About MLPerf

MLPerf is a community-led broad ML benchmark suite for measuring the performance of ML software frameworks, ML hardware accelerators, and ML cloud and edge platforms. For additional information on MLPerf, please visit mlperf.org.

To be involved in the activities surrounding MLPerf and its discussions, you may be interested in joining the following MLPerf working groups, which are identified below:

https://groups.google.com/forum/#!forum/mlperf
https://groups.google.com/forum/#!forum/mlperf-inference-submitters

The Zenodo MLPerf community hosts the models that will be used for inference benchmarking.

If you wish to submit a model, please follow the guidelines provided below when you upload a model to the MLPerf community. The submitted model will be reviewed by the curator to ensure it follows the standards/conventions and approved/rejected. So, we strongly recommend following the below guidelines.

Upload Guidelines

Please follow the following guidelines when you upload the files:

Enter all the information required on the upload form.
Follow the guidelines provided below for Step 1.
Please upload only .zip files.

On the upload form, there are multiple sections (e.g. Upload type, Basic information, etc.). For each of those sections, kindly follow the guidelines provided below.

Not following the guidelines can cause the curator to reject your upload.

Upload Type

Choose “Other”

Basic Information

Title: "Trained Model for <NN Model name> for MLPerf Inference"

Authors: Enter all the authors that have helped contribute to the work.

Description: We recommend filling in the "Description" field with the below information:

Application: <e.g. Image Classification, Object Detection, Speech To Text, ...>
ML Task: <e.g. MobileNet, ResNet-50, Deep Speech 2, ...>
Framework: <The name and version of the framework where the model is trained as follows: <Framework_name> <version>>
Training Information:
Quality: <target>
Precision: <single-precision float>/<half-precision float>/<...>
Is Quantized: <yes>/<no>
Is ONNX: <yes>/<no>
Dataset: <the dataset that is used to train the model>

Version: MLPerf v0.5 <Training>/<Inference>

Keywords: Add the followings to keywords <Application>, <ML Task>, <Framework>, <Dataset>, <Training/Inference>, Pretrained Model

Additional Notes: (This is optional, put whatever additional information here).

License

Please select the following: