:166 161002 6 5 1
index/apply_top_k_top_p_with_sorted/docs/aclnnApplyTopKTopP_en.md:202 161002 8 3 5
index/index_put_v2/docs/aclnnIndexPutImpl_en.md:193 161002 6 5 1
index/scatter_elements_v2/docs/aclnnScatterValue&aclnnInplaceScatterValue_en.md:231 161002 9 8 1
index/scatter_elements_v2/docs/aclnnScatterValue&aclnnInplaceScatterValue_en.md:432 161002 7 6 1
loss/binary_cross_entropy_grad/docs/aclnnBinaryCrossEntropyBackward_en.md:216 161002 6 5 1
loss/fused_cross_entropy_loss_with_max_sum/docs/aclnnFusedCrossEntropyLossWithMaxSum_en.md:227 161002 7 6 1
loss/logit/docs/aclnnLogit_en.md:173 161002 8 1 7
loss/logit_grad/docs/aclnnLogitGrad_en.md:178 161002 8 1 7
matmul/batch_mat_mul_v3/docs/aclnnBatchMatMul_en.md:194 161002 5 4 1
matmul/dual_level_quant_batch_matmul/docs/aclnnDualLevelQuantMatmulWeightNz_en.md:299 161002 5 4 1
matmul/flat_quant/docs/aclnnFlatQuantV2_en.md:307 161002 12 11 1
matmul/flat_quant/docs/aclnnFlatQuant_en.md:248 161002 12 11 1
matmul/mat_mul_v3/docs/aclnnMatmulWeightNz_en.md:195 161002 5 3 2
matmul/mv/docs/aclnnMv_en.md:198 161002 7 4 3
matmul/quant_matmul/docs/aclnnQuantMatmulV2_en.md:203 161002 7 5 2
matmul/quant_matmul/docs/aclnnQuantMatmul_en.md:182 161002 7 5 2
matmul/quant_matmul_dequant/docs/aclnnQuantMatmulDequant_en.md:283 561002 3 1 2
matmul/sparse4to2quant_matmul/docs/aclnnSparse4to2QuantMatmulWeightNz_en.md:193 161002 8 1 7
matmul/sparse4to2quant_matmul/docs/aclnnTransSparse4to2Para_en.md:167 161002 8 1 7
matmul/transpose_batch_mat_mul/docs/aclnnTransposeBatchMatMul_en.md:265 161002 6 5 1
matmul/transpose_quant_batch_mat_mul/docs/aclnnTransposeQuantBatchMatMulWeightNz_en.md:270 161002 8 6 2
matmul/transpose_quant_batch_mat_mul/docs/aclnnTransposeQuantBatchMatMul_en.md:284 161002 8 5 3
norm/add_rms_norm_quant_v2/docs/aclnnAddRmsNormQuantV2_en.md:344 161002 3 1 2
norm/lp_norm_v2/docs/aclnnLinalgVectorNorm_en.md:192 161002 7 6 1
pooling/adaptive_avg_pool2d_grad/docs/aclnnAdaptiveAvgPool2dBackward_en.md:155 161002 6 5 1
pooling/max_pool3d_grad_with_argmax/docs/aclnnMaxPool2dWithIndicesBackward_en.md:250 161002 11 10 1
pooling/max_pool3d_grad_with_argmax/docs/aclnnMaxPool2dWithMaskBackward_en.md:216 161002 11 10 1
pooling/max_pool3d_grad_with_argmax/docs/aclnnMaxPool3dWithArgmaxBackward_en.md:221 161002 12 11 1
pooling/max_pool3d_with_argmax_v2/docs/aclnnMaxPool3dWithArgmax_en.md:237 161002 13 7 6
pooling/max_pool_v3/docs/aclnnMaxPool_en.md:236 161002 11 10 1
quant/anti_mx_quant/docs/aclnnAntiMxQuant_en.md:176 161001 2 1 1
quant/ascend_quant_v2/docs/aclnnAscendQuant_en.md:224 161002 7 6 1
quant/dynamic_block_quant/docs/aclnnDynamicBlockQuantV2_en.md:255 161002 5 2 3
quant/dynamic_block_quant/docs/aclnnDynamicBlockQuant_en.md:243 161002 5 1 4
quant/group_quant/docs/aclnnGroupQuant_en.md:192 161002 6 5 1
quant/grouped_quant_max/docs/aclnnGroupedQuantMax_en.md:215 161002 5 4 1
quant/mx_to_block_mx_quant/docs/aclnnMxToBlockMxQuant_en.md:200 161001 2 1 1
quant/quantize/docs/aclnnQuantize_en.md:209 161002 7 6 1
rnn/bidirection_lstm/docs/aclnnBidirectionLSTM_en.md:343 561002 3 1 2
rnn/bidirection_lstmv2/docs/aclnnBidirectionLSTMV2_en.md:365 561002 3 1 2
rnn/dynamic_rnn/docs/aclnnLSTM_en.md:419 161002 12 4 8
rnn/gru_grad/docs/aclnnGRUBackward_en.md:359 161002 12 4 8
rnn/single_layer_lstm_grad/docs/aclnnLstmBackward_en.md:603 161002 12 4 8
rnn/thnn_fused_lstm_cell_grad/docs/aclnnThnnFusedLstmCellBackward_en.md:256 161002 12 3 9
likedislike
liuhanwen8liuhanwen8
16 days ago Change title to "[Documentation|文档反馈]: ops-nn 本仓 123 处错误码表 rowspan 声明行数多于实际描述行(缺 381 条子项,正式目录 43 处 / 37 个算子)", originally titled "[Documentation|文档反馈]: 四仓 204 处错误码表 rowspan 声明行数多于实际描述行(缺 615 条子项,正式目录涉及 64 个算子)".
liuhanwen8liuhanwen8
16 days ago Modify description of issue
liuhanwen8liuhanwen8
16 days ago Modify description of issue
liuhanwen8
liuhanwen8
16 days ago Comments:

完整清单(cann/ops-nn 共 123 处 / 381 条缺失)

列说明:file:line / 错误码 / rowspan 声明值 / 实际描述条数 / 缺失条数。_en.md 与中文文档是同一处缺陷的两份副本,修的时候建议同步。

按算子汇总(正式目录·中文,43 处 / 96 条 / 37 个算子)

算子目录 不符处数 缺失条目数
activation/leaky_relu 2 11
rnn/thnn_fused_lstm_cell_grad 1 9
activation/mish 1 6
activation/sigmoid 1 6
pooling/max_pool3d_with_argmax_v2 1 6
activation/swish 1 5
index/apply_top_k_top_p_with_sorted 1 5
activation/celu_v2 1 4
matmul/quant_matmul 2 4
matmul/mv 1 3
pooling/max_pool3d_grad_with_argmax 3 3
activation/gelu_quant 1 2
index/scatter_elements_v2 2 2
matmul/flat_quant 2 2
matmul/mat_mul_v3 1 2
matmul/quant_matmul_dequant 1 2
norm/add_rms_norm_quant_v2 1 2
rnn/bidirection_lstm 1 2
rnn/bidirection_lstmv2 1 2
activation/log_softmax_v2 1 1
conv/convolution_backward 1 1
foreach/foreach_div_scalar 1 1
foreach/foreach_maximum_scalar 1 1
index/index_put_v2 1 1
loss/binary_cross_entropy_grad 1 1
loss/fused_cross_entropy_loss_with_max_sum 1 1
matmul/batch_mat_mul_v3 1 1
matmul/dual_level_quant_batch_matmul 1 1
norm/lp_norm_v2 1 1
pooling/adaptive_avg_pool2d_grad 1 1
pooling/max_pool_v3 1 1
quant/anti_mx_quant 1 1
quant/ascend_quant_v2 1 1
quant/grouped_quant_max 1 1
quant/group_quant 1 1
quant/mx_to_block_mx_quant 1 1
quant/quantize 1 1

正式目录·中文(43 处 / 96 条)

file:line code rowspan actual missing
activation/celu_v2/docs/aclnnCelu&aclnnInplaceCelu.md:344 161002 8 4 4
activation/gelu_quant/docs/aclnnGeluQuant.md:272 361001 3 1 2
activation/leaky_relu/docs/aclnnLeakyRelu&aclnnInplaceLeakyRelu.md:338 161002 8 1 7
activation/leaky_relu/docs/aclnnRReluWithNoise&aclnnInplaceRReluWithNoise.md:462 161002 8 4 4
activation/log_softmax_v2/docs/aclnnLogSoftmax.md:164 161002 3 2 1
activation/mish/docs/aclnnMish&aclnnInplaceMish.md:316 161002 8 2 6
activation/sigmoid/docs/aclnnSigmoid&aclnnInplaceSigmoid.md:165 161002 8 2 6
activation/swish/docs/aclnnSilu.md:151 161002 8 3 5
conv/convolution_backward/docs/aclnnConvolutionBackward.md:373 161001 2 1 1
foreach/foreach_div_scalar/docs/aclnnForeachDivScalarV2.md:162 161002 6 5 1
foreach/foreach_maximum_scalar/docs/aclnnForeachMaximumScalarV2.md:164 161002 6 5 1
index/apply_top_k_top_p_with_sorted/docs/aclnnApplyTopKTopP.md:200 161002 8 3 5
index/index_put_v2/docs/aclnnIndexPutImpl.md:191 161002 6 5 1
index/scatter_elements_v2/docs/aclnnScatterValue&aclnnInplaceScatterValue.md:229 161002 9 8 1
index/scatter_elements_v2/docs/aclnnScatterValue&aclnnInplaceScatterValue.md:430 161002 7 6 1
loss/binary_cross_entropy_grad/docs/aclnnBinaryCrossEntropyBackward.md:214 161002 6 5 1
loss/fused_cross_entropy_loss_with_max_sum/docs/aclnnFusedCrossEntropyLossWithMaxSum.md:225 161002 7 6 1
matmul/batch_mat_mul_v3/docs/aclnnBatchMatMul.md:192 161002 5 4 1
matmul/dual_level_quant_batch_matmul/docs/aclnnDualLevelQuantMatmulWeightNz.md:308 161002 5 4 1
matmul/flat_quant/docs/aclnnFlatQuant.md:246 161002 12 11 1
matmul/flat_quant/docs/aclnnFlatQuantV2.md:305 161002 12 11 1
matmul/mat_mul_v3/docs/aclnnMatmulWeightNz.md:193 161002 5 3 2
matmul/mv/docs/aclnnMv.md:196 161002 7 4 3
matmul/quant_matmul/docs/aclnnQuantMatmul.md:180 161002 7 5 2
matmul/quant_matmul/docs/aclnnQuantMatmulV2.md:201 161002 7 5 2
matmul/quant_matmul_dequant/docs/aclnnQuantMatmulDequant.md:281 561002 3 1 2
norm/add_rms_norm_quant_v2/docs/aclnnAddRmsNormQuantV2.md:342 161002 3 1 2
norm/lp_norm_v2/docs/aclnnLinalgVectorNorm.md:190 161002 7 6 1
pooling/adaptive_avg_pool2d_grad/docs/aclnnAdaptiveAvgPool2dBackward.md:153 161002 6 5 1
pooling/max_pool3d_grad_with_argmax/docs/aclnnMaxPool2dWithIndicesBackward.md:248 161002 11 10 1
pooling/max_pool3d_grad_with_argmax/docs/aclnnMaxPool2dWithMaskBackward.md:214 161002 11 10 1
pooling/max_pool3d_grad_with_argmax/docs/aclnnMaxPool3dWithArgmaxBackward.md:219 161002 12 11 1
pooling/max_pool3d_with_argmax_v2/docs/aclnnMaxPool3dWithArgmax.md:235 161002 13 7 6
pooling/max_pool_v3/docs/aclnnMaxPool.md:234 161002 11 10 1
quant/anti_mx_quant/docs/aclnnAntiMxQuant.md:174 161001 2 1 1
quant/ascend_quant_v2/docs/aclnnAscendQuant.md:222 161002 7 6 1
quant/group_quant/docs/aclnnGroupQuant.md:190 161002 6 5 1
quant/grouped_quant_max/docs/aclnnGroupedQuantMax.md:213 161002 5 4 1
quant/mx_to_block_mx_quant/docs/aclnnMxToBlockMxQuant.md:198 161001 2 1 1
quant/quantize/docs/aclnnQuantize.md:207 161002 7 6 1
rnn/bidirection_lstm/docs/aclnnBidirectionLSTM.md:341 561002 3 1 2
rnn/bidirection_lstmv2/docs/aclnnBidirectionLSTMV2.md:363 561002 3 1 2
rnn/thnn_fused_lstm_cell_grad/docs/aclnnThnnFusedLstmCellBackward.md:254 161002 12 3 9

experimental·中文(22 处 / 110 条)

file:line code rowspan actual missing
experimental/activation/gelu/docs/aclnnGelu.md:132 161002 8 4 4
experimental/activation/gelu_grad/docs/aclnnGeluBackward.md:165 161002 8 3 5
experimental/activation/hard_sigmoid_grad/docs/aclnnHardsigmoidBackward.md:144 161002 8 3 5
experimental/activation/hardtanh_grad/docs/aclnnHardtanhBackward.md:171 161002 8 4 4
experimental/activation/mish_grad/docs/aclnnMishBackward.md:135 161002 8 3 5
experimental/activation/relu_grad/docs/aclnnThresholdBackward.md:153 161002 8 3 5
experimental/activation/relu_v2/docs/aclnnAddRelu&aclnnInplaceAddRelu.md:184 161002 10 5 5
experimental/activation/relu_v2/docs/aclnnAddRelu&aclnnInplaceAddRelu.md:349 161002 10 5 5
experimental/activation/relu_v2/docs/aclnnRelu&aclnnInplaceRelu.md:157 161002 8 3 5
experimental/activation/relu_v2/docs/aclnnRelu&aclnnInplaceRelu.md:298 161002 8 2 6
experimental/activation/sigmoid/docs/aclnnSigmoid&aclnnInplaceSigmoid.md:169 161002 8 2 6
experimental/activation/sigmoid_grad/docs/aclnnSigmoidBackward.md:146 161002 8 2 6
experimental/activation/silu_grad/docs/aclnnSiluBackward.md:158 161002 8 3 5
experimental/activation/silu_mul/docs/aclnnSiluMul.md:153 161002 8 1 7
experimental/activation/softplus_v2/docs/aclnnSoftplus.md:163 161002 8 3 5
experimental/activation/softplus_v2_grad/docs/aclnnSoftplusBackward.md:165 161002 8 2 6
experimental/activation/swish/docs/aclnnSilu.md:134 161002 8 3 5
experimental/activation/swish/docs/aclnnSwish.md:151 161002 8 3 5
experimental/activation/swish_grad/docs/aclnnSwishBackward.md:171 161002 8 3 5
experimental/activation/threshold_grad_v2_d/docs/aclnnThresholdBackward.md:156 161002 8 3 5
experimental/norm/sync_batch_norm_gather_stats/docs/aclnnSyncBatchNormGatherStats.md:230 161002 8 3 5
experimental/quant/ascend_anti_quant_v2/docs/aclnnAscendAntiQuant.md:193 161002 4 3 1

英文文档(58 处 / 175 条)

file:line code rowspan actual missing
activation/celu_v2/docs/aclnnCelu&aclnnInplaceCelu_en.md:346 161002 8 4 4
activation/fatrelu_mul/docs/aclnnFatreluMul_en.md:186 161002 8 1 7
activation/gelu_quant/docs/aclnnGeluQuant_en.md:274 361001 3 1 2
activation/hard_shrink/docs/aclnnHardshrink_en.md:171 161002 8 3 5
activation/leaky_relu/docs/aclnnLeakyRelu&aclnnInplaceLeakyRelu_en.md:338 161002 8 1 7
activation/leaky_relu/docs/aclnnRReluWithNoise&aclnnInplaceRReluWithNoise_en.md:462 161002 8 4 4
activation/log_softmax_v2/docs/aclnnLogSoftmax_en.md:166 161002 3 2 1
activation/mish/docs/aclnnMish&aclnnInplaceMish_en.md:318 161002 8 2 6
activation/sigmoid/docs/aclnnSigmoid&aclnnInplaceSigmoid_en.md:168 161002 8 2 6
activation/swish/docs/aclnnSilu_en.md:153 161002 8 3 5
conv/convolution_backward/docs/aclnnConvolutionBackward_en.md:361 161001 2 1 1
foreach/foreach_add_scalar/docs/aclnnForeachAddScalarV2_en.md:183 161002 6 4 2
foreach/foreach_div_scalar/docs/aclnnForeachDivScalarV2_en.md:164 161002 6 5 1
foreach/foreach_maximum_scalar/docs/aclnnForeachMaximumScalarV2_en.md:166 161002 6 5 1
index/apply_top_k_top_p_with_sorted/docs/aclnnApplyTopKTopP_en.md:202 161002 8 3 5
index/index_put_v2/docs/aclnnIndexPutImpl_en.md:193 161002 6 5 1
index/scatter_elements_v2/docs/aclnnScatterValue&aclnnInplaceScatterValue_en.md:231 161002 9 8 1
index/scatter_elements_v2/docs/aclnnScatterValue&aclnnInplaceScatterValue_en.md:432 161002 7 6 1
loss/binary_cross_entropy_grad/docs/aclnnBinaryCrossEntropyBackward_en.md:216 161002 6 5 1
loss/fused_cross_entropy_loss_with_max_sum/docs/aclnnFusedCrossEntropyLossWithMaxSum_en.md:227 161002 7 6 1
loss/logit/docs/aclnnLogit_en.md:173 161002 8 1 7
loss/logit_grad/docs/aclnnLogitGrad_en.md:178 161002 8 1 7
matmul/batch_mat_mul_v3/docs/aclnnBatchMatMul_en.md:194 161002 5 4 1
matmul/dual_level_quant_batch_matmul/docs/aclnnDualLevelQuantMatmulWeightNz_en.md:299 161002 5 4 1
matmul/flat_quant/docs/aclnnFlatQuantV2_en.md:307 161002 12 11 1
matmul/flat_quant/docs/aclnnFlatQuant_en.md:248 161002 12 11 1
matmul/mat_mul_v3/docs/aclnnMatmulWeightNz_en.md:195 161002 5 3 2
matmul/mv/docs/aclnnMv_en.md:198 161002 7 4 3
matmul/quant_matmul/docs/aclnnQuantMatmulV2_en.md:203 161002 7 5 2
matmul/quant_matmul/docs/aclnnQuantMatmul_en.md:182 161002 7 5 2
matmul/quant_matmul_dequant/docs/aclnnQuantMatmulDequant_en.md:283 561002 3 1 2
matmul/sparse4to2quant_matmul/docs/aclnnSparse4to2QuantMatmulWeightNz_en.md:193 161002 8 1 7
matmul/sparse4to2quant_matmul/docs/aclnnTransSparse4to2Para_en.md:167 161002 8 1 7
matmul/transpose_batch_mat_mul/docs/aclnnTransposeBatchMatMul_en.md:265 161002 6 5 1
matmul/transpose_quant_batch_mat_mul/docs/aclnnTransposeQuantBatchMatMulWeightNz_en.md:270 161002 8 6 2
matmul/transpose_quant_batch_mat_mul/docs/aclnnTransposeQuantBatchMatMul_en.md:284 161002 8 5 3
norm/add_rms_norm_quant_v2/docs/aclnnAddRmsNormQuantV2_en.md:344 161002 3 1 2
norm/lp_norm_v2/docs/aclnnLinalgVectorNorm_en.md:192 161002 7 6 1
pooling/adaptive_avg_pool2d_grad/docs/aclnnAdaptiveAvgPool2dBackward_en.md:155 161002 6 5 1
pooling/max_pool3d_grad_with_argmax/docs/aclnnMaxPool2dWithIndicesBackward_en.md:250 161002 11 10 1
pooling/max_pool3d_grad_with_argmax/docs/aclnnMaxPool2dWithMaskBackward_en.md:216 161002 11 10 1
pooling/max_pool3d_grad_with_argmax/docs/aclnnMaxPool3dWithArgmaxBackward_en.md:221 161002 12 11 1
pooling/max_pool3d_with_argmax_v2/docs/aclnnMaxPool3dWithArgmax_en.md:237 161002 13 7 6
pooling/max_pool_v3/docs/aclnnMaxPool_en.md:236 161002 11 10 1
quant/anti_mx_quant/docs/aclnnAntiMxQuant_en.md:176 161001 2 1 1
quant/ascend_quant_v2/docs/aclnnAscendQuant_en.md:224 161002 7 6 1
quant/dynamic_block_quant/docs/aclnnDynamicBlockQuantV2_en.md:255 161002 5 2 3
quant/dynamic_block_quant/docs/aclnnDynamicBlockQuant_en.md:243 161002 5 1 4
quant/group_quant/docs/aclnnGroupQuant_en.md:192 161002 6 5 1
quant/grouped_quant_max/docs/aclnnGroupedQuantMax_en.md:215 161002 5 4 1
quant/mx_to_block_mx_quant/docs/aclnnMxToBlockMxQuant_en.md:200 161001 2 1 1
quant/quantize/docs/aclnnQuantize_en.md:209 161002 7 6 1
rnn/bidirection_lstm/docs/aclnnBidirectionLSTM_en.md:343 561002 3 1 2
rnn/bidirection_lstmv2/docs/aclnnBidirectionLSTMV2_en.md:365 561002 3 1 2
rnn/dynamic_rnn/docs/aclnnLSTM_en.md:419 161002 12 4 8
rnn/gru_grad/docs/aclnnGRUBackward_en.md:359 161002 12 4 8
rnn/single_layer_lstm_grad/docs/aclnnLstmBackward_en.md:603 161002 12 4 8
rnn/thnn_fused_lstm_cell_grad/docs/aclnnThnnFusedLstmCellBackward_en.md:256 161002 12 3 9
likedislike
liuhanwen8liuhanwen8
16 days ago Modify description of issue
yuning_chenyuning_chenCollaborators
15 days ago Assign gitee-duhuiping as responsible person
yuning_chen
yuning_chenCollaborators
15 days ago Comments:

初步分析

感谢反馈,我们已开始核对这个问题。

现象理解:[Documentation|文档反馈]: ops-nn 本仓 123 处错误码表 rowspan 声明行数多于实际描述行(缺 381 条子项,正式目录 43 处 / 37 个算子)。

当前判断:初步判断,待复现确认,还不能当成已确认根因。

涉及范围:根据标题与描述,先对照对应算子的实现/资料/校验路径排查。

下一步:我们会按描述构造用例复现,并对照代码与资料给出修复或文档修正方向。

likedislike
yuning_chen
yuning_chenCollaborators
15 days ago Comments:

您好,该问题已收到,当前问题 @gitee-duhuiping 正在处理中

likedislike
gitee-duhuipinggitee-duhuipingCollaborators
12 days ago Assign fengying555682 as responsible person
gitee-duhuipinggitee-duhuipingCollaborators
12 days ago Assign yanglu-1 as responsible person
gitee-duhuipinggitee-duhuipingCollaborators
11 days ago Associated pull request
fengying555682fengying555682Collaborators
10 days ago Associated pull request
chenjiayi_chenjiayi_Collaborators
5 days ago Assign chenjiayi_ as responsible person
chenjiayi_
chenjiayi_Collaborators
5 days ago Comments:

补充说明本问题的处理进展:

当前状态:
关联 PR PR#14176 已合入,合入时间为 2026-11-11 14:56;该 issue 仍为 open,指派后评论区保持静止。

依据:

  • PR 状态:merged。
  • 合入时间:2026-11-11 14:56。
  • 当前责任人:gitee-duhuiping, fengying555682, yanglu-1。
  • 指派时间:2026-11-09 10:17。

后续:
请责任人复核 master 的合入情况;确认修复后关闭 issue。

likedislike
yanglu-1yanglu-1Collaborators
2 days ago Associated pull request
CANN-robotCANN-robotCollaborators
2 days ago Close issue
CANN-robotCANN-robotCollaborators
2 days ago changed custom state from 进行中 to 已完成
CANN-robotCANN-robotCollaborators
2 days ago Added label: resolved
chenjiayi_chenjiayi_Collaborators
12 hours ago Removed responsible person chenjiayi_