sgundabo

Results 8 comments of sgundabo

> @junliume Can you help me checking the failing reason of "Window Build" and "Jenkins - Fp32 Hip Debug gfx90a"? I think these are the issues causing the CI to...

**Perf Raw Data gfx90a** [BatchNormInferFusedInfo.zip](https://github.com/user-attachments/files/16431290/BatchNormInferFusedInfo.zip) **Perf FP32** ![BatchNormInferFusedExtraInfo_FP32](https://github.com/user-attachments/assets/d4605080-5786-4da6-b599-a799bb64db8f)

> > @CAHEK7 @amberhassaan > > Please find the profiling results attached below. > > @sgundabo Just for a reference - what king of gpu did you use to get...

### PerfData [DropoutPerf_large_5.csv](https://github.com/user-attachments/files/16386909/DropoutPerf_large_5.csv) [DropoutPerf_large_4.csv](https://github.com/user-attachments/files/16386919/DropoutPerf_large_4.csv) [DropoutPerf_large_3.csv](https://github.com/user-attachments/files/16386922/DropoutPerf_large_3.csv) [DropoutPerf_large_2.csv](https://github.com/user-attachments/files/16386923/DropoutPerf_large_2.csv) [DropoutPerf_large_1.csv](https://github.com/user-attachments/files/16386924/DropoutPerf_large_1.csv) HW tested: gfx90a ### FP32 Perf ![DropoutPerf_large_FP32](https://github.com/user-attachments/assets/758940c9-5085-4fa5-b4da-87ec4d74599d) ### FP16 Perf ![DropoutPerf_large_FP16](https://github.com/user-attachments/assets/4645574c-0783-4b50-9268-8427c8666786)

**Raw Perf data with detailed kernel information** [DropoutPerf_FP16.zip](https://github.com/user-attachments/files/16434930/DropoutPerf_FP16.zip) [DropoutPerf_FP32.zip](https://github.com/user-attachments/files/16434931/DropoutPerf_FP32.zip) HW tested: gfx90a **FP32 Perf** ![DropoutPerfRaw_FP32](https://github.com/user-attachments/assets/9ae85c96-97fa-4e5b-962f-1d90f8daa597) **FP16 Perf** ![DropoutPerfRaw_FP16](https://github.com/user-attachments/assets/ffe6984d-af7f-4155-9ac0-f076ae5be2a9)

**RawPerf Data** [DropoutPerf_smalltensors.zip](https://github.com/user-attachments/files/16463730/DropoutPerf_smalltensors.zip) **Perf FP16** ![DropoutPerf_smalltensors_FP16](https://github.com/user-attachments/assets/8f5f78b9-4b1a-47d4-a841-cd3cfdc61c36) **Perf FP32** ![DropoutPerf_smalltensors_FP32](https://github.com/user-attachments/assets/2061decf-c220-4cda-9230-1771e957de74)

**FP32 Perf Analysis** **Boxplot dropout0.50 mask0** ![Boxplot_min_exec_time_ratio](https://github.com/user-attachments/assets/61a15cc1-0ba4-459c-90c2-73f615119ec1) **Boxplot all dropouts and mask combinations** ![Boxplot_all_combinations](https://github.com/user-attachments/assets/a2b825bc-cf2f-49cb-ac0a-6463db7293bb)

**Raw Perf Data** [BNFwdInferRawPerfData.zip](https://github.com/user-attachments/files/16440356/BNFwdInferRawPerfData.zip) HW info: gfx90a **FP32 Perf** ![BNFwdInferRawPerfData_FP32](https://github.com/user-attachments/assets/47b23f2a-2fa7-4a09-bde9-fa45d792c0fb) **FP16 Perf** ![BNFwdInferRawPerfData_FP16](https://github.com/user-attachments/assets/262443a2-b162-4059-a967-b295e30d204f)