From a9667afbfb3fc93b778b22739cdb34ced7ff52ae Mon Sep 17 00:00:00 2001 From: Marcel Date: Thu, 24 Mar 2022 12:33:57 +0100 Subject: [PATCH 1/8] Add proposal for course automation task --- .../segatz-marcelj/README.md | 23 +++++++++++++++++++ 1 file changed, 23 insertions(+) create mode 100644 contributions/course-automation/segatz-marcelj/README.md diff --git a/contributions/course-automation/segatz-marcelj/README.md b/contributions/course-automation/segatz-marcelj/README.md new file mode 100644 index 0000000000..4cb574e3ed --- /dev/null +++ b/contributions/course-automation/segatz-marcelj/README.md @@ -0,0 +1,23 @@ +# Assignment Proposal + +## Title + +_Automatic reading suggestions of earlier similar contributions_ + +## Names and KTH ID + - Fabian Segatz (segatz@kth.se) + - Marcel Juschak (marcelj@kth.se) + +## Deadline + +Task 1 + +## Category + +Course Automation + +## Description + +As the course is repeatedly taught, more and more brilliant contributions from previous students are accumulated that can be leveraged to enhance the quality of new students' contributions. + +We will create a Github Action that compares the title of an accepted assignment proposal to the selected student works of previous years and create reading suggestions for good contributions on similar topics. The output of the action will be given as a comment to the pull request. We anticipate that this will help students that feel overwhelmed to gain a quick overview of the field and show what approaches are particularly suitable for this kind of topic. From 6663b46f3239928d451eefa409bec1167f7df409 Mon Sep 17 00:00:00 2001 From: Neproxx Date: Fri, 25 Mar 2022 14:51:43 +0100 Subject: [PATCH 2/8] Rework proposal --- contributions/course-automation/segatz-marcelj/README.md | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/contributions/course-automation/segatz-marcelj/README.md b/contributions/course-automation/segatz-marcelj/README.md index 4cb574e3ed..e0004585a4 100644 --- a/contributions/course-automation/segatz-marcelj/README.md +++ b/contributions/course-automation/segatz-marcelj/README.md @@ -2,7 +2,7 @@ ## Title -_Automatic reading suggestions of earlier similar contributions_ +_Automatic comparison of proposals with previous contributions_ ## Names and KTH ID - Fabian Segatz (segatz@kth.se) @@ -20,4 +20,4 @@ Course Automation As the course is repeatedly taught, more and more brilliant contributions from previous students are accumulated that can be leveraged to enhance the quality of new students' contributions. -We will create a Github Action that compares the title of an accepted assignment proposal to the selected student works of previous years and create reading suggestions for good contributions on similar topics. The output of the action will be given as a comment to the pull request. We anticipate that this will help students that feel overwhelmed to gain a quick overview of the field and show what approaches are particularly suitable for this kind of topic. +We will create a Github Action that compares the title of an assignment proposal to the selected student works of previous years and create reading suggestions for good contributions on similar topics. In addition, we will check whether a comparison with **all** previous pull requests leads to an even more useful result. The output of the action will be given as a comment to the pull request. We anticipate that this will help students that feel overwhelmed to gain a quick overview of the field and show what approaches are particularly suitable for this kind of topic. Furthermore, this supports the TAs in assessing whether the proposal is too similar to previous ones. From 01116eb2f094f347837f132d745e00cbd434110d Mon Sep 17 00:00:00 2001 From: Neproxx Date: Fri, 25 Mar 2022 16:59:11 +0100 Subject: [PATCH 3/8] Slight modification --- contributions/course-automation/segatz-marcelj/README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/contributions/course-automation/segatz-marcelj/README.md b/contributions/course-automation/segatz-marcelj/README.md index e0004585a4..39fd0b60c0 100644 --- a/contributions/course-automation/segatz-marcelj/README.md +++ b/contributions/course-automation/segatz-marcelj/README.md @@ -20,4 +20,4 @@ Course Automation As the course is repeatedly taught, more and more brilliant contributions from previous students are accumulated that can be leveraged to enhance the quality of new students' contributions. -We will create a Github Action that compares the title of an assignment proposal to the selected student works of previous years and create reading suggestions for good contributions on similar topics. In addition, we will check whether a comparison with **all** previous pull requests leads to an even more useful result. The output of the action will be given as a comment to the pull request. We anticipate that this will help students that feel overwhelmed to gain a quick overview of the field and show what approaches are particularly suitable for this kind of topic. Furthermore, this supports the TAs in assessing whether the proposal is too similar to previous ones. +We will create a Github Action that compares the title of an assignment proposal to the selected student works of previous years and create reading suggestions for good contributions on similar topics. In addition, we will check whether a comparison with **all previously merged** pull requests leads to an even more useful result. The output of the action will be given as a comment to the pull request. We anticipate that this will help students that feel overwhelmed to gain a quick overview of the field and show what approaches are particularly suitable for this kind of topic. Furthermore, this supports the TAs in assessing whether the proposal is too similar to previous ones. From 42a5b3a5727ab24a12567da47a223f15ab504cf4 Mon Sep 17 00:00:00 2001 From: Khashayar Etemadi Date: Mon, 28 Mar 2022 16:41:16 +0200 Subject: [PATCH 4/8] Update README.md --- contributions/course-automation/segatz-marcelj/README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/contributions/course-automation/segatz-marcelj/README.md b/contributions/course-automation/segatz-marcelj/README.md index 39fd0b60c0..bd012982af 100644 --- a/contributions/course-automation/segatz-marcelj/README.md +++ b/contributions/course-automation/segatz-marcelj/README.md @@ -10,7 +10,7 @@ _Automatic comparison of proposals with previous contributions_ ## Deadline -Task 1 +Deadline for Task 1 ## Category From 9e3a49c47f979d38e6c1b424a4df48e1723416dc Mon Sep 17 00:00:00 2001 From: Marcel Date: Mon, 4 Apr 2022 11:44:46 +0200 Subject: [PATCH 5/8] Add essay proposal --- contributions/essay/marcelj/README.md | 22 ++++++++++++++++++++++ 1 file changed, 22 insertions(+) create mode 100644 contributions/essay/marcelj/README.md diff --git a/contributions/essay/marcelj/README.md b/contributions/essay/marcelj/README.md new file mode 100644 index 0000000000..a8de43a143 --- /dev/null +++ b/contributions/essay/marcelj/README.md @@ -0,0 +1,22 @@ +# Assignment Proposal + +## Title + +Leveraging Kubernetes with Kubeflow to Streamline Machine Learning Processes + +## Names and KTH ID + - Marcel Juschak (marcelj@kth.se) + +## Deadline + +Task 2 + +## Category + +Essay + +## Description + +The term MLOps gained significance because of the many steps that are involved in the machine learning (ML) development and deployment process. These complicated processes pose a great challenge which is adressed by Kubeflow that leverages Kubernetes, which is a highly demanded technology in the ML field. + +Firstly I want to give an overview of what Kuebernetes is and why it became so popular. Secondly, I want to to describe the Kubeflow platform that builds on Kubernetes and aims at enhancing the process of developing and deploying ML systems. There is a natural hierachy: Kubeflow builds on Kubernetes, Kubernetes builds on Docker. Therefore, I will explain all three technologies, dedicating roughly 50% of the content to Docker+Kubernetes and Kubeflow respectively. The essay will give a coherent picture of how DevOps technologies can simplify the machine learning process. From 40e39730dfb86abf06e47d15c1bc9908ac96af77 Mon Sep 17 00:00:00 2001 From: Marcel Date: Tue, 5 Apr 2022 21:18:44 +0200 Subject: [PATCH 6/8] Add Khalid to proposal --- .../executable-tutorial/marcelj/README.md | 23 +++++++++++++++++++ 1 file changed, 23 insertions(+) create mode 100644 contributions/executable-tutorial/marcelj/README.md diff --git a/contributions/executable-tutorial/marcelj/README.md b/contributions/executable-tutorial/marcelj/README.md new file mode 100644 index 0000000000..c305b5c7f9 --- /dev/null +++ b/contributions/executable-tutorial/marcelj/README.md @@ -0,0 +1,23 @@ +# Assignment Proposal + +## Title + +Massively cutting server costs for model training with spot instances on Azure. + +## Names and KTH ID + - Marcel Juschak (marcelj@kth.se) + - Khalid El Yaacoub (khalidey@kth.se) + +## Deadline + +Task 3 + +## Category + +Executable Tutorial + +## Description + +Training a machine learning (ML) model is one of the core components of MLOps, e.g. for continuous deployment. However, training a model can require high-end hardware resources over many days which in turn leads to high monetary costs. Spot instances on cloud platforms like Azure are servers that can be rented for usually 10% to 25% of the original price. As a downside, access to the instance may be withdrawn at any point after a 30 second notice. Model training can take advantage of spot instances by checkpointing the training state so that training can be resumed from a checkpoint after termination and restart of the server. + +In this tutorial we will show how to create a docker container that trains an ML model with checkpointing and resumes training after random termination + automatic restart of the Azure spot instance. From 25dd254b0fbbecc9e369041d58756ae8afa797be Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?C=C3=A9sar=20Soto=20Valero?= Date: Wed, 6 Apr 2022 11:36:19 +0200 Subject: [PATCH 7/8] Delete README.md --- contributions/essay/marcelj/README.md | 22 ---------------------- 1 file changed, 22 deletions(-) delete mode 100644 contributions/essay/marcelj/README.md diff --git a/contributions/essay/marcelj/README.md b/contributions/essay/marcelj/README.md deleted file mode 100644 index a8de43a143..0000000000 --- a/contributions/essay/marcelj/README.md +++ /dev/null @@ -1,22 +0,0 @@ -# Assignment Proposal - -## Title - -Leveraging Kubernetes with Kubeflow to Streamline Machine Learning Processes - -## Names and KTH ID - - Marcel Juschak (marcelj@kth.se) - -## Deadline - -Task 2 - -## Category - -Essay - -## Description - -The term MLOps gained significance because of the many steps that are involved in the machine learning (ML) development and deployment process. These complicated processes pose a great challenge which is adressed by Kubeflow that leverages Kubernetes, which is a highly demanded technology in the ML field. - -Firstly I want to give an overview of what Kuebernetes is and why it became so popular. Secondly, I want to to describe the Kubeflow platform that builds on Kubernetes and aims at enhancing the process of developing and deploying ML systems. There is a natural hierachy: Kubeflow builds on Kubernetes, Kubernetes builds on Docker. Therefore, I will explain all three technologies, dedicating roughly 50% of the content to Docker+Kubernetes and Kubeflow respectively. The essay will give a coherent picture of how DevOps technologies can simplify the machine learning process. From a7fefb068e9a0cb740f9ea6b00ea3392a137d271 Mon Sep 17 00:00:00 2001 From: Marcel Date: Wed, 6 Apr 2022 22:07:35 +0200 Subject: [PATCH 8/8] Correct naming of proposal folder --- .../executable-tutorial/{marcelj => marcelj-khalidey}/README.md | 0 1 file changed, 0 insertions(+), 0 deletions(-) rename contributions/executable-tutorial/{marcelj => marcelj-khalidey}/README.md (100%) diff --git a/contributions/executable-tutorial/marcelj/README.md b/contributions/executable-tutorial/marcelj-khalidey/README.md similarity index 100% rename from contributions/executable-tutorial/marcelj/README.md rename to contributions/executable-tutorial/marcelj-khalidey/README.md