Using the IBM analog in-memory hardware acceleration kit for neural network training and inference
Analog In-Memory Computing (AIMC) is a promising approach to reduce the latency and energy consumption of Deep Neural Network (DNN) inference and training. However, the noisy and non-linear device characteristics and the non-ideal peripheral circuitry in AIMC chips require adapting DNNs to be deploy...
| Main Authors: | , , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
AIP Publishing LLC
2023-12-01
|
| Series: | APL Machine Learning |
| Online Access: | http://dx.doi.org/10.1063/5.0168089 |
| _version_ | 1797367098795622400 |
|---|---|
| author | Manuel Le Gallo Corey Lammie Julian Büchel Fabio Carta Omobayode Fagbohungbe Charles Mackin Hsinyu Tsai Vijay Narayanan Abu Sebastian Kaoutar El Maghraoui Malte J. Rasch |
| author_facet | Manuel Le Gallo Corey Lammie Julian Büchel Fabio Carta Omobayode Fagbohungbe Charles Mackin Hsinyu Tsai Vijay Narayanan Abu Sebastian Kaoutar El Maghraoui Malte J. Rasch |
| author_sort | Manuel Le Gallo |
| collection | DOAJ |
| description | Analog In-Memory Computing (AIMC) is a promising approach to reduce the latency and energy consumption of Deep Neural Network (DNN) inference and training. However, the noisy and non-linear device characteristics and the non-ideal peripheral circuitry in AIMC chips require adapting DNNs to be deployed on such hardware to achieve equivalent accuracy to digital computing. In this Tutorial, we provide a deep dive into how such adaptations can be achieved and evaluated using the recently released IBM Analog Hardware Acceleration Kit (AIHWKit), freely available at https://github.com/IBM/aihwkit. AIHWKit is a Python library that simulates inference and training of DNNs using AIMC. We present an in-depth description of the AIHWKit design, functionality, and best practices to properly perform inference and training. We also present an overview of the Analog AI Cloud Composer, a platform that provides the benefits of using the AIHWKit simulation in a fully managed cloud setting along with physical AIMC hardware access, freely available at https://aihw-composer.draco.res.ibm.com. Finally, we show examples of how users can expand and customize AIHWKit for their own needs. This Tutorial is accompanied by comprehensive Jupyter Notebook code examples that can be run using AIHWKit, which can be downloaded from https://github.com/IBM/aihwkit/tree/master/notebooks/tutorial. |
| first_indexed | 2024-03-08T17:12:26Z |
| format | Article |
| id | doaj.art-fa2ca9179db74df2aabd9c81cc0a6597 |
| institution | Directory Open Access Journal |
| issn | 2770-9019 |
| language | English |
| last_indexed | 2024-03-08T17:12:26Z |
| publishDate | 2023-12-01 |
| publisher | AIP Publishing LLC |
| record_format | Article |
| series | APL Machine Learning |
| spelling | doaj.art-fa2ca9179db74df2aabd9c81cc0a65972024-01-03T19:54:29ZengAIP Publishing LLCAPL Machine Learning2770-90192023-12-0114041102041102-3610.1063/5.0168089Using the IBM analog in-memory hardware acceleration kit for neural network training and inferenceManuel Le Gallo0Corey Lammie1Julian Büchel2Fabio Carta3Omobayode Fagbohungbe4Charles Mackin5Hsinyu Tsai6Vijay Narayanan7Abu Sebastian8Kaoutar El Maghraoui9Malte J. Rasch10IBM Research Europe, 8803 Rüschlikon, SwitzerlandIBM Research Europe, 8803 Rüschlikon, SwitzerlandIBM Research Europe, 8803 Rüschlikon, SwitzerlandIBM Research - Yorktown Heights, Yorktown Heights, New York 10598, USAIBM Research - Yorktown Heights, Yorktown Heights, New York 10598, USAIBM Research - Almaden, San Jose, California 95120, USAIBM Research - Almaden, San Jose, California 95120, USAIBM Research - Yorktown Heights, Yorktown Heights, New York 10598, USAIBM Research Europe, 8803 Rüschlikon, SwitzerlandIBM Research - Yorktown Heights, Yorktown Heights, New York 10598, USAIBM Research - Yorktown Heights, Yorktown Heights, New York 10598, USAAnalog In-Memory Computing (AIMC) is a promising approach to reduce the latency and energy consumption of Deep Neural Network (DNN) inference and training. However, the noisy and non-linear device characteristics and the non-ideal peripheral circuitry in AIMC chips require adapting DNNs to be deployed on such hardware to achieve equivalent accuracy to digital computing. In this Tutorial, we provide a deep dive into how such adaptations can be achieved and evaluated using the recently released IBM Analog Hardware Acceleration Kit (AIHWKit), freely available at https://github.com/IBM/aihwkit. AIHWKit is a Python library that simulates inference and training of DNNs using AIMC. We present an in-depth description of the AIHWKit design, functionality, and best practices to properly perform inference and training. We also present an overview of the Analog AI Cloud Composer, a platform that provides the benefits of using the AIHWKit simulation in a fully managed cloud setting along with physical AIMC hardware access, freely available at https://aihw-composer.draco.res.ibm.com. Finally, we show examples of how users can expand and customize AIHWKit for their own needs. This Tutorial is accompanied by comprehensive Jupyter Notebook code examples that can be run using AIHWKit, which can be downloaded from https://github.com/IBM/aihwkit/tree/master/notebooks/tutorial.http://dx.doi.org/10.1063/5.0168089 |
| spellingShingle | Manuel Le Gallo Corey Lammie Julian Büchel Fabio Carta Omobayode Fagbohungbe Charles Mackin Hsinyu Tsai Vijay Narayanan Abu Sebastian Kaoutar El Maghraoui Malte J. Rasch Using the IBM analog in-memory hardware acceleration kit for neural network training and inference APL Machine Learning |
| title | Using the IBM analog in-memory hardware acceleration kit for neural network training and inference |
| title_full | Using the IBM analog in-memory hardware acceleration kit for neural network training and inference |
| title_fullStr | Using the IBM analog in-memory hardware acceleration kit for neural network training and inference |
| title_full_unstemmed | Using the IBM analog in-memory hardware acceleration kit for neural network training and inference |
| title_short | Using the IBM analog in-memory hardware acceleration kit for neural network training and inference |
| title_sort | using the ibm analog in memory hardware acceleration kit for neural network training and inference |
| url | http://dx.doi.org/10.1063/5.0168089 |
| work_keys_str_mv | AT manuellegallo usingtheibmanaloginmemoryhardwareaccelerationkitforneuralnetworktrainingandinference AT coreylammie usingtheibmanaloginmemoryhardwareaccelerationkitforneuralnetworktrainingandinference AT julianbuchel usingtheibmanaloginmemoryhardwareaccelerationkitforneuralnetworktrainingandinference AT fabiocarta usingtheibmanaloginmemoryhardwareaccelerationkitforneuralnetworktrainingandinference AT omobayodefagbohungbe usingtheibmanaloginmemoryhardwareaccelerationkitforneuralnetworktrainingandinference AT charlesmackin usingtheibmanaloginmemoryhardwareaccelerationkitforneuralnetworktrainingandinference AT hsinyutsai usingtheibmanaloginmemoryhardwareaccelerationkitforneuralnetworktrainingandinference AT vijaynarayanan usingtheibmanaloginmemoryhardwareaccelerationkitforneuralnetworktrainingandinference AT abusebastian usingtheibmanaloginmemoryhardwareaccelerationkitforneuralnetworktrainingandinference AT kaoutarelmaghraoui usingtheibmanaloginmemoryhardwareaccelerationkitforneuralnetworktrainingandinference AT maltejrasch usingtheibmanaloginmemoryhardwareaccelerationkitforneuralnetworktrainingandinference |