SlideShare a Scribd company logo
MuSe2023 from practical view
Ganghyun Kim
2023 / 08 / 18
2/30
3/30
MuSe2023 overview
• Goal: Predict subject’s emotional state using multimodal data
• Subject is under stressed condition (simulated job interview)
• Emotional state can be represented in 2 axis, Arousal and Valence
4/30
https://www.mdpi.com/1424-8220/23/3/1598
Input data (feature)
• Audio: DeepSpectrum, eGeMAPS, Wav2vec2.0
• Video: ViT, FaceNet, FAU, Pose
• Other: BERT, Biosignals
• All of them tabular data extracted with corresponding feature extractor.
5/30
6/30
2022
7/30
2023
8/30
Big Titan server: Titan Xp X2
Small Titan server: Titan Xp X2
Student server 1: RTX 1080 X2
Student server 2: RTX2080 X2
Student server 3: A5000 X2
TA server: RTX3090 X3
Code Management
6F
6F
IGC IT CENTER
IGC IT CENTER
IGC IT CENTER
IGC IT CENTER
Execute commands via
6F
Tmux
11/30
Termius
12/30
Termius
13/30
Wandb Sweep
Generates HP tuning task list
Start signal
Check remaining task,
assign 1 HP combination run to server,
repeat when server finishes
Wandb Sweep
• wandb sweep sweep.yaml # returns your SWEEP_ID
• wandb agent mcg70107
• Done!
15/30
Previously..
• PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=python CUDA_VISIBLE_DEVICES=1 python main_rnn.py --cache --task stress --emo_dim valence --d_rnn 64 --rnn_n_layers 2 --
lr 0.005 --feature mfcc_combined --win_len 200 --rnn_bi --n_seeds 20 --seed 101 --batch_size 256 --early_stopping_patience 15 --reduce_lr_patience 5 --project_name
Muse_20221026_early_fusion_RNN_valence_mfcc_combined --use_gpu
• PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=python CUDA_VISIBLE_DEVICES=1 python main_rnn.py --cache --task stress --emo_dim valence --d_rnn 64 --rnn_n_layers 2 --
lr 0.005 --feature mfcc_combined --win_len 300 --rnn_bi --n_seeds 20 --seed 101 --batch_size 256 --early_stopping_patience 15 --reduce_lr_patience 5 --project_name
Muse_20221026_early_fusion_RNN_valence_mfcc_combined --use_gpu
• PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=python CUDA_VISIBLE_DEVICES=1 python main_rnn.py --cache --task stress --emo_dim valence --d_rnn 64 --rnn_n_layers 2 --
lr 0.005 --feature mfcc_combined --win_len 400 --rnn_bi --n_seeds 20 --seed 101 --batch_size 256 --early_stopping_patience 15 --reduce_lr_patience 5 --project_name
Muse_20221026_early_fusion_RNN_valence_mfcc_combined --use_gpu
• PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=python CUDA_VISIBLE_DEVICES=1 python main_rnn.py --cache --task stress --emo_dim valence --d_rnn 64 --rnn_n_layers 4 --
lr 0.001 --feature mfcc_combined --win_len 200 --rnn_bi --n_seeds 20 --seed 101 --batch_size 256 --early_stopping_patience 15 --reduce_lr_patience 5 --project_name
Muse_20221026_early_fusion_RNN_valence_mfcc_combined --use_gpu
• PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=python CUDA_VISIBLE_DEVICES=1 python main_rnn.py --cache --task stress --emo_dim valence --d_rnn 64 --rnn_n_layers 4 --
lr 0.001 --feature mfcc_combined --win_len 300 --rnn_bi --n_seeds 20 --seed 101 --batch_size 256 --early_stopping_patience 15 --reduce_lr_patience 5 --project_name
Muse_20221026_early_fusion_RNN_valence_mfcc_combined --use_gpu
• PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=python CUDA_VISIBLE_DEVICES=1 python main_rnn.py --cache --task stress --emo_dim valence --d_rnn 64 --rnn_n_layers 4 --
lr 0.001 --feature mfcc_combined --win_len 400 --rnn_bi --n_seeds 20 --seed 101 --batch_size 256 --early_stopping_patience 15 --reduce_lr_patience 5 --project_name
Muse_20221026_early_fusion_RNN_valence_mfcc_combined --use_gpu
• PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=python CUDA_VISIBLE_DEVICES=1 python main_rnn.py --cache --task stress --emo_dim valence --d_rnn 64 --rnn_n_layers 4 --
lr 0.002 --feature mfcc_combined --win_len 200 --rnn_bi --n_seeds 20 --seed 101 --batch_size 256 --early_stopping_patience 15 --reduce_lr_patience 5 --project_name
Muse_20221026_early_fusion_RNN_valence_mfcc_combined --use_gpu
• PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=python CUDA_VISIBLE_DEVICES=1 python main_rnn.py --cache --task stress --emo_dim valence --d_rnn 64 --rnn_n_layers 4 --
lr 0.002 --feature mfcc_combined --win_len 300 --rnn_bi --n_seeds 20 --seed 101 --batch_size 256 --early_stopping_patience 15 --reduce_lr_patience 5 --project_name
Muse_20221026_early_fusion_RNN_valence_mfcc_combined --use_gpu
• PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=python CUDA_VISIBLE_DEVICES=1 python main_rnn.py --cache --task stress --emo_dim valence --d_rnn 64 --rnn_n_layers 4 --
lr 0.002 --feature mfcc_combined --win_len 400 --rnn_bi --n_seeds 20 --seed 101 --batch_size 256 --early_stopping_patience 15 --reduce_lr_patience 5 --project_name
Muse_20221026_early_fusion_RNN_valence_mfcc_combined --use_gpu
16/30
Previously..
• Command generator writes combinations of HP tuning runs
• Split them in .sh file, in appropriate length considering GPU power of server, # of GPU
• Upload .sh file to each server, run them
Problem
• Hard to restart from crashed run. Needed to locate which run crashed, edit .sh file to start from
there
• Tedious work, considering number of experiments I had to run
17/30
Previously..
19/30
20/30
Parallel coordinates plot
21/30
22/30
23/30
Result
24/30
Result
25/30

More Related Content

Similar to MuSe2023.pptx

jRecruiter - The AJUG Job Posting Service
jRecruiter - The AJUG Job Posting ServicejRecruiter - The AJUG Job Posting Service
jRecruiter - The AJUG Job Posting Service
Gunnar Hillert
 
Spring Boot Actuator 2.0 & Micrometer
Spring Boot Actuator 2.0 & MicrometerSpring Boot Actuator 2.0 & Micrometer
Spring Boot Actuator 2.0 & Micrometer
Toshiaki Maki
 
OSMC 2015: Linux Performance Profiling and Monitoring by Werner Fischer
OSMC 2015: Linux Performance Profiling and Monitoring by Werner FischerOSMC 2015: Linux Performance Profiling and Monitoring by Werner Fischer
OSMC 2015: Linux Performance Profiling and Monitoring by Werner Fischer
NETWAYS
 
OSMC 2015 | Linux Performance Profiling and Monitoring by Werner Fischer
OSMC 2015 | Linux Performance Profiling and Monitoring by Werner FischerOSMC 2015 | Linux Performance Profiling and Monitoring by Werner Fischer
OSMC 2015 | Linux Performance Profiling and Monitoring by Werner Fischer
NETWAYS
 
Verified AZ-104 Exam Dumps (V26.02) - Pass Microsoft AZ-104 Exam (2024)
Verified AZ-104 Exam Dumps (V26.02) - Pass Microsoft AZ-104 Exam (2024)Verified AZ-104 Exam Dumps (V26.02) - Pass Microsoft AZ-104 Exam (2024)
Verified AZ-104 Exam Dumps (V26.02) - Pass Microsoft AZ-104 Exam (2024)
yarusun
 
BPF: Tracing and more
BPF: Tracing and moreBPF: Tracing and more
BPF: Tracing and more
Brendan Gregg
 
NVIDIA深度學習教育機構 (DLI): Object detection with jetson
NVIDIA深度學習教育機構 (DLI): Object detection with jetsonNVIDIA深度學習教育機構 (DLI): Object detection with jetson
NVIDIA深度學習教育機構 (DLI): Object detection with jetson
NVIDIA Taiwan
 
MCSoC'13 Keynote Talk "Taming Big Data Streams"
MCSoC'13 Keynote Talk "Taming Big Data Streams"MCSoC'13 Keynote Talk "Taming Big Data Streams"
MCSoC'13 Keynote Talk "Taming Big Data Streams"
Hideyuki Kawashima
 
Io 120404052713-phpapp01
Io 120404052713-phpapp01Io 120404052713-phpapp01
Io 120404052713-phpapp01
coxchen
 
了解IO协议栈
了解IO协议栈了解IO协议栈
了解IO协议栈
Feng Yu
 
Msc sylabus ptu
Msc sylabus ptuMsc sylabus ptu
Msc sylabus ptu
James West
 
BRKRST-3066 - Troubleshooting Nexus 7000 (2013 Melbourne) - 2 Hours.pdf
BRKRST-3066 - Troubleshooting Nexus 7000 (2013 Melbourne) - 2 Hours.pdfBRKRST-3066 - Troubleshooting Nexus 7000 (2013 Melbourne) - 2 Hours.pdf
BRKRST-3066 - Troubleshooting Nexus 7000 (2013 Melbourne) - 2 Hours.pdf
aaajjj4
 
OSDC 2015: Georg Schönberger | Linux Performance Profiling and Monitoring
OSDC 2015: Georg Schönberger | Linux Performance Profiling and MonitoringOSDC 2015: Georg Schönberger | Linux Performance Profiling and Monitoring
OSDC 2015: Georg Schönberger | Linux Performance Profiling and Monitoring
NETWAYS
 
Linux Performance Profiling and Monitoring
Linux Performance Profiling and MonitoringLinux Performance Profiling and Monitoring
Linux Performance Profiling and Monitoring
Georg Schönberger
 
20180811 coscup
20180811 coscup20180811 coscup
20180811 coscup
Quey-Liang Kao
 
High Performance Erlang - Pitfalls and Solutions
High Performance Erlang - Pitfalls and SolutionsHigh Performance Erlang - Pitfalls and Solutions
High Performance Erlang - Pitfalls and Solutions
Yinghai Lu
 
Mac authentication amigopod radius
Mac authentication amigopod radiusMac authentication amigopod radius
Mac authentication amigopod radius
Aruba, a Hewlett Packard Enterprise company
 
Ekon22 tensorflow machinelearning2
Ekon22 tensorflow machinelearning2Ekon22 tensorflow machinelearning2
Ekon22 tensorflow machinelearning2
Max Kleiner
 
BTS Key Mgt
BTS Key MgtBTS Key Mgt
BTS Key Mgt
Mahmudul Hassan
 
VMworld 2013: vSphere Data Protection (VDP) Technical Deep Dive and Troublesh...
VMworld 2013: vSphere Data Protection (VDP) Technical Deep Dive and Troublesh...VMworld 2013: vSphere Data Protection (VDP) Technical Deep Dive and Troublesh...
VMworld 2013: vSphere Data Protection (VDP) Technical Deep Dive and Troublesh...
VMworld
 

Similar to MuSe2023.pptx (20)

jRecruiter - The AJUG Job Posting Service
jRecruiter - The AJUG Job Posting ServicejRecruiter - The AJUG Job Posting Service
jRecruiter - The AJUG Job Posting Service
 
Spring Boot Actuator 2.0 & Micrometer
Spring Boot Actuator 2.0 & MicrometerSpring Boot Actuator 2.0 & Micrometer
Spring Boot Actuator 2.0 & Micrometer
 
OSMC 2015: Linux Performance Profiling and Monitoring by Werner Fischer
OSMC 2015: Linux Performance Profiling and Monitoring by Werner FischerOSMC 2015: Linux Performance Profiling and Monitoring by Werner Fischer
OSMC 2015: Linux Performance Profiling and Monitoring by Werner Fischer
 
OSMC 2015 | Linux Performance Profiling and Monitoring by Werner Fischer
OSMC 2015 | Linux Performance Profiling and Monitoring by Werner FischerOSMC 2015 | Linux Performance Profiling and Monitoring by Werner Fischer
OSMC 2015 | Linux Performance Profiling and Monitoring by Werner Fischer
 
Verified AZ-104 Exam Dumps (V26.02) - Pass Microsoft AZ-104 Exam (2024)
Verified AZ-104 Exam Dumps (V26.02) - Pass Microsoft AZ-104 Exam (2024)Verified AZ-104 Exam Dumps (V26.02) - Pass Microsoft AZ-104 Exam (2024)
Verified AZ-104 Exam Dumps (V26.02) - Pass Microsoft AZ-104 Exam (2024)
 
BPF: Tracing and more
BPF: Tracing and moreBPF: Tracing and more
BPF: Tracing and more
 
NVIDIA深度學習教育機構 (DLI): Object detection with jetson
NVIDIA深度學習教育機構 (DLI): Object detection with jetsonNVIDIA深度學習教育機構 (DLI): Object detection with jetson
NVIDIA深度學習教育機構 (DLI): Object detection with jetson
 
MCSoC'13 Keynote Talk "Taming Big Data Streams"
MCSoC'13 Keynote Talk "Taming Big Data Streams"MCSoC'13 Keynote Talk "Taming Big Data Streams"
MCSoC'13 Keynote Talk "Taming Big Data Streams"
 
Io 120404052713-phpapp01
Io 120404052713-phpapp01Io 120404052713-phpapp01
Io 120404052713-phpapp01
 
了解IO协议栈
了解IO协议栈了解IO协议栈
了解IO协议栈
 
Msc sylabus ptu
Msc sylabus ptuMsc sylabus ptu
Msc sylabus ptu
 
BRKRST-3066 - Troubleshooting Nexus 7000 (2013 Melbourne) - 2 Hours.pdf
BRKRST-3066 - Troubleshooting Nexus 7000 (2013 Melbourne) - 2 Hours.pdfBRKRST-3066 - Troubleshooting Nexus 7000 (2013 Melbourne) - 2 Hours.pdf
BRKRST-3066 - Troubleshooting Nexus 7000 (2013 Melbourne) - 2 Hours.pdf
 
OSDC 2015: Georg Schönberger | Linux Performance Profiling and Monitoring
OSDC 2015: Georg Schönberger | Linux Performance Profiling and MonitoringOSDC 2015: Georg Schönberger | Linux Performance Profiling and Monitoring
OSDC 2015: Georg Schönberger | Linux Performance Profiling and Monitoring
 
Linux Performance Profiling and Monitoring
Linux Performance Profiling and MonitoringLinux Performance Profiling and Monitoring
Linux Performance Profiling and Monitoring
 
20180811 coscup
20180811 coscup20180811 coscup
20180811 coscup
 
High Performance Erlang - Pitfalls and Solutions
High Performance Erlang - Pitfalls and SolutionsHigh Performance Erlang - Pitfalls and Solutions
High Performance Erlang - Pitfalls and Solutions
 
Mac authentication amigopod radius
Mac authentication amigopod radiusMac authentication amigopod radius
Mac authentication amigopod radius
 
Ekon22 tensorflow machinelearning2
Ekon22 tensorflow machinelearning2Ekon22 tensorflow machinelearning2
Ekon22 tensorflow machinelearning2
 
BTS Key Mgt
BTS Key MgtBTS Key Mgt
BTS Key Mgt
 
VMworld 2013: vSphere Data Protection (VDP) Technical Deep Dive and Troublesh...
VMworld 2013: vSphere Data Protection (VDP) Technical Deep Dive and Troublesh...VMworld 2013: vSphere Data Protection (VDP) Technical Deep Dive and Troublesh...
VMworld 2013: vSphere Data Protection (VDP) Technical Deep Dive and Troublesh...
 

Recently uploaded

Microservice Teams - How the cloud changes the way we work
Microservice Teams - How the cloud changes the way we workMicroservice Teams - How the cloud changes the way we work
Microservice Teams - How the cloud changes the way we work
Sven Peters
 
Unlock the Secrets to Effortless Video Creation with Invideo: Your Ultimate G...
Unlock the Secrets to Effortless Video Creation with Invideo: Your Ultimate G...Unlock the Secrets to Effortless Video Creation with Invideo: Your Ultimate G...
Unlock the Secrets to Effortless Video Creation with Invideo: Your Ultimate G...
The Third Creative Media
 
Modelling Up - DDDEurope 2024 - Amsterdam
Modelling Up - DDDEurope 2024 - AmsterdamModelling Up - DDDEurope 2024 - Amsterdam
Modelling Up - DDDEurope 2024 - Amsterdam
Alberto Brandolini
 
Malibou Pitch Deck For Its €3M Seed Round
Malibou Pitch Deck For Its €3M Seed RoundMalibou Pitch Deck For Its €3M Seed Round
Malibou Pitch Deck For Its €3M Seed Round
sjcobrien
 
UI5con 2024 - Boost Your Development Experience with UI5 Tooling Extensions
UI5con 2024 - Boost Your Development Experience with UI5 Tooling ExtensionsUI5con 2024 - Boost Your Development Experience with UI5 Tooling Extensions
UI5con 2024 - Boost Your Development Experience with UI5 Tooling Extensions
Peter Muessig
 
Safelyio Toolbox Talk Softwate & App (How To Digitize Safety Meetings)
Safelyio Toolbox Talk Softwate & App (How To Digitize Safety Meetings)Safelyio Toolbox Talk Softwate & App (How To Digitize Safety Meetings)
Safelyio Toolbox Talk Softwate & App (How To Digitize Safety Meetings)
safelyiotech
 
What’s New in Odoo 17 – A Complete Roadmap
What’s New in Odoo 17 – A Complete RoadmapWhat’s New in Odoo 17 – A Complete Roadmap
What’s New in Odoo 17 – A Complete Roadmap
Envertis Software Solutions
 
ppt on the brain chip neuralink.pptx
ppt  on   the brain  chip neuralink.pptxppt  on   the brain  chip neuralink.pptx
ppt on the brain chip neuralink.pptx
Reetu63
 
Project Management: The Role of Project Dashboards.pdf
Project Management: The Role of Project Dashboards.pdfProject Management: The Role of Project Dashboards.pdf
Project Management: The Role of Project Dashboards.pdf
Karya Keeper
 
Using Query Store in Azure PostgreSQL to Understand Query Performance
Using Query Store in Azure PostgreSQL to Understand Query PerformanceUsing Query Store in Azure PostgreSQL to Understand Query Performance
Using Query Store in Azure PostgreSQL to Understand Query Performance
Grant Fritchey
 
Oracle 23c New Features For DBAs and Developers.pptx
Oracle 23c New Features For DBAs and Developers.pptxOracle 23c New Features For DBAs and Developers.pptx
Oracle 23c New Features For DBAs and Developers.pptx
Remote DBA Services
 
Enums On Steroids - let's look at sealed classes !
Enums On Steroids - let's look at sealed classes !Enums On Steroids - let's look at sealed classes !
Enums On Steroids - let's look at sealed classes !
Marcin Chrost
 
Measures in SQL (SIGMOD 2024, Santiago, Chile)
Measures in SQL (SIGMOD 2024, Santiago, Chile)Measures in SQL (SIGMOD 2024, Santiago, Chile)
Measures in SQL (SIGMOD 2024, Santiago, Chile)
Julian Hyde
 
Webinar On-Demand: Using Flutter for Embedded
Webinar On-Demand: Using Flutter for EmbeddedWebinar On-Demand: Using Flutter for Embedded
Webinar On-Demand: Using Flutter for Embedded
ICS
 
一比一原版(USF毕业证)旧金山大学毕业证如何办理
一比一原版(USF毕业证)旧金山大学毕业证如何办理一比一原版(USF毕业证)旧金山大学毕业证如何办理
一比一原版(USF毕业证)旧金山大学毕业证如何办理
dakas1
 
一比一原版(UMN毕业证)明尼苏达大学毕业证如何办理
一比一原版(UMN毕业证)明尼苏达大学毕业证如何办理一比一原版(UMN毕业证)明尼苏达大学毕业证如何办理
一比一原版(UMN毕业证)明尼苏达大学毕业证如何办理
dakas1
 
Energy consumption of Database Management - Florina Jonuzi
Energy consumption of Database Management - Florina JonuziEnergy consumption of Database Management - Florina Jonuzi
Energy consumption of Database Management - Florina Jonuzi
Green Software Development
 
ACE - Team 24 Wrapup event at ahmedabad.
ACE - Team 24 Wrapup event at ahmedabad.ACE - Team 24 Wrapup event at ahmedabad.
ACE - Team 24 Wrapup event at ahmedabad.
Maitrey Patel
 
E-commerce Development Services- Hornet Dynamics
E-commerce Development Services- Hornet DynamicsE-commerce Development Services- Hornet Dynamics
E-commerce Development Services- Hornet Dynamics
Hornet Dynamics
 
8 Best Automated Android App Testing Tool and Framework in 2024.pdf
8 Best Automated Android App Testing Tool and Framework in 2024.pdf8 Best Automated Android App Testing Tool and Framework in 2024.pdf
8 Best Automated Android App Testing Tool and Framework in 2024.pdf
kalichargn70th171
 

Recently uploaded (20)

Microservice Teams - How the cloud changes the way we work
Microservice Teams - How the cloud changes the way we workMicroservice Teams - How the cloud changes the way we work
Microservice Teams - How the cloud changes the way we work
 
Unlock the Secrets to Effortless Video Creation with Invideo: Your Ultimate G...
Unlock the Secrets to Effortless Video Creation with Invideo: Your Ultimate G...Unlock the Secrets to Effortless Video Creation with Invideo: Your Ultimate G...
Unlock the Secrets to Effortless Video Creation with Invideo: Your Ultimate G...
 
Modelling Up - DDDEurope 2024 - Amsterdam
Modelling Up - DDDEurope 2024 - AmsterdamModelling Up - DDDEurope 2024 - Amsterdam
Modelling Up - DDDEurope 2024 - Amsterdam
 
Malibou Pitch Deck For Its €3M Seed Round
Malibou Pitch Deck For Its €3M Seed RoundMalibou Pitch Deck For Its €3M Seed Round
Malibou Pitch Deck For Its €3M Seed Round
 
UI5con 2024 - Boost Your Development Experience with UI5 Tooling Extensions
UI5con 2024 - Boost Your Development Experience with UI5 Tooling ExtensionsUI5con 2024 - Boost Your Development Experience with UI5 Tooling Extensions
UI5con 2024 - Boost Your Development Experience with UI5 Tooling Extensions
 
Safelyio Toolbox Talk Softwate & App (How To Digitize Safety Meetings)
Safelyio Toolbox Talk Softwate & App (How To Digitize Safety Meetings)Safelyio Toolbox Talk Softwate & App (How To Digitize Safety Meetings)
Safelyio Toolbox Talk Softwate & App (How To Digitize Safety Meetings)
 
What’s New in Odoo 17 – A Complete Roadmap
What’s New in Odoo 17 – A Complete RoadmapWhat’s New in Odoo 17 – A Complete Roadmap
What’s New in Odoo 17 – A Complete Roadmap
 
ppt on the brain chip neuralink.pptx
ppt  on   the brain  chip neuralink.pptxppt  on   the brain  chip neuralink.pptx
ppt on the brain chip neuralink.pptx
 
Project Management: The Role of Project Dashboards.pdf
Project Management: The Role of Project Dashboards.pdfProject Management: The Role of Project Dashboards.pdf
Project Management: The Role of Project Dashboards.pdf
 
Using Query Store in Azure PostgreSQL to Understand Query Performance
Using Query Store in Azure PostgreSQL to Understand Query PerformanceUsing Query Store in Azure PostgreSQL to Understand Query Performance
Using Query Store in Azure PostgreSQL to Understand Query Performance
 
Oracle 23c New Features For DBAs and Developers.pptx
Oracle 23c New Features For DBAs and Developers.pptxOracle 23c New Features For DBAs and Developers.pptx
Oracle 23c New Features For DBAs and Developers.pptx
 
Enums On Steroids - let's look at sealed classes !
Enums On Steroids - let's look at sealed classes !Enums On Steroids - let's look at sealed classes !
Enums On Steroids - let's look at sealed classes !
 
Measures in SQL (SIGMOD 2024, Santiago, Chile)
Measures in SQL (SIGMOD 2024, Santiago, Chile)Measures in SQL (SIGMOD 2024, Santiago, Chile)
Measures in SQL (SIGMOD 2024, Santiago, Chile)
 
Webinar On-Demand: Using Flutter for Embedded
Webinar On-Demand: Using Flutter for EmbeddedWebinar On-Demand: Using Flutter for Embedded
Webinar On-Demand: Using Flutter for Embedded
 
一比一原版(USF毕业证)旧金山大学毕业证如何办理
一比一原版(USF毕业证)旧金山大学毕业证如何办理一比一原版(USF毕业证)旧金山大学毕业证如何办理
一比一原版(USF毕业证)旧金山大学毕业证如何办理
 
一比一原版(UMN毕业证)明尼苏达大学毕业证如何办理
一比一原版(UMN毕业证)明尼苏达大学毕业证如何办理一比一原版(UMN毕业证)明尼苏达大学毕业证如何办理
一比一原版(UMN毕业证)明尼苏达大学毕业证如何办理
 
Energy consumption of Database Management - Florina Jonuzi
Energy consumption of Database Management - Florina JonuziEnergy consumption of Database Management - Florina Jonuzi
Energy consumption of Database Management - Florina Jonuzi
 
ACE - Team 24 Wrapup event at ahmedabad.
ACE - Team 24 Wrapup event at ahmedabad.ACE - Team 24 Wrapup event at ahmedabad.
ACE - Team 24 Wrapup event at ahmedabad.
 
E-commerce Development Services- Hornet Dynamics
E-commerce Development Services- Hornet DynamicsE-commerce Development Services- Hornet Dynamics
E-commerce Development Services- Hornet Dynamics
 
8 Best Automated Android App Testing Tool and Framework in 2024.pdf
8 Best Automated Android App Testing Tool and Framework in 2024.pdf8 Best Automated Android App Testing Tool and Framework in 2024.pdf
8 Best Automated Android App Testing Tool and Framework in 2024.pdf
 

MuSe2023.pptx

  • 1. MuSe2023 from practical view Ganghyun Kim 2023 / 08 / 18
  • 4. MuSe2023 overview • Goal: Predict subject’s emotional state using multimodal data • Subject is under stressed condition (simulated job interview) • Emotional state can be represented in 2 axis, Arousal and Valence 4/30 https://www.mdpi.com/1424-8220/23/3/1598
  • 5. Input data (feature) • Audio: DeepSpectrum, eGeMAPS, Wav2vec2.0 • Video: ViT, FaceNet, FAU, Pose • Other: BERT, Biosignals • All of them tabular data extracted with corresponding feature extractor. 5/30
  • 9. Big Titan server: Titan Xp X2 Small Titan server: Titan Xp X2 Student server 1: RTX 1080 X2 Student server 2: RTX2080 X2 Student server 3: A5000 X2 TA server: RTX3090 X3 Code Management
  • 10. 6F 6F IGC IT CENTER IGC IT CENTER IGC IT CENTER IGC IT CENTER Execute commands via 6F
  • 14. Wandb Sweep Generates HP tuning task list Start signal Check remaining task, assign 1 HP combination run to server, repeat when server finishes
  • 15. Wandb Sweep • wandb sweep sweep.yaml # returns your SWEEP_ID • wandb agent mcg70107 • Done! 15/30
  • 16. Previously.. • PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=python CUDA_VISIBLE_DEVICES=1 python main_rnn.py --cache --task stress --emo_dim valence --d_rnn 64 --rnn_n_layers 2 -- lr 0.005 --feature mfcc_combined --win_len 200 --rnn_bi --n_seeds 20 --seed 101 --batch_size 256 --early_stopping_patience 15 --reduce_lr_patience 5 --project_name Muse_20221026_early_fusion_RNN_valence_mfcc_combined --use_gpu • PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=python CUDA_VISIBLE_DEVICES=1 python main_rnn.py --cache --task stress --emo_dim valence --d_rnn 64 --rnn_n_layers 2 -- lr 0.005 --feature mfcc_combined --win_len 300 --rnn_bi --n_seeds 20 --seed 101 --batch_size 256 --early_stopping_patience 15 --reduce_lr_patience 5 --project_name Muse_20221026_early_fusion_RNN_valence_mfcc_combined --use_gpu • PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=python CUDA_VISIBLE_DEVICES=1 python main_rnn.py --cache --task stress --emo_dim valence --d_rnn 64 --rnn_n_layers 2 -- lr 0.005 --feature mfcc_combined --win_len 400 --rnn_bi --n_seeds 20 --seed 101 --batch_size 256 --early_stopping_patience 15 --reduce_lr_patience 5 --project_name Muse_20221026_early_fusion_RNN_valence_mfcc_combined --use_gpu • PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=python CUDA_VISIBLE_DEVICES=1 python main_rnn.py --cache --task stress --emo_dim valence --d_rnn 64 --rnn_n_layers 4 -- lr 0.001 --feature mfcc_combined --win_len 200 --rnn_bi --n_seeds 20 --seed 101 --batch_size 256 --early_stopping_patience 15 --reduce_lr_patience 5 --project_name Muse_20221026_early_fusion_RNN_valence_mfcc_combined --use_gpu • PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=python CUDA_VISIBLE_DEVICES=1 python main_rnn.py --cache --task stress --emo_dim valence --d_rnn 64 --rnn_n_layers 4 -- lr 0.001 --feature mfcc_combined --win_len 300 --rnn_bi --n_seeds 20 --seed 101 --batch_size 256 --early_stopping_patience 15 --reduce_lr_patience 5 --project_name Muse_20221026_early_fusion_RNN_valence_mfcc_combined --use_gpu • PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=python CUDA_VISIBLE_DEVICES=1 python main_rnn.py --cache --task stress --emo_dim valence --d_rnn 64 --rnn_n_layers 4 -- lr 0.001 --feature mfcc_combined --win_len 400 --rnn_bi --n_seeds 20 --seed 101 --batch_size 256 --early_stopping_patience 15 --reduce_lr_patience 5 --project_name Muse_20221026_early_fusion_RNN_valence_mfcc_combined --use_gpu • PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=python CUDA_VISIBLE_DEVICES=1 python main_rnn.py --cache --task stress --emo_dim valence --d_rnn 64 --rnn_n_layers 4 -- lr 0.002 --feature mfcc_combined --win_len 200 --rnn_bi --n_seeds 20 --seed 101 --batch_size 256 --early_stopping_patience 15 --reduce_lr_patience 5 --project_name Muse_20221026_early_fusion_RNN_valence_mfcc_combined --use_gpu • PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=python CUDA_VISIBLE_DEVICES=1 python main_rnn.py --cache --task stress --emo_dim valence --d_rnn 64 --rnn_n_layers 4 -- lr 0.002 --feature mfcc_combined --win_len 300 --rnn_bi --n_seeds 20 --seed 101 --batch_size 256 --early_stopping_patience 15 --reduce_lr_patience 5 --project_name Muse_20221026_early_fusion_RNN_valence_mfcc_combined --use_gpu • PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=python CUDA_VISIBLE_DEVICES=1 python main_rnn.py --cache --task stress --emo_dim valence --d_rnn 64 --rnn_n_layers 4 -- lr 0.002 --feature mfcc_combined --win_len 400 --rnn_bi --n_seeds 20 --seed 101 --batch_size 256 --early_stopping_patience 15 --reduce_lr_patience 5 --project_name Muse_20221026_early_fusion_RNN_valence_mfcc_combined --use_gpu 16/30
  • 17. Previously.. • Command generator writes combinations of HP tuning runs • Split them in .sh file, in appropriate length considering GPU power of server, # of GPU • Upload .sh file to each server, run them Problem • Hard to restart from crashed run. Needed to locate which run crashed, edit .sh file to start from there • Tedious work, considering number of experiments I had to run 17/30
  • 19. 19/30
  • 20. 20/30
  • 22. 22/30
  • 23. 23/30

Editor's Notes

  1. Consists of 3 different challenges, we did stress
  2. Consists of 3 different challenges, we did stress