amdinfer 0.4.0

Features

• Supports client requests using HTTP/REST, gRPC and websocket protocols using an API based on KServe’s v2 specification

• Custom applications can directly call the backend bypassing the other protocols using the native C++ API

• C++ library with Python bindings to simplify making requests to the server

• Incoming requests are transparently batched based on the user specifications

• Users can define how many models, and how many instances of each, to run in parallel

The AMD Inference Server is integrated with the following libraries out of the gate:

• TensorFlow and PyTorch models with ZenDNN on CPUs (optimized for AMD CPUs)

• ONNX models with MIGraphX on AMD GPUs

• XModel models with Vitis AI on AMD FPGAs

• A graph of computation including as pre- and post-processing can be written using AKS on AMD FPGAs for end-to-end inference

Quick Start Deployment and Inference

The following example demonstrates how to deploy the server locally and run a sample inference. This example runs on the CPU and does not require any special hardware. You can see a more detailed version of this example in the quickstart.

# Step 1: Download the example files and create a model repository
wget https://github.com/Xilinx/inference-server/raw/main/examples/resnet50/quickstart-setup.sh
chmod +x ./quickstart-setup.sh
./quickstart-setup.sh

# Step 2: Launch the AMD Inference Server
docker run -d --net=host -v ${PWD}/model_repository:/mnt/models:rw amdih/serve:uif1.1_zendnn_amdinfer_0.3.0 amdinfer-server --enable-repository-watcher

# Step 3: Install the Python client library
pip install amdinfer

# Step 4: Send an inference request
python3 tfzendnn.py --endpoint resnet50 --image ./dog-3619020_640.jpg --labels ./imagenet_classes.txt

# Inference should print the following:
#
#     Running the TF+ZenDNN example for ResNet50 in Python
#     Waiting until the server is ready...
#     Making inferences...
#     Top 5 classes for ../../tests/assets/dog-3619020_640.jpg:
#       n02112018 Pomeranian
#       n02112350 keeshond
#       n02086079 Pekinese, Pekingese, Peke
#       n02112137 chow, chow chow
#       n02113023 Pembroke, Pembroke Welsh corgi

Learn more

The documentation for the AMD Inference Server is available online.

Check out the quickstart online to help you get started.

Support

Raise issues if you find a bug or need help. Refer to Contributing for more information.

License

The AMD Inference Server is licensed under the terms of Apache 2.0 (see LICENSE). The LICENSE file contains additional license information for third-party files distributed with this work. More license information can be seen in the dependencies.

IMPORTANT NOTICE CONCERNING OPEN-SOURCE SOFTWARE

Materials in this release may be licensed by Xilinx or third parties and may be subject to the GNU General Public License, the GNU Lesser General License, or other licenses.

Licenses and source files may be downloaded from:

• libamdinfer

• libbrotlicommon

• libbrotlidec

• libbrotlienc

• libcares

• libcom_err

• libcrypto

• libdrogon

• libefsw

• libgssapi_krb5

• libjsoncpp

• libk5crypto

• libkeyutils

• libkrb5

• libkrb5support

• libopencv_core

• libopencv_imgcodecs

• libopencv_imgproc

• libossp-uuid

• libpcre

• libprometheus-cpp-core

• libprotobuf

• libselinux

• libssl

• libtrantor

Note: You are solely responsible for checking the header files and other accompanying source files (i) provided within, in support of, or that otherwise accompanies these materials or (ii) created from the use of third party software and tools (and associated libraries and utilities) that are supplied with these materials, because such header and/or source files may contain or describe various copyright notices and license terms and conditions governing such files, which vary from case to case based on your usage and are beyond the control of Xilinx. You are solely responsible for complying with the terms and conditions imposed by third parties as applicable to your software applications created from the use of third party software and tools (and associated libraries and utilities) that are supplied with the materials.