SlideShare una empresa de Scribd logo
1 de 35
Descargar para leer sin conexión
Meetup "IA à portée de la main »
20 mai 2019 – IBM France Lab@Paris-Saclay
Deep Learning ?
© 2019 IBM Corporation
Classification
Counting
Detection
{
"webAPIId":"8ace1c11-0f7e-4a17-a54d-63c741c7b31d",
"imageUrl":"http://powerai-vision-portal:9080/powerai-vision-
api/uploads/temp/8ace1c11-0f7e-4a17-a54d-63c741c7b31d/3f3db18c-eb02-4dca-8015-
8a68a830eb5b.jpg",
"imageMd5":"c13e24226dbd67404a4b464f18a231ac",
"classified":[
{
"confidence":0.9997112154960632,
"label":"up",
"ymax":608,"xmax":600,"xmin":503,"ymin":430
},
{
"confidence":0.9993865489959717,
"label":"up",
"ymax":543, "xmax":949,"xmin":837,"ymin":372
}
],
"result":"success"
}
Le véritable objectif
© 2019 IBM Corporation
Cycle de vie d’une application cognitive
© 2019 IBM Corporation
Idée / ROI
Application
Services
Développement
Déploiement Utilisation
© 2019 IBM Corporation
Suivi du parcours client
SERVER
PERFORMANCE
RETAIL
E-COMMERCE
Suivi des clients d’un supermarché
© 2019 IBM Corporation
Surveillance des tunnels d’autoroute
SERVER
PERFORMANCE
Détection des incidents
Application
© 2019 IBM Corporation
IBM Cloud Private
DevOps
Microservices
Services
© 2019 IBM Corporation
Deux approches
Prêt-à-porter
IBM
One AI
Sur-mesure
Watson
© 2019 IBM Corporation
Discovery
Knowledge
Studio
Knowledge
Catalog
Discovery
Conversation
Assistant
Empathy
Personality
Insights
Tone
Analyser
Language
Language
Understanding
Language
Classifier
Language
Translator
Voice
Speech
to Text
Text to
Speech
Perception
Visual
Recognition
IBM One AI
© 2019 IBM Corporation
Data Scientist App Developer AI Ops
Build AI Run AI Operate AI
Watson OpenScale
Fairness & Explainability
Inputs for Continuous Evolution
Business KPIs and production
metrics
Watson Studio
Watson Machine
Learning
Build
Deploy
and run
Operate
trusted AI
Business user
Consume AI
Data Exploration
Data Preparation
Model Development
Model Deployment
Model Management
Retraining
Watson Knowledge
Catalog
Data Profiling
Quality and Lineage
Data Governance
Organize and
Govern data
Data Engineer
Organize
Data for AI
Watson Machine Learning AcceleratorWatson Machine Learning Accelerator
Power AI
© 2019 IBM Corporation
Les frameworks opensource habituels optimisés
Plus de capacités, plus de possibilités, plus de
scalabilité
Des serveurs conçus pour le calcul accéléré
5.6x plus de bande passante qu’avec PCIe gen3
Solutions de partage de ressources
Des solutions pour aider à la conception et la mise à
disposition de vos services d’IA
AC922
Watson Machine Learning Accelerator
Surveiller l’usage
© 2019 IBM Corporation
Watson OpenScale
Partenariats
© 2019 IBM Corporation
Our clients
H2O.ai IBM IVAPlanetANEO
OKTAL SE Cyclope.ai
Aide à la réflexion sur
votre stratégie d’IA
Preuves de concepts
Collaboration avec des
partenaires
Accès à des experts
HPC / IA
Planification
d’infrastructure IA
© 2019 IBM Corporation
Ne restez pas tout seuls !
Partenaire
© 2019 IBM Corporation
© 2019 IBM Corporation
Power AI Vision
Démo
© 2019 IBM Corporation
Une stack logicielle basée sur l’opensource
© 2019 IBM Corporation
Watson Machine Learning (WML)
PowerAI
Watson ML
Watson ML AcceleratorWatson Studio
AI OpenScale
Development
Environment
Train Models
Runtime
Environment
Train, Deploy, & Manage Models
Operation Fabric
Monitor & Improve
Deployed Models
SnapML
Previous Names:
WML Accelerator = PowerAI Enterprise
GPU-Accelerated
Power Servers
Storage
PowerAI
Des besoins de plus en plus « gros »
© 2019 IBM Corporation
Volume of data generated
by a single car = 1 TB+ / h
Exemples de datasets de haute résolution
© 2019 IBM Corporation
•CityScapes : 1024x2048 5k images.
Semantic Segmentation, Benchmark
results.
•Mapillary Vistas : various sizes - split into
18,000 for training, 2,000 for validation,
5,000 for testing (average image
resolution of ~9 megapixels)
•Synthia : Large volume of data & ground
truth: +200,000 HD images from video
streams and +20,000 HD images from
independent snapshots
Impact de la résolution sur la précision des modèles
© 2019 IBM Corporation
La solution d’IBM à la limitation de la mémoire des GPUs
© 2019 IBM Corporation
Large Model Support
POWER
CPU
DDR4
GPU
NVLink
Graphics
Memory
Traditional Model Support
CPUDDR4
GPU
PCIe
Graphics
Memory
Limited memory on GPU forces trade-off in
model size / data resolution
Use system memory and GPU to support more
complex and higher resolution data
32GB max
2TB max
2TB max
32GB max
• Leveraging NVLink and CPU-GPU memory coherence enables larger and more complex models
• Improves model accuracy with more images and higher resolution images
Large Memory Support (LMS)
© 2019 IBM Corporation
LMS Concept
1. Treat GPU memory as an application-level cache
w.r.t the host main memory.
2. All data reside on host memory and are copied to
GPU memory only when needed.
3. After the GPU memory is used, depending on
whether it has been modified or has future usage, it
can be either copied back to the system memory or
simply discarded.
4. To efficiently utilize GPU memory, LMS
implementation keeps a large memory pool so that
different memory pieces from CPU memory can
share the same GPU memory chunk. Therefore, at
any moment, the GPU only holds data necessary to
process one operation, for example, the forward
propagation of one operation in a neural network
(Feature maps).
If the memory requirement from any operation is larger
than the GPU memory, then even LMS will fail as well. In
theory, LMS should be able to handle a deep neural
network of an arbitrary capacity, as long as all the data
from the largest operation can fit into the GPU memory.
à As LMS uses system memory to store features map
tensors, the training performance limiting factor is no
longer the PCIe bus but the CPU-RAM high-bandwidth.
Ex: Feature Maps tensor backward reuse for the gradient
update with AlexNet
Without LMS: forward temporary data are kept in
GPU memory for the backward computation
LMS: Caffe
© 2019 IBM Corporation
Caffe is one of the most popular open-source deep learning frameworks, developed by the Berke-
ley AI Research. It clearly separates the network description (made with a Protocol Buffer file,
similar to
JSON), and the training run, where we have a second file to specify the parameters, how we optimize
the
network, what will be logged.
This is called "define-AND-run", and most of the frame-works work that way. In addition, it provides
a Python API to write for example new layer types, which is useful for loading complex data and
feeding it to the network.
Caffe LMS Command line Flags:
-lms : to activate LMS for the current training.
LMS: Chainer
© 2019 IBM Corporation
Chainer Chainer is a lesser-known open-source deep learning framework in Python, popular in Japan but
not really beyond. It is developed by a Japanese com-
pany called Preferred Networks, in partnership with several companies including IBM.
Unlike Caffe, it has a "define-BY-run" approach, where network description and training are "mixed" in
the same code. This is more flexible and allows to dynamically update the
network if needed (although it is useful on very specific deep learning algorithms I haven’t used during
the internship).
Light python code change for LMS activation :
with chainer.out_of_core_mode(fine_granularity=True, async=False,
devices=gpu):
trainer.run()
LMS: Tensorflow
© 2019 IBM Corporation
Keras API
from tensorflow.contrib.lms import LMSKerasCallback lms_callback =
LMSKerasCallback()
…
model.fit_generator(generator=training_gen,
callbacks=[lms_callback])
Estimator API
with tf.name_scope(‘gradientscope'):
optimizer = tf.train.GradientDescentOptimizer(learning_rate=0.001)
train_op = optimizer.minimize(loss=loss,
global_step=tf.train.get_global_step())
from tensorflow.contrib.lms import LMSHook
lms_hook = LMSHook({‘gradientscope '})
mnist_classifier.train(input_fn=train_input_fn, steps=20000,
hooks=[logging_hook, lms_hook])
IBM AC922
4 GPUs @150GB/s
CPU ßà GPU bandwidth
6 GPUs @100GB/s
CPU ßà GPU bandwidth
Coherent access to system memory
PCIe Gen 4 and CAPI 2.0 to InfiniBand
Air and Water cooled options
Coherent access to system memory
PCIe Gen 4 and CAPI 2.0 to InfiniBand
Water cooled only
NVLink
100GB/s
NVLink
100GB/s
NVDIA V100
Coherent
access to
system memory
(2TB)
NVLink
100GB/s
NVLink
100GB/s
NVLink
100GB/s
170GB/s
CPU
PCIe Gen 4
CAPI 2.0
NVDIA V100NVDIA V100
DDR4
I
B
Coherent
access to
system memory
(2TB)
NVLink
150GB/s
NVLink
150GB/s
170GB/s
CP
U
PCIe Gen 4
CAPI 2.0
NVLink
150GB/s NVDIA V100NVDIA V100
DDR4
IB
© 2019 IBM Corporation
AC922: Des capacités uniques pour l’IA
28
Large Model Support
Only server that can leverage System Memory
from GPU side for 2TB+ per node!
Distributed Deep Learning
Most efficient scaling, due to better I/O
between nodes (95% across 256 GPUS)
3.1 Hours
49 Mins
0
2000
4000
6000
8000
10000
12000
Xeon x86 2640v4 w/
4x V100 GPUs
Power AC922 w/ 4x
V100 GPUs
Time(secs)
Caffe with LMS (Large Model Support)
Runtime of 1000 Iterations
GoogleNet model on Enlarged
ImageNet Dataset (2240x2240)
Near Ideal Scaling to 256 GPUs
1
2
4
8
16
32
64
128
256
4 16 64 256
Speedup
Number of GPUs
Ideal Scaling
DDL Actual Scaling
Caffe with PowerAI DDL, Running on S822LC Power System
ResNet-50, ImageNet-1K
~4x faster insights
Train More Build More Know More
© 2019 IBM Corporation
LMS: impact du NVLink entre CPU et GPU
© 2019 IBM Corporation
NVLink 2.0 connected GPU training one high res 3D MRI with large model support
LMS: Impact sur le temps d’une epoch
© 2019 IBM Corporation
0
250
500
750
1000
1250
1500
1750
2000
2250
AC922
with NVLink 2.0
GPU server with PCI GPU server with PCI
contention
Seconds
Epoch times at max resolution with swapping
LMS: Impact sur le taux d’utilisation des GPUs
© 2019 IBM Corporation
GB/s
10 GB/s
20 GB/s
30 GB/s
40 GB/s
50 GB/s
60 GB/s
70 GB/s
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
AC922
with NVLink 2.0
GPU server with PCI GPU server with PCI
contention
MemoryCopyThroughput
GPUUtilization
GPU utilization
Memory copy throughput
LMS: Overhead par rapport à la mémoire du GPU
© 2019 IBM Corporation
3%
13%
25%
0%
20%
40%
60%
80%
100%
0
100
200
300
400
500
600
700
1.4x 1.8x 2.4x
OverheadPercentage
Epochtimeinseconds Swapping overhead using resolutions above the 16GB GPU
memory limit
32 GB V100 GPU, no swapping 16GB V100 GPU, swapping Overhead
Une architecture évolutive
© 2019 IBM Corporation
Experimentation
Single Tenant
Stabilization & Production
Secure Multitenant
Expansion
Enterprise Scale / Multiple Lines of Business
Data
Scientist’s
workstations
Internal
SAS
drives &
NVM’s
IBM
Power
Systems
AC922
High-Speed
Network
Subsystem
Existing
Organization
Infrastructure
IBM
Elastic
Storage
Server
(ESS)
Training & Inference Cluster
IBM Power Systems AC922, LC921 & LC922
Master & Failover Master
Nodes IBM Power Systems
LC921 & LC922
Login Nodes
IBM Power Systems
LC921 & LC922
Training Cluster
IBM Power Systems AC922
IBM
Elastic
Storage
Server
(ESS)
High-Speed
Network
Subsystem
Existing
Organization
Infrastructure
One software stack from experimentation to expansion
IBM PowerAI Enterprise
Red Hat Enterprise Linux (RHEL)
IBM Power System & x86 Servers
Services&
Support
IBM Spectrum Scale / IBM Elastic Storage Server (ESS)
Aide à la réflexion sur
votre stratégie d’IA
Preuves de concepts
Collaboration avec des
partenaires
Accès à des experts
HPC / IA
Planification
d’infrastructure IA
© 2019 IBM Corporation
Ne restez pas tout seuls !
Merci !
IBM AI Portfolio
Dédiaboliser les projets d’IA…

Más contenido relacionado

La actualidad más candente

Transparent Hardware Acceleration for Deep Learning
Transparent Hardware Acceleration for Deep LearningTransparent Hardware Acceleration for Deep Learning
Transparent Hardware Acceleration for Deep LearningIndrajit Poddar
 
OpenPOWER/POWER9 Webinar from MIT and IBM
OpenPOWER/POWER9 Webinar from MIT and IBM OpenPOWER/POWER9 Webinar from MIT and IBM
OpenPOWER/POWER9 Webinar from MIT and IBM Ganesan Narayanasamy
 
Fast Scalable Easy Machine Learning with OpenPOWER, GPUs and Docker
Fast Scalable Easy Machine Learning with OpenPOWER, GPUs and DockerFast Scalable Easy Machine Learning with OpenPOWER, GPUs and Docker
Fast Scalable Easy Machine Learning with OpenPOWER, GPUs and DockerIndrajit Poddar
 
OpenPOWER Webinar on Machine Learning for Academic Research
OpenPOWER Webinar on Machine Learning for Academic Research OpenPOWER Webinar on Machine Learning for Academic Research
OpenPOWER Webinar on Machine Learning for Academic Research Ganesan Narayanasamy
 
Gschwind, PowerAI: A Co-Optimized Software Stack for AI on Power
Gschwind, PowerAI: A Co-Optimized Software Stack for AI on PowerGschwind, PowerAI: A Co-Optimized Software Stack for AI on Power
Gschwind, PowerAI: A Co-Optimized Software Stack for AI on PowerMichael Gschwind
 
Large Model support and Distribute deep learning
Large Model support and Distribute deep learningLarge Model support and Distribute deep learning
Large Model support and Distribute deep learningGanesan Narayanasamy
 
Big Memory for HPC
Big Memory for HPCBig Memory for HPC
Big Memory for HPCMemVerge
 
Power AI introduction
Power AI introductionPower AI introduction
Power AI introductionSnowy Chen
 
AIST Super Green Cloud: lessons learned from the operation and the performanc...
AIST Super Green Cloud: lessons learned from the operation and the performanc...AIST Super Green Cloud: lessons learned from the operation and the performanc...
AIST Super Green Cloud: lessons learned from the operation and the performanc...Ryousei Takano
 
Deep Learning Computer Build
Deep Learning Computer BuildDeep Learning Computer Build
Deep Learning Computer BuildPetteriTeikariPhD
 
IBM Power Systems E850C and S824
IBM Power Systems E850C and S824IBM Power Systems E850C and S824
IBM Power Systems E850C and S824David Spurway
 
High Performance Computing (HPC) on AWS 101
High Performance Computing (HPC) on AWS 101High Performance Computing (HPC) on AWS 101
High Performance Computing (HPC) on AWS 101Amazon Web Services
 
GPUDirect RDMA and Green Multi-GPU Architectures
GPUDirect RDMA and Green Multi-GPU ArchitecturesGPUDirect RDMA and Green Multi-GPU Architectures
GPUDirect RDMA and Green Multi-GPU Architecturesinside-BigData.com
 
Distributed deep learning reference architecture v3.2l
Distributed deep learning reference architecture v3.2lDistributed deep learning reference architecture v3.2l
Distributed deep learning reference architecture v3.2lGanesan Narayanasamy
 
Deep Learning Frameworks Using Spark on YARN by Vartika Singh
Deep Learning Frameworks Using Spark on YARN by Vartika SinghDeep Learning Frameworks Using Spark on YARN by Vartika Singh
Deep Learning Frameworks Using Spark on YARN by Vartika SinghData Con LA
 

La actualidad más candente (20)

OpenPOWER Seminar at IIT Madras
OpenPOWER Seminar at IIT MadrasOpenPOWER Seminar at IIT Madras
OpenPOWER Seminar at IIT Madras
 
Open power ddl and lms
Open power ddl and lmsOpen power ddl and lms
Open power ddl and lms
 
Transparent Hardware Acceleration for Deep Learning
Transparent Hardware Acceleration for Deep LearningTransparent Hardware Acceleration for Deep Learning
Transparent Hardware Acceleration for Deep Learning
 
OpenPOWER/POWER9 Webinar from MIT and IBM
OpenPOWER/POWER9 Webinar from MIT and IBM OpenPOWER/POWER9 Webinar from MIT and IBM
OpenPOWER/POWER9 Webinar from MIT and IBM
 
Fast Scalable Easy Machine Learning with OpenPOWER, GPUs and Docker
Fast Scalable Easy Machine Learning with OpenPOWER, GPUs and DockerFast Scalable Easy Machine Learning with OpenPOWER, GPUs and Docker
Fast Scalable Easy Machine Learning with OpenPOWER, GPUs and Docker
 
OpenPOWER Webinar on Machine Learning for Academic Research
OpenPOWER Webinar on Machine Learning for Academic Research OpenPOWER Webinar on Machine Learning for Academic Research
OpenPOWER Webinar on Machine Learning for Academic Research
 
Gschwind, PowerAI: A Co-Optimized Software Stack for AI on Power
Gschwind, PowerAI: A Co-Optimized Software Stack for AI on PowerGschwind, PowerAI: A Co-Optimized Software Stack for AI on Power
Gschwind, PowerAI: A Co-Optimized Software Stack for AI on Power
 
Large Model support and Distribute deep learning
Large Model support and Distribute deep learningLarge Model support and Distribute deep learning
Large Model support and Distribute deep learning
 
Big Memory for HPC
Big Memory for HPCBig Memory for HPC
Big Memory for HPC
 
Power AI introduction
Power AI introductionPower AI introduction
Power AI introduction
 
AIST Super Green Cloud: lessons learned from the operation and the performanc...
AIST Super Green Cloud: lessons learned from the operation and the performanc...AIST Super Green Cloud: lessons learned from the operation and the performanc...
AIST Super Green Cloud: lessons learned from the operation and the performanc...
 
Deep Learning Computer Build
Deep Learning Computer BuildDeep Learning Computer Build
Deep Learning Computer Build
 
IBM Power Systems E850C and S824
IBM Power Systems E850C and S824IBM Power Systems E850C and S824
IBM Power Systems E850C and S824
 
2017 04-13-google-tpu-04
2017 04-13-google-tpu-042017 04-13-google-tpu-04
2017 04-13-google-tpu-04
 
High Performance Computing (HPC) on AWS 101
High Performance Computing (HPC) on AWS 101High Performance Computing (HPC) on AWS 101
High Performance Computing (HPC) on AWS 101
 
GPUDirect RDMA and Green Multi-GPU Architectures
GPUDirect RDMA and Green Multi-GPU ArchitecturesGPUDirect RDMA and Green Multi-GPU Architectures
GPUDirect RDMA and Green Multi-GPU Architectures
 
WML SNAP ML
WML SNAP MLWML SNAP ML
WML SNAP ML
 
OpenPOWER Boot camp in Zurich
OpenPOWER Boot camp in ZurichOpenPOWER Boot camp in Zurich
OpenPOWER Boot camp in Zurich
 
Distributed deep learning reference architecture v3.2l
Distributed deep learning reference architecture v3.2lDistributed deep learning reference architecture v3.2l
Distributed deep learning reference architecture v3.2l
 
Deep Learning Frameworks Using Spark on YARN by Vartika Singh
Deep Learning Frameworks Using Spark on YARN by Vartika SinghDeep Learning Frameworks Using Spark on YARN by Vartika Singh
Deep Learning Frameworks Using Spark on YARN by Vartika Singh
 

Similar a IBM Cloud Paris Meetup - 20190520 - IA & Power

How to Optimize Hortonworks Apache Spark ML Workloads on Modern Processors
How to Optimize Hortonworks Apache Spark ML Workloads on Modern Processors How to Optimize Hortonworks Apache Spark ML Workloads on Modern Processors
How to Optimize Hortonworks Apache Spark ML Workloads on Modern Processors DataWorks Summit/Hadoop Summit
 
Backend.AI Technical Introduction (19.09 / 2019 Autumn)
Backend.AI Technical Introduction (19.09 / 2019 Autumn)Backend.AI Technical Introduction (19.09 / 2019 Autumn)
Backend.AI Technical Introduction (19.09 / 2019 Autumn)Lablup Inc.
 
AI in Health Care using IBM Systems/OpenPOWER systems
AI in Health Care using IBM Systems/OpenPOWER systemsAI in Health Care using IBM Systems/OpenPOWER systems
AI in Health Care using IBM Systems/OpenPOWER systemsGanesan Narayanasamy
 
AI in Healh Care using IBM POWER systems
AI in Healh Care using IBM POWER systems AI in Healh Care using IBM POWER systems
AI in Healh Care using IBM POWER systems Ganesan Narayanasamy
 
Harnessing the virtual realm for successful real world artificial intelligence
Harnessing the virtual realm for successful real world artificial intelligenceHarnessing the virtual realm for successful real world artificial intelligence
Harnessing the virtual realm for successful real world artificial intelligenceAlison B. Lowndes
 
Capi snap overview
Capi snap overviewCapi snap overview
Capi snap overviewYutaka Kawai
 
IBM Power Systems - enabling cloud solutions
IBM Power Systems - enabling cloud solutionsIBM Power Systems - enabling cloud solutions
IBM Power Systems - enabling cloud solutionsDavid Spurway
 
Cisco connect montreal 2018 compute v final
Cisco connect montreal 2018   compute v finalCisco connect montreal 2018   compute v final
Cisco connect montreal 2018 compute v finalCisco Canada
 
Sundar Ranganathan, NetApp + Vinod Iyengar, H2O.ai - Driverless AI integratio...
Sundar Ranganathan, NetApp + Vinod Iyengar, H2O.ai - Driverless AI integratio...Sundar Ranganathan, NetApp + Vinod Iyengar, H2O.ai - Driverless AI integratio...
Sundar Ranganathan, NetApp + Vinod Iyengar, H2O.ai - Driverless AI integratio...Sri Ambati
 
MIKE BY DHI SAAS - New features as of October 2014
MIKE BY DHI SAAS - New features as of October 2014MIKE BY DHI SAAS - New features as of October 2014
MIKE BY DHI SAAS - New features as of October 2014Stephen Flood
 
Amazon EC2 instances: Customizable cloud computing across workloads - DEM20-S...
Amazon EC2 instances: Customizable cloud computing across workloads - DEM20-S...Amazon EC2 instances: Customizable cloud computing across workloads - DEM20-S...
Amazon EC2 instances: Customizable cloud computing across workloads - DEM20-S...Amazon Web Services
 
Scalable Multi-Node Deep Learning Training in the Cloud (CMP368-R1) - AWS re:...
Scalable Multi-Node Deep Learning Training in the Cloud (CMP368-R1) - AWS re:...Scalable Multi-Node Deep Learning Training in the Cloud (CMP368-R1) - AWS re:...
Scalable Multi-Node Deep Learning Training in the Cloud (CMP368-R1) - AWS re:...Amazon Web Services
 
Deep Dive on Amazon EC2 Accelerated Computing - AWS Online Tech Talks
Deep Dive on Amazon EC2 Accelerated Computing - AWS Online Tech TalksDeep Dive on Amazon EC2 Accelerated Computing - AWS Online Tech Talks
Deep Dive on Amazon EC2 Accelerated Computing - AWS Online Tech TalksAmazon Web Services
 

Similar a IBM Cloud Paris Meetup - 20190520 - IA & Power (20)

BSC LMS DDL
BSC LMS DDL BSC LMS DDL
BSC LMS DDL
 
How to Optimize Hortonworks Apache Spark ML Workloads on Modern Processors
How to Optimize Hortonworks Apache Spark ML Workloads on Modern Processors How to Optimize Hortonworks Apache Spark ML Workloads on Modern Processors
How to Optimize Hortonworks Apache Spark ML Workloads on Modern Processors
 
Backend.AI Technical Introduction (19.09 / 2019 Autumn)
Backend.AI Technical Introduction (19.09 / 2019 Autumn)Backend.AI Technical Introduction (19.09 / 2019 Autumn)
Backend.AI Technical Introduction (19.09 / 2019 Autumn)
 
AI in Health Care using IBM Systems/OpenPOWER systems
AI in Health Care using IBM Systems/OpenPOWER systemsAI in Health Care using IBM Systems/OpenPOWER systems
AI in Health Care using IBM Systems/OpenPOWER systems
 
AI in Healh Care using IBM POWER systems
AI in Healh Care using IBM POWER systems AI in Healh Care using IBM POWER systems
AI in Healh Care using IBM POWER systems
 
Harnessing the virtual realm for successful real world artificial intelligence
Harnessing the virtual realm for successful real world artificial intelligenceHarnessing the virtual realm for successful real world artificial intelligence
Harnessing the virtual realm for successful real world artificial intelligence
 
2018 bsc power9 and power ai
2018   bsc power9 and power ai 2018   bsc power9 and power ai
2018 bsc power9 and power ai
 
Deeplearningusingcloudpakfordata
DeeplearningusingcloudpakfordataDeeplearningusingcloudpakfordata
Deeplearningusingcloudpakfordata
 
Capi snap overview
Capi snap overviewCapi snap overview
Capi snap overview
 
What Can HPC on AWS Do?
What Can HPC on AWS Do?What Can HPC on AWS Do?
What Can HPC on AWS Do?
 
IBM Power Systems - enabling cloud solutions
IBM Power Systems - enabling cloud solutionsIBM Power Systems - enabling cloud solutions
IBM Power Systems - enabling cloud solutions
 
WML OpenPOWER presentation
WML OpenPOWER presentationWML OpenPOWER presentation
WML OpenPOWER presentation
 
IBM AI at Scale
IBM AI at ScaleIBM AI at Scale
IBM AI at Scale
 
Cisco connect montreal 2018 compute v final
Cisco connect montreal 2018   compute v finalCisco connect montreal 2018   compute v final
Cisco connect montreal 2018 compute v final
 
Sundar Ranganathan, NetApp + Vinod Iyengar, H2O.ai - Driverless AI integratio...
Sundar Ranganathan, NetApp + Vinod Iyengar, H2O.ai - Driverless AI integratio...Sundar Ranganathan, NetApp + Vinod Iyengar, H2O.ai - Driverless AI integratio...
Sundar Ranganathan, NetApp + Vinod Iyengar, H2O.ai - Driverless AI integratio...
 
MIKE BY DHI SAAS - New features as of October 2014
MIKE BY DHI SAAS - New features as of October 2014MIKE BY DHI SAAS - New features as of October 2014
MIKE BY DHI SAAS - New features as of October 2014
 
Amazon EC2 instances: Customizable cloud computing across workloads - DEM20-S...
Amazon EC2 instances: Customizable cloud computing across workloads - DEM20-S...Amazon EC2 instances: Customizable cloud computing across workloads - DEM20-S...
Amazon EC2 instances: Customizable cloud computing across workloads - DEM20-S...
 
SAP HANA on POWER9 systems
SAP HANA on POWER9 systemsSAP HANA on POWER9 systems
SAP HANA on POWER9 systems
 
Scalable Multi-Node Deep Learning Training in the Cloud (CMP368-R1) - AWS re:...
Scalable Multi-Node Deep Learning Training in the Cloud (CMP368-R1) - AWS re:...Scalable Multi-Node Deep Learning Training in the Cloud (CMP368-R1) - AWS re:...
Scalable Multi-Node Deep Learning Training in the Cloud (CMP368-R1) - AWS re:...
 
Deep Dive on Amazon EC2 Accelerated Computing - AWS Online Tech Talks
Deep Dive on Amazon EC2 Accelerated Computing - AWS Online Tech TalksDeep Dive on Amazon EC2 Accelerated Computing - AWS Online Tech Talks
Deep Dive on Amazon EC2 Accelerated Computing - AWS Online Tech Talks
 

Más de IBM France Lab

20200113 - IBM Cloud Côte d'Azur - DeepDive Kubernetes
20200113 - IBM Cloud Côte d'Azur - DeepDive Kubernetes20200113 - IBM Cloud Côte d'Azur - DeepDive Kubernetes
20200113 - IBM Cloud Côte d'Azur - DeepDive KubernetesIBM France Lab
 
20200114 - IBM Cloud Paris Meetup - DevOps
20200114 - IBM Cloud Paris Meetup - DevOps20200114 - IBM Cloud Paris Meetup - DevOps
20200114 - IBM Cloud Paris Meetup - DevOpsIBM France Lab
 
20200128 - Meetup Nice Côte d'Azur - Agile Mindset
20200128 - Meetup Nice Côte d'Azur - Agile Mindset20200128 - Meetup Nice Côte d'Azur - Agile Mindset
20200128 - Meetup Nice Côte d'Azur - Agile MindsetIBM France Lab
 
Défis de l'IA : droits, devoirs, enjeux économiques et éthiques
Défis de l'IA : droits, devoirs, enjeux économiques et éthiquesDéfis de l'IA : droits, devoirs, enjeux économiques et éthiques
Défis de l'IA : droits, devoirs, enjeux économiques et éthiquesIBM France Lab
 
Meetup ibm abakus banque postale
Meetup ibm abakus banque postaleMeetup ibm abakus banque postale
Meetup ibm abakus banque postaleIBM France Lab
 
20190613 - IBM Cloud Côte d'Azur meetup - "Cloud & Containers"
20190613 - IBM Cloud Côte d'Azur meetup - "Cloud & Containers"20190613 - IBM Cloud Côte d'Azur meetup - "Cloud & Containers"
20190613 - IBM Cloud Côte d'Azur meetup - "Cloud & Containers"IBM France Lab
 
20190613 - IBM Cloud Côte d'Azur meetup - "Cloud & Containers"
20190613 - IBM Cloud Côte d'Azur meetup - "Cloud & Containers"20190613 - IBM Cloud Côte d'Azur meetup - "Cloud & Containers"
20190613 - IBM Cloud Côte d'Azur meetup - "Cloud & Containers"IBM France Lab
 
IBM Watson IOT - Acoustic or Visual Insights
IBM Watson IOT - Acoustic or Visual InsightsIBM Watson IOT - Acoustic or Visual Insights
IBM Watson IOT - Acoustic or Visual InsightsIBM France Lab
 
Retour expérience Track & Trace - IBM using Sigfox.
Retour expérience Track & Trace - IBM using Sigfox.Retour expérience Track & Trace - IBM using Sigfox.
Retour expérience Track & Trace - IBM using Sigfox.IBM France Lab
 
20190520 - IBM Cloud Paris-Saclay Meetup - Hardis Group
20190520  - IBM Cloud Paris-Saclay Meetup - Hardis Group20190520  - IBM Cloud Paris-Saclay Meetup - Hardis Group
20190520 - IBM Cloud Paris-Saclay Meetup - Hardis GroupIBM France Lab
 
IBM Cloud Côte d'Azur Meetup - 20190328 - Optimisation
IBM Cloud Côte d'Azur Meetup - 20190328 - OptimisationIBM Cloud Côte d'Azur Meetup - 20190328 - Optimisation
IBM Cloud Côte d'Azur Meetup - 20190328 - OptimisationIBM France Lab
 
IBM Cloud Côte d'Azur Meetup - 20190328 - Optimisation
IBM Cloud Côte d'Azur Meetup - 20190328 - OptimisationIBM Cloud Côte d'Azur Meetup - 20190328 - Optimisation
IBM Cloud Côte d'Azur Meetup - 20190328 - OptimisationIBM France Lab
 
IBM Cloud Bordeaux Meetup - 20190325 - Software Factory
IBM Cloud Bordeaux Meetup - 20190325 - Software FactoryIBM Cloud Bordeaux Meetup - 20190325 - Software Factory
IBM Cloud Bordeaux Meetup - 20190325 - Software FactoryIBM France Lab
 
IBM Cloud Paris Meetup - 20190129 - Assima
IBM Cloud Paris Meetup - 20190129 - AssimaIBM Cloud Paris Meetup - 20190129 - Assima
IBM Cloud Paris Meetup - 20190129 - AssimaIBM France Lab
 
IBM Cloud Paris Meetup - 20190129 - Myrtea
IBM Cloud Paris Meetup - 20190129 - MyrteaIBM Cloud Paris Meetup - 20190129 - Myrtea
IBM Cloud Paris Meetup - 20190129 - MyrteaIBM France Lab
 
IBM Cloud Paris Meetup - 20181016 - L'agilité à l'échelle
IBM Cloud Paris Meetup - 20181016 - L'agilité à l'échelleIBM Cloud Paris Meetup - 20181016 - L'agilité à l'échelle
IBM Cloud Paris Meetup - 20181016 - L'agilité à l'échelleIBM France Lab
 
IBM Cloud Côte d'Azur Meetup - Blockchain Business Processes & Rule-based Sm...
IBM Cloud Côte d'Azur Meetup - Blockchain Business Processes &  Rule-based Sm...IBM Cloud Côte d'Azur Meetup - Blockchain Business Processes &  Rule-based Sm...
IBM Cloud Côte d'Azur Meetup - Blockchain Business Processes & Rule-based Sm...IBM France Lab
 
IBM Cloud Côte D'Azur Meetup - 20181004 - Blockchain Hyperledger Workshop
IBM Cloud Côte D'Azur Meetup - 20181004 - Blockchain Hyperledger WorkshopIBM Cloud Côte D'Azur Meetup - 20181004 - Blockchain Hyperledger Workshop
IBM Cloud Côte D'Azur Meetup - 20181004 - Blockchain Hyperledger WorkshopIBM France Lab
 
IBM Cloud Paris Meetup - 20180911 - Common Ledger for Public Administration
IBM Cloud Paris Meetup - 20180911 - Common Ledger for Public AdministrationIBM Cloud Paris Meetup - 20180911 - Common Ledger for Public Administration
IBM Cloud Paris Meetup - 20180911 - Common Ledger for Public AdministrationIBM France Lab
 
IBM Cloud Paris Meetup - 20180911 - Smart Citizen Bot
IBM Cloud Paris Meetup - 20180911 - Smart Citizen BotIBM Cloud Paris Meetup - 20180911 - Smart Citizen Bot
IBM Cloud Paris Meetup - 20180911 - Smart Citizen BotIBM France Lab
 

Más de IBM France Lab (20)

20200113 - IBM Cloud Côte d'Azur - DeepDive Kubernetes
20200113 - IBM Cloud Côte d'Azur - DeepDive Kubernetes20200113 - IBM Cloud Côte d'Azur - DeepDive Kubernetes
20200113 - IBM Cloud Côte d'Azur - DeepDive Kubernetes
 
20200114 - IBM Cloud Paris Meetup - DevOps
20200114 - IBM Cloud Paris Meetup - DevOps20200114 - IBM Cloud Paris Meetup - DevOps
20200114 - IBM Cloud Paris Meetup - DevOps
 
20200128 - Meetup Nice Côte d'Azur - Agile Mindset
20200128 - Meetup Nice Côte d'Azur - Agile Mindset20200128 - Meetup Nice Côte d'Azur - Agile Mindset
20200128 - Meetup Nice Côte d'Azur - Agile Mindset
 
Défis de l'IA : droits, devoirs, enjeux économiques et éthiques
Défis de l'IA : droits, devoirs, enjeux économiques et éthiquesDéfis de l'IA : droits, devoirs, enjeux économiques et éthiques
Défis de l'IA : droits, devoirs, enjeux économiques et éthiques
 
Meetup ibm abakus banque postale
Meetup ibm abakus banque postaleMeetup ibm abakus banque postale
Meetup ibm abakus banque postale
 
20190613 - IBM Cloud Côte d'Azur meetup - "Cloud & Containers"
20190613 - IBM Cloud Côte d'Azur meetup - "Cloud & Containers"20190613 - IBM Cloud Côte d'Azur meetup - "Cloud & Containers"
20190613 - IBM Cloud Côte d'Azur meetup - "Cloud & Containers"
 
20190613 - IBM Cloud Côte d'Azur meetup - "Cloud & Containers"
20190613 - IBM Cloud Côte d'Azur meetup - "Cloud & Containers"20190613 - IBM Cloud Côte d'Azur meetup - "Cloud & Containers"
20190613 - IBM Cloud Côte d'Azur meetup - "Cloud & Containers"
 
IBM Watson IOT - Acoustic or Visual Insights
IBM Watson IOT - Acoustic or Visual InsightsIBM Watson IOT - Acoustic or Visual Insights
IBM Watson IOT - Acoustic or Visual Insights
 
Retour expérience Track & Trace - IBM using Sigfox.
Retour expérience Track & Trace - IBM using Sigfox.Retour expérience Track & Trace - IBM using Sigfox.
Retour expérience Track & Trace - IBM using Sigfox.
 
20190520 - IBM Cloud Paris-Saclay Meetup - Hardis Group
20190520  - IBM Cloud Paris-Saclay Meetup - Hardis Group20190520  - IBM Cloud Paris-Saclay Meetup - Hardis Group
20190520 - IBM Cloud Paris-Saclay Meetup - Hardis Group
 
IBM Cloud Côte d'Azur Meetup - 20190328 - Optimisation
IBM Cloud Côte d'Azur Meetup - 20190328 - OptimisationIBM Cloud Côte d'Azur Meetup - 20190328 - Optimisation
IBM Cloud Côte d'Azur Meetup - 20190328 - Optimisation
 
IBM Cloud Côte d'Azur Meetup - 20190328 - Optimisation
IBM Cloud Côte d'Azur Meetup - 20190328 - OptimisationIBM Cloud Côte d'Azur Meetup - 20190328 - Optimisation
IBM Cloud Côte d'Azur Meetup - 20190328 - Optimisation
 
IBM Cloud Bordeaux Meetup - 20190325 - Software Factory
IBM Cloud Bordeaux Meetup - 20190325 - Software FactoryIBM Cloud Bordeaux Meetup - 20190325 - Software Factory
IBM Cloud Bordeaux Meetup - 20190325 - Software Factory
 
IBM Cloud Paris Meetup - 20190129 - Assima
IBM Cloud Paris Meetup - 20190129 - AssimaIBM Cloud Paris Meetup - 20190129 - Assima
IBM Cloud Paris Meetup - 20190129 - Assima
 
IBM Cloud Paris Meetup - 20190129 - Myrtea
IBM Cloud Paris Meetup - 20190129 - MyrteaIBM Cloud Paris Meetup - 20190129 - Myrtea
IBM Cloud Paris Meetup - 20190129 - Myrtea
 
IBM Cloud Paris Meetup - 20181016 - L'agilité à l'échelle
IBM Cloud Paris Meetup - 20181016 - L'agilité à l'échelleIBM Cloud Paris Meetup - 20181016 - L'agilité à l'échelle
IBM Cloud Paris Meetup - 20181016 - L'agilité à l'échelle
 
IBM Cloud Côte d'Azur Meetup - Blockchain Business Processes & Rule-based Sm...
IBM Cloud Côte d'Azur Meetup - Blockchain Business Processes &  Rule-based Sm...IBM Cloud Côte d'Azur Meetup - Blockchain Business Processes &  Rule-based Sm...
IBM Cloud Côte d'Azur Meetup - Blockchain Business Processes & Rule-based Sm...
 
IBM Cloud Côte D'Azur Meetup - 20181004 - Blockchain Hyperledger Workshop
IBM Cloud Côte D'Azur Meetup - 20181004 - Blockchain Hyperledger WorkshopIBM Cloud Côte D'Azur Meetup - 20181004 - Blockchain Hyperledger Workshop
IBM Cloud Côte D'Azur Meetup - 20181004 - Blockchain Hyperledger Workshop
 
IBM Cloud Paris Meetup - 20180911 - Common Ledger for Public Administration
IBM Cloud Paris Meetup - 20180911 - Common Ledger for Public AdministrationIBM Cloud Paris Meetup - 20180911 - Common Ledger for Public Administration
IBM Cloud Paris Meetup - 20180911 - Common Ledger for Public Administration
 
IBM Cloud Paris Meetup - 20180911 - Smart Citizen Bot
IBM Cloud Paris Meetup - 20180911 - Smart Citizen BotIBM Cloud Paris Meetup - 20180911 - Smart Citizen Bot
IBM Cloud Paris Meetup - 20180911 - Smart Citizen Bot
 

Último

08448380779 Call Girls In Civil Lines Women Seeking Men
08448380779 Call Girls In Civil Lines Women Seeking Men08448380779 Call Girls In Civil Lines Women Seeking Men
08448380779 Call Girls In Civil Lines Women Seeking MenDelhi Call girls
 
The Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdf
The Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdfThe Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdf
The Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdfEnterprise Knowledge
 
Real Time Object Detection Using Open CV
Real Time Object Detection Using Open CVReal Time Object Detection Using Open CV
Real Time Object Detection Using Open CVKhem
 
Automating Google Workspace (GWS) & more with Apps Script
Automating Google Workspace (GWS) & more with Apps ScriptAutomating Google Workspace (GWS) & more with Apps Script
Automating Google Workspace (GWS) & more with Apps Scriptwesley chun
 
Powerful Google developer tools for immediate impact! (2023-24 C)
Powerful Google developer tools for immediate impact! (2023-24 C)Powerful Google developer tools for immediate impact! (2023-24 C)
Powerful Google developer tools for immediate impact! (2023-24 C)wesley chun
 
A Domino Admins Adventures (Engage 2024)
A Domino Admins Adventures (Engage 2024)A Domino Admins Adventures (Engage 2024)
A Domino Admins Adventures (Engage 2024)Gabriella Davis
 
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...apidays
 
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...Drew Madelung
 
A Year of the Servo Reboot: Where Are We Now?
A Year of the Servo Reboot: Where Are We Now?A Year of the Servo Reboot: Where Are We Now?
A Year of the Servo Reboot: Where Are We Now?Igalia
 
Histor y of HAM Radio presentation slide
Histor y of HAM Radio presentation slideHistor y of HAM Radio presentation slide
Histor y of HAM Radio presentation slidevu2urc
 
GenCyber Cyber Security Day Presentation
GenCyber Cyber Security Day PresentationGenCyber Cyber Security Day Presentation
GenCyber Cyber Security Day PresentationMichael W. Hawkins
 
What Are The Drone Anti-jamming Systems Technology?
What Are The Drone Anti-jamming Systems Technology?What Are The Drone Anti-jamming Systems Technology?
What Are The Drone Anti-jamming Systems Technology?Antenna Manufacturer Coco
 
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024The Digital Insurer
 
Presentation on how to chat with PDF using ChatGPT code interpreter
Presentation on how to chat with PDF using ChatGPT code interpreterPresentation on how to chat with PDF using ChatGPT code interpreter
Presentation on how to chat with PDF using ChatGPT code interpreternaman860154
 
Axa Assurance Maroc - Insurer Innovation Award 2024
Axa Assurance Maroc - Insurer Innovation Award 2024Axa Assurance Maroc - Insurer Innovation Award 2024
Axa Assurance Maroc - Insurer Innovation Award 2024The Digital Insurer
 
EIS-Webinar-Prompt-Knowledge-Eng-2024-04-08.pptx
EIS-Webinar-Prompt-Knowledge-Eng-2024-04-08.pptxEIS-Webinar-Prompt-Knowledge-Eng-2024-04-08.pptx
EIS-Webinar-Prompt-Knowledge-Eng-2024-04-08.pptxEarley Information Science
 
Driving Behavioral Change for Information Management through Data-Driven Gree...
Driving Behavioral Change for Information Management through Data-Driven Gree...Driving Behavioral Change for Information Management through Data-Driven Gree...
Driving Behavioral Change for Information Management through Data-Driven Gree...Enterprise Knowledge
 
Advantages of Hiring UIUX Design Service Providers for Your Business
Advantages of Hiring UIUX Design Service Providers for Your BusinessAdvantages of Hiring UIUX Design Service Providers for Your Business
Advantages of Hiring UIUX Design Service Providers for Your BusinessPixlogix Infotech
 
Factors to Consider When Choosing Accounts Payable Services Providers.pptx
Factors to Consider When Choosing Accounts Payable Services Providers.pptxFactors to Consider When Choosing Accounts Payable Services Providers.pptx
Factors to Consider When Choosing Accounts Payable Services Providers.pptxKatpro Technologies
 
A Call to Action for Generative AI in 2024
A Call to Action for Generative AI in 2024A Call to Action for Generative AI in 2024
A Call to Action for Generative AI in 2024Results
 

Último (20)

08448380779 Call Girls In Civil Lines Women Seeking Men
08448380779 Call Girls In Civil Lines Women Seeking Men08448380779 Call Girls In Civil Lines Women Seeking Men
08448380779 Call Girls In Civil Lines Women Seeking Men
 
The Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdf
The Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdfThe Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdf
The Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdf
 
Real Time Object Detection Using Open CV
Real Time Object Detection Using Open CVReal Time Object Detection Using Open CV
Real Time Object Detection Using Open CV
 
Automating Google Workspace (GWS) & more with Apps Script
Automating Google Workspace (GWS) & more with Apps ScriptAutomating Google Workspace (GWS) & more with Apps Script
Automating Google Workspace (GWS) & more with Apps Script
 
Powerful Google developer tools for immediate impact! (2023-24 C)
Powerful Google developer tools for immediate impact! (2023-24 C)Powerful Google developer tools for immediate impact! (2023-24 C)
Powerful Google developer tools for immediate impact! (2023-24 C)
 
A Domino Admins Adventures (Engage 2024)
A Domino Admins Adventures (Engage 2024)A Domino Admins Adventures (Engage 2024)
A Domino Admins Adventures (Engage 2024)
 
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
 
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
 
A Year of the Servo Reboot: Where Are We Now?
A Year of the Servo Reboot: Where Are We Now?A Year of the Servo Reboot: Where Are We Now?
A Year of the Servo Reboot: Where Are We Now?
 
Histor y of HAM Radio presentation slide
Histor y of HAM Radio presentation slideHistor y of HAM Radio presentation slide
Histor y of HAM Radio presentation slide
 
GenCyber Cyber Security Day Presentation
GenCyber Cyber Security Day PresentationGenCyber Cyber Security Day Presentation
GenCyber Cyber Security Day Presentation
 
What Are The Drone Anti-jamming Systems Technology?
What Are The Drone Anti-jamming Systems Technology?What Are The Drone Anti-jamming Systems Technology?
What Are The Drone Anti-jamming Systems Technology?
 
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
 
Presentation on how to chat with PDF using ChatGPT code interpreter
Presentation on how to chat with PDF using ChatGPT code interpreterPresentation on how to chat with PDF using ChatGPT code interpreter
Presentation on how to chat with PDF using ChatGPT code interpreter
 
Axa Assurance Maroc - Insurer Innovation Award 2024
Axa Assurance Maroc - Insurer Innovation Award 2024Axa Assurance Maroc - Insurer Innovation Award 2024
Axa Assurance Maroc - Insurer Innovation Award 2024
 
EIS-Webinar-Prompt-Knowledge-Eng-2024-04-08.pptx
EIS-Webinar-Prompt-Knowledge-Eng-2024-04-08.pptxEIS-Webinar-Prompt-Knowledge-Eng-2024-04-08.pptx
EIS-Webinar-Prompt-Knowledge-Eng-2024-04-08.pptx
 
Driving Behavioral Change for Information Management through Data-Driven Gree...
Driving Behavioral Change for Information Management through Data-Driven Gree...Driving Behavioral Change for Information Management through Data-Driven Gree...
Driving Behavioral Change for Information Management through Data-Driven Gree...
 
Advantages of Hiring UIUX Design Service Providers for Your Business
Advantages of Hiring UIUX Design Service Providers for Your BusinessAdvantages of Hiring UIUX Design Service Providers for Your Business
Advantages of Hiring UIUX Design Service Providers for Your Business
 
Factors to Consider When Choosing Accounts Payable Services Providers.pptx
Factors to Consider When Choosing Accounts Payable Services Providers.pptxFactors to Consider When Choosing Accounts Payable Services Providers.pptx
Factors to Consider When Choosing Accounts Payable Services Providers.pptx
 
A Call to Action for Generative AI in 2024
A Call to Action for Generative AI in 2024A Call to Action for Generative AI in 2024
A Call to Action for Generative AI in 2024
 

IBM Cloud Paris Meetup - 20190520 - IA & Power

  • 1. Meetup "IA à portée de la main » 20 mai 2019 – IBM France Lab@Paris-Saclay
  • 2. Deep Learning ? © 2019 IBM Corporation Classification Counting Detection { "webAPIId":"8ace1c11-0f7e-4a17-a54d-63c741c7b31d", "imageUrl":"http://powerai-vision-portal:9080/powerai-vision- api/uploads/temp/8ace1c11-0f7e-4a17-a54d-63c741c7b31d/3f3db18c-eb02-4dca-8015- 8a68a830eb5b.jpg", "imageMd5":"c13e24226dbd67404a4b464f18a231ac", "classified":[ { "confidence":0.9997112154960632, "label":"up", "ymax":608,"xmax":600,"xmin":503,"ymin":430 }, { "confidence":0.9993865489959717, "label":"up", "ymax":543, "xmax":949,"xmin":837,"ymin":372 } ], "result":"success" }
  • 3. Le véritable objectif © 2019 IBM Corporation
  • 4. Cycle de vie d’une application cognitive © 2019 IBM Corporation Idée / ROI Application Services Développement Déploiement Utilisation
  • 5. © 2019 IBM Corporation Suivi du parcours client SERVER PERFORMANCE RETAIL E-COMMERCE Suivi des clients d’un supermarché
  • 6. © 2019 IBM Corporation Surveillance des tunnels d’autoroute SERVER PERFORMANCE Détection des incidents
  • 7. Application © 2019 IBM Corporation IBM Cloud Private DevOps Microservices
  • 8. Services © 2019 IBM Corporation Deux approches Prêt-à-porter IBM One AI Sur-mesure
  • 9. Watson © 2019 IBM Corporation Discovery Knowledge Studio Knowledge Catalog Discovery Conversation Assistant Empathy Personality Insights Tone Analyser Language Language Understanding Language Classifier Language Translator Voice Speech to Text Text to Speech Perception Visual Recognition
  • 10. IBM One AI © 2019 IBM Corporation Data Scientist App Developer AI Ops Build AI Run AI Operate AI Watson OpenScale Fairness & Explainability Inputs for Continuous Evolution Business KPIs and production metrics Watson Studio Watson Machine Learning Build Deploy and run Operate trusted AI Business user Consume AI Data Exploration Data Preparation Model Development Model Deployment Model Management Retraining Watson Knowledge Catalog Data Profiling Quality and Lineage Data Governance Organize and Govern data Data Engineer Organize Data for AI Watson Machine Learning AcceleratorWatson Machine Learning Accelerator
  • 11. Power AI © 2019 IBM Corporation Les frameworks opensource habituels optimisés Plus de capacités, plus de possibilités, plus de scalabilité Des serveurs conçus pour le calcul accéléré 5.6x plus de bande passante qu’avec PCIe gen3 Solutions de partage de ressources Des solutions pour aider à la conception et la mise à disposition de vos services d’IA AC922 Watson Machine Learning Accelerator
  • 12. Surveiller l’usage © 2019 IBM Corporation Watson OpenScale
  • 13. Partenariats © 2019 IBM Corporation Our clients H2O.ai IBM IVAPlanetANEO OKTAL SE Cyclope.ai
  • 14. Aide à la réflexion sur votre stratégie d’IA Preuves de concepts Collaboration avec des partenaires Accès à des experts HPC / IA Planification d’infrastructure IA © 2019 IBM Corporation Ne restez pas tout seuls !
  • 15. Partenaire © 2019 IBM Corporation
  • 16. © 2019 IBM Corporation Power AI Vision
  • 17. Démo © 2019 IBM Corporation
  • 18. Une stack logicielle basée sur l’opensource © 2019 IBM Corporation Watson Machine Learning (WML) PowerAI Watson ML Watson ML AcceleratorWatson Studio AI OpenScale Development Environment Train Models Runtime Environment Train, Deploy, & Manage Models Operation Fabric Monitor & Improve Deployed Models SnapML Previous Names: WML Accelerator = PowerAI Enterprise GPU-Accelerated Power Servers Storage PowerAI
  • 19. Des besoins de plus en plus « gros » © 2019 IBM Corporation Volume of data generated by a single car = 1 TB+ / h
  • 20. Exemples de datasets de haute résolution © 2019 IBM Corporation •CityScapes : 1024x2048 5k images. Semantic Segmentation, Benchmark results. •Mapillary Vistas : various sizes - split into 18,000 for training, 2,000 for validation, 5,000 for testing (average image resolution of ~9 megapixels) •Synthia : Large volume of data & ground truth: +200,000 HD images from video streams and +20,000 HD images from independent snapshots
  • 21. Impact de la résolution sur la précision des modèles © 2019 IBM Corporation
  • 22. La solution d’IBM à la limitation de la mémoire des GPUs © 2019 IBM Corporation Large Model Support POWER CPU DDR4 GPU NVLink Graphics Memory Traditional Model Support CPUDDR4 GPU PCIe Graphics Memory Limited memory on GPU forces trade-off in model size / data resolution Use system memory and GPU to support more complex and higher resolution data 32GB max 2TB max 2TB max 32GB max • Leveraging NVLink and CPU-GPU memory coherence enables larger and more complex models • Improves model accuracy with more images and higher resolution images
  • 23. Large Memory Support (LMS) © 2019 IBM Corporation LMS Concept 1. Treat GPU memory as an application-level cache w.r.t the host main memory. 2. All data reside on host memory and are copied to GPU memory only when needed. 3. After the GPU memory is used, depending on whether it has been modified or has future usage, it can be either copied back to the system memory or simply discarded. 4. To efficiently utilize GPU memory, LMS implementation keeps a large memory pool so that different memory pieces from CPU memory can share the same GPU memory chunk. Therefore, at any moment, the GPU only holds data necessary to process one operation, for example, the forward propagation of one operation in a neural network (Feature maps). If the memory requirement from any operation is larger than the GPU memory, then even LMS will fail as well. In theory, LMS should be able to handle a deep neural network of an arbitrary capacity, as long as all the data from the largest operation can fit into the GPU memory. à As LMS uses system memory to store features map tensors, the training performance limiting factor is no longer the PCIe bus but the CPU-RAM high-bandwidth. Ex: Feature Maps tensor backward reuse for the gradient update with AlexNet Without LMS: forward temporary data are kept in GPU memory for the backward computation
  • 24. LMS: Caffe © 2019 IBM Corporation Caffe is one of the most popular open-source deep learning frameworks, developed by the Berke- ley AI Research. It clearly separates the network description (made with a Protocol Buffer file, similar to JSON), and the training run, where we have a second file to specify the parameters, how we optimize the network, what will be logged. This is called "define-AND-run", and most of the frame-works work that way. In addition, it provides a Python API to write for example new layer types, which is useful for loading complex data and feeding it to the network. Caffe LMS Command line Flags: -lms : to activate LMS for the current training.
  • 25. LMS: Chainer © 2019 IBM Corporation Chainer Chainer is a lesser-known open-source deep learning framework in Python, popular in Japan but not really beyond. It is developed by a Japanese com- pany called Preferred Networks, in partnership with several companies including IBM. Unlike Caffe, it has a "define-BY-run" approach, where network description and training are "mixed" in the same code. This is more flexible and allows to dynamically update the network if needed (although it is useful on very specific deep learning algorithms I haven’t used during the internship). Light python code change for LMS activation : with chainer.out_of_core_mode(fine_granularity=True, async=False, devices=gpu): trainer.run()
  • 26. LMS: Tensorflow © 2019 IBM Corporation Keras API from tensorflow.contrib.lms import LMSKerasCallback lms_callback = LMSKerasCallback() … model.fit_generator(generator=training_gen, callbacks=[lms_callback]) Estimator API with tf.name_scope(‘gradientscope'): optimizer = tf.train.GradientDescentOptimizer(learning_rate=0.001) train_op = optimizer.minimize(loss=loss, global_step=tf.train.get_global_step()) from tensorflow.contrib.lms import LMSHook lms_hook = LMSHook({‘gradientscope '}) mnist_classifier.train(input_fn=train_input_fn, steps=20000, hooks=[logging_hook, lms_hook])
  • 27. IBM AC922 4 GPUs @150GB/s CPU ßà GPU bandwidth 6 GPUs @100GB/s CPU ßà GPU bandwidth Coherent access to system memory PCIe Gen 4 and CAPI 2.0 to InfiniBand Air and Water cooled options Coherent access to system memory PCIe Gen 4 and CAPI 2.0 to InfiniBand Water cooled only NVLink 100GB/s NVLink 100GB/s NVDIA V100 Coherent access to system memory (2TB) NVLink 100GB/s NVLink 100GB/s NVLink 100GB/s 170GB/s CPU PCIe Gen 4 CAPI 2.0 NVDIA V100NVDIA V100 DDR4 I B Coherent access to system memory (2TB) NVLink 150GB/s NVLink 150GB/s 170GB/s CP U PCIe Gen 4 CAPI 2.0 NVLink 150GB/s NVDIA V100NVDIA V100 DDR4 IB © 2019 IBM Corporation
  • 28. AC922: Des capacités uniques pour l’IA 28 Large Model Support Only server that can leverage System Memory from GPU side for 2TB+ per node! Distributed Deep Learning Most efficient scaling, due to better I/O between nodes (95% across 256 GPUS) 3.1 Hours 49 Mins 0 2000 4000 6000 8000 10000 12000 Xeon x86 2640v4 w/ 4x V100 GPUs Power AC922 w/ 4x V100 GPUs Time(secs) Caffe with LMS (Large Model Support) Runtime of 1000 Iterations GoogleNet model on Enlarged ImageNet Dataset (2240x2240) Near Ideal Scaling to 256 GPUs 1 2 4 8 16 32 64 128 256 4 16 64 256 Speedup Number of GPUs Ideal Scaling DDL Actual Scaling Caffe with PowerAI DDL, Running on S822LC Power System ResNet-50, ImageNet-1K ~4x faster insights Train More Build More Know More © 2019 IBM Corporation
  • 29. LMS: impact du NVLink entre CPU et GPU © 2019 IBM Corporation NVLink 2.0 connected GPU training one high res 3D MRI with large model support
  • 30. LMS: Impact sur le temps d’une epoch © 2019 IBM Corporation 0 250 500 750 1000 1250 1500 1750 2000 2250 AC922 with NVLink 2.0 GPU server with PCI GPU server with PCI contention Seconds Epoch times at max resolution with swapping
  • 31. LMS: Impact sur le taux d’utilisation des GPUs © 2019 IBM Corporation GB/s 10 GB/s 20 GB/s 30 GB/s 40 GB/s 50 GB/s 60 GB/s 70 GB/s 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% AC922 with NVLink 2.0 GPU server with PCI GPU server with PCI contention MemoryCopyThroughput GPUUtilization GPU utilization Memory copy throughput
  • 32. LMS: Overhead par rapport à la mémoire du GPU © 2019 IBM Corporation 3% 13% 25% 0% 20% 40% 60% 80% 100% 0 100 200 300 400 500 600 700 1.4x 1.8x 2.4x OverheadPercentage Epochtimeinseconds Swapping overhead using resolutions above the 16GB GPU memory limit 32 GB V100 GPU, no swapping 16GB V100 GPU, swapping Overhead
  • 33. Une architecture évolutive © 2019 IBM Corporation Experimentation Single Tenant Stabilization & Production Secure Multitenant Expansion Enterprise Scale / Multiple Lines of Business Data Scientist’s workstations Internal SAS drives & NVM’s IBM Power Systems AC922 High-Speed Network Subsystem Existing Organization Infrastructure IBM Elastic Storage Server (ESS) Training & Inference Cluster IBM Power Systems AC922, LC921 & LC922 Master & Failover Master Nodes IBM Power Systems LC921 & LC922 Login Nodes IBM Power Systems LC921 & LC922 Training Cluster IBM Power Systems AC922 IBM Elastic Storage Server (ESS) High-Speed Network Subsystem Existing Organization Infrastructure One software stack from experimentation to expansion IBM PowerAI Enterprise Red Hat Enterprise Linux (RHEL) IBM Power System & x86 Servers Services& Support IBM Spectrum Scale / IBM Elastic Storage Server (ESS)
  • 34. Aide à la réflexion sur votre stratégie d’IA Preuves de concepts Collaboration avec des partenaires Accès à des experts HPC / IA Planification d’infrastructure IA © 2019 IBM Corporation Ne restez pas tout seuls !
  • 35. Merci ! IBM AI Portfolio Dédiaboliser les projets d’IA…