February 2019 - Candle-aesp - mailman3.cels.anl.gov

Relevant to our effort
by Stevens, Rick L. 27 Feb '19

27 Feb '19

https://s3-eu-west-1.amazonaws.com/pstorage-chemrxiv-899408398289/10662715/…

1 0

Fwd: the Canadian-US-Mexico workshop
by Stevens, Rick L. 26 Feb '19

26 Feb '19

Worth reading this Sent from my iPad Begin forwarded message: From: "Foster, Ian T." <foster(a)anl.gov<mailto:[email protected]>> To: "Stevens, Rick L." <stevens(a)anl.gov<mailto:[email protected]>> Subject: the Canadian-US-Mexico workshop

1 0

Fwd: So much to catch up on, so little time...
by Stevens, Rick L. 26 Feb '19

26 Feb '19

Do we have anything that might be suitable ? Anyone intrrsted in working with John and student s? Sent from my iPhone Begin forwarded message: From: John Santerre <john.santerre(a)gmail.com<mailto:[email protected]>> Date: February 21, 2019 at 12:13:33 EST To: "Stevens, Rick L." <stevens(a)anl.gov<mailto:[email protected]>>, "Rick L. Stevens" <edwinrickstevens(a)gmail.com<mailto:[email protected]>> Subject: So much to catch up on, so little time... Rick, Dude, hope and trust all is well in your world... While i have 5+ irons in the fire ATM, one in particular might be of interest to you. I'm advising 6 Masters students(divided into pairs) for their capstone at SMU over the next 6 months. One in particular is interested in further exploring the blurring the line between CPU and GPU for deep learning. This is building on that work I did in conjunction with Intel and Google at SAP. It occurred to me that you might have some analysis scraps that haven't yet been followed thru on yet. The scope needs to be short and to the point, ie, it needs to be pretty clear that the result will turn out as they don't have a lot of flexibility in terms of adjusting as they progress. For instance, another group is hoping to implement a solution on this work, CheXpert: A Large Dataset of Chest X-Rays and Competition for Automated Chest X-Ray Interpretation. While the last is considering different techniques for this: http://science.sciencemag.org/content/360/6393/1133 So if you had a data set that your team doesn't have time to process etc, I think we could target a nice use case/ collaboration. I may try and go back to the AMR use case, but it would be nice to use some new data... Again hope all is well in CHI, so far LA is treating me very nicely! J

1 0

Fwd: NSF HDR Frameworks
by Stevens, Rick L. 26 Feb '19

26 Feb '19

Did anyone get. Chance to look at this call ? Trying to decide if worth us going after ? Sent from my iPhone Begin forwarded message: From: "Dietrich, Emily M." <edietrich(a)anl.gov<mailto:[email protected]>> Date: February 24, 2019 at 08:45:35 EST To: "Stevens, Rick L." <stevens(a)anl.gov<mailto:[email protected]>> Subject: Re: NSF HDR Frameworks Let me know if you want me to join the call on Monday; always happy to help out. Emily — Emily Dietrich Coordinating Writer / Editor Computing, Environment and Life Sciences Argonne National Laboratory 9700 S. Cass Avenue Argonne, IL 60439 Phone: 630.252.5815 From: "Stevens, Rick L." <stevens(a)anl.gov<mailto:[email protected]>> Date: Friday, February 22, 2019 at 2:13 PM To: Emily Dietrich <edietrich(a)anl.gov<mailto:[email protected]>> Subject: Fwd: NSF HDR Frameworks Sent from my iPad Begin forwarded message: From: "Stevens, Rick L." <stevens(a)anl.gov<mailto:[email protected]>> Date: February 23, 2019 at 4:33:36 AM GMT+9 To: "Brettin, Thomas S." <brettin(a)anl.gov<mailto:[email protected]>>, "Xia, Fangfang" <fangfang(a)anl.gov<mailto:[email protected]>>, "Zhu, Yitan" <yitan.zhu(a)anl.gov<mailto:[email protected]>>, "Clyde, Austin Robert" <aclyde(a)anl.gov<mailto:[email protected]>>, "Partin, Alexander" <apartin(a)anl.gov<mailto:[email protected]>>, "Ramanathan, Arvind" <ramanathana(a)anl.gov<mailto:[email protected]>>, "Shukla, Maulik" <mshukla(a)anl.gov<mailto:[email protected]>>, "hyunseung.yoo(a)gmail.com<mailto:[email protected]>" <hyunseung.yoo(a)gmail.com<mailto:[email protected]>> Subject: Fwd: NSF HDR Frameworks Hi Please take a look at this call. Julia and I will discuss on Monday what a response might look like.. looking for ideas and folks that might want to help. —rick Sent from my iPad Begin forwarded message: From: Julia Lane <jlane2(a)uchicago.edu<mailto:[email protected]>> Date: February 23, 2019 at 12:27:17 AM GMT+9 To: "Stevens, Rick L." <stevens(a)anl.gov<mailto:[email protected]>> Subject: Re: NSF HDR Frameworks Hi Rick, I wanted to circle back to my email below and see if you would be interested in this opportunity. Mike, Dan, and I were discussing and we thought you would the natural leader for something like this. However, if this isn't the right time that's no problem. Julia Julia Lane Executive Director, Center for Data and Computing (CDAC) University of Chicago John Crerar Library Building jlane2(a)uchicago.edu<mailto:[email protected]> 773-834-2985 ________________________________ From: Julia Lane Sent: Monday, February 18, 2019 3:50:48 PM To: Stevens, Rick L. Subject: NSF HDR Frameworks Hi Rick, You may have seen that NSF has recently released the HDR Frameworks solicitation for data-intensive research in science and engineering; given the topic I wanted to flag it for you. While I know you have a lot on your plate already, I thought this might be of interest to you. Let me know if you are interested and if CDAC can be of help. NSF HDR Frameworks<https://www.nsf.gov/pubs/2019/nsf19549/nsf19549.htm?WT.mc_id=USNSF_25&WT.mc…> Deadline: May 7, 2019 Description: The overarching goal of the HDR Institutes DIRSE Frameworks solicitation is to foster convergent approaches to data-driven research in science and engineering. Frameworks will consist of interdisciplinary teams to conceptualize and pilot new modalities for collaboration and convergence that go beyond institutional walls and traditional disciplinary boundaries, to build innovative connections between scientific groups and data scientists and engineers, to integrate research infrastructure and education infrastructure. The Frameworks should focus on science and engineering areas that: (1) are at a "tipping point" where a timely investment in data-intensive approaches has the maximum potential for a transformative effect, (2) have needs that can benefit from interdisciplinary investments in data analytics infrastructure, and (3) represent investment priorities for the participating NSF directorates during, and beyond, the lifetime of the HDR Big Idea. Specific outcomes expected from the Frameworks include identification of frontier science and engineering challenge problems and the associated data and data-science barriers or tipping points, as well as development of new strategies and innovative approaches to foster scientific breakthroughs involving researchers from diverse scientific backgrounds. Julia Julia Lane Executive Director, Center for Data and Computing (CDAC) University of Chicago John Crerar Library Building jlane2(a)uchicago.edu<mailto:[email protected]> 773-834-2985

1 0

Rescheduling the AESP Jam Session
by Mulligan, Silvia P. 20 Feb '19

20 Feb '19

Greetings, We’d like to reschedule the AESP Jam Session that was canceled last month. Please advise by c.o.b. this Friday, if you’re available on Tuesday, April 16th, 11:15am – 3:15pm. This meeting will be held at the University of Chicago. If this doesn’t work, I will send a doodle poll with options in May. Thanks, Silvia ********************* Silvia Mulligan Executive Secretary to Rick Stevens Computing, Environment and Life Sciences Argonne National Laboratory 9700 S. Cass Avenue Argonne, IL 60439 630-252-6188

1 0

Re: [Candle-aesp] [Pilot1-anl] Getting Started - Comet.ml
by Ramanathan, Arvind 10 Feb '19

10 Feb '19

I usually like to use: https://www.pachyderm.io/index.html This has been more robust compared to Comet. Provides all the capabilities that comet provides but additionally allows version control of data which is cool. Best, Arvind On Feb 10, 2019, at 9:29 AM, Rick Stevens via Pilot1-anl <pilot1-anl(a)lists.cels.anl.gov<mailto:[email protected]>> wrote: I noticed that Austin was using Comet.ml It might be something we want to tinker with and decide if it makese sense to integrate with CANDLE. https://www.comet.ml/docs/python-sdk/getting-started/ Getting Started The core class of Comet.ml is an Experiment, a specific run of a script that generated a result such as training a model on a single set of hyper parameters. An Experiment<x-msg://31/Experiment/#experiment>. will automatically log scripts output (stdout/stderr), code, and command line arguments on any script and for the supported libraries will also log hyper parameters, metrics and model configuration. Here is the Experiment object: from comet_ml import Experiment experiment = Experiment(api_key="YOUR_API_KEY") You can also set the API_KEY as an environment variable COMET_API_KEY then instantiate your experiment like this: from comet_ml import Experiment experiment = Experiment() The Experiment object logs various parameters of your experiment to Comet.ml from comet_ml import Experiment experiment = Experiment(api_key="YOUR_API_KEY") batch_size = 4 experiment.log_parameter("batch_size", batch_size) By default your experiment will be added to the project Uncategorized Experiments. You can also log your experiment to a specific project. from comet_ml import Experiment experiment = Experiment(api_key="YOUR_API_KEY", project_name="my project name") batch_size = 4 experiment.log_parameter("batch_size", batch_size) You can also log a custom list of hyperparameters to your experiment via a dictionary. from comet_ml import Experiment experiment = Experiment(api_key="YOUR_API_KEY", project_name="my project name", auto_param_logging=False) batch_size = 128 num_classes = 10 epochs = 20 params={ "batch_size":batch_size, "epochs":epochs, "num_classes":num_classes} experiment.log_parameters(params) We all strive to be data driven and yet every day valuable experiments results are just lost and forgotten. Comet.ml provides a dead simple way of fixing that. Works with any workflow, any ML task, any machine and any piece of code. For a more in-depth tutorial about Comet.ml, you can check out: * Getting started with the Experiment class<x-msg://31/Experiment/#experiment> ________________________________ Installation * Sign up (free) on comet.ml and obtain an API key<https://www.comet.ml/> * Install Comet.ml from PyPI: pip install comet_ml Comet.ml python SDK is compatible with: Python 2.7-3.6. Checking Comet.ml version pip show comet_ml _______________________________________________ Pilot1-anl mailing list Pilot1-anl(a)lists.cels.anl.gov<mailto:[email protected]> https://lists.cels.anl.gov/mailman/listinfo/pilot1-anl

1 0

Getting Started - Comet.ml
by Rick Stevens 10 Feb '19

10 Feb '19

I noticed that Austin was using Comet.ml It might be something we want to tinker with and decide if it makese sense to integrate with CANDLE. > > https://www.comet.ml/docs/python-sdk/getting-started/ <https://www.comet.ml/docs/python-sdk/getting-started/> > > Getting Started > > The core class of Comet.ml is an Experiment, a specific run of a script that generated a result such as training a model on a single set of hyper parameters. An Experiment <../Experiment/#experiment>. will automatically log scripts output (stdout/stderr), code, and command line arguments on any script and for the supported libraries will also log hyper parameters, metrics and model configuration. > > Here is the Experiment object: > > from comet_ml import Experiment > experiment = Experiment(api_key="YOUR_API_KEY") > > > You can also set the API_KEY as an environment variable COMET_API_KEY then instantiate your experiment like this: > > from comet_ml import Experiment > experiment = Experiment() > > > The Experiment object logs various parameters of your experiment to Comet.ml > > from comet_ml import Experiment > experiment = Experiment(api_key="YOUR_API_KEY") > batch_size = 4 > > experiment.log_parameter("batch_size", batch_size) > By default your experiment will be added to the project Uncategorized Experiments. You can also log your experiment to a specific project. > > from comet_ml import Experiment > > > experiment = Experiment(api_key="YOUR_API_KEY", > project_name="my project name") > batch_size = 4 > > experiment.log_parameter("batch_size", batch_size) > You can also log a custom list of hyperparameters to your experiment via a dictionary. > > from comet_ml import Experiment > experiment = Experiment(api_key="YOUR_API_KEY", > project_name="my project name", > auto_param_logging=False) > batch_size = 128 > num_classes = 10 > epochs = 20 > > params={ > "batch_size":batch_size, > "epochs":epochs, > "num_classes":num_classes} > > experiment.log_parameters(params) > We all strive to be data driven and yet every day valuable experiments results are just lost and forgotten. Comet.ml provides a dead simple way of fixing that. Works with any workflow, any ML task, any machine and any piece of code. > > For a more in-depth tutorial about Comet.ml, you can check out: > > Getting started with the Experiment class <../Experiment/#experiment> > Installation > > Sign up (free) on comet.ml and obtain an API key <https://www.comet.ml/> > Install Comet.ml from PyPI: > > pip install comet_ml > Comet.ml python SDK is compatible with: Python 2.7-3.6. > > Checking Comet.ml version > > pip show comet_ml

1 0

FW: You're Invited: Aurora ESP Webinar on February 20
by Dietrich, Emily M. 07 Feb '19

07 Feb '19

FYI Emily — Emily Dietrich Coordinating Writer / Editor Computing, Environment and Life Sciences Argonne National Laboratory 9700 S. Cass Avenue Argonne, IL 60439 Phone: 630.252.5815 On 2/7/19, 9:34 AM, "Tim Williams" <tjwilliams(a)anl.gov> wrote: Dear Colleague, Please join us on Wednesday, February 20, as we continue our quarterly Aurora ESP webinar series, with a session on performance tuning with Intel's Advisor and VTune Amplifier tools. The webinar, which will be held from 11:30 a.m. - 2 p.m. (CT), will provide an introduction to using these tools for profiling and tuning applications. Submission scripts customized for Theta, and easily modified for other systems, will be provided as part of the session. For more information and to register, visit https://esp-training-series-feb-2019.eventbrite.com Let me know if you have any questions. We hope to see you there. Best regards, Tim Williams, Aurora ESP manager

1 0

interesting paper I found by searching for descriptors that turned up in the Top6 models.
by Rick Stevens 03 Feb '19

03 Feb '19

> http://images.biomedsearch.com/24473174/marinedrugs-12-00757.pdf?AWSAccessK… <http://images.biomedsearch.com/24473174/marinedrugs-12-00757.pdf?AWSAccessK…> >

1 0