Difference between revisions of "Community/training/2018"

From Nordic Language Processing Laboratory
Jump to: navigation, search
(Background)
(Background)
 
(One intermediate revision by the same user not shown)
Line 12: Line 12:
 
The winter school had a duration of two days, i.e. ended with
 
The winter school had a duration of two days, i.e. ended with
 
lunch on Wednesday, January 31.
 
lunch on Wednesday, January 31.
There were around 20–25 participants, where the majority came from the NLPL partner sites.
+
There were around 25 participants, where the majority came from the NLPL partner sites.
  
 
NLPL team members, associates at partner sites (e.g. doctoral and post-doctoral fellows,
 
NLPL team members, associates at partner sites (e.g. doctoral and post-doctoral fellows,
Line 76: Line 76:
 
'''Dinner 19:30 Monday and Tuesday'''
 
'''Dinner 19:30 Monday and Tuesday'''
  
 +
= Software Environment =
 +
 +
The deep learning sessions will include hands-on laboratory sessions following the
 +
[https://github.com/LxMLS/lxmls_guide LxMLS tutorial], although switching over
 +
to PyTorch (instead Theano).
 +
Stephan Oepen has prepared a Python virtual environment will all LxLMS dependencies
 +
installed, including PyTorch with GPU support:
 +
 +
module use -a /projects/nlpl/software/modulefiles/
 +
module add nlpl-lxmls
 +
 +
There is a one-time preparaptory step to obtain the ‘student’ starting package from LXMLS:
 +
 +
git clone https://github.com/LxMLS/lxmls-toolkit.git
 +
cd lxmls-toolkit
 +
git checkout student
 +
 +
Just now at least, there is a minor incompatibility with the standard version of SciPy
 +
installed on Abel and a newer version that the LXMLS toolkit assumes.
 +
To accomodate for this mismatch, all occurence of <tt>scipy.special</tt>
 +
must be changed to <tt>scipy.misc</tt>, including in the LXMLS library file
 +
<tt>lxmls/deep_learning/numpy_models/mlp.py</tt>.
 +
 +
Finally, the LXMLS code assumes that its top-level directory is visible for Python <tt>import</tt> commands;
 +
when inside the <tt>lxmls-toolkit</tt> directory, the following shell command should have
 +
the right effect:
 +
 +
export PYTHONPATH=$PYTHONPATH:.
 +
 +
 +
= Logistics and Registration =
 +
 +
The bus to Skeikampen will leave at 09:30 (CET). We will be informed about which pier on Monday morning. This is due to the traffic conditions at OSL.
 +
Please meet at the OSL Arrival Area Meeting Point at 09:00.
 +
 +
[https://avinor.no/en/airport/oslo-airport/plan-your-trip/map/  Complete Map of OSL]
 +
 +
NLPL will provide bus transfer from and to Oslo Aiport Gardermoen (OSL),
 +
leaving at 9:30 on Monday morning and returning to OSL around 16:30 on
 +
Wedensday.
 +
The transfer and accomodation at Skeikampen will be directly covered
 +
by NeIC.
 +
Travel expenses to and from Oslo Airport will need to be submitted
 +
to NeIC for reimbursement after completion of the winter school.
 +
Note that reimbursement for the travel to OSL is for NLPL Project Team members only.
 +
 +
Please record your intent to participate in the winter school
 +
through the on-line registration form.
 +
A limited number of hotel rooms and travel stipends are available.
 +
For reasons of fairness, we will want to make sure that all NLPL partner sites can send
 +
some participants (three per site, on average).
 +
We will post confirmed registrations on this page by Monday, December 18;
 +
please make your own travel arrangements once your participation has been
 +
confirmed.
 +
 +
= Participants =
 +
 +
 +
{|
 +
|-
 +
| Ramon Fernandez Astudillo, Unbabel and ''Instituto de Engenharia de Sistemas e Computadores''
 +
|-
 +
| Isabelle Augenstein, University of Copenhagen
 +
|-
 +
| Johannes Bjerva, University of Copenhagen
 +
|-
 +
| Murhaf Fares, University of Oslo
 +
|-
 +
| Mareike Hartmann, University of Copenhagen
 +
|-
 +
| Taraka Rama Kasicheyanula, University of Oslo
 +
|-
 +
| Andrei Kutuzov, University of Oslo
 +
|-
 +
| Miryam de Lhoneux, Uppsala University
 +
|-
 +
| Bjørn Lindi, Nordic e-Infrastructure Collaboration
 +
|-
 +
| Andre Martins, Unbabel and Instituto Superior Técnico
 +
|-
 +
| Farhad Nooralahzadeh, University of Oslo
 +
|-
 +
| Stephan Oepen, University of Oslo
 +
|-
 +
| Barbara Plank, IT University of Copenhagen
 +
|-
 +
| Yves Scherrer, University of Helsinki
 +
|-
 +
| Natalie Schluter, IT University of Copenhagen
 +
|-
 +
| Yan Shao, Uppsala University
 +
|-
 +
| Aaron Smith, Uppsala University
 +
|-
 +
| Jörg Tiedemann, University of Helsinki
 +
|-
 +
| Aleksi Vesanto, University of Turku
 +
|-
 +
| Maria Barrett, University of Copenhagen
 +
|-
 +
| Manex Zabaleta, University of Copenhagen
 +
|-
 +
| Senka Droba, University of Helsinki
 +
|-
 +
| Mika Hämäläinen, University of Helsinki
 +
|-
 +
|Yova Kementchedjhieva, University of Copenhagen
 +
|}
  
 
[[File:skeikampen.2018.jpg|center]]
 
[[File:skeikampen.2018.jpg|center]]

Latest revision as of 16:29, 12 September 2018

Background

The 2018 NLPL Winter School was co-located and partially overlapping with the NeIC All-Hands Meeting (AHM), which some NLPL team members attended. The NeIC meeting was held from January 29 to February 1, 2018, in the Norwegian mountain resort at Skeikampen.

Starting already after lunch on Monday, January 29, NLPL kicked off its winter school in E-Infrastructure and Scientific Computing for Nordic Natural Language Processing Research. The winter school had a duration of two days, i.e. ended with lunch on Wednesday, January 31. There were around 25 participants, where the majority came from the NLPL partner sites.

NLPL team members, associates at partner sites (e.g. doctoral and post-doctoral fellows, possibly also some MSc students), and other prospective users of the infrastructure were invited to attend; NLPL partners(i.e. members of the project team) had their cost of participation covered by the project.

Programme

The winter school will comprise different types of activities, including (a) overview talks, for example on GPU programming (with emphasis on Taito and Abel), ‘deep’ learning paradigms and toolkits, or other scientific programming and HPC techniques; (b) in-depth tutorials on parts of the NLPL infrastructure (e.g. translation, parsing, and extrinsic evaluation software, as well as corpora and embeddings) and other topics of relevance to the project (e.g. ‘containerization’); and (c) hands-on hackathons, i.e. collective programming and experimentation in specific toolkits (e.g. environments like DyNet, MPI, or CUDA). Participation in the hackathons will require preparation prior to the winter school, which NLPL partner sites will organize locally.

Two of the co-organizers of the Lisbon Machine Learning School (LxML), André Martins and Ramon Fernandez Astudillo will work through selections of the LxML programme, with a special focus on deep learning approaches to NLP. Additionally, NLPL team members will provide a walk-through of the emerging infrastructure, there will be a high-level tutorial on the use of Singularity containers, and representatives of the Finnish and Norwegian national providers will provide a survey of available and upcoming computing infrastructures, with emphasis on GPU resources.

Monday, January 29, 2018 Tuesday, January 30, 2018 Wednesday, January 31, 2018
09:00 12:00 Travel to Skeikampen. Lunch at 12:30 09:00 10:30 Session 4: Deep Learning for NLP (LxMLS) Room: Olav 2 09:00 10:00 Session 8: Deep Learning for NLP Wrap-Up (LxMLS) Room: Olav 2
10:30 15:00 Break and lunch, which is served 13:00 - 14:30 10:30 12:00 Session 9: Tutorial on Containers (Abdulrahman Azab, UiO/NeIC) Room: Olav 2
13:30 15:00 Session 1: Deep Learning for NLP (LxMLS) 15:00 16:30 Session 5: Deep Learning for NLP (LxMLS) Room: Vinterhagen 12:30 13:30 Lunch
15:00 15:30 Coffee Break 16:30 17:00 Coffee Break 13:45 14:00 Get on the bus. The Bus leaves for OSL Airport at 14:00
15:30 17:00 Session 2: Deep Learning for NLP (LxMLS) 17:00 18:30 Session 6: Deep Learning for NLP(LxMLS) Room: Vinterhagen
17:00 17:30 Coffee Break 18:30 18:45 Coffee Break
17:30 19:00 Session 3: Walk-Through of NLPL Infrastructure 18:45 19:30 Session 7: Experience Exchange on DL and GPUs, Room: Vinterhagen

Dinner 19:30 Monday and Tuesday

Software Environment

The deep learning sessions will include hands-on laboratory sessions following the LxMLS tutorial, although switching over to PyTorch (instead Theano). Stephan Oepen has prepared a Python virtual environment will all LxLMS dependencies installed, including PyTorch with GPU support:

module use -a /projects/nlpl/software/modulefiles/
module add nlpl-lxmls

There is a one-time preparaptory step to obtain the ‘student’ starting package from LXMLS:

git clone https://github.com/LxMLS/lxmls-toolkit.git
cd lxmls-toolkit
git checkout student

Just now at least, there is a minor incompatibility with the standard version of SciPy installed on Abel and a newer version that the LXMLS toolkit assumes. To accomodate for this mismatch, all occurence of scipy.special must be changed to scipy.misc, including in the LXMLS library file lxmls/deep_learning/numpy_models/mlp.py.

Finally, the LXMLS code assumes that its top-level directory is visible for Python import commands; when inside the lxmls-toolkit directory, the following shell command should have the right effect:

export PYTHONPATH=$PYTHONPATH:.


Logistics and Registration

The bus to Skeikampen will leave at 09:30 (CET). We will be informed about which pier on Monday morning. This is due to the traffic conditions at OSL. Please meet at the OSL Arrival Area Meeting Point at 09:00.

Complete Map of OSL

NLPL will provide bus transfer from and to Oslo Aiport Gardermoen (OSL), leaving at 9:30 on Monday morning and returning to OSL around 16:30 on Wedensday. The transfer and accomodation at Skeikampen will be directly covered by NeIC. Travel expenses to and from Oslo Airport will need to be submitted to NeIC for reimbursement after completion of the winter school. Note that reimbursement for the travel to OSL is for NLPL Project Team members only.

Please record your intent to participate in the winter school through the on-line registration form. A limited number of hotel rooms and travel stipends are available. For reasons of fairness, we will want to make sure that all NLPL partner sites can send some participants (three per site, on average). We will post confirmed registrations on this page by Monday, December 18; please make your own travel arrangements once your participation has been confirmed.

Participants

Ramon Fernandez Astudillo, Unbabel and Instituto de Engenharia de Sistemas e Computadores
Isabelle Augenstein, University of Copenhagen
Johannes Bjerva, University of Copenhagen
Murhaf Fares, University of Oslo
Mareike Hartmann, University of Copenhagen
Taraka Rama Kasicheyanula, University of Oslo
Andrei Kutuzov, University of Oslo
Miryam de Lhoneux, Uppsala University
Bjørn Lindi, Nordic e-Infrastructure Collaboration
Andre Martins, Unbabel and Instituto Superior Técnico
Farhad Nooralahzadeh, University of Oslo
Stephan Oepen, University of Oslo
Barbara Plank, IT University of Copenhagen
Yves Scherrer, University of Helsinki
Natalie Schluter, IT University of Copenhagen
Yan Shao, Uppsala University
Aaron Smith, Uppsala University
Jörg Tiedemann, University of Helsinki
Aleksi Vesanto, University of Turku
Maria Barrett, University of Copenhagen
Manex Zabaleta, University of Copenhagen
Senka Droba, University of Helsinki
Mika Hämäläinen, University of Helsinki
Yova Kementchedjhieva, University of Copenhagen
Skeikampen.2018.jpg