Join this masterclass on ‘Speed up deep learning inference with Intel® Neural Compressor’
Intel®, in collaboration with Analytics India Magazine, is organising an oneAPI workshop on Intel® Neural Compressor on May 13th from 3:00 PM to 5:00 PM.
Quantisation is an important acceleration method, and to support it in hardware, Intel developed Intel® Deep Learning Boost. Additionally, to help developers quantise AI models easily and quickly, Intel has developed a tool called Intel® Neural Compressor.
The workshop will introduce Intel® Optimisation for Tensorflow to help Tensorflow users get better performance on Intel platforms. Intel® Optimisation for TensorFlow is the binary distribution of TensorFlow with Intel® oneAPI deep neural network (oneDNN) library primitives.
oneDNN is an open-source, cross-platform performance library for deep learning applications. The optimisations are directly upstreamed and made available in the official TensorFlow release via a simple flag update, enabling developers to benefit from the Intel® optimisations seamlessly.
Intel® has already released Intel® Optimization for Tensorflow.
Click here to view the installation guide.
The workshop will cover
- oneAPI AI Analytics Toolkit overview
- Introduction to Intel® Optimization for Tensorflow
- Intel optimisations for Tensorflow
- Performance boost numbers
- Hands-on demo to showcase usage and performance boost on DevCloud
The workshop will also offer a demo on how to speed up the AI BERT model by quantisation using Intel®’s Neural Compressor, alongside comparing the performance increase online with Intel® Deep Learning Boost (AMX, VNNI).
The demo will walk you through an end-to-end pipeline to train a BERT model to classify sentences with a small customer dataset and speed up the model based on quantisation by Intel® Neural Compressor, including:
- Train a BERT model by Keras and Intel Optimization for Tensorflow.
- Get an INT8 model by Intel® Neural Compressor.
- Compare the performance of FP32 and INT8 models with the same script.
Agenda:
| Session | Content | Duration | Owner |
| Introduction | Developer Ecosystem Program | 3:00 – 3:10 PM | Kavita Aroor |
| Overview | oneAPI AI Analytics Toolkit Overview, DevCloud Setup | 3:10 – 3:25 PM | Aditya |
| Intel® Extension for TensorFlow | Introduction of Intel® Optimization for TensorFlow + Intel® Extension for TensorFlow hands-on | 3:30 – 4:10 PM | Aditya |
| Intel Neural Compressor Hands-on +QA | Introduction of Intel Neural CompressorHands-on with quantization workload | 4:10 – 5:00 PM | Jianyu Zhang |
Who should attend
- Software Developers
- IT managers
- ML developers
- AI engineers
- Data science professionals
- IT, technology, and software architects
- Senior managers of technology/engineering/software
___________________________________________________
Exclusive Contests — Participate & Win!
Lucky Draw Contest
- Analytics India Magazine is running a Lucky Draw wherein at the end of the workshop 10 lucky participants will get a chance to WIN Amazon Vouchers worth INR 2000/- each.
Note: The winners will be selected based on their engagement on Discord throughout the workshop.
_______________________________________________________
Speaker details
Zhang Jianyu
Zhang Jianyu (Neo) is a senior AI software solution engineer (SSE) of SATG AIA in the PRC. Focus on optimising, consulting and supporting the AI frameworks on the Intel® Platform. Jianyu graduated from Northwestern Polytechnical University (China) with a master’s in Pattern Recognition and Intelligence System. Senior software engineer with rich experience in AI, virtualisation, high concurrency communication and embedded systems.
Kavita Aroor
Kavita Aroor is a developer marketing lead – Asia Pacific & Japan at Intel. She has over 16+ years of experience in marketing.
Aditya Sirvaiya
Aditya Sirvaiya is an AI Software Solutions Engineer at Intel. He specialises in Intel Optimized AI frameworks and OpenVINO Toolkit.
