Skip to content

add 730 performance with 2 diff configs #72

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
wenliang wants to merge 1 commit into from
Closed

add 730 performance with 2 diff configs #72

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
113 changes: 113 additions & 0 deletions docs/toolchain/appendix/performance.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -91,3 +91,116 @@
| Unet | 384x384 | 3.903 | 256.1 ms | No |
| Lite-HRNet | 256x192 | 200.8 | 4.980 ms | No |
| CenterNet | 512x512 | 6.657 | 150.2 ms | No |

*More Performance simulation result on NPU KDP730 (default 8bit settings):*
| Model | Size | FPS (npu only) | Time(npu only) | Has CPU node(s)? |
Copy link
Member

@MrWhoami MrWhoami Jan 8, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Please leave a blank line before the table.

| ------------------------------------------------------ | ---------------------------------- | -------------- | -------------- | ---------------- |
| baai_bert_en.opt | 384, 384, 384 | 16.9512 | 58.9928 | Yes |
| baai_small-en-v1.5.opt | 384, 384, 384 | 17.0124 | 58.7807 | Yes |
| bert_opset11_with_weight_041823_kneron | 384, 384 | 18.2599 | 54.7649 | Yes |
| centernet_res18_pytorch_3c_512w_512h | 3x512x512 | 18.6832 | 53.524 | No |
| conformer-tiny-p16_8xb128_in1k_kneron_optimized | 3x224x224 | 52.9233 | 18.8953 | No |
| decoder_modified_2 | 1024x512,1024x512,1024x1024,1024 | 5.19971 | 192.318 | No |
| deeplabv3plus_mobv2_1x0_keras_3c_512w_512h | 3x512x512 | 40.1408 | 24.9123 | No |
| deit-tiny_pt-4xb256_in1k_kneron_optimized | 3x224x224 | 181.023 | 5.52417 | No |
| efficientnet_lite4_rw_pytorch_3c_300w_300h | 3x300x300 | 180.42 | 5.54263 | No |
| Facedetection_SSD_without_BN | 3x200x200 | 928.681 | 1.0768 | No |
| Facelandmark_kneron_shufflenet_68 | 3x240x240 | 1850.3 | 0.540452 | No |
| inception_resnetv2_pytorch_3c_299w_299h | 3x299x299 | 36.7164 | 27.2358 | No |
| inceptionv3_pytorch_3c_299w_299h | 3x299x299 | 92.9287 | 10.7609 | No |
| inceptionv4_keras_3c_299w_299h | 3x299x299 | 46.2884 | 21.6037 | No |
| mlp-mixer-base-p16_64xb64_in1k_kneron_optimized | 3x224x224 | 27.7357 | 36.0546 | No |
| mobilenetv2_100_rw_pytorch_3c_224w_224h | 3x224x224 | 894.72 | 1.11767 | No |
| mobilenetv3_large_minimalistic_x1_0_keras_3c_224w_224h | 3x224x224 | 883.495 | 1.13187 | No |
| model_704_nocut | 3x256x640 | 158.468 | 6.31042 | No |
| model_706_nocut | 48x64x160 | 86.0932 | 11.6153 | No |
| model_707_nocut | 256x640,3x256x640 | 59.2393 | 16.8807 | No |
| model_9096_nocut | 80x3000 | 0.689647 | 1450.02 | No |
| model_9096_nocut_layer0 | 1500x768 | 14.3383 | 69.7431 | No |
| model_9097_nocut | 1500x768,4x768,4x768,4x4 | 10.97 | 91.158 | No |
| model_9232_nocut | 12x223x64,12x223x64,12x1500x64,... | 18.664 | 53.5792 | No |
| openclip_vit_base_patch_16_SigLIP_vision_encoder.opt | 3x224x224 | 21.6036 | 46.2887 | No |
| resnet101_pytorch_3c_224w_224h | 3x224x224 | 59.7392 | 16.7394 | No |
| resnet50_80_224x224-pytorch | 3x224x224 | 94.1109 | 10.6258 | No |
| retinanet_pvt-t_fpn_1x_coco_kneron_optimized | 3x224x224 | 41.2293 | 24.2546 | No |
| retinanet_pvtv2-b0_fpn_1x_coco_kneron_optimized | 3x224x224 | 54.5772 | 18.3227 | No |
| retinanet_swin-t-p4-w7_fpn_1x_coco_kneron_optimized | 3x224x224 | 24.7706 | 40.3704 | No |
| shufflenetv2_x1_0_pytorch_3c_224w_224h | 3x224x224 | 869.66 | 1.14987 | No |
| swin-cut | 3136x96 | 553.717 | 1.80598 | No |
| swin-tiny_16xb64_in1k_kneron_optimized | 3x224x224 | 38.646 | 25.8759 | No |
| swin_tiny_classify_head | 3x224x224 | 28.3665 | 35.2529 | No |
| swinv2_tiny_classify_head | 3x256x256 | 15.9521 | 62.6876 | Yes |
| t2t-vit-t-14_8xb64_in1k_kneron_optimized | 3x224x224 | 26.0946 | 38.3221 | Yes |
| tiny_yolov3_keras_3c_608w_608h | 3x608x608 | 130.218 | 7.67941 | No |
| tnt-s-p16_16xb64_in1k_kneron_optimized | 3x224x224 | 23.1818 | 43.1373 | Yes |
| vit-base-p16_pt-64xb64_in1k-224_kneron_optimized | 3x224x224 | 21.6917 | 46.1007 | No |
| vit_base_classify_head_kneron | 3x224x224 | 19.5094 | 51.2575 | No |
| wenet_u2_decoder | 100x512,99x512,99x512 | 95.2782 | 10.4956 | No |
| wenet_u2_encoder | 100x80 | 42.3604 | 23.607 | No |
| yolo11m_640x640.opt | 3x640x640 | 11.5716 | 86.4183 | No |
| yolo11n_640x640.opt | 3x640x640 | 64.921 | 15.4033 | No |
| yolov3_keras_3c_416w_416h | 3x416x416 | 21.498 | 46.5159 | No |
| yolov4_keras_3c_416w_416h | 3x416x416 | 23.9978 | 41.6705 | No |
| yolov5s_pytorch_12c_336w_336h | 12x336x336 | 61.9733 | 16.136 | No |
| yolov7_640x640.opt | 3x640x640 | 10.4601 | 95.6015 | No |
| yolov8m_640x640.opt | 3x640x640 | 14.9085 | 67.0759 | No |
| yolov8n_640x640.opt | 3x640x640 | 69.8148 | 14.3236 | No |
| yolov9m_640x640.opt | 3x640x640 | 12.801 | 78.1191 | No |
| yolov9t_640x640.opt | 3x640x640 | 64.4753 | 15.5098 | No |

*More Performance simulation result on NPU KDP730 (input 8 bit，output 16bit， data mixlight, weight 8bit):*

| Model | Size | FPS (npu only) | Time(npu only) | Has CPU node(s)? |
| ------------------------------------------------------ | ---------------------------------- | -------------- | -------------- | ---------------- |
| baai_bert_en.opt | 384, 384, 384 | 12.8872 | 77.5966 | Yes |
| baai_small-en-v1.5.opt | 384, 384, 384 | 15.8069 | 63.2637 | Yes |
| bert_opset11_with_weight_041823_kneron | 384, 384 | 13.158 | 75.9992 | Yes |
| centernet_res18_pytorch_3c_512w_512h | 3x512x512 | 17.0192 | 58.7572 | No |
| conformer-tiny-p16_8xb128_in1k_kneron_optimized | 3x224x224 | 35.8217 | 27.9161 | No |
| decoder_modified_2 | 1024x512,1024x512,1024x1024,1024 | 4.85533 | 205.959 | No |
| deeplabv3plus_mobv2_1x0_keras_3c_512w_512h | 3x512x512 | 36.7646 | 27.2001 | No |
| deit-tiny_pt-4xb256_in1k_kneron_optimized | 3x224x224 | 156.979 | 6.37026 | No |
| efficientnet_lite4_rw_pytorch_3c_300w_300h | 3x300x300 | 162.867 | 6.13998 | No |
| Facedetection_SSD_without_BN | 3x200x200 | 849.861 | 1.17666 | No |
| Facelandmark_kneron_shufflenet_68 | 3x240x240 | 1536.22 | 0.650948 | No |
| inception_resnetv2_pytorch_3c_299w_299h | 3x299x299 | 29.2921 | 34.1389 | No |
| inceptionv3_pytorch_3c_299w_299h | 3x299x299 | 76.0408 | 13.1508 | No |
| inceptionv4_keras_3c_299w_299h | 3x299x299 | 36.4444 | 27.439 | No |
| mlp-mixer-base-p16_64xb64_in1k_kneron_optimized | 3x224x224 | 18.9444 | 52.7859 | No |
| mobilenetv2_100_rw_pytorch_3c_224w_224h | 3x224x224 | 663.951 | 1.50613 | No |
| mobilenetv3_large_minimalistic_x1_0_keras_3c_224w_224h | 3x224x224 | 689.478 | 1.45037 | No |
| model_704_nocut | 3x256x640 | 76.2507 | 13.1146 | No |
| model_706_nocut | 48x64x160 | 48.9827 | 20.4154 | No |
| model_707_nocut | 256x640,3x256x640 | 27.0065 | 37.0282 | No |
| model_9096_nocut | 80x3000 | 0.595831 | 1678.33 | No |
| model_9096_nocut_layer0 | 1500x768 | 11.5566 | 86.5305 | No |
| model_9097_nocut | 1500x768,4x768,4x768,4x4 | 7.33319 | 136.366 | No |
| model_9232_nocut | 12x223x64,12x223x64,12x1500x64,... | 18.436 | 54.2416 | No |
| openclip_vit_base_patch_16_SigLIP_vision_encoder.opt | 3x224x224 | 14.276 | 70.0475 | No |
| resnet101_pytorch_3c_224w_224h | 3x224x224 | 47.4349 | 21.0815 | No |
| resnet50_80_224x224-pytorch | 3x224x224 | 71.9074 | 13.9068 | No |
| retinanet_pvt-t_fpn_1x_coco_kneron_optimized | 3x224x224 | 38.2791 | 26.1239 | No |
| retinanet_pvtv2-b0_fpn_1x_coco_kneron_optimized | 3x224x224 | 50.7597 | 19.7007 | No |
| retinanet_swin-t-p4-w7_fpn_1x_coco_kneron_optimized | 3x224x224 | 22.2134 | 45.018 | No |
| shufflenetv2_x1_0_pytorch_3c_224w_224h | 3x224x224 | 573.333 | 1.74419 | No |
| swin-cut | 3136x96 | 341.628 | 2.92716 | No |
| swin-tiny_16xb64_in1k_kneron_optimized | 3x224x224 | 31.8914 | 31.3564 | No |
| swin_tiny_classify_head | 3x224x224 | 29.5016 | 33.8965 | No |
| swinv2_tiny_classify_head | 3x256x256 | 16.3202 | 61.2736 | Yes |
| t2t-vit-t-14_8xb64_in1k_kneron_optimized | 3x224x224 | 20.0674 | 49.832 | Yes |
| tiny_yolov3_keras_3c_608w_608h | 3x608x608 | 117.58 | 8.50487 | No |
| tnt-s-p16_16xb64_in1k_kneron_optimized | 3x224x224 | 19.3189 | 51.7628 | Yes |
| vit-base-p16_pt-64xb64_in1k-224_kneron_optimized | 3x224x224 | 13.9005 | 71.9398 | No |
| vit_base_classify_head_kneron | 3x224x224 | 15.2754 | 65.4649 | No |
| wenet_u2_decoder | 100x512,99x512,99x512 | 80.2259 | 12.4648 | No |
| wenet_u2_encoder | 100x80 | 37.5957 | 26.5988 | No |
| yolo11m_640x640.opt | 3x640x640 | 9.59189 | 104.255 | No |
| yolo11n_640x640.opt | 3x640x640 | 53.7206 | 18.6148 | No |
| yolov3_keras_3c_416w_416h | 3x416x416 | 19.4 | 51.5464 | No |
| yolov4_keras_3c_416w_416h | 3x416x416 | 21.4122 | 46.7023 | No |
| yolov5s_pytorch_12c_336w_336h | 12x336x336 | 50.6076 | 19.7599 | No |
| yolov7_640x640.opt | 3x640x640 | 8.60255 | 116.245 | No |
| yolov8m_640x640.opt | 3x640x640 | 12.6344 | 79.1488 | No |
| yolov8n_640x640.opt | 3x640x640 | 54.9075 | 18.2125 | No |
| yolov9m_640x640.opt | 3x640x640 | 11.0915 | 90.1591 | No |
| yolov9t_640x640.opt | 3x640x640 | 50.1821 | 19.9274 | No |
Loading