添加回退校验+迭代循环+[修改input_case 为从json读取]

[chopper0126]

变更说明

[添加回退校验+迭代循环]
[add batch_run_performance.sh]
[修改input_case 为从json读取]

变更类型

• 新增算子 / Skill
• 性能优化
• Bug 修复
• Benchmark（新增 case / 修改评测逻辑）
• Agent / 框架改动
• 文档 / 基础设施

影响范围

• Triton 侧
• AscendC 侧
• 共享（router / benchmark-scheduler）

性能数据（涉及算子生成/优化时必填）

Benchmark 评测（与 BASELINE.md 对比）

指标	BASELINE	本次评测
编译通过数	4	13
精度通过数	4	13
平均 Speedup

测试环境

• 设备型号：
• CANN 版本：
• PyTorch 版本：

冒烟测试（涉及算子生成/框架改动时必填）

• Triton 通路：✅ / ❌（失败原因：）
• AscendC 通路：✅ / ❌（失败原因：）

验证清单

• 双通路冒烟测试通过
• 通过率不退化（编译、精度均 >= BASELINE）
• 平均 Speedup 不退化（>= BASELINE × 0.95）
• 性能优化类：已跑全量评测、逐任务无退化、至少 1 个提升 >= 5%

退化说明（如有通过率下跌）

[chopper0126]

>>> 锁定设备 1 并启动算子: 10_SwigluQuant
>>> 锁定设备 2 并启动算子: 13_InterleaveRope
>>> 锁定设备 3 并启动算子: 14_AdaptiveInstanceNormalization2DBackward
>>> 锁定设备 4 并启动算子: 18_FusedAddRmsnorm
>>> 锁定设备 5 并启动算子: 1_RotaryMul
>>> 锁定设备 6 并启动算子: 21_GaussianTopkSparseActivation
>>> 锁定设备 7 并启动算子: 22_HybridAttentionMaskPreparation
Performance Report
========================================================================================
Operator  : /home/y00889327/AscendOpGenAgent/output_0410_l2_ziji/13_InterleaveRope
Task Dir  : /home/y00889327/AscendOpGenAgent/output_0410_l2_ziji/13_InterleaveRope
Device    : npu
Warmup    : 5
Repeat    : 10
Seed      : 0
----------------------------------------------------------------------------------------
Inputs
----------------------------------------------------------------------------------------
inputs[0]: list[3]
inputs[0][0]: Tensor(shape=(1, 1, 1, 64), dtype=torch.float16, device=npu:0)
inputs[0][1]: Tensor(shape=(1, 1, 1, 64), dtype=torch.float16, device=npu:0)
inputs[0][2]: Tensor(shape=(1, 1, 1, 64), dtype=torch.float16, device=npu:0)
inputs[1]: list[3]
inputs[1][0]: Tensor(shape=(1, 1, 64, 64), dtype=torch.float16, device=npu:0)
inputs[1][1]: Tensor(shape=(1, 1, 64, 64), dtype=torch.float16, device=npu:0)
inputs[1][2]: Tensor(shape=(1, 1, 64, 64), dtype=torch.float16, device=npu:0)
inputs[2]: list[3]
inputs[2][0]: Tensor(shape=(1, 1, 128, 64), dtype=torch.float16, device=npu:0)
inputs[2][1]: Tensor(shape=(1, 1, 128, 64), dtype=torch.float16, device=npu:0)
inputs[2][2]: Tensor(shape=(1, 1, 128, 64), dtype=torch.float16, device=npu:0)
inputs[3]: list[3]
inputs[3][0]: Tensor(shape=(1, 1, 256, 64), dtype=torch.float16, device=npu:0)
inputs[3][1]: Tensor(shape=(1, 1, 1, 64), dtype=torch.float16, device=npu:0)
inputs[3][2]: Tensor(shape=(1, 1, 1, 64), dtype=torch.float16, device=npu:0)
inputs[4]: list[3]
inputs[4][0]: Tensor(shape=(1, 1, 512, 64), dtype=torch.float16, device=npu:0)
inputs[4][1]: Tensor(shape=(1, 1, 512, 64), dtype=torch.float16, device=npu:0)
inputs[4][2]: Tensor(shape=(1, 1, 512, 64), dtype=torch.float16, device=npu:0)
inputs[5]: list[3]
inputs[5][0]: Tensor(shape=(1, 1, 1, 64), dtype=torch.bfloat16, device=npu:0)
inputs[5][1]: Tensor(shape=(1, 1, 1, 64), dtype=torch.bfloat16, device=npu:0)
inputs[5][2]: Tensor(shape=(1, 1, 1, 64), dtype=torch.bfloat16, device=npu:0)
inputs[6]: list[3]
inputs[6][0]: Tensor(shape=(1, 1, 64, 64), dtype=torch.bfloat16, device=npu:0)
inputs[6][1]: Tensor(shape=(1, 1, 1, 64), dtype=torch.bfloat16, device=npu:0)
inputs[6][2]: Tensor(shape=(1, 1, 1, 64), dtype=torch.bfloat16, device=npu:0)
inputs[7]: list[3]
inputs[7][0]: Tensor(shape=(1, 1, 128, 64), dtype=torch.bfloat16, device=npu:0)
inputs[7][1]: Tensor(shape=(1, 1, 128, 64), dtype=torch.bfloat16, device=npu:0)
inputs[7][2]: Tensor(shape=(1, 1, 128, 64), dtype=torch.bfloat16, device=npu:0)
inputs[8]: list[3]
inputs[8][0]: Tensor(shape=(1, 1, 256, 64), dtype=torch.bfloat16, device=npu:0)
inputs[8][1]: Tensor(shape=(1, 1, 256, 64), dtype=torch.bfloat16, device=npu:0)
inputs[8][2]: Tensor(shape=(1, 1, 256, 64), dtype=torch.bfloat16, device=npu:0)
inputs[9]: list[3]
inputs[9][0]: Tensor(shape=(1, 1, 512, 64), dtype=torch.bfloat16, device=npu:0)
inputs[9][1]: Tensor(shape=(1, 1, 1, 64), dtype=torch.bfloat16, device=npu:0)
inputs[9][2]: Tensor(shape=(1, 1, 1, 64), dtype=torch.bfloat16, device=npu:0)
inputs[10]: list[3]
inputs[10][0]: Tensor(shape=(2, 1, 1, 64), dtype=torch.float16, device=npu:0)
inputs[10][1]: Tensor(shape=(2, 1, 1, 64), dtype=torch.float16, device=npu:0)
inputs[10][2]: Tensor(shape=(2, 1, 1, 64), dtype=torch.float16, device=npu:0)
inputs[11]: list[3]
inputs[11][0]: Tensor(shape=(2, 1, 32, 64), dtype=torch.float16, device=npu:0)
inputs[11][1]: Tensor(shape=(2, 1, 32, 64), dtype=torch.float16, device=npu:0)
inputs[11][2]: Tensor(shape=(2, 1, 32, 64), dtype=torch.float16, device=npu:0)
inputs[12]: list[3]
inputs[12][0]: Tensor(shape=(2, 1, 64, 64), dtype=torch.float16, device=npu:0)
inputs[12][1]: Tensor(shape=(2, 1, 1, 64), dtype=torch.float16, device=npu:0)
inputs[12][2]: Tensor(shape=(2, 1, 1, 64), dtype=torch.float16, device=npu:0)
inputs[13]: list[3]
inputs[13][0]: Tensor(shape=(2, 1, 128, 64), dtype=torch.float16, device=npu:0)
inputs[13][1]: Tensor(shape=(2, 1, 128, 64), dtype=torch.float16, device=npu:0)
inputs[13][2]: Tensor(shape=(2, 1, 128, 64), dtype=torch.float16, device=npu:0)
inputs[14]: list[3]
inputs[14][0]: Tensor(shape=(2, 1, 256, 64), dtype=torch.float16, device=npu:0)
inputs[14][1]: Tensor(shape=(2, 1, 256, 64), dtype=torch.float16, device=npu:0)
inputs[14][2]: Tensor(shape=(2, 1, 256, 64), dtype=torch.float16, device=npu:0)
inputs[15]: list[3]
inputs[15][0]: Tensor(shape=(2, 1, 1, 64), dtype=torch.bfloat16, device=npu:0)
inputs[15][1]: Tensor(shape=(2, 1, 1, 64), dtype=torch.bfloat16, device=npu:0)
inputs[15][2]: Tensor(shape=(2, 1, 1, 64), dtype=torch.bfloat16, device=npu:0)
inputs[16]: list[3]
inputs[16][0]: Tensor(shape=(2, 1, 32, 64), dtype=torch.bfloat16, device=npu:0)
inputs[16][1]: Tensor(shape=(2, 1, 1, 64), dtype=torch.bfloat16, device=npu:0)
inputs[16][2]: Tensor(shape=(2, 1, 1, 64), dtype=torch.bfloat16, device=npu:0)
inputs[17]: list[3]
inputs[17][0]: Tensor(shape=(2, 1, 64, 64), dtype=torch.bfloat16, device=npu:0)
inputs[17][1]: Tensor(shape=(2, 1, 64, 64), dtype=torch.bfloat16, device=npu:0)
inputs[17][2]: Tensor(shape=(2, 1, 64, 64), dtype=torch.bfloat16, device=npu:0)
inputs[18]: list[3]
inputs[18][0]: Tensor(shape=(2, 1, 128, 64), dtype=torch.bfloat16, device=npu:0)
inputs[18][1]: Tensor(shape=(2, 1, 128, 64), dtype=torch.bfloat16, device=npu:0)
inputs[18][2]: Tensor(shape=(2, 1, 128, 64), dtype=torch.bfloat16, device=npu:0)
inputs[19]: list[3]
inputs[19][0]: Tensor(shape=(2, 1, 256, 64), dtype=torch.bfloat16, device=npu:0)
inputs[19][1]: Tensor(shape=(2, 1, 1, 64), dtype=torch.bfloat16, device=npu:0)
inputs[19][2]: Tensor(shape=(2, 1, 1, 64), dtype=torch.bfloat16, device=npu:0)
inputs[20]: list[3]
inputs[20][0]: Tensor(shape=(4, 1, 1, 64), dtype=torch.float16, device=npu:0)
inputs[20][1]: Tensor(shape=(4, 1, 1, 64), dtype=torch.float16, device=npu:0)
inputs[20][2]: Tensor(shape=(4, 1, 1, 64), dtype=torch.float16, device=npu:0)
inputs[21]: list[3]
inputs[21][0]: Tensor(shape=(4, 1, 16, 64), dtype=torch.float16, device=npu:0)
inputs[21][1]: Tensor(shape=(4, 1, 16, 64), dtype=torch.float16, device=npu:0)
inputs[21][2]: Tensor(shape=(4, 1, 16, 64), dtype=torch.float16, device=npu:0)
inputs[22]: list[3]
inputs[22][0]: Tensor(shape=(4, 1, 32, 64), dtype=torch.float16, device=npu:0)
inputs[22][1]: Tensor(shape=(4, 1, 1, 64), dtype=torch.float16, device=npu:0)
inputs[22][2]: Tensor(shape=(4, 1, 1, 64), dtype=torch.float16, device=npu:0)
inputs[23]: list[3]
inputs[23][0]: Tensor(shape=(4, 1, 64, 64), dtype=torch.float16, device=npu:0)
inputs[23][1]: Tensor(shape=(4, 1, 64, 64), dtype=torch.float16, device=npu:0)
inputs[23][2]: Tensor(shape=(4, 1, 64, 64), dtype=torch.float16, device=npu:0)
inputs[24]: list[3]
inputs[24][0]: Tensor(shape=(4, 1, 128, 64), dtype=torch.float16, device=npu:0)
inputs[24][1]: Tensor(shape=(4, 1, 128, 64), dtype=torch.float16, device=npu:0)
inputs[24][2]: Tensor(shape=(4, 1, 128, 64), dtype=torch.float16, device=npu:0)
inputs[25]: list[3]
inputs[25][0]: Tensor(shape=(4, 1, 1, 64), dtype=torch.bfloat16, device=npu:0)
inputs[25][1]: Tensor(shape=(4, 1, 1, 64), dtype=torch.bfloat16, device=npu:0)
inputs[25][2]: Tensor(shape=(4, 1, 1, 64), dtype=torch.bfloat16, device=npu:0)
inputs[26]: list[3]
inputs[26][0]: Tensor(shape=(4, 1, 16, 64), dtype=torch.bfloat16, device=npu:0)
inputs[26][1]: Tensor(shape=(4, 1, 1, 64), dtype=torch.bfloat16, device=npu:0)
inputs[26][2]: Tensor(shape=(4, 1, 1, 64), dtype=torch.bfloat16, device=npu:0)
inputs[27]: list[3]
inputs[27][0]: Tensor(shape=(4, 1, 32, 64), dtype=torch.bfloat16, device=npu:0)
inputs[27][1]: Tensor(shape=(4, 1, 32, 64), dtype=torch.bfloat16, device=npu:0)
inputs[27][2]: Tensor(shape=(4, 1, 32, 64), dtype=torch.bfloat16, device=npu:0)
inputs[28]: list[3]
inputs[28][0]: Tensor(shape=(4, 1, 64, 64), dtype=torch.bfloat16, device=npu:0)
inputs[28][1]: Tensor(shape=(4, 1, 64, 64), dtype=torch.bfloat16, device=npu:0)
inputs[28][2]: Tensor(shape=(4, 1, 64, 64), dtype=torch.bfloat16, device=npu:0)
inputs[29]: list[3]
inputs[29][0]: Tensor(shape=(4, 1, 128, 64), dtype=torch.bfloat16, device=npu:0)
inputs[29][1]: Tensor(shape=(4, 1, 1, 64), dtype=torch.bfloat16, device=npu:0)
inputs[29][2]: Tensor(shape=(4, 1, 1, 64), dtype=torch.bfloat16, device=npu:0)
inputs[30]: list[3]
inputs[30][0]: Tensor(shape=(8, 1, 1, 64), dtype=torch.float16, device=npu:0)
inputs[30][1]: Tensor(shape=(8, 1, 1, 64), dtype=torch.float16, device=npu:0)
inputs[30][2]: Tensor(shape=(8, 1, 1, 64), dtype=torch.float16, device=npu:0)
inputs[31]: list[3]
inputs[31][0]: Tensor(shape=(8, 1, 8, 64), dtype=torch.float16, device=npu:0)
inputs[31][1]: Tensor(shape=(8, 1, 8, 64), dtype=torch.float16, device=npu:0)
inputs[31][2]: Tensor(shape=(8, 1, 8, 64), dtype=torch.float16, device=npu:0)
inputs[32]: list[3]
inputs[32][0]: Tensor(shape=(8, 1, 16, 64), dtype=torch.float16, device=npu:0)
inputs[32][1]: Tensor(shape=(8, 1, 1, 64), dtype=torch.float16, device=npu:0)
inputs[32][2]: Tensor(shape=(8, 1, 1, 64), dtype=torch.float16, device=npu:0)
inputs[33]: list[3]
inputs[33][0]: Tensor(shape=(8, 1, 32, 64), dtype=torch.float16, device=npu:0)
inputs[33][1]: Tensor(shape=(8, 1, 32, 64), dtype=torch.float16, device=npu:0)
inputs[33][2]: Tensor(shape=(8, 1, 32, 64), dtype=torch.float16, device=npu:0)
inputs[34]: list[3]
inputs[34][0]: Tensor(shape=(8, 1, 64, 64), dtype=torch.float16, device=npu:0)
inputs[34][1]: Tensor(shape=(8, 1, 64, 64), dtype=torch.float16, device=npu:0)
inputs[34][2]: Tensor(shape=(8, 1, 64, 64), dtype=torch.float16, device=npu:0)
inputs[35]: list[3]
inputs[35][0]: Tensor(shape=(8, 1, 1, 64), dtype=torch.bfloat16, device=npu:0)
inputs[35][1]: Tensor(shape=(8, 1, 1, 64), dtype=torch.bfloat16, device=npu:0)
inputs[35][2]: Tensor(shape=(8, 1, 1, 64), dtype=torch.bfloat16, device=npu:0)
inputs[36]: list[3]
inputs[36][0]: Tensor(shape=(8, 1, 8, 64), dtype=torch.bfloat16, device=npu:0)
inputs[36][1]: Tensor(shape=(8, 1, 1, 64), dtype=torch.bfloat16, device=npu:0)
inputs[36][2]: Tensor(shape=(8, 1, 1, 64), dtype=torch.bfloat16, device=npu:0)
inputs[37]: list[3]
inputs[37][0]: Tensor(shape=(8, 1, 16, 64), dtype=torch.bfloat16, device=npu:0)
inputs[37][1]: Tensor(shape=(8, 1, 16, 64), dtype=torch.bfloat16, device=npu:0)
inputs[37][2]: Tensor(shape=(8, 1, 16, 64), dtype=torch.bfloat16, device=npu:0)
inputs[38]: list[3]
inputs[38][0]: Tensor(shape=(8, 1, 32, 64), dtype=torch.bfloat16, device=npu:0)
inputs[38][1]: Tensor(shape=(8, 1, 32, 64), dtype=torch.bfloat16, device=npu:0)
inputs[38][2]: Tensor(shape=(8, 1, 32, 64), dtype=torch.bfloat16, device=npu:0)
inputs[39]: list[3]
inputs[39][0]: Tensor(shape=(8, 1, 64, 64), dtype=torch.bfloat16, device=npu:0)
inputs[39][1]: Tensor(shape=(8, 1, 1, 64), dtype=torch.bfloat16, device=npu:0)
inputs[39][2]: Tensor(shape=(8, 1, 1, 64), dtype=torch.bfloat16, device=npu:0)
inputs[40]: list[3]
inputs[40][0]: Tensor(shape=(16, 1, 1, 64), dtype=torch.float16, device=npu:0)
inputs[40][1]: Tensor(shape=(16, 1, 1, 64), dtype=torch.float16, device=npu:0)
inputs[40][2]: Tensor(shape=(16, 1, 1, 64), dtype=torch.float16, device=npu:0)
inputs[41]: list[3]
inputs[41][0]: Tensor(shape=(16, 1, 4, 64), dtype=torch.float16, device=npu:0)
inputs[41][1]: Tensor(shape=(16, 1, 4, 64), dtype=torch.float16, device=npu:0)
inputs[41][2]: Tensor(shape=(16, 1, 4, 64), dtype=torch.float16, device=npu:0)
inputs[42]: list[3]
inputs[42][0]: Tensor(shape=(16, 1, 8, 64), dtype=torch.float16, device=npu:0)
inputs[42][1]: Tensor(shape=(16, 1, 1, 64), dtype=torch.float16, device=npu:0)
inputs[42][2]: Tensor(shape=(16, 1, 1, 64), dtype=torch.float16, device=npu:0)
inputs[43]: list[3]
inputs[43][0]: Tensor(shape=(16, 1, 16, 64), dtype=torch.float16, device=npu:0)
inputs[43][1]: Tensor(shape=(16, 1, 16, 64), dtype=torch.float16, device=npu:0)
inputs[43][2]: Tensor(shape=(16, 1, 16, 64), dtype=torch.float16, device=npu:0)
inputs[44]: list[3]
inputs[44][0]: Tensor(shape=(16, 1, 32, 64), dtype=torch.float16, device=npu:0)
inputs[44][1]: Tensor(shape=(16, 1, 32, 64), dtype=torch.float16, device=npu:0)
inputs[44][2]: Tensor(shape=(16, 1, 32, 64), dtype=torch.float16, device=npu:0)
inputs[45]: list[3]
inputs[45][0]: Tensor(shape=(16, 1, 1, 64), dtype=torch.bfloat16, device=npu:0)
inputs[45][1]: Tensor(shape=(16, 1, 1, 64), dtype=torch.bfloat16, device=npu:0)
inputs[45][2]: Tensor(shape=(16, 1, 1, 64), dtype=torch.bfloat16, device=npu:0)
inputs[46]: list[3]
inputs[46][0]: Tensor(shape=(16, 1, 4, 64), dtype=torch.bfloat16, device=npu:0)
inputs[46][1]: Tensor(shape=(16, 1, 1, 64), dtype=torch.bfloat16, device=npu:0)
inputs[46][2]: Tensor(shape=(16, 1, 1, 64), dtype=torch.bfloat16, device=npu:0)
inputs[47]: list[3]
inputs[47][0]: Tensor(shape=(16, 1, 8, 64), dtype=torch.bfloat16, device=npu:0)
inputs[47][1]: Tensor(shape=(16, 1, 8, 64), dtype=torch.bfloat16, device=npu:0)
inputs[47][2]: Tensor(shape=(16, 1, 8, 64), dtype=torch.bfloat16, device=npu:0)
inputs[48]: list[3]
inputs[48][0]: Tensor(shape=(16, 1, 16, 64), dtype=torch.bfloat16, device=npu:0)
inputs[48][1]: Tensor(shape=(16, 1, 16, 64), dtype=torch.bfloat16, device=npu:0)
inputs[48][2]: Tensor(shape=(16, 1, 16, 64), dtype=torch.bfloat16, device=npu:0)
inputs[49]: list[3]
inputs[49][0]: Tensor(shape=(16, 1, 32, 64), dtype=torch.bfloat16, device=npu:0)
inputs[49][1]: Tensor(shape=(16, 1, 1, 64), dtype=torch.bfloat16, device=npu:0)
inputs[49][2]: Tensor(shape=(16, 1, 1, 64), dtype=torch.bfloat16, device=npu:0)
----------------------------------------------------------------------------------------
Impl         Status       Mean(ms)       Median          Min          Max          Std
----------------------------------------------------------------------------------------
reference    OK              0.080        0.044        0.029        5.230        0.345
tilelang     OK              0.269        0.253        0.225        1.091        0.066
ascendc      OK              0.187        0.174        0.153        0.912        0.066
----------------------------------------------------------------------------------------
reference -> /home/y00889327/AscendOpGenAgent/output_0410_l2_ziji/13_InterleaveRope/model.py
case[0] mean=0.062 ms, median=0.055 ms, samples(ms): [0.109, 0.055, 0.062, 0.055, 0.052, 0.057, 0.084, 0.055, 0.046, 0.045]
case[1] mean=0.044 ms, median=0.045 ms, samples(ms): [0.046, 0.044, 0.047, 0.044, 0.037, 0.045, 0.039, 0.045, 0.050, 0.043]
case[2] mean=0.049 ms, median=0.048 ms, samples(ms): [0.050, 0.053, 0.044, 0.068, 0.048, 0.049, 0.039, 0.047, 0.045, 0.047]
case[3] mean=0.340 ms, median=0.049 ms, samples(ms): [0.044, 2.832, 0.162, 0.056, 0.043, 0.073, 0.050, 0.044, 0.045, 0.047]
case[4] mean=0.044 ms, median=0.044 ms, samples(ms): [0.042, 0.048, 0.045, 0.044, 0.041, 0.043, 0.045, 0.041, 0.046, 0.040]
case[5] mean=0.039 ms, median=0.039 ms, samples(ms): [0.042, 0.039, 0.038, 0.040, 0.039, 0.037, 0.037, 0.036, 0.039, 0.043]
case[6] mean=0.042 ms, median=0.041 ms, samples(ms): [0.041, 0.041, 0.045, 0.042, 0.046, 0.046, 0.040, 0.044, 0.040, 0.038]
case[7] mean=0.043 ms, median=0.042 ms, samples(ms): [0.042, 0.037, 0.043, 0.061, 0.040, 0.045, 0.040, 0.043, 0.042, 0.039]
case[8] mean=0.043 ms, median=0.043 ms, samples(ms): [0.044, 0.040, 0.046, 0.047, 0.040, 0.044, 0.040, 0.047, 0.040, 0.042]
case[9] mean=0.044 ms, median=0.043 ms, samples(ms): [0.048, 0.045, 0.044, 0.045, 0.042, 0.042, 0.043, 0.049, 0.041, 0.041]
case[10] mean=0.039 ms, median=0.038 ms, samples(ms): [0.044, 0.038, 0.037, 0.035, 0.037, 0.035, 0.043, 0.042, 0.039, 0.040]
case[11] mean=0.042 ms, median=0.042 ms, samples(ms): [0.044, 0.037, 0.047, 0.040, 0.039, 0.045, 0.043, 0.043, 0.042, 0.037]
case[12] mean=0.045 ms, median=0.043 ms, samples(ms): [0.045, 0.042, 0.040, 0.058, 0.041, 0.045, 0.043, 0.048, 0.042, 0.043]
case[13] mean=0.042 ms, median=0.042 ms, samples(ms): [0.046, 0.045, 0.039, 0.046, 0.041, 0.043, 0.048, 0.039, 0.038, 0.039]
case[14] mean=0.043 ms, median=0.042 ms, samples(ms): [0.040, 0.044, 0.042, 0.040, 0.050, 0.040, 0.041, 0.044, 0.041, 0.045]
case[15] mean=0.041 ms, median=0.041 ms, samples(ms): [0.045, 0.041, 0.041, 0.044, 0.042, 0.042, 0.039, 0.037, 0.038, 0.043]
case[16] mean=0.045 ms, median=0.044 ms, samples(ms): [0.039, 0.041, 0.040, 0.046, 0.043, 0.049, 0.048, 0.051, 0.045, 0.041]
case[17] mean=0.044 ms, median=0.042 ms, samples(ms): [0.056, 0.039, 0.048, 0.042, 0.042, 0.039, 0.040, 0.042, 0.044, 0.043]
case[18] mean=0.044 ms, median=0.044 ms, samples(ms): [0.047, 0.050, 0.041, 0.044, 0.045, 0.040, 0.042, 0.046, 0.046, 0.042]
case[19] mean=0.047 ms, median=0.045 ms, samples(ms): [0.056, 0.045, 0.042, 0.044, 0.040, 0.047, 0.056, 0.045, 0.052, 0.045]
case[20] mean=0.045 ms, median=0.042 ms, samples(ms): [0.046, 0.038, 0.056, 0.044, 0.030, 0.039, 0.043, 0.040, 0.072, 0.039]
case[21] mean=0.041 ms, median=0.039 ms, samples(ms): [0.047, 0.037, 0.044, 0.036, 0.037, 0.036, 0.037, 0.043, 0.051, 0.041]
case[22] mean=0.050 ms, median=0.046 ms, samples(ms): [0.056, 0.049, 0.080, 0.041, 0.046, 0.047, 0.042, 0.045, 0.055, 0.042]
case[23] mean=0.043 ms, median=0.041 ms, samples(ms): [0.041, 0.042, 0.039, 0.039, 0.032, 0.040, 0.045, 0.049, 0.050, 0.049]
case[24] mean=0.055 ms, median=0.049 ms, samples(ms): [0.049, 0.057, 0.047, 0.094, 0.052, 0.048, 0.044, 0.046, 0.069, 0.044]
case[25] mean=0.050 ms, median=0.047 ms, samples(ms): [0.072, 0.048, 0.047, 0.044, 0.043, 0.056, 0.047, 0.044, 0.059, 0.043]
case[26] mean=0.050 ms, median=0.044 ms, samples(ms): [0.058, 0.044, 0.059, 0.043, 0.042, 0.051, 0.041, 0.041, 0.076, 0.043]
case[27] mean=0.049 ms, median=0.049 ms, samples(ms): [0.056, 0.049, 0.050, 0.047, 0.045, 0.049, 0.043, 0.054, 0.059, 0.041]
case[28] mean=0.049 ms, median=0.046 ms, samples(ms): [0.070, 0.055, 0.047, 0.043, 0.042, 0.054, 0.044, 0.044, 0.049, 0.044]
case[29] mean=0.052 ms, median=0.051 ms, samples(ms): [0.051, 0.055, 0.046, 0.050, 0.052, 0.053, 0.050, 0.052, 0.066, 0.046]
case[30] mean=0.074 ms, median=0.057 ms, samples(ms): [0.049, 0.051, 0.045, 0.120, 0.176, 0.062, 0.052, 0.061, 0.073, 0.050]
case[31] mean=0.051 ms, median=0.050 ms, samples(ms): [0.063, 0.062, 0.057, 0.052, 0.044, 0.047, 0.043, 0.045, 0.049, 0.050]
case[32] mean=0.051 ms, median=0.048 ms, samples(ms): [0.058, 0.068, 0.043, 0.045, 0.057, 0.047, 0.052, 0.043, 0.050, 0.046]
case[33] mean=0.052 ms, median=0.046 ms, samples(ms): [0.044, 0.055, 0.046, 0.044, 0.055, 0.044, 0.058, 0.044, 0.081, 0.047]
case[34] mean=0.053 ms, median=0.051 ms, samples(ms): [0.055, 0.059, 0.056, 0.068, 0.047, 0.046, 0.047, 0.052, 0.050, 0.047]
case[35] mean=0.052 ms, median=0.048 ms, samples(ms): [0.047, 0.054, 0.043, 0.079, 0.043, 0.043, 0.071, 0.042, 0.049, 0.050]
case[36] mean=1.239 ms, median=0.098 ms, samples(ms): [0.056, 5.230, 0.140, 0.046, 0.040, 2.910, 0.270, 0.044, 0.034, 3.622]
case[37] mean=0.045 ms, median=0.043 ms, samples(ms): [0.045, 0.063, 0.043, 0.044, 0.040, 0.050, 0.040, 0.041, 0.040, 0.047]
case[38] mean=0.046 ms, median=0.041 ms, samples(ms): [0.038, 0.049, 0.068, 0.040, 0.040, 0.040, 0.042, 0.038, 0.051, 0.050]
case[39] mean=0.050 ms, median=0.050 ms, samples(ms): [0.048, 0.053, 0.051, 0.050, 0.050, 0.044, 0.052, 0.045, 0.053, 0.049]
case[40] mean=0.038 ms, median=0.037 ms, samples(ms): [0.038, 0.040, 0.042, 0.037, 0.036, 0.036, 0.037, 0.033, 0.032, 0.045]
case[41] mean=0.038 ms, median=0.037 ms, samples(ms): [0.040, 0.040, 0.036, 0.035, 0.034, 0.035, 0.037, 0.037, 0.036, 0.046]
case[42] mean=0.040 ms, median=0.038 ms, samples(ms): [0.041, 0.046, 0.038, 0.038, 0.035, 0.038, 0.038, 0.039, 0.037, 0.047]
case[43] mean=0.252 ms, median=0.040 ms, samples(ms): [0.040, 0.051, 0.040, 0.040, 0.039, 0.038, 0.039, 0.048, 2.124, 0.058]
case[44] mean=0.044 ms, median=0.041 ms, samples(ms): [0.044, 0.050, 0.041, 0.042, 0.040, 0.041, 0.040, 0.041, 0.040, 0.055]
case[45] mean=0.045 ms, median=0.044 ms, samples(ms): [0.043, 0.041, 0.054, 0.041, 0.042, 0.045, 0.046, 0.040, 0.046, 0.050]
case[46] mean=0.043 ms, median=0.041 ms, samples(ms): [0.044, 0.043, 0.050, 0.041, 0.040, 0.041, 0.041, 0.040, 0.040, 0.047]
case[47] mean=0.047 ms, median=0.041 ms, samples(ms): [0.043, 0.041, 0.049, 0.041, 0.040, 0.040, 0.041, 0.039, 0.047, 0.086]
case[48] mean=0.035 ms, median=0.035 ms, samples(ms): [0.038, 0.037, 0.040, 0.029, 0.037, 0.034, 0.032, 0.032, 0.034, 0.037]
case[49] mean=0.048 ms, median=0.047 ms, samples(ms): [0.046, 0.046, 0.053, 0.044, 0.048, 0.047, 0.051, 0.052, 0.046, 0.050]
----------------------------------------------------------------------------------------
tilelang -> /home/y00889327/AscendOpGenAgent/output_0410_l2_ziji/13_InterleaveRope/model_new_tilelang.py
case[0] mean=0.499 ms, median=0.360 ms, samples(ms): [0.358, 0.331, 0.379, 1.091, 0.404, 0.362, 0.347, 0.334, 0.350, 1.032]
case[1] mean=0.382 ms, median=0.396 ms, samples(ms): [0.351, 0.344, 0.336, 0.349, 0.418, 0.397, 0.428, 0.399, 0.395, 0.401]
case[2] mean=0.372 ms, median=0.378 ms, samples(ms): [0.321, 0.323, 0.319, 0.362, 0.365, 0.402, 0.392, 0.393, 0.446, 0.393]
case[3] mean=0.433 ms, median=0.434 ms, samples(ms): [0.457, 0.505, 0.389, 0.370, 0.443, 0.425, 0.423, 0.448, 0.420, 0.449]
case[4] mean=0.259 ms, median=0.252 ms, samples(ms): [0.253, 0.246, 0.241, 0.256, 0.257, 0.253, 0.248, 0.252, 0.248, 0.340]
case[5] mean=0.232 ms, median=0.231 ms, samples(ms): [0.234, 0.235, 0.225, 0.233, 0.230, 0.227, 0.241, 0.229, 0.236, 0.226]
case[6] mean=0.278 ms, median=0.271 ms, samples(ms): [0.278, 0.266, 0.272, 0.260, 0.259, 0.334, 0.277, 0.270, 0.268, 0.297]
case[7] mean=0.277 ms, median=0.264 ms, samples(ms): [0.267, 0.280, 0.283, 0.351, 0.321, 0.257, 0.251, 0.262, 0.252, 0.241]
case[8] mean=0.251 ms, median=0.253 ms, samples(ms): [0.254, 0.241, 0.262, 0.246, 0.257, 0.253, 0.243, 0.262, 0.252, 0.242]
case[9] mean=0.265 ms, median=0.265 ms, samples(ms): [0.266, 0.262, 0.266, 0.256, 0.278, 0.270, 0.257, 0.274, 0.264, 0.261]
case[10] mean=0.252 ms, median=0.251 ms, samples(ms): [0.259, 0.257, 0.254, 0.248, 0.254, 0.249, 0.247, 0.252, 0.247, 0.251]
case[11] mean=0.249 ms, median=0.249 ms, samples(ms): [0.249, 0.251, 0.252, 0.254, 0.245, 0.249, 0.245, 0.242, 0.255, 0.244]
case[12] mean=0.299 ms, median=0.264 ms, samples(ms): [0.270, 0.261, 0.264, 0.263, 0.257, 0.608, 0.271, 0.256, 0.265, 0.272]
case[13] mean=0.252 ms, median=0.249 ms, samples(ms): [0.248, 0.252, 0.272, 0.246, 0.247, 0.248, 0.252, 0.258, 0.251, 0.245]
case[14] mean=0.251 ms, median=0.250 ms, samples(ms): [0.252, 0.250, 0.252, 0.247, 0.251, 0.247, 0.246, 0.262, 0.249, 0.256]
case[15] mean=0.251 ms, median=0.250 ms, samples(ms): [0.256, 0.248, 0.247, 0.251, 0.258, 0.245, 0.249, 0.255, 0.248, 0.251]
case[16] mean=0.267 ms, median=0.267 ms, samples(ms): [0.261, 0.272, 0.258, 0.278, 0.267, 0.256, 0.270, 0.280, 0.257, 0.268]
case[17] mean=0.264 ms, median=0.264 ms, samples(ms): [0.262, 0.248, 0.267, 0.254, 0.248, 0.274, 0.274, 0.296, 0.268, 0.249]
case[18] mean=0.251 ms, median=0.248 ms, samples(ms): [0.255, 0.250, 0.266, 0.256, 0.257, 0.246, 0.247, 0.242, 0.246, 0.246]
case[19] mean=0.264 ms, median=0.262 ms, samples(ms): [0.267, 0.279, 0.261, 0.271, 0.268, 0.255, 0.262, 0.263, 0.257, 0.260]
case[20] mean=0.252 ms, median=0.251 ms, samples(ms): [0.254, 0.258, 0.248, 0.256, 0.271, 0.241, 0.259, 0.241, 0.247, 0.243]
case[21] mean=0.250 ms, median=0.250 ms, samples(ms): [0.251, 0.249, 0.255, 0.254, 0.248, 0.252, 0.254, 0.246, 0.244, 0.245]
case[22] mean=0.263 ms, median=0.262 ms, samples(ms): [0.262, 0.261, 0.269, 0.263, 0.261, 0.264, 0.262, 0.261, 0.265, 0.258]
case[23] mean=0.254 ms, median=0.252 ms, samples(ms): [0.268, 0.259, 0.253, 0.249, 0.258, 0.252, 0.247, 0.251, 0.259, 0.248]
case[24] mean=0.248 ms, median=0.248 ms, samples(ms): [0.248, 0.254, 0.249, 0.253, 0.246, 0.242, 0.255, 0.243, 0.251, 0.243]
case[25] mean=0.245 ms, median=0.244 ms, samples(ms): [0.251, 0.244, 0.248, 0.244, 0.249, 0.244, 0.245, 0.250, 0.241, 0.238]
case[26] mean=0.258 ms, median=0.255 ms, samples(ms): [0.270, 0.262, 0.254, 0.268, 0.255, 0.256, 0.253, 0.253, 0.255, 0.255]
case[27] mean=0.241 ms, median=0.241 ms, samples(ms): [0.243, 0.244, 0.240, 0.239, 0.242, 0.244, 0.238, 0.243, 0.239, 0.237]
case[28] mean=0.251 ms, median=0.244 ms, samples(ms): [0.244, 0.243, 0.242, 0.243, 0.247, 0.243, 0.308, 0.253, 0.246, 0.244]
case[29] mean=0.256 ms, median=0.254 ms, samples(ms): [0.263, 0.257, 0.259, 0.261, 0.254, 0.252, 0.249, 0.254, 0.254, 0.251]
case[30] mean=0.243 ms, median=0.243 ms, samples(ms): [0.250, 0.245, 0.242, 0.247, 0.245, 0.243, 0.240, 0.240, 0.239, 0.241]
case[31] mean=0.244 ms, median=0.243 ms, samples(ms): [0.255, 0.244, 0.239, 0.243, 0.247, 0.243, 0.240, 0.244, 0.242, 0.244]
case[32] mean=0.259 ms, median=0.259 ms, samples(ms): [0.258, 0.262, 0.265, 0.253, 0.263, 0.256, 0.260, 0.261, 0.256, 0.252]
case[33] mean=0.241 ms, median=0.243 ms, samples(ms): [0.247, 0.243, 0.243, 0.235, 0.252, 0.243, 0.238, 0.234, 0.235, 0.243]
case[34] mean=0.241 ms, median=0.240 ms, samples(ms): [0.242, 0.240, 0.246, 0.243, 0.244, 0.239, 0.238, 0.241, 0.240, 0.236]
case[35] mean=0.246 ms, median=0.245 ms, samples(ms): [0.257, 0.250, 0.243, 0.255, 0.241, 0.239, 0.248, 0.242, 0.246, 0.241]
case[36] mean=0.258 ms, median=0.255 ms, samples(ms): [0.262, 0.254, 0.254, 0.255, 0.265, 0.265, 0.253, 0.253, 0.254, 0.266]
case[37] mean=0.253 ms, median=0.246 ms, samples(ms): [0.302, 0.265, 0.246, 0.246, 0.248, 0.246, 0.244, 0.246, 0.247, 0.244]
case[38] mean=0.245 ms, median=0.244 ms, samples(ms): [0.250, 0.240, 0.243, 0.249, 0.247, 0.246, 0.242, 0.247, 0.243, 0.242]
case[39] mean=0.259 ms, median=0.260 ms, samples(ms): [0.266, 0.252, 0.262, 0.255, 0.256, 0.256, 0.261, 0.264, 0.259, 0.262]
case[40] mean=0.248 ms, median=0.247 ms, samples(ms): [0.248, 0.246, 0.254, 0.248, 0.239, 0.252, 0.247, 0.243, 0.254, 0.247]
case[41] mean=0.246 ms, median=0.244 ms, samples(ms): [0.251, 0.250, 0.259, 0.248, 0.238, 0.240, 0.243, 0.243, 0.239, 0.245]
case[42] mean=0.259 ms, median=0.257 ms, samples(ms): [0.258, 0.257, 0.270, 0.255, 0.256, 0.255, 0.264, 0.256, 0.252, 0.272]
case[43] mean=0.248 ms, median=0.246 ms, samples(ms): [0.246, 0.252, 0.246, 0.245, 0.248, 0.242, 0.247, 0.243, 0.245, 0.264]
case[44] mean=0.251 ms, median=0.247 ms, samples(ms): [0.245, 0.259, 0.246, 0.245, 0.248, 0.250, 0.243, 0.242, 0.265, 0.265]
case[45] mean=0.246 ms, median=0.246 ms, samples(ms): [0.245, 0.247, 0.248, 0.247, 0.248, 0.242, 0.242, 0.244, 0.245, 0.254]
case[46] mean=0.257 ms, median=0.255 ms, samples(ms): [0.258, 0.255, 0.256, 0.254, 0.254, 0.261, 0.253, 0.259, 0.271, 0.253]
case[47] mean=0.315 ms, median=0.314 ms, samples(ms): [0.321, 0.313, 0.313, 0.313, 0.320, 0.311, 0.313, 0.322, 0.314, 0.314]
case[48] mean=0.261 ms, median=0.249 ms, samples(ms): [0.319, 0.298, 0.250, 0.257, 0.248, 0.246, 0.245, 0.247, 0.243, 0.257]
case[49] mean=0.262 ms, median=0.260 ms, samples(ms): [0.260, 0.260, 0.269, 0.262, 0.259, 0.257, 0.257, 0.259, 0.274, 0.261]
----------------------------------------------------------------------------------------
ascendc -> /home/y00889327/AscendOpGenAgent/output_0410_l2_ziji/13_InterleaveRope/model_new_ascendc.py
case[0] mean=0.166 ms, median=0.162 ms, samples(ms): [0.174, 0.180, 0.176, 0.166, 0.163, 0.159, 0.162, 0.158, 0.159, 0.162]
case[1] mean=0.170 ms, median=0.169 ms, samples(ms): [0.177, 0.169, 0.176, 0.174, 0.173, 0.165, 0.169, 0.166, 0.167, 0.167]
case[2] mean=0.176 ms, median=0.174 ms, samples(ms): [0.175, 0.173, 0.185, 0.172, 0.190, 0.177, 0.172, 0.175, 0.167, 0.171]
case[3] mean=0.183 ms, median=0.182 ms, samples(ms): [0.184, 0.190, 0.182, 0.183, 0.179, 0.178, 0.177, 0.188, 0.182, 0.188]
case[4] mean=0.173 ms, median=0.170 ms, samples(ms): [0.178, 0.169, 0.170, 0.174, 0.171, 0.173, 0.190, 0.170, 0.168, 0.169]
case[5] mean=0.164 ms, median=0.159 ms, samples(ms): [0.170, 0.174, 0.188, 0.167, 0.159, 0.156, 0.158, 0.157, 0.153, 0.157]
case[6] mean=0.184 ms, median=0.182 ms, samples(ms): [0.199, 0.186, 0.183, 0.187, 0.181, 0.183, 0.182, 0.182, 0.181, 0.178]
case[7] mean=0.173 ms, median=0.170 ms, samples(ms): [0.190, 0.175, 0.169, 0.168, 0.168, 0.183, 0.171, 0.170, 0.172, 0.165]
case[8] mean=0.173 ms, median=0.171 ms, samples(ms): [0.179, 0.173, 0.177, 0.171, 0.170, 0.171, 0.170, 0.178, 0.171, 0.168]
case[9] mean=0.186 ms, median=0.186 ms, samples(ms): [0.190, 0.187, 0.187, 0.185, 0.182, 0.183, 0.185, 0.188, 0.186, 0.185]
case[10] mean=0.186 ms, median=0.176 ms, samples(ms): [0.185, 0.176, 0.176, 0.171, 0.172, 0.171, 0.172, 0.198, 0.220, 0.218]
case[11] mean=0.215 ms, median=0.212 ms, samples(ms): [0.217, 0.239, 0.216, 0.211, 0.215, 0.212, 0.211, 0.209, 0.211, 0.211]
case[12] mean=0.199 ms, median=0.192 ms, samples(ms): [0.226, 0.227, 0.220, 0.190, 0.198, 0.187, 0.193, 0.188, 0.181, 0.179]
case[13] mean=0.170 ms, median=0.170 ms, samples(ms): [0.170, 0.172, 0.169, 0.169, 0.166, 0.17

[Just-it]

1、验证移到skill里面
2、缺少对torch_npu的约束
3、这两个脚本是可以归一的
参照一下这个PR：#81

[Just-it]

This is an automatic reply, confirming that your email was received.I will reply as soon as possible.Thank you! —Chen Shushu