AMDMIGraphX
AMDMIGraphX copied to clipboard
ROCmSoftwarePlatform
Gridsample bicubic mode
This is a follow up on https://github.com/ROCm/AMDMIGraphX/pull/2909 addresses: https://github.com/ROCm/AMDMIGraphX/issues/2960
Moved this to draft until https://github.com/ROCm/AMDMIGraphX/pull/2909 is merged.
Codecov Report
Attention: Patch coverage is 89.69697% with 17 lines in your changes missing coverage. Please review.
Project coverage is 92.12%. Comparing base (
b1f64cc) to head (925da3c). Report is 2 commits behind head on develop.
| Files with missing lines | Patch % | Lines |
|---|---|---|
| src/onnx/parse_gridsample.cpp | 89.69% | 17 Missing :warning: |
Additional details and impacted files
@@ Coverage Diff @@
## develop #2971 +/- ##
===========================================
- Coverage 92.14% 92.12% -0.03%
===========================================
Files 504 504
Lines 20508 20656 +148
===========================================
+ Hits 18898 19029 +131
- Misses 1610 1627 +17
:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.
![codecov[bot] avatar](https://avatars.githubusercontent.com/in/254?v=4 "Published by a pub.dev verified publisher"){kind=small}
@TedThemistokleous @kahmed10 I've updated the PR and now it's ready for review
| Test | Batch | Rate new f36038 |
Rate old 3843a9 |
Diff | Compare |
|---|---|---|---|---|---|
| torchvision-resnet50 | 64 | 3,252.53 | 3,250.51 | 0.06% | :white_check_mark: |
| torchvision-resnet50_fp16 | 64 | 6,995.92 | 6,988.22 | 0.11% | :white_check_mark: |
| torchvision-densenet121 | 32 | 2,432.02 | 2,432.18 | -0.01% | :white_check_mark: |
| torchvision-densenet121_fp16 | 32 | 4,090.13 | 4,115.71 | -0.62% | :white_check_mark: |
| torchvision-inceptionv3 | 32 | 1,636.15 | 1,635.81 | 0.02% | :white_check_mark: |
| torchvision-inceptionv3_fp16 | 32 | 2,738.62 | 2,737.92 | 0.03% | :white_check_mark: |
| cadene-inceptionv4 | 16 | 776.39 | 776.17 | 0.03% | :white_check_mark: |
| cadene-resnext64x4 | 16 | 808.50 | 808.70 | -0.02% | :white_check_mark: |
| slim-mobilenet | 64 | 7,460.64 | 7,451.95 | 0.12% | :white_check_mark: |
| slim-nasnetalarge | 64 | 208.01 | 207.83 | 0.09% | :white_check_mark: |
| slim-resnet50v2 | 64 | 3,345.76 | 3,344.41 | 0.04% | :white_check_mark: |
| bert-mrpc-onnx | 8 | 1,149.21 | 1,151.34 | -0.18% | :white_check_mark: |
| bert-mrpc-tf | 1 | 316.99 | 314.48 | 0.80% | :white_check_mark: |
| pytorch-examples-wlang-gru | 1 | 412.47 | 406.30 | 1.52% | :white_check_mark: |
| pytorch-examples-wlang-lstm | 1 | 380.30 | 381.84 | -0.41% | :white_check_mark: |
| torchvision-resnet50_1 | 1 | 814.70 | 773.61 | 5.31% | :high_brightness: |
| cadene-dpn92_1 | 1 | 435.08 | 435.31 | -0.05% | :white_check_mark: |
| cadene-resnext101_1 | 1 | 375.03 | 381.03 | -1.58% | :white_check_mark: |
| onnx-taau-downsample | 1 | 344.49 | 343.81 | 0.20% | :white_check_mark: |
| dlrm-criteoterabyte | 1 | 35.08 | 35.04 | 0.12% | :white_check_mark: |
| dlrm-criteoterabyte_fp16 | 1 | 58.09 | 58.09 | -0.00% | :white_check_mark: |
| agentmodel | 1 | 7,870.05 | 8,125.79 | -3.15% | :red_circle: |
| unet_fp16 | 2 | 57.87 | 57.80 | 0.12% | :white_check_mark: |
| resnet50v1_fp16 | 1 | 981.27 | 946.54 | 3.67% | :high_brightness: |
| resnet50v1_int8 | 1 | 935.19 | 930.81 | 0.47% | :white_check_mark: |
| bert_base_cased_fp16 | 64 | 1,153.99 | 1,153.31 | 0.06% | :white_check_mark: |
| bert_large_uncased_fp16 | 32 | 355.68 | 355.88 | -0.06% | :white_check_mark: |
| bert_large_fp16 | 1 | 210.06 | 212.11 | -0.97% | :white_check_mark: |
| distilgpt2_fp16 | 16 | 2,161.11 | 2,163.67 | -0.12% | :white_check_mark: |
| yolov5s | 1 | 538.16 | 537.62 | 0.10% | :white_check_mark: |
| tinyllama | 1 | 43.58 | 43.44 | 0.32% | :white_check_mark: |
| vicuna-fastchat | 1 | 171.10 | 171.56 | -0.26% | :white_check_mark: |
| whisper-tiny-encoder | 1 | 408.00 | 407.39 | 0.15% | :white_check_mark: |
| whisper-tiny-decoder | 1 | 435.18 | 430.43 | 1.10% | :white_check_mark: |
This build is not recommended to merge :red_circle:
:red_circle:bert_large_uncased_fp16: FAILED: MIGraphX is not within tolerance - check verbose output