@cannbot-plugin/cannbot-harness 0.0.0-stage → 0.1.4

package/assets/ops-direct-invoke/skills/ascendc-regbase-best-practice/references/reference-ops/open_source_operator_table.md

@@ -0,0 +1,740 @@
+---
+title: 开源算子参考表
+purpose: 提供候选算子目录索引，并标注需要进一步确认是否包含 RegBase body 或 RegBase route。
+read_when:
+  - 需要为新 RegBase 任务选择真实参考实现。
+  - 需要按算子类型定位可能相关的开源实现目录。
+not_for:
+  - 证明某个算子目录一定可复用。
+  - 替代源码阅读或当前任务的 RegBase 适配判断。
+keywords:
+  - reference operator
+  - open source
+  - RegBase
+  - arch35
+depth: reference
+topic_type: reference
+---
+
+# 开源算子参考表
+
+该表是轻量开源参考表，保留常见算子名、基础类型和代码路径，用于在设计或编码前快速定位值得阅读的现有实现。请把它当作候选索引，而不是可直接复用的证明；表中不是每个算子都已经有 RegBase body。
+
+## 使用方式
+
+- 想确认开源仓中是否已有值得检查的真实参考实现时使用本表。
+- 命中只表示“建议阅读该候选目录”，不表示目标任务可以直接复用，也不证明该目录一定包含 RegBase 实现。
+- 查看 `代码路径`，定位最接近候选实现所在的仓库子树。
+- 使用某个条目前，必须在对应目录下确认是否存在 `op_kernel/arch35/*regbase*`、`*_apt.cpp`、`__VEC_SCOPE__`、`__simd_vf__`、`RegTensor` 或 `AscendC::MicroAPI::*` 等 RegBase 证据。
+
+## 优先验证的 RegBase 参考路径示例
+
+下面这些路径通常包含 `arch35/*regbase*` 或 RegBase tiling / body 文件，适合作为第一批参考入口。仍需按当前任务重新判断 shape、dtype、tail 和 API 适配性。
+
+| 场景 | 代码路径 | 可重点阅读 |
+|---|---|---|
+| foreach 逐元素 / scalar | `ops-nn/foreach/foreach_add_scalar/` | `_apt.cpp` route、`op_kernel/arch35/foreach_add_scalar_regbase.h` |
+| foreach 融合三元链 | `ops-nn/foreach/foreach_addcmul_scalar/` | 多输入 VF 链、scalar 参数处理 |
+| foreach reduce / norm | `ops-nn/foreach/foreach_norm/` | reduction、tail mask、partial result |
+| norm 融合与量化 | `ops-nn/norm/add_rms_norm_dynamic_quant/` | `__VEC_SCOPE__`、`RegTensor`、fp32 中间值、量化写回 |
+| norm split reduce | `ops-nn/norm/add_rms_norm_quant/` | `regbase_split_reduce` 与 common helper |
+| group norm | `ops-nn/norm/group_norm_v2/` | welford / two-pass route、generalized body |
+| pooling 大 kernel | `ops-nn/pooling/max_pool3d_with_argmax_v2/` | 多 route、复杂 tiling、arch35 RegBase body |
+
+## 算子表
+
+| 算子 | 基础类型 | 代码路径 |
+|---|---|---|
+| `abs` | `math` | `ops-math/math/abs/` |
+| `acos` | `math` | `ops-math/experimental/math/acos/` |
+| `ada_layer_norm` | `norm` | `ops-nn/norm/ada_layer_norm/` |
+| `ada_layer_norm_quant` | `norm` | `ops-nn/norm/ada_layer_norm_quant/` |
+| `ada_layer_norm_v2` | `norm` | `ops-nn/norm/ada_layer_norm_v2/` |
+| `adam_apply_one` | `optim` | `ops-nn/optim/adam_apply_one/` |
+| `adam_apply_one_with_decay` | `optim` | `ops-nn/optim/adam_apply_one_with_decay/` |
+| `adam_apply_one_with_decay_assign` | `optim` | `ops-nn/optim/adam_apply_one_with_decay_assign/` |
+| `adaptive_avg_pool3d` | `pooling` | `ops-nn/pooling/adaptive_avg_pool3d/` |
+| `adaptive_avg_pool3d_grad` | `pooling` | `ops-nn/pooling/adaptive_avg_pool3d_grad/` |
+| `adaptive_max_pool3d` | `pooling` | `ops-nn/pooling/adaptive_max_pool3d/` |
+| `adaptive_max_pool3d_grad` | `pooling` | `ops-nn/pooling/adaptive_max_pool3d_grad/` |
+| `add` | `math` | `ops-math/math/add/` |
+| `add_layer_norm_grad` | `norm` | `ops-nn/norm/add_layer_norm_grad/` |
+| `add_layer_norm_quant` | `norm` | `ops-nn/norm/add_layer_norm_quant/` |
+| `add_lora` | `math` | `ops-math/math/add_lora/` |
+| `add_n` | `math` | `ops-math/math/add_n/` |
+| `add_rms_norm_cast` | `norm` | `ops-nn/norm/add_rms_norm_cast/` |
+| `add_rms_norm_dynamic_quant` | `norm` | `ops-nn/norm/add_rms_norm_dynamic_quant/` |
+| `add_rms_norm_dynamic_quant_v2` | `norm` | `ops-nn/norm/add_rms_norm_dynamic_quant_v2/` |
+| `add_rms_norm_quant` | `norm` | `ops-nn/norm/add_rms_norm_quant/` |
+| `add_rms_norm_quant_v2` | `norm` | `ops-nn/norm/add_rms_norm_quant_v2/` |
+| `add_v2` | `math` | `ops-math/experimental/math/add_v2/` |
+| `addcdiv` | `math` | `ops-math/math/addcdiv/` |
+| `addcmul` | `math` | `ops-math/math/addcmul/` |
+| `addr` | `math` | `ops-math/math/addr/` |
+| `adjacent_difference` | `math` | `ops-math/math/adjacent_difference/` |
+| `advance_step` | `optim` | `ops-nn/optim/advance_step/` |
+| `aipp` | `image` | `ops-cv/image/aipp/` |
+| `all_gather_matmul` | `mc2` | `ops-transformer/mc2/all_gather_matmul/` |
+| `all_gather_matmul_v2` | `mc2` | `ops-transformer/mc2/all_gather_matmul_v2/` |
+| `allto_all_all_gather_batch_mat_mul` | `mc2` | `ops-transformer/mc2/allto_all_all_gather_batch_mat_mul/` |
+| `allto_all_matmul` | `mc2` | `ops-transformer/mc2/allto_all_matmul/` |
+| `allto_allv_grouped_mat_mul` | `mc2` | `ops-transformer/mc2/allto_allv_grouped_mat_mul/` |
+| `angle_v2` | `math` | `ops-math/math/angle_v2/` |
+| `apply_adagrad_d` | `optim` | `ops-nn/optim/apply_adagrad_d/` |
+| `apply_adam_d` | `optim` | `ops-nn/optim/apply_adam_d/` |
+| `apply_adam_w` | `optim` | `ops-nn/optim/apply_adam_w/` |
+| `apply_adam_wquant` | `optim` | `ops-nn/optim/apply_adam_w_quant/` |
+| `apply_adam_wv2` | `optim` | `ops-nn/optim/apply_adam_w_v2/` |
+| `apply_ftrl` | `optim` | `ops-nn/optim/apply_ftrl/` |
+| `apply_fused_ema_adam` | `optim` | `ops-nn/optim/apply_fused_ema_adam/` |
+| `apply_gradient_descent` | `optim` | `ops-nn/optim/apply_gradient_descent/` |
+| `apply_momentum` | `optim` | `ops-nn/optim/apply_momentum/` |
+| `apply_rotary_pos_emb` | `posembedding` | `ops-transformer/posembedding/apply_rotary_pos_emb/` |
+| `apply_top_ktop_pwith_sorted` | `index` | `ops-nn/index/apply_top_k_top_p_with_sorted/` |
+| `arange` | `math` | `ops-math/experimental/math/arange/` |
+| `arg_max_v2` | `math` | `ops-math/math/arg_max_v2/` |
+| `arg_max_with_value` | `math` | `ops-math/math/arg_max_with_value/` |
+| `arg_min` | `math` | `ops-math/math/arg_min/` |
+| `arg_min_with_value` | `math` | `ops-math/math/arg_min_with_value/` |
+| `as_strided` | `conversion` | `ops-math/conversion/as_strided/` |
+| `ascend_quant` | `quant` | `ops-nn/quant/ascend_quant/` |
+| `ascend_quant_v2` | `quant` | `ops-nn/quant/ascend_quant_v2/` |
+| `asinh` | `math` | `ops-math/experimental/math/asinh/` |
+| `assign` | `conversion` | `ops-math/conversion/assign/` |
+| `assign_add` | `math` | `ops-math/math/assign_add/` |
+| `assign_sub` | `math` | `ops-math/math/assign_sub/` |
+| `atan` | `math` | `ops-math/experimental/math/atan/` |
+| `attention_to_ffn` | `mc2` | `ops-transformer/mc2/attention_to_ffn/` |
+| `attention_update` | `attention` | `ops-transformer/attention/attention_update/` |
+| `attention_worker_combine` | `attention` | `ops-transformer/attention/attention_worker_combine/` |
+| `avg_pool3d` | `pooling` | `ops-nn/pooling/avg_pool3_d/` |
+| `avg_pool3d_grad` | `pooling` | `ops-nn/pooling/avg_pool3_d_grad/` |
+| `avg_pool_v2grad` | `pooling` | `ops-nn/pooling/avg_pool_v2_grad/` |
+| `axpy` | `math` | `ops-math/math/axpy/` |
+| `axpy_v2` | `math` | `ops-math/experimental/math/axpy_v2/` |
+| `axpy_v3` | `math` | `ops-math/experimental/math/axpy_v3/` |
+| `background_replace` | `objdetect` | `ops-cv/objdetect/background_replace/` |
+| `batch_mat_mul_reduce_scatter_allto_all` | `mc2` | `ops-transformer/mc2/batch_mat_mul_reduce_scatter_allto_all/` |
+| `batch_mat_mul_v3` | `matmul` | `ops-nn/matmul/batch_mat_mul_v3/` |
+| `batch_norm_grad` | `norm` | `ops-nn/norm/batch_norm_grad/` |
+| `batch_norm_grad_v3` | `norm` | `ops-nn/norm/batch_norm_grad_v3/` |
+| `batch_norm_v3` | `norm` | `ops-nn/norm/batch_norm_v3/` |
+| `bias_add` | `math` | `ops-math/math/bias_add/` |
+| `bias_add_grad` | `math` | `ops-math/math/bias_add_grad/` |
+| `bidirection_lstm` | `rnn` | `ops-nn/rnn/bidirection_lstm/` |
+| `bidirection_lstmv2` | `rnn` | `ops-nn/rnn/bidirection_lstmv2/` |
+| `binary_cross_entropy_grad_v2` | `loss` | `ops-nn/experimental/loss/binary_cross_entropy_grad_v2/` |
+| `bincount` | `math` | `ops-math/math/bincount/` |
+| `bitwise_and` | `math` | `ops-math/experimental/math/bitwise_and/` |
+| `bitwise_or` | `math` | `ops-math/math/bitwise_or/` |
+| `bitwise_xor` | `math` | `ops-math/math/bitwise_xor/` |
+| `blend_images_custom` | `objdetect` | `ops-cv/objdetect/blend_images_custom/` |
+| `broadcast_to` | `conversion` | `ops-math/conversion/broadcast_to/` |
+| `broadcast_v2` | `conversion` | `ops-math/experimental/conversion/broadcast_v2/` |
+| `cast` | `math` | `ops-math/math/cast/` |
+| `cast_v2` | `math` | `ops-math/experimental/math/cast_v2/` |
+| `cdist` | `math` | `ops-math/math/cdist/` |
+| `ceil` | `math` | `ops-math/math/ceil/` |
+| `ceil_v2` | `math` | `ops-math/experimental/math/ceil_v2/` |
+| `celu_v2` | `activation` | `ops-nn/experimental/activation/celu_v2/` |
+| `chamfer_distance_grad` | `loss` | `ops-nn/loss/chamfer_distance_grad/` |
+| `cholesky` | `math` | `ops-math/math/cholesky/` |
+| `circular_pad` | `conversion` | `ops-math/conversion/circular_pad/` |
+| `circular_pad_grad` | `conversion` | `ops-math/conversion/circular_pad_grad/` |
+| `clip_by_value` | `conversion` | `ops-math/conversion/clip_by_value/` |
+| `clip_by_value_v2` | `conversion` | `ops-math/conversion/clip_by_value_v2/` |
+| `clipped_swiglu` | `activation` | `ops-nn/activation/clipped_swiglu/` |
+| `coalesce_sparse` | `conversion` | `ops-math/conversion/coalesce_sparse/` |
+| `col2im` | `conversion` | `ops-math/experimental/conversion/col2_im/` |
+| `complex_v2` | `math` | `ops-math/experimental/math/complex_v2/` |
+| `concat` | `conversion` | `ops-math/conversion/concat/` |
+| `concat_d` | `conversion` | `ops-math/conversion/concat_d/` |
+| `concat_dv2` | `conversion` | `ops-math/conversion/concat_dv2/` |
+| `concat_v2` | `conversion` | `ops-math/conversion/concat_v2/` |
+| `confusion_transpose_d` | `conversion` | `ops-math/conversion/confusion_transpose_d/` |
+| `conv2d_v2` | `conv` | `ops-nn/conv/conv2d_v2/` |
+| `conv3d_backprop_filter_v2` | `conv` | `ops-nn/conv/conv3d_backprop_filter_v2/` |
+| `conv3d_backprop_input_v2` | `conv` | `ops-nn/conv/conv3d_backprop_input_v2/` |
+| `conv3d_transpose_v2` | `conv` | `ops-nn/conv/conv3d_transpose_v2/` |
+| `conv3d_v2` | `conv` | `ops-nn/conv/conv3d_v2/` |
+| `cos` | `math` | `ops-math/math/cos/` |
+| `cosh` | `math` | `ops-math/experimental/math/cosh/` |
+| `cross_entropy_loss` | `loss` | `ops-nn/loss/cross_entropy_loss/` |
+| `cross_entropy_loss_grad` | `loss` | `ops-nn/loss/cross_entropy_loss_grad/` |
+| `cross_v2` | `loss` | `ops-nn/loss/cross_v2/` |
+| `ctc_loss_v2` | `loss` | `ops-nn/loss/ctc_loss_v2/` |
+| `ctc_loss_v2grad` | `loss` | `ops-nn/loss/ctc_loss_v2_grad/` |
+| `ctc_loss_v3` | `loss` | `ops-nn/loss/ctc_loss_v3/` |
+| `ctc_loss_v3grad` | `loss` | `ops-nn/loss/ctc_loss_v3_grad/` |
+| `cumsum` | `math` | `ops-math/math/cumsum/` |
+| `deep_norm_grad` | `norm` | `ops-nn/norm/deep_norm_grad/` |
+| `deformable_conv2d` | `conv` | `ops-nn/conv/deformable_conv2d/` |
+| `deformable_offsets` | `conv` | `ops-nn/conv/deformable_offsets/` |
+| `deformable_offsets_grad` | `image` | `ops-cv/image/deformable_offsets_grad/` |
+| `dense_lightning_indexer_grad_kl_loss` | `attention` | `ops-transformer/attention/dense_lightning_indexer_grad_kl_loss/` |
+| `dense_lightning_indexer_softmax_lse` | `attention` | `ops-transformer/attention/dense_lightning_indexer_softmax_lse/` |
+| `depth_to_space` | `conversion` | `ops-math/conversion/depth_to_space/` |
+| `dequant_bias` | `quant` | `ops-nn/quant/dequant_bias/` |
+| `dequant_rope_quant_kvcache` | `posembedding` | `ops-transformer/posembedding/dequant_rope_quant_kvcache/` |
+| `dequant_swiglu_quant` | `quant` | `ops-nn/quant/dequant_swiglu_quant/` |
+| `diag_flat` | `conversion` | `ops-math/conversion/diag_flat/` |
+| `diag_part` | `math` | `ops-math/math/diag_part/` |
+| `diag_v2` | `conversion` | `ops-math/conversion/diag_v2/` |
+| `distribute_barrier` | `mc2` | `ops-transformer/mc2/distribute_barrier/` |
+| `div` | `math` | `ops-math/math/div/` |
+| `div_no_nan` | `math` | `ops-math/math/div_no_nan/` |
+| `div_v2` | `math` | `ops-math/experimental/math/div_v2/` |
+| `dot` | `math` | `ops-math/math/dot/` |
+| `dot_v2` | `math` | `ops-math/experimental/math/dot_v2/` |
+| `drop_out_do_mask` | `random` | `ops-math/random/drop_out_do_mask/` |
+| `drop_out_v3` | `random` | `ops-math/random/drop_out_v3/` |
+| `dua_quantize_add_layer_norm` | `norm` | `ops-nn/norm/dua_quantize_add_layer_norm/` |
+| `dual_level_quant_batch_matmul` | `matmul` | `ops-nn/matmul/dual_level_quant_batch_matmul/` |
+| `dynamic_block_quant` | `quant` | `ops-nn/quant/dynamic_block_quant/` |
+| `dynamic_dual_level_mx_quant` | `quant` | `ops-nn/quant/dynamic_dual_level_mx_quant/` |
+| `dynamic_mx_quant` | `quant` | `ops-nn/quant/dynamic_mx_quant/` |
+| `dynamic_mx_quant_with_dual_axis` | `quant` | `ops-nn/quant/dynamic_mx_quant_with_dual_axis/` |
+| `dynamic_partition` | `conversion` | `ops-math/conversion/dynamic_partition/` |
+| `dynamic_quant` | `quant` | `ops-nn/quant/dynamic_quant/` |
+| `dynamic_quant_update_scatter` | `quant` | `ops-nn/quant/dynamic_quant_update_scatter/` |
+| `dynamic_quant_update_scatter_v2` | `quant` | `ops-nn/quant/dynamic_quant_update_scatter_v2/` |
+| `dynamic_quant_v2` | `quant` | `ops-nn/quant/dynamic_quant_v2/` |
+| `dynamic_rnn` | `rnn` | `ops-nn/rnn/dynamic_rnn/` |
+| `dynamic_rnnv2` | `rnn` | `ops-nn/rnn/dynamic_rnnv2/` |
+| `elu` | `activation` | `ops-nn/activation/elu/` |
+| `elu_grad` | `activation` | `ops-nn/activation/elu_grad/` |
+| `elu_grad_v2` | `activation` | `ops-nn/activation/elu_grad_v2/` |
+| `embedding_bag` | `index` | `ops-nn/index/embedding_bag/` |
+| `embedding_dense_grad` | `index` | `ops-nn/index/embedding_dense_grad/` |
+| `embedding_dense_grad_v2` | `index` | `ops-nn/index/embedding_dense_grad_v2/` |
+| `embedding_hash_table_apply_adam_w` | `hash` | `ops-nn/hash/embedding_hash_table_apply_adam_w/` |
+| `embedding_hash_table_export` | `hash` | `ops-nn/hash/embedding_hash_table_export/` |
+| `embedding_hash_table_import` | `hash` | `ops-nn/hash/embedding_hash_table_import/` |
+| `embedding_hash_table_lookup_or_insert` | `hash` | `ops-nn/hash/embedding_hash_table_lookup_or_insert/` |
+| `equal` | `math` | `ops-math/math/equal/` |
+| `erf` | `math` | `ops-math/experimental/math/erf/` |
+| `exp` | `math` | `ops-math/math/exp/` |
+| `exp_segsum_grad` | `math` | `ops-math/math/exp_segsum_grad/` |
+| `expand_into_jagged_permute` | `index` | `ops-nn/index/expand_into_jagged_permute/` |
+| `extend_conv2d` | `conv` | `ops-nn/conv/extend_conv2d/` |
+| `eye` | `math` | `ops-math/math/eye/` |
+| `eye_v2` | `math` | `ops-math/experimental/math/eye_v2/` |
+| `fake_quant_affine_cachemask` | `quant` | `ops-nn/quant/fake_quant_affine_cachemask/` |
+| `fast_gelu` | `activation` | `ops-nn/activation/fast_gelu/` |
+| `fast_gelu_grad` | `activation` | `ops-nn/activation/fast_gelu_grad/` |
+| `fatrelu_mul` | `activation` | `ops-nn/activation/fatrelu_mul/` |
+| `feeds_repeat` | `conversion` | `ops-math/conversion/feeds_repeat/` |
+| `ffn` | `ffn` | `ops-transformer/ffn/ffn/` |
+| `ffn_to_attention` | `mc2` | `ops-transformer/mc2/ffn_to_attention/` |
+| `ffn_worker_batching` | `ffn` | `ops-transformer/ffn/ffn_worker_batching/` |
+| `fill` | `conversion` | `ops-math/conversion/fill/` |
+| `fill_diagonal_v2` | `conversion` | `ops-math/conversion/fill_diagonal_v2/` |
+| `fill_v2` | `conversion` | `ops-math/experimental/conversion/fill_v2/` |
+| `fills` | `conversion` | `ops-math/conversion/fills/` |
+| `flash_attention_score` | `attention` | `ops-transformer/attention/flash_attention_score/` |
+| `flash_attention_score_grad` | `attention` | `ops-transformer/attention/flash_attention_score_grad/` |
+| `flat_quant` | `quant` | `ops-nn/quant/flat_quant/` |
+| `floor` | `math` | `ops-math/math/floor/` |
+| `floor_div` | `math` | `ops-math/experimental/math/floor_div/` |
+| `floor_mod` | `math` | `ops-math/experimental/math/floor_mod/` |
+| `foreach_abs` | `foreach` | `ops-nn/foreach/foreach_abs/` |
+| `foreach_acos` | `foreach` | `ops-nn/foreach/foreach_acos/` |
+| `foreach_add_list` | `foreach` | `ops-nn/foreach/foreach_add_list/` |
+| `foreach_add_scalar` | `foreach` | `ops-nn/foreach/foreach_add_scalar/` |
+| `foreach_add_scalar_list` | `foreach` | `ops-nn/foreach/foreach_add_scalar_list/` |
+| `foreach_addcdiv_list` | `foreach` | `ops-nn/foreach/foreach_addcdiv_list/` |
+| `foreach_addcdiv_scalar` | `foreach` | `ops-nn/foreach/foreach_addcdiv_scalar/` |
+| `foreach_addcdiv_scalar_list` | `foreach` | `ops-nn/foreach/foreach_addcdiv_scalar_list/` |
+| `foreach_addcmul_list` | `foreach` | `ops-nn/foreach/foreach_addcmul_list/` |
+| `foreach_addcmul_scalar` | `foreach` | `ops-nn/foreach/foreach_addcmul_scalar/` |
+| `foreach_addcmul_scalar_list` | `foreach` | `ops-nn/foreach/foreach_addcmul_scalar_list/` |
+| `foreach_asin` | `foreach` | `ops-nn/foreach/foreach_asin/` |
+| `foreach_atan` | `foreach` | `ops-nn/foreach/foreach_atan/` |
+| `foreach_copy` | `foreach` | `ops-nn/foreach/foreach_copy/` |
+| `foreach_cos` | `foreach` | `ops-nn/foreach/foreach_cos/` |
+| `foreach_cosh` | `foreach` | `ops-nn/foreach/foreach_cosh/` |
+| `foreach_div_list` | `foreach` | `ops-nn/foreach/foreach_div_list/` |
+| `foreach_div_scalar` | `foreach` | `ops-nn/foreach/foreach_div_scalar/` |
+| `foreach_div_scalar_list` | `foreach` | `ops-nn/foreach/foreach_div_scalar_list/` |
+| `foreach_erf` | `foreach` | `ops-nn/foreach/foreach_erf/` |
+| `foreach_erfc` | `foreach` | `ops-nn/foreach/foreach_erfc/` |
+| `foreach_exp` | `foreach` | `ops-nn/foreach/foreach_exp/` |
+| `foreach_expm1` | `foreach` | `ops-nn/foreach/foreach_expm1/` |
+| `foreach_lerp_list` | `foreach` | `ops-nn/foreach/foreach_lerp_list/` |
+| `foreach_lerp_scalar` | `foreach` | `ops-nn/foreach/foreach_lerp_scalar/` |
+| `foreach_log` | `foreach` | `ops-nn/foreach/foreach_log/` |
+| `foreach_log10` | `foreach` | `ops-nn/foreach/foreach_log10/` |
+| `foreach_log1p` | `foreach` | `ops-nn/foreach/foreach_log1p/` |
+| `foreach_log2` | `foreach` | `ops-nn/foreach/foreach_log2/` |
+| `foreach_maximum_list` | `foreach` | `ops-nn/foreach/foreach_maximum_list/` |
+| `foreach_maximum_scalar` | `foreach` | `ops-nn/foreach/foreach_maximum_scalar/` |
+| `foreach_maximum_scalar_list` | `foreach` | `ops-nn/foreach/foreach_maximum_scalar_list/` |
+| `foreach_minimum_list` | `foreach` | `ops-nn/foreach/foreach_minimum_list/` |
+| `foreach_minimum_scalar` | `foreach` | `ops-nn/foreach/foreach_minimum_scalar/` |
+| `foreach_minimum_scalar_list` | `foreach` | `ops-nn/foreach/foreach_minimum_scalar_list/` |
+| `foreach_mul_list` | `foreach` | `ops-nn/foreach/foreach_mul_list/` |
+| `foreach_mul_scalar` | `foreach` | `ops-nn/foreach/foreach_mul_scalar/` |
+| `foreach_mul_scalar_list` | `foreach` | `ops-nn/foreach/foreach_mul_scalar_list/` |
+| `foreach_neg` | `foreach` | `ops-nn/foreach/foreach_neg/` |
+| `foreach_non_finite_check_and_unscale` | `foreach` | `ops-nn/foreach/foreach_non_finite_check_and_unscale/` |
+| `foreach_norm` | `foreach` | `ops-nn/foreach/foreach_norm/` |
+| `foreach_pow_list` | `foreach` | `ops-nn/foreach/foreach_pow_list/` |
+| `foreach_pow_scalar` | `foreach` | `ops-nn/foreach/foreach_pow_scalar/` |
+| `foreach_pow_scalar_and_tensor` | `foreach` | `ops-nn/foreach/foreach_pow_scalar_and_tensor/` |
+| `foreach_pow_scalar_list` | `foreach` | `ops-nn/foreach/foreach_pow_scalar_list/` |
+| `foreach_reciprocal` | `foreach` | `ops-nn/foreach/foreach_reciprocal/` |
+| `foreach_round_off_number` | `foreach` | `ops-nn/foreach/foreach_round_off_number/` |
+| `foreach_sigmoid` | `foreach` | `ops-nn/foreach/foreach_sigmoid/` |
+| `foreach_sign` | `foreach` | `ops-nn/foreach/foreach_sign/` |
+| `foreach_sin` | `foreach` | `ops-nn/foreach/foreach_sin/` |
+| `foreach_sinh` | `foreach` | `ops-nn/foreach/foreach_sinh/` |
+| `foreach_sqrt` | `foreach` | `ops-nn/foreach/foreach_sqrt/` |
+| `foreach_sub_list` | `foreach` | `ops-nn/foreach/foreach_sub_list/` |
+| `foreach_sub_scalar` | `foreach` | `ops-nn/foreach/foreach_sub_scalar/` |
+| `foreach_sub_scalar_list` | `foreach` | `ops-nn/foreach/foreach_sub_scalar_list/` |
+| `foreach_tan` | `foreach` | `ops-nn/foreach/foreach_tan/` |
+| `foreach_tanh` | `foreach` | `ops-nn/foreach/foreach_tanh/` |
+| `foreach_zero_inplace` | `foreach` | `ops-nn/foreach/foreach_zero_inplace/` |
+| `fused_cross_entropy_loss_with_max_sum` | `loss` | `ops-nn/loss/fused_cross_entropy_loss_with_max_sum/` |
+| `fused_floyd_attention` | `attention` | `ops-transformer/attention/fused_floyd_attention/` |
+| `fused_floyd_attention_grad` | `attention` | `ops-transformer/attention/fused_floyd_attention_grad/` |
+| `fused_linear_cross_entropy_loss_grad` | `matmul` | `ops-nn/matmul/fused_linear_cross_entropy_loss_grad/` |
+| `fused_linear_online_max_sum` | `matmul` | `ops-nn/matmul/fused_linear_online_max_sum/` |
+| `fused_mat_mul` | `matmul` | `ops-nn/matmul/fused_mat_mul/` |
+| `fused_mul_add_n` | `math` | `ops-math/math/fused_mul_add_n/` |
+| `fused_quant_mat_mul` | `matmul` | `ops-nn/matmul/fused_quant_mat_mul/` |
+| `gather_elements` | `index` | `ops-nn/index/gather_elements/` |
+| `gather_elements_v2` | `index` | `ops-nn/index/gather_elements_v2/` |
+| `gather_elements_v3` | `index` | `ops-nn/experimental/index/gather_elements_v3/` |
+| `gather_nd` | `index` | `ops-nn/index/gather_nd/` |
+| `gather_pa_kv_cache` | `attention` | `ops-transformer/attention/gather_pa_kv_cache/` |
+| `gather_v2` | `index` | `ops-nn/index/gather_v2/` |
+| `ge_glu_grad_v2` | `activation` | `ops-nn/activation/ge_glu_grad_v2/` |
+| `ge_glu_v2` | `activation` | `ops-nn/activation/ge_glu_v2/` |
+| `gelu` | `activation` | `ops-nn/activation/gelu/` |
+| `gelu_grad` | `activation` | `ops-nn/activation/gelu_grad/` |
+| `gelu_quant` | `activation` | `ops-nn/activation/gelu_quant/` |
+| `gelu_v2` | `activation` | `ops-nn/activation/gelu_v2/` |
+| `gemm_v2` | `matmul` | `ops-nn/matmul/gemm_v2/` |
+| `gemm_v3` | `matmul` | `ops-nn/matmul/gemm_v3/` |
+| `gemma_rms_norm` | `norm` | `ops-nn/norm/gemma_rms_norm/` |
+| `ger_v2` | `math` | `ops-math/experimental/math/ger_v2/` |
+| `glu` | `activation` | `ops-nn/activation/glu/` |
+| `greater` | `math` | `ops-math/math/greater/` |
+| `greater_equal` | `math` | `ops-math/math/greater_equal/` |
+| `grid_sample` | `image` | `ops-cv/image/grid_sample/` |
+| `grid_sampler2d_grad` | `image` | `ops-cv/image/grid_sampler2_d_grad/` |
+| `grid_sampler3d_grad` | `image` | `ops-cv/image/grid_sampler3_d_grad/` |
+| `group_norm_grad` | `norm` | `ops-nn/norm/group_norm_grad/` |
+| `group_norm_silu` | `norm` | `ops-nn/norm/group_norm_silu/` |
+| `group_norm_silu_quant` | `norm` | `ops-nn/norm/group_norm_silu_quant/` |
+| `group_norm_swish` | `norm` | `ops-nn/norm/group_norm_swish/` |
+| `group_norm_swish_grad` | `norm` | `ops-nn/norm/group_norm_swish_grad/` |
+| `group_norm_v2` | `norm` | `ops-nn/norm/group_norm_v2/` |
+| `group_quant` | `quant` | `ops-nn/quant/group_quant/` |
+| `grouped_bias_add_grad` | `math` | `ops-math/math/grouped_bias_add_grad/` |
+| `grouped_dynamic_block_quant` | `quant` | `ops-nn/quant/grouped_dynamic_block_quant/` |
+| `grouped_dynamic_mx_quant` | `quant` | `ops-nn/quant/grouped_dynamic_mx_quant/` |
+| `grouped_mat_mul_all_reduce` | `mc2` | `ops-transformer/mc2/grouped_mat_mul_all_reduce/` |
+| `grouped_mat_mul_allto_allv` | `mc2` | `ops-transformer/mc2/grouped_mat_mul_allto_allv/` |
+| `grouped_matmul` | `gmm` | `ops-transformer/gmm/grouped_matmul/` |
+| `grouped_matmul_add` | `gmm` | `ops-transformer/gmm/grouped_matmul_add/` |
+| `grouped_matmul_finalize_routing` | `gmm` | `ops-transformer/gmm/grouped_matmul_finalize_routing/` |
+| `grouped_matmul_swiglu_quant` | `gmm` | `ops-transformer/gmm/grouped_matmul_swiglu_quant/` |
+| `grouped_matmul_swiglu_quant_v2` | `gmm` | `ops-transformer/gmm/grouped_matmul_swiglu_quant_v2/` |
+| `hans_decode` | `math` | `ops-math/math/hans_decode/` |
+| `hans_encode` | `math` | `ops-math/math/hans_encode/` |
+| `hard_swish_grad_v2` | `activation` | `ops-nn/activation/hard_swish_grad_v2/` |
+| `hard_swish_v2` | `activation` | `ops-nn/experimental/activation/hard_swish_v2/` |
+| `hardtanh_grad` | `activation` | `ops-nn/activation/hardtanh_grad/` |
+| `heaviside` | `activation` | `ops-nn/activation/heaviside/` |
+| `histogram_v2` | `math` | `ops-math/math/histogram_v2/` |
+| `im2col` | `conversion` | `ops-math/conversion/im2col/` |
+| `incre_flash_attention` | `attention` | `ops-transformer/attention/incre_flash_attention/` |
+| `index` | `index` | `ops-nn/index/index/` |
+| `index_fill` | `index` | `ops-nn/index/index_fill/` |
+| `index_fill_d` | `index` | `ops-nn/index/index_fill_d/` |
+| `index_put_v2` | `index` | `ops-nn/index/index_put_v2/` |
+| `index_put_with_sort` | `index` | `ops-nn/index/index_put_with_sort/` |
+| `index_put_with_sort_v2` | `index` | `ops-nn/index/index_put_with_sort_v2/` |
+| `init_embedding_hash_table` | `hash` | `ops-nn/hash/init_embedding_hash_table/` |
+| `inplace_add_layer_norm` | `norm` | `ops-nn/norm/inplace_add_layer_norm/` |
+| `inplace_add_rms_norm` | `norm` | `ops-nn/norm/inplace_add_rms_norm/` |
+| `inplace_index_add` | `index` | `ops-nn/index/inplace_index_add/` |
+| `inplace_index_add_with_sorted` | `index` | `ops-nn/index/inplace_index_add_with_sorted/` |
+| `inplace_matmul_all_reduce_add_rms_norm` | `mc2` | `ops-transformer/mc2/inplace_matmul_all_reduce_add_rms_norm/` |
+| `inplace_scatter_add` | `index` | `ops-nn/index/inplace_scatter_add/` |
+| `interleave_rope` | `posembedding` | `ops-transformer/posembedding/interleave_rope/` |
+| `invert` | `math` | `ops-math/math/invert/` |
+| `invert_v2` | `math` | `ops-math/experimental/math/invert_v2/` |
+| `iou_v2` | `objdetect` | `ops-cv/objdetect/iou_v2/` |
+| `is_close` | `math` | `ops-math/math/is_close/` |
+| `is_finite` | `math` | `ops-math/math/is_finite/` |
+| `is_inf` | `math` | `ops-math/math/is_inf/` |
+| `is_nan` | `math` | `ops-math/math/is_nan/` |
+| `is_neg_inf` | `math` | `ops-math/math/is_neg_inf/` |
+| `is_pos_inf` | `math` | `ops-math/math/is_pos_inf/` |
+| `kl_div_loss_grad` | `loss` | `ops-nn/loss/kl_div_loss_grad/` |
+| `kl_div_v2` | `math` | `ops-math/math/kl_div_v2/` |
+| `kv_quant_sparse_flash_attention` | `attention` | `ops-transformer/attention/kv_quant_sparse_flash_attention/` |
+| `kv_rms_norm_rope_cache` | `posembedding` | `ops-transformer/posembedding/kv_rms_norm_rope_cache/` |
+| `l1loss_grad` | `loss` | `ops-nn/loss/l1_loss_grad/` |
+| `l2loss` | `loss` | `ops-nn/loss/l2_loss/` |
+| `layer_norm_grad` | `norm` | `ops-nn/norm/layer_norm_grad/` |
+| `layer_norm_grad_v3` | `norm` | `ops-nn/norm/layer_norm_grad_v3/` |
+| `layer_norm_v3` | `norm` | `ops-nn/norm/layer_norm_v3/` |
+| `leaky_relu` | `activation` | `ops-nn/activation/leaky_relu/` |
+| `leaky_relu_grad` | `activation` | `ops-nn/activation/leaky_relu_grad/` |
+| `leakyrelu_v2` | `activation` | `ops-nn/experimental/activation/leakyrelu_v2/` |
+| `lerp` | `math` | `ops-math/math/lerp/` |
+| `less` | `math` | `ops-math/math/less/` |
+| `less_equal` | `math` | `ops-math/math/less_equal/` |
+| `lightning_indexer` | `attention` | `ops-transformer/attention/lightning_indexer/` |
+| `lightning_indexer_grad` | `attention` | `ops-transformer/attention/lightning_indexer_grad/` |
+| `lin_space` | `math` | `ops-math/math/lin_space/` |
+| `lin_space_d` | `math` | `ops-math/experimental/math/lin_space_d/` |
+| `linear_index` | `index` | `ops-nn/index/linear_index/` |
+| `linear_index_v2` | `index` | `ops-nn/index/linear_index_v2/` |
+| `log` | `math` | `ops-math/experimental/math/log/` |
+| `log1p` | `math` | `ops-math/math/log1p/` |
+| `log1p_v2` | `math` | `ops-math/experimental/math/log1p_v2/` |
+| `log_softmax_v2` | `activation` | `ops-nn/activation/log_softmax_v2/` |
+| `logical_and` | `math` | `ops-math/experimental/math/logical_and/` |
+| `logical_not` | `math` | `ops-math/math/logical_not/` |
+| `logical_or` | `math` | `ops-math/math/logical_or/` |
+| `logit` | `loss` | `ops-nn/loss/logit/` |
+| `logit_grad` | `loss` | `ops-nn/loss/logit_grad/` |
+| `lp_loss` | `loss` | `ops-nn/loss/lp_loss/` |
+| `lp_norm_v2` | `norm` | `ops-nn/norm/lp_norm_v2/` |
+| `map_index` | `index` | `ops-nn/index/map_index/` |
+| `masked_fill` | `conversion` | `ops-math/conversion/masked_fill/` |
+| `masked_scale` | `math` | `ops-math/math/masked_scale/` |
+| `masked_scatter` | `index` | `ops-nn/index/masked_scatter/` |
+| `masked_scatter_with_position` | `index` | `ops-nn/index/masked_scatter_with_position/` |
+| `masked_select_v3` | `conversion` | `ops-math/conversion/masked_select_v3/` |
+| `masked_softmax_with_rel_pos_bias` | `norm` | `ops-nn/norm/masked_softmax_with_rel_pos_bias/` |
+| `mat_mul_v3` | `matmul` | `ops-nn/matmul/mat_mul_v3/` |
+| `math_axpy_v2` | `math` | `ops-math/math/axpy_v2/` |
+| `math_bitwise_and` | `math` | `ops-math/math/bitwise_and/` |
+| `math_concat_v2` | `conversion` | `ops-math/experimental/conversion/concat_v2/` |
+| `math_floor_div` | `math` | `ops-math/math/floor_div/` |
+| `math_floor_mod` | `math` | `ops-math/math/floor_mod/` |
+| `math_im2col` | `conversion` | `ops-math/experimental/conversion/im2_col/` |
+| `math_log` | `math` | `ops-math/math/math_log/` |
+| `math_logical_and` | `math` | `ops-math/math/logical_and/` |
+| `math_masked_fill` | `math` | `ops-math/experimental/math/masked_fill/` |
+| `math_minimum` | `math` | `ops-math/math/minimum/` |
+| `math_ones_like` | `math` | `ops-math/math/ones_like/` |
+| `math_pow` | `math` | `ops-math/math/pow/` |
+| `math_range` | `math` | `ops-math/math/range/` |
+| `math_real_div` | `math` | `ops-math/math/real_div/` |
+| `math_round` | `math` | `ops-math/math/round/` |
+| `math_select_v2` | `math` | `ops-math/math/select_v2/` |
+| `math_sqrt` | `math` | `ops-math/math/sqrt/` |
+| `math_strided_slice` | `conversion` | `ops-math/experimental/conversion/strided_slice/` |
+| `math_tril` | `conversion` | `ops-math/experimental/conversion/tril/` |
+| `math_triu` | `conversion` | `ops-math/experimental/conversion/triu/` |
+| `matmul_all_reduce` | `mc2` | `ops-transformer/mc2/matmul_all_reduce/` |
+| `matmul_all_reduce_add_rms_norm` | `mc2` | `ops-transformer/mc2/matmul_all_reduce_add_rms_norm/` |
+| `matmul_allto_all` | `mc2` | `ops-transformer/mc2/matmul_allto_all/` |
+| `matmul_fp32` | `matmul` | `ops-nn/experimental/matmul/matmul_fp32/` |
+| `matmul_reduce_scatter` | `mc2` | `ops-transformer/mc2/matmul_reduce_scatter/` |
+| `matmul_reduce_scatter_v2` | `mc2` | `ops-transformer/mc2/matmul_reduce_scatter_v2/` |
+| `matrix_diag` | `conversion` | `ops-math/conversion/matrix_diag/` |
+| `max_pool3d` | `pooling` | `ops-nn/pooling/max_pool3_d/` |
+| `max_pool3d_grad_with_argmax` | `pooling` | `ops-nn/pooling/max_pool3d_grad_with_argmax/` |
+| `max_pool3d_with_argmax_v2` | `pooling` | `ops-nn/pooling/max_pool3d_with_argmax_v2/` |
+| `max_pool_grad_with_argmax` | `pooling` | `ops-nn/pooling/max_pool_grad_with_argmax/` |
+| `max_pool_grad_with_argmax_v3` | `pooling` | `ops-nn/pooling/max_pool_grad_with_argmax_v3/` |
+| `max_pool_v2` | `pooling` | `ops-nn/pooling/max_pool_v2/` |
+| `max_pool_v3` | `pooling` | `ops-nn/pooling/max_pool_v3/` |
+| `max_pool_with_argmax` | `pooling` | `ops-nn/pooling/max_pool_with_argmax/` |
+| `max_pool_with_argmax_v3` | `pooling` | `ops-nn/pooling/max_pool_with_argmax_v3/` |
+| `maximum` | `math` | `ops-math/math/maximum/` |
+| `maximum_v2` | `math` | `ops-math/experimental/math/maximum_v2/` |
+| `mc2batch_mat_mul_v3` | `mc2` | `ops-transformer/mc2/3rd/batch_mat_mul_v3/` |
+| `mc2mat_mul_v3` | `mc2` | `ops-transformer/mc2/3rd/mat_mul_v3/` |
+| `mem_set` | `conversion` | `ops-math/conversion/mem_set/` |
+| `mem_set_v2` | `conversion` | `ops-math/conversion/mem_set_v2/` |
+| `minimum` | `math` | `ops-math/experimental/math/minimum/` |
+| `mirror_pad` | `conversion` | `ops-math/conversion/mirror_pad/` |
+| `mish` | `activation` | `ops-nn/activation/mish/` |
+| `mish_grad` | `activation` | `ops-nn/experimental/activation/mish_grad/` |
+| `mla_preprocess` | `attention` | `ops-transformer/attention/mla_preprocess/` |
+| `mla_preprocess_v2` | `attention` | `ops-transformer/attention/mla_preprocess_v2/` |
+| `mod` | `math` | `ops-math/math/mod/` |
+| `modulate` | `vfusion` | `ops-nn/vfusion/modulate/` |
+| `modulate_grad` | `vfusion` | `ops-nn/vfusion/modulate_grad/` |
+| `moe_compute_expert_tokens` | `moe` | `ops-transformer/moe/moe_compute_expert_tokens/` |
+| `moe_distribute_combine` | `mc2` | `ops-transformer/mc2/moe_distribute_combine/` |
+| `moe_distribute_combine_add_rms_norm` | `mc2` | `ops-transformer/mc2/moe_distribute_combine_add_rms_norm/` |
+| `moe_distribute_combine_shmem` | `mc2` | `ops-transformer/experimental/mc2/moe_distribute_combine_shmem/` |
+| `moe_distribute_combine_v2` | `mc2` | `ops-transformer/mc2/moe_distribute_combine_v2/` |
+| `moe_distribute_dispatch` | `mc2` | `ops-transformer/mc2/moe_distribute_dispatch/` |
+| `moe_distribute_dispatch_shmem` | `mc2` | `ops-transformer/experimental/mc2/moe_distribute_dispatch_shmem/` |
+| `moe_distribute_dispatch_v2` | `mc2` | `ops-transformer/mc2/moe_distribute_dispatch_v2/` |
+| `moe_finalize_routing` | `moe` | `ops-transformer/moe/moe_finalize_routing/` |
+| `moe_finalize_routing_v2` | `moe` | `ops-transformer/moe/moe_finalize_routing_v2/` |
+| `moe_finalize_routing_v2grad` | `moe` | `ops-transformer/moe/moe_finalize_routing_v2_grad/` |
+| `moe_fused_topk` | `moe` | `ops-transformer/moe/moe_fused_topk/` |
+| `moe_gating_top_k` | `moe` | `ops-transformer/moe/moe_gating_top_k/` |
+| `moe_gating_top_ksoftmax` | `moe` | `ops-transformer/moe/moe_gating_top_k_softmax/` |
+| `moe_gating_top_ksoftmax_v2` | `moe` | `ops-transformer/moe/moe_gating_top_k_softmax_v2/` |
+| `moe_init_routing` | `moe` | `ops-transformer/moe/moe_init_routing/` |
+| `moe_init_routing_quant` | `moe` | `ops-transformer/moe/moe_init_routing_quant/` |
+| `moe_init_routing_quant_v2` | `moe` | `ops-transformer/moe/moe_init_routing_quant_v2/` |
+| `moe_init_routing_v2` | `moe` | `ops-transformer/moe/moe_init_routing_v2/` |
+| `moe_init_routing_v2grad` | `moe` | `ops-transformer/moe/moe_init_routing_v2_grad/` |
+| `moe_init_routing_v3` | `moe` | `ops-transformer/moe/moe_init_routing_v3/` |
+| `moe_inplace_index_add_with_sorted` | `moe` | `ops-transformer/moe/3rd/moe_inplace_index_add_with_sorted/` |
+| `moe_re_routing` | `moe` | `ops-transformer/moe/moe_re_routing/` |
+| `moe_token_permute` | `moe` | `ops-transformer/moe/moe_token_permute/` |
+| `moe_token_permute_grad` | `moe` | `ops-transformer/moe/moe_token_permute_grad/` |
+| `moe_token_permute_with_ep` | `moe` | `ops-transformer/moe/moe_token_permute_with_ep/` |
+| `moe_token_permute_with_ep_grad` | `moe` | `ops-transformer/moe/moe_token_permute_with_ep_grad/` |
+| `moe_token_permute_with_routing_map` | `moe` | `ops-transformer/moe/moe_token_permute_with_routing_map/` |
+| `moe_token_permute_with_routing_map_grad` | `moe` | `ops-transformer/moe/moe_token_permute_with_routing_map_grad/` |
+| `moe_token_unpermute` | `moe` | `ops-transformer/moe/moe_token_unpermute/` |
+| `moe_token_unpermute_grad` | `moe` | `ops-transformer/moe/moe_token_unpermute_grad/` |
+| `moe_token_unpermute_with_ep` | `moe` | `ops-transformer/moe/moe_token_unpermute_with_ep/` |
+| `moe_token_unpermute_with_ep_grad` | `moe` | `ops-transformer/moe/moe_token_unpermute_with_ep_grad/` |
+| `moe_token_unpermute_with_routing_map` | `moe` | `ops-transformer/moe/moe_token_unpermute_with_routing_map/` |
+| `moe_token_unpermute_with_routing_map_grad` | `moe` | `ops-transformer/moe/moe_token_unpermute_with_routing_map_grad/` |
+| `moe_update_expert` | `mc2` | `ops-transformer/mc2/moe_update_expert/` |
+| `mrgba_custom` | `objdetect` | `ops-cv/objdetect/mrgba_custom/` |
+| `mse_loss` | `loss` | `ops-nn/loss/mse_loss/` |
+| `mse_loss_grad` | `loss` | `ops-nn/loss/mse_loss_grad/` |
+| `mse_loss_grad_v2` | `loss` | `ops-nn/loss/mse_loss_grad_v2/` |
+| `mse_loss_v2` | `loss` | `ops-nn/loss/mse_loss_v2/` |
+| `mul` | `math` | `ops-math/math/mul/` |
+| `mul_addn` | `math` | `ops-math/math/mul_addn/` |
+| `mul_v2` | `math` | `ops-math/experimental/math/mul_v2/` |
+| `muls` | `math` | `ops-math/math/muls/` |
+| `multi_add_rms_norm_dynamic_quant` | `norm` | `ops-nn/norm/multi_add_rms_norm_dynamic_quant/` |
+| `multi_scale_deformable_attention_grad` | `vfusion` | `ops-nn/vfusion/multi_scale_deformable_attention_grad/` |
+| `multi_scale_deformable_attn_function` | `vfusion` | `ops-nn/vfusion/multi_scale_deformable_attn_function/` |
+| `neg` | `math` | `ops-math/math/neg/` |
+| `neg_v2` | `math` | `ops-math/math/neg_v2/` |
+| `nll_loss` | `loss` | `ops-nn/loss/nll_loss/` |
+| `nll_loss_grad` | `loss` | `ops-nn/loss/nll_loss_grad/` |
+| `nms_with_mask` | `image` | `ops-cv/experimental/image/nms_with_mask/` |
+| `nn_relu_v2` | `activation` | `ops-nn/experimental/activation/relu_v2/` |
+| `nn_swish_grad` | `activation` | `ops-nn/experimental/activation/swish_grad/` |
+| `non_finite_check` | `math` | `ops-math/math/non_finite_check/` |
+| `non_zero` | `index` | `ops-nn/index/non_zero/` |
+| `norm_rope_concat` | `posembedding` | `ops-transformer/posembedding/norm_rope_concat/` |
+| `norm_rope_concat_grad` | `posembedding` | `ops-transformer/posembedding/norm_rope_concat_grad/` |
+| `not_equal` | `math` | `ops-math/math/not_equal/` |
+| `nsa_compress` | `attention` | `ops-transformer/attention/nsa_compress/` |
+| `nsa_compress_attention` | `attention` | `ops-transformer/attention/nsa_compress_attention/` |
+| `nsa_compress_attention_infer` | `attention` | `ops-transformer/attention/nsa_compress_attention_infer/` |
+| `nsa_compress_grad` | `attention` | `ops-transformer/attention/nsa_compress_grad/` |
+| `nsa_compress_with_cache` | `attention` | `ops-transformer/attention/nsa_compress_with_cache/` |
+| `nsa_selected_attention` | `attention` | `ops-transformer/attention/nsa_selected_attention/` |
+| `nsa_selected_attention_grad` | `attention` | `ops-transformer/attention/nsa_selected_attention_grad/` |
+| `nsa_selected_attention_infer` | `attention` | `ops-transformer/attention/nsa_selected_attention_infer/` |
+| `one_hot` | `math` | `ops-math/math/one_hot/` |
+| `onehot` | `conversion` | `ops-math/experimental/conversion/onehot/` |
+| `ones_like` | `math` | `ops-math/experimental/math/ones_like/` |
+| `pack` | `conversion` | `ops-math/conversion/pack/` |
+| `pad` | `conversion` | `ops-math/conversion/pad/` |
+| `pad_v3` | `conversion` | `ops-math/conversion/pad_v3/` |
+| `pad_v3grad_replicate` | `conversion` | `ops-math/conversion/pad_v3_grad_replicate/` |
+| `pad_v3grad_replication` | `conversion` | `ops-math/conversion/pad_v3_grad_replication/` |
+| `pad_v4grad` | `conversion` | `ops-math/conversion/pad_v4_grad/` |
+| `pow` | `math` | `ops-math/experimental/math/pow/` |
+| `pows` | `math` | `ops-math/math/pows/` |
+| `qkv_rms_norm_rope_cache` | `posembedding` | `ops-transformer/posembedding/qkv_rms_norm_rope_cache/` |
+| `quant_all_reduce` | `mc2` | `ops-transformer/mc2/quant_all_reduce/` |
+| `quant_batch_matmul_inplace_add` | `matmul` | `ops-nn/matmul/quant_batch_matmul_inplace_add/` |
+| `quant_batch_matmul_v3` | `matmul` | `ops-nn/matmul/quant_batch_matmul_v3/` |
+| `quant_batch_matmul_v4` | `matmul` | `ops-nn/matmul/quant_batch_matmul_v4/` |
+| `quant_conv3d` | `conv` | `ops-nn/conv/quant_conv3d/` |
+| `quant_grouped_matmul_dequant` | `gmm` | `ops-transformer/gmm/quant_grouped_matmul_dequant/` |
+| `quant_grouped_matmul_inplace_add` | `gmm` | `ops-transformer/gmm/quant_grouped_matmul_inplace_add/` |
+| `quant_lightning_indexer` | `attention` | `ops-transformer/attention/quant_lightning_indexer/` |
+| `quant_matmul_reduce_sum` | `matmul` | `ops-nn/matmul/quant_matmul_reduce_sum/` |
+| `quant_reduce_scatter` | `mc2` | `ops-transformer/mc2/quant_reduce_scatter/` |
+| `quant_update_scatter` | `index` | `ops-nn/index/quant_update_scatter/` |
+| `quantize` | `quant` | `ops-nn/quant/quantize/` |
+| `quantize_add_layer_norm` | `norm` | `ops-nn/norm/quantize_add_layer_norm/` |
+| `quantized_batch_norm` | `norm` | `ops-nn/norm/quantized_batch_norm/` |
+| `rain_fusion_attention` | `attention` | `ops-transformer/attention/rain_fusion_attention/` |
+| `random_standard_normal_v2` | `random` | `ops-math/random/random_standard_normal_v2/` |
+| `random_uniform_int_v2` | `random` | `ops-math/random/random_uniform_int_v2/` |
+| `random_uniform_v2` | `random` | `ops-math/random/random_uniform_v2/` |
+| `range` | `math` | `ops-math/experimental/math/range/` |
+| `real_div` | `math` | `ops-math/experimental/math/real_div/` |
+| `reciprocal` | `math` | `ops-math/math/reciprocal/` |
+| `recurrent_gated_delta_rule` | `attention` | `ops-transformer/attention/recurrent_gated_delta_rule/` |
+| `reduce_all` | `math` | `ops-math/math/reduce_all/` |
+| `reduce_any` | `math` | `ops-math/math/reduce_any/` |
+| `reduce_log_sum_exp` | `math` | `ops-math/math/reduce_log_sum_exp/` |
+| `reduce_max` | `math` | `ops-math/math/reduce_max/` |
+| `reduce_max_v2` | `math` | `ops-math/experimental/math/reduce_max_v2/` |
+| `reduce_mean` | `math` | `ops-math/math/reduce_mean/` |
+| `reduce_min` | `math` | `ops-math/math/reduce_min/` |
+| `reduce_min_v2` | `math` | `ops-math/experimental/math/reduce_min_v2/` |
+| `reduce_prod` | `math` | `ops-math/math/reduce_prod/` |
+| `reduce_std_v2` | `math` | `ops-math/math/reduce_std_v2/` |
+| `reduce_sum` | `math` | `ops-math/math/reduce_sum/` |
+| `reduce_sum_v2` | `math` | `ops-math/experimental/math/reduce_sum_v2/` |
+| `reduce_var` | `math` | `ops-math/math/reduce_var/` |
+| `reflection_pad1d_v2` | `math` | `ops-math/experimental/math/reflection_pad1d_v2/` |
+| `reflection_pad3d_grad` | `conversion` | `ops-math/conversion/reflection_pad3d_grad/` |
+| `relu` | `activation` | `ops-nn/activation/relu/` |
+| `relu_grad` | `activation` | `ops-nn/activation/relu_grad/` |
+| `relu_grad_v2` | `activation` | `ops-nn/activation/relu_grad_v2/` |
+| `relu_v2` | `activation` | `ops-nn/activation/relu_v2/` |
+| `repeat_interleave` | `index` | `ops-nn/index/repeat_interleave/` |
+| `repeat_interleave_grad` | `index` | `ops-nn/index/repeat_interleave_grad/` |
+| `resize_bicubic_v2` | `image` | `ops-cv/image/resize_bicubic_v2/` |
+| `resize_bicubic_v2grad` | `image` | `ops-cv/image/resize_bicubic_v2_grad/` |
+| `resize_bilinear_v2` | `image` | `ops-cv/image/resize_bilinear_v2/` |
+| `resize_bilinear_v2grad` | `image` | `ops-cv/image/resize_bilinear_v2_grad/` |
+| `resize_linear` | `image` | `ops-cv/image/resize_linear/` |
+| `resize_linear_grad` | `image` | `ops-cv/image/resize_linear_grad/` |
+| `resize_nearest_neighbor_v2` | `image` | `ops-cv/image/resize_nearest_neighbor_v2/` |
+| `resize_nearest_neighbor_v2grad` | `image` | `ops-cv/image/resize_nearest_neighbor_v2_grad/` |
+| `resize_upsample_trilinear` | `image` | `ops-cv/image/resize_upsample_trilinear/` |
+| `reverse_sequence` | `index` | `ops-nn/index/reverse_sequence/` |
+| `rfft1d` | `math` | `ops-math/math/rfft1_d/` |
+| `ring_attention_update` | `attention` | `ops-transformer/attention/ring_attention_update/` |
+| `rms_norm` | `norm` | `ops-nn/norm/rms_norm/` |
+| `rms_norm_grad` | `norm` | `ops-nn/norm/rms_norm_grad/` |
+| `rms_norm_quant` | `norm` | `ops-nn/norm/rms_norm_quant/` |
+| `rms_norm_quant_v2` | `norm` | `ops-nn/norm/rms_norm_quant_v2/` |
+| `roi_align_rotated` | `objdetect` | `ops-cv/objdetect/roi_align_rotated/` |
+| `roi_align_rotated_grad` | `objdetect` | `ops-cv/objdetect/roi_align_rotated_grad/` |
+| `roll` | `conversion` | `ops-math/conversion/roll/` |
+| `rope_quant_kvcache` | `posembedding` | `ops-transformer/posembedding/rope_quant_kvcache/` |
+| `rope_with_sin_cos_cache` | `posembedding` | `ops-transformer/posembedding/rope_with_sin_cos_cache/` |
+| `rotary_position_embedding` | `posembedding` | `ops-transformer/posembedding/rotary_position_embedding/` |
+| `rotary_position_embedding_grad` | `posembedding` | `ops-transformer/posembedding/rotary_position_embedding_grad/` |
+| `round` | `math` | `ops-math/experimental/math/round/` |
+| `rsqrt_grad` | `math` | `ops-math/math/rsqrt_grad/` |
+| `scaled_masked_softmax_grad_v2` | `vfusion` | `ops-nn/vfusion/scaled_masked_softmax_grad_v2/` |
+| `scaled_masked_softmax_v2` | `vfusion` | `ops-nn/vfusion/scaled_masked_softmax_v2/` |
+| `scatter` | `index` | `ops-nn/index/scatter/` |
+| `scatter_add` | `index` | `ops-nn/index/scatter_add/` |
+| `scatter_add_with_sorted` | `index` | `ops-nn/index/scatter_add_with_sorted/` |
+| `scatter_elements` | `index` | `ops-nn/index/scatter_elements/` |
+| `scatter_elements_v2` | `index` | `ops-nn/index/scatter_elements_v2/` |
+| `scatter_list` | `index` | `ops-nn/index/scatter_list/` |
+| `scatter_nd` | `index` | `ops-nn/index/scatter_nd/` |
+| `scatter_nd_add` | `index` | `ops-nn/index/scatter_nd_add/` |
+| `scatter_pa_cache` | `attention` | `ops-transformer/attention/scatter_pa_cache/` |
+| `scatter_pa_kv_cache` | `attention` | `ops-transformer/attention/scatter_pa_kv_cache/` |
+| `scatter_sub` | `index` | `ops-nn/index/scatter_sub/` |
+| `segment_sum` | `index` | `ops-nn/index/segment_sum/` |
+| `segsum` | `math` | `ops-math/math/segsum/` |
+| `select` | `math` | `ops-math/math/select/` |
+| `select_v2` | `math` | `ops-math/experimental/math/select_v2/` |
+| `select_v3` | `math` | `ops-math/experimental/math/select_v3/` |
+| `sigmoid` | `activation` | `ops-nn/activation/sigmoid/` |
+| `sigmoid_cross_entropy_with_logits_v2` | `loss` | `ops-nn/loss/sigmoid_cross_entropy_with_logits_v2/` |
+| `sigmoid_grad` | `activation` | `ops-nn/activation/sigmoid_grad/` |
+| `sign` | `math` | `ops-math/math/sign/` |
+| `silu_grad` | `activation` | `ops-nn/activation/silu_grad/` |
+| `silu_mul` | `activation` | `ops-nn/experimental/activation/silu_mul/` |
+| `sim_thread_exponential` | `random` | `ops-math/random/sim_thread_exponential/` |
+| `single_layer_lstm_grad` | `rnn` | `ops-nn/rnn/single_layer_lstm_grad/` |
+| `sinh` | `math` | `ops-math/experimental/math/sinh/` |
+| `sinkhorn` | `math` | `ops-math/math/sinkhorn/` |
+| `slice` | `conversion` | `ops-math/conversion/slice/` |
+| `slice_v3` | `conversion` | `ops-math/experimental/conversion/slice_v3/` |
+| `smooth_l1loss_grad_v2` | `loss` | `ops-nn/loss/smooth_l1_loss_grad_v2/` |
+| `smooth_l1loss_v2` | `loss` | `ops-nn/loss/smooth_l1_loss_v2/` |
+| `softmax_cross_entropy_with_logits` | `activation` | `ops-nn/activation/softmax_cross_entropy_with_logits/` |
+| `softmax_grad` | `activation` | `ops-nn/activation/softmax_grad/` |
+| `softmax_v2` | `activation` | `ops-nn/activation/softmax_v2/` |
+| `softplus` | `activation` | `ops-nn/activation/softplus/` |
+| `sort` | `math` | `ops-math/math/sort/` |
+| `sort_with_index` | `math` | `ops-math/math/sort_with_index/` |
+| `space_to_depth` | `conversion` | `ops-math/conversion/space_to_depth/` |
+| `sparse4to2quant_matmul` | `matmul` | `ops-nn/matmul/sparse4to2quant_matmul/` |
+| `sparse_flash_attention` | `attention` | `ops-transformer/attention/sparse_flash_attention/` |
+| `sparse_flash_attention_grad` | `attention` | `ops-transformer/attention/sparse_flash_attention_grad/` |
+| `sparse_lightning_indexer_grad_kl_loss` | `attention` | `ops-transformer/attention/sparse_lightning_indexer_grad_kl_loss/` |
+| `sparse_softmax_cross_entropy_with_logits` | `loss` | `ops-nn/loss/sparse_softmax_cross_entropy_with_logits/` |
+| `sparse_tensor_dense_mat_mul` | `matmul` | `ops-nn/matmul/sparse_tensor_dense_mat_mul/` |
+| `sparse_to_dense` | `index` | `ops-nn/index/sparse_to_dense/` |
+| `split` | `conversion` | `ops-math/conversion/split/` |
+| `split_v` | `conversion` | `ops-math/conversion/split_v/` |
+| `sqrt` | `math` | `ops-math/experimental/math/sqrt/` |
+| `sqrt_grad` | `math` | `ops-math/math/sqrt_grad/` |
+| `square` | `math` | `ops-math/math/square/` |
+| `square_sum_v1` | `math` | `ops-math/math/square_sum_v1/` |
+| `square_v2` | `math` | `ops-math/experimental/math/square_v2/` |
+| `squared_difference` | `math` | `ops-math/math/squared_difference/` |
+| `stack_ball_query` | `conversion` | `ops-math/conversion/stack_ball_query/` |
+| `stack_group_points` | `objdetect` | `ops-cv/objdetect/stack_group_points/` |
+| `stateless_bernoulli` | `random` | `ops-math/random/stateless_bernoulli/` |
+| `stateless_drop_out_gen_mask` | `random` | `ops-math/random/stateless_drop_out_gen_mask/` |
+| `stateless_random_choice_with_mask` | `random` | `ops-math/random/stateless_random_choice_with_mask/` |
+| `stateless_random_normal_v2` | `random` | `ops-math/random/stateless_random_normal_v2/` |
+| `stateless_random_uniform_v2` | `random` | `ops-math/random/stateless_random_uniform_v2/` |
+| `stft` | `math` | `ops-math/math/stft/` |
+| `strided_slice` | `conversion` | `ops-math/conversion/strided_slice/` |
+| `strided_slice_assign_v2` | `conversion` | `ops-math/conversion/strided_slice_assign_v2/` |
+| `strided_slice_grad` | `conversion` | `ops-math/conversion/strided_slice_grad/` |
+| `sub` | `math` | `ops-math/math/sub/` |
+| `sub_v2` | `math` | `ops-math/experimental/math/sub_v2/` |
+| `swi_glu` | `activation` | `ops-nn/activation/swi_glu/` |
+| `swi_glu_grad` | `activation` | `ops-nn/activation/swi_glu_grad/` |
+| `swi_glu_quant` | `quant` | `ops-nn/quant/swi_glu_quant/` |
+| `swin_attention_ffn` | `ffn` | `ops-transformer/ffn/swin_attention_ffn/` |
+| `swin_attention_score_quant` | `attention` | `ops-transformer/attention/swin_attention_score_quant/` |
+| `swin_transformer_ln_qkv` | `ffn` | `ops-transformer/ffn/swin_transformer_ln_qkv/` |
+| `swin_transformer_ln_qkv_quant` | `ffn` | `ops-transformer/ffn/swin_transformer_ln_qkv_quant/` |
+| `swish` | `activation` | `ops-nn/activation/swish/` |
+| `swish_grad` | `activation` | `ops-nn/activation/swish_grad/` |
+| `tan_v2` | `math` | `ops-math/experimental/math/tan_v2/` |
+| `tanh` | `math` | `ops-math/math/tanh/` |
+| `tanh_grad` | `math` | `ops-math/math/tanh_grad/` |
+| `tensor_equal` | `math` | `ops-math/math/tensor_equal/` |
+| `tensor_move` | `conversion` | `ops-math/conversion/tensor_move/` |
+| `thnn_fused_lstm_cell_grad` | `rnn` | `ops-nn/rnn/thnn_fused_lstm_cell_grad/` |
+| `three_interpolate_backward` | `image` | `ops-cv/image/three_interpolate_backward/` |
+| `tile` | `math` | `ops-math/math/tile/` |
+| `top_ktop_psample` | `index` | `ops-nn/index/top_k_top_p_sample/` |
+| `top_ktop_psample_v2` | `index` | `ops-nn/index/top_k_top_p_sample_v2/` |
+| `top_kv2` | `math` | `ops-math/math/top_k_v2/` |
+| `trace_v2` | `math` | `ops-math/experimental/math/trace_v2/` |
+| `trans_quant_param_v2` | `quant` | `ops-nn/quant/trans_quant_param_v2/` |
+| `transform_bias_rescale_qkv` | `math` | `ops-math/math/transform_bias_rescale_qkv/` |
+| `transpose` | `conversion` | `ops-math/conversion/transpose/` |
+| `transpose_batch_mat_mul` | `matmul` | `ops-nn/matmul/transpose_batch_mat_mul/` |
+| `transpose_quant_batch_mat_mul` | `matmul` | `ops-nn/matmul/transpose_quant_batch_mat_mul/` |
+| `transpose_v2` | `conversion` | `ops-math/conversion/transpose_v2/` |
+| `transposev` | `conversion` | `ops-math/experimental/conversion/transposev/` |
+| `tril` | `conversion` | `ops-math/conversion/tril/` |
+| `triu` | `conversion` | `ops-math/conversion/triu/` |
+| `trunc` | `math` | `ops-math/math/trunc/` |
+| `trunc_v2` | `math` | `ops-math/experimental/math/trunc_v2/` |
+| `truncated_normal_v2` | `random` | `ops-math/random/truncated_normal_v2/` |
+| `unfold_grad` | `conversion` | `ops-math/conversion/unfold_grad/` |
+| `unique_consecutive` | `index` | `ops-nn/index/unique_consecutive/` |
+| `unpack` | `conversion` | `ops-math/conversion/unpack/` |
+| `unsorted_segment_min` | `index` | `ops-nn/index/unsorted_segment_min/` |
+| `upsample_bicubic2d` | `image` | `ops-cv/image/upsample_bicubic2d/` |
+| `upsample_bicubic2d_aa` | `image` | `ops-cv/image/upsample_bicubic2d_aa/` |
+| `upsample_bicubic2d_aa_grad` | `image` | `ops-cv/image/upsample_bicubic2d_aa_grad/` |
+| `upsample_bicubic2d_grad` | `image` | `ops-cv/image/upsample_bicubic2d_grad/` |
+| `upsample_bilinear2d` | `image` | `ops-cv/image/upsample_bilinear2d/` |
+| `upsample_bilinear2d_aa` | `image` | `ops-cv/image/upsample_bilinear2d_aa/` |
+| `upsample_bilinear2d_aa_backward` | `image` | `ops-cv/image/upsample_bilinear2d_aa_backward/` |
+| `upsample_bilinear2d_grad` | `image` | `ops-cv/image/upsample_bilinear2d_grad/` |
+| `upsample_linear1d` | `image` | `ops-cv/image/upsample_linear1d/` |
+| `upsample_nearest` | `image` | `ops-cv/image/upsample_nearest/` |
+| `upsample_nearest2d_grad` | `image` | `ops-cv/image/upsample_nearest2d_grad/` |
+| `upsample_nearest3d` | `image` | `ops-cv/image/upsample_nearest3d/` |
+| `upsample_nearest3d_grad` | `image` | `ops-cv/image/upsample_nearest3d_grad/` |
+| `upsample_nearest_exact2d_grad` | `image` | `ops-cv/image/upsample_nearest_exact2d_grad/` |
+| `upsample_nearest_exact3d` | `image` | `ops-cv/image/upsample_nearest_exact3d/` |
+| `upsample_nearest_exact3d_grad` | `image` | `ops-cv/image/upsample_nearest_exact3d_grad/` |
+| `upsample_trilinear3d_backward` | `image` | `ops-cv/image/upsample_trilinear3d_backward/` |
+| `view_copy` | `conversion` | `ops-math/conversion/view_copy/` |
+| `weight_quant_batch_matmul_experiment` | `matmul` | `ops-nn/experimental/matmul/weight_quant_batch_matmul_experiment/` |
+| `weight_quant_batch_matmul_v2` | `matmul` | `ops-nn/matmul/weight_quant_batch_matmul_v2/` |
+| `zeros_like` | `conversion` | `ops-math/conversion/zeros_like/` |