AMDMIGraphX
AMDMIGraphX copied to clipboard
ROCmSoftwarePlatform
Modified documentation
Changes include:
- Updating the landing page
- Removing Search and Index links
- Updating the What Is using text from "Inference optimization with MIGraphX"
- Creating an install and build section
- Changing the title of "contributing to MIGraphX"
- Changing the intro of "contributing" to smooth it out
- Updated the index file
- Updated the toc
Codecov Report
All modified and coverable lines are covered by tests :white_check_mark:
Project coverage is 92.02%. Comparing base (
b6b228c) to head (5498b39). Report is 192 commits behind head on develop.
Additional details and impacted files
@@ Coverage Diff @@
## develop #3481 +/- ##
===========================================
- Coverage 92.14% 92.02% -0.13%
===========================================
Files 504 509 +5
Lines 20508 21005 +497
===========================================
+ Hits 18898 19330 +432
- Misses 1610 1675 +65
: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}
I added some comments but if you are going to spend time modifying the documentation it would be good to make a clear distinction between the developer and user use of the product. Separate out the installing and building. A developer is interested in building the code, a user is interested in installing the code. Mixing them together creates confusion.
This was a first pass at exactly that. This is why "build" and "install" are separate sections. I'll make it clear at the top of the build sections that this is for developing for and contributing to MIGraphX rather than for using the MIGraphX API.
There is only one way to build the code since all three use rbuild. Having two other ways is pointless.
Which steps from which doc would you prefer me to keep?
Good starting point for updating the documentation. I agree with the changes Chris proposed. Main change would be simplifying the installation section to just use
rbuild. Hopefully we find some time later to add in a section to teach how to use MIGraphX if you just want to run models. Approving, but please make the changes Chris mentioned.
I really do want to add that section too. Hopefully soon.
I've removed all the build sections except build with rbuild, removed the prerequisites, and moved What is to the front of the page. I also added a link to the github repo at the top.
| Test | Batch | Rate new 268367 |
Rate old 1ab830 |
Diff | Compare |
|---|---|---|---|---|---|
| torchvision-resnet50 | 64 | 3,248.15 | 3,258.24 | -0.31% | :white_check_mark: |
| torchvision-resnet50_fp16 | 64 | 6,986.91 | 6,986.28 | 0.01% | :white_check_mark: |
| torchvision-densenet121 | 32 | 2,431.89 | 2,435.48 | -0.15% | :white_check_mark: |
| torchvision-densenet121_fp16 | 32 | 4,097.43 | 4,094.65 | 0.07% | :white_check_mark: |
| torchvision-inceptionv3 | 32 | 1,635.18 | 1,636.30 | -0.07% | :white_check_mark: |
| torchvision-inceptionv3_fp16 | 32 | 2,738.44 | 2,745.30 | -0.25% | :white_check_mark: |
| cadene-inceptionv4 | 16 | 776.33 | 779.12 | -0.36% | :white_check_mark: |
| cadene-resnext64x4 | 16 | 807.48 | 808.04 | -0.07% | :white_check_mark: |
| slim-mobilenet | 64 | 7,460.10 | 7,462.28 | -0.03% | :white_check_mark: |
| slim-nasnetalarge | 64 | 207.82 | 208.07 | -0.12% | :white_check_mark: |
| slim-resnet50v2 | 64 | 3,342.79 | 3,438.00 | -2.77% | :white_check_mark: |
| bert-mrpc-onnx | 8 | 1,147.91 | 1,154.53 | -0.57% | :white_check_mark: |
| bert-mrpc-tf | 1 | 306.92 | 320.51 | -4.24% | :red_circle: |
| pytorch-examples-wlang-gru | 1 | 408.53 | 391.03 | 4.48% | :high_brightness: |
| pytorch-examples-wlang-lstm | 1 | 374.57 | 383.67 | -2.37% | :white_check_mark: |
| torchvision-resnet50_1 | 1 | 774.13 | 786.92 | -1.63% | :white_check_mark: |
| cadene-dpn92_1 | 1 | 397.27 | 400.89 | -0.90% | :white_check_mark: |
| cadene-resnext101_1 | 1 | 381.47 | 382.29 | -0.21% | :white_check_mark: |
| onnx-taau-downsample | 1 | 344.47 | 343.92 | 0.16% | :white_check_mark: |
| dlrm-criteoterabyte | 1 | 35.09 | 35.06 | 0.08% | :white_check_mark: |
| dlrm-criteoterabyte_fp16 | 1 | 58.13 | 58.15 | -0.04% | :white_check_mark: |
| agentmodel | 1 | 8,096.58 | 8,244.43 | -1.79% | :white_check_mark: |
| unet_fp16 | 2 | 57.85 | 58.03 | -0.31% | :white_check_mark: |
| resnet50v1_fp16 | 1 | 913.32 | 935.05 | -2.32% | :white_check_mark: |
| resnet50v1_int8 | 1 | 937.35 | 972.22 | -3.59% | :red_circle: |
| bert_base_cased_fp16 | 64 | 1,153.52 | 1,172.19 | -1.59% | :white_check_mark: |
| bert_large_uncased_fp16 | 32 | 355.82 | 362.83 | -1.93% | :white_check_mark: |
| bert_large_fp16 | 1 | 212.04 | 211.45 | 0.28% | :white_check_mark: |
| distilgpt2_fp16 | 16 | 2,162.02 | 2,204.90 | -1.94% | :white_check_mark: |
| yolov5s | 1 | 544.27 | 527.91 | 3.10% | :high_brightness: |
| tinyllama | 1 | 43.38 | 43.44 | -0.13% | :white_check_mark: |
| vicuna-fastchat | 1 | 176.07 | 172.02 | 2.36% | :white_check_mark: |
| whisper-tiny-encoder | 1 | 406.78 | 415.69 | -2.14% | :white_check_mark: |
| whisper-tiny-decoder | 1 | 434.32 | 424.19 | 2.39% | :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