From 3b2f6aeae8dd05b1ea22e9a94958168e10e6d2ba Mon Sep 17 00:00:00 2001 From: Andrea Cumpelik Date: Fri, 13 Mar 2026 15:55:19 -0400 Subject: [PATCH 1/3] add QOS-based GPU allocation info for job submissions --- docs/Resources/hpc.md | 3 +++ 1 file changed, 3 insertions(+) diff --git a/docs/Resources/hpc.md b/docs/Resources/hpc.md index 9d32c8d..45911aa 100644 --- a/docs/Resources/hpc.md +++ b/docs/Resources/hpc.md @@ -49,6 +49,9 @@ To get started visit this page for [some info](https://docs.ycrc.yale.edu/cluste 5. `mpi`: for highly-parallelized code 6. `pi_NAME`: PI and lab specific nodes available for purchace from YCPC +### QOS-Based GPU Allocation +By default, jobs are submitted with standard priority. This means that when cluster resources are scarce, these jobs can be preempted, or terminated, in favor of high-priority jobs. This isn't generally a problem for shorter jobs, but for long jobs that are resource-intensive, it may make sense to submit a high-priority job using a SLURM mechanism called QOS (Quality of Service). Our lab has a certain number of guaranteed resources which you can gain access by using the `--qos=qos_levenstein` along with `salloc` or `sbatch`. See the [Misha Guide](https://ood-misha.ycrc.yale.edu/public/misha/04-qos.html) for more info. It's also recommended to use checkpoints in case a job does get terminated. + ### Cheat Sheet * Interactive jobs: From 32904bc38b39fba77a61c13399a541e8b02f9214 Mon Sep 17 00:00:00 2001 From: Dan Levenstein Date: Mon, 16 Mar 2026 14:58:09 -0400 Subject: [PATCH 2/3] Add tip for presenting to the least knowledgeable audience Added a tip about presenting to the least knowledgeable person in the room to enhance audience understanding. --- docs/Resources/presentations.md | 2 ++ 1 file changed, 2 insertions(+) diff --git a/docs/Resources/presentations.md b/docs/Resources/presentations.md index 7d6a762..41783c3 100644 --- a/docs/Resources/presentations.md +++ b/docs/Resources/presentations.md @@ -53,6 +53,8 @@ When presenting, your job is not to get through your slides. It's to make sure e **Watch good talks and PAY ATTENTION.** You'll learn more about presenting from actively watching great speakers than from any list of tips (including this one). Pay attention to how they structure their story, how they use their slides, how they handle transitions, how they explain figures. +**Present to the least knowledgable person in the room.** A common mistaken assumption most students have is that everyone else knows everything they know, and more. However, most of your audience (including professors!) will not know everything you know about your project or its background. If the students in the audience come away understanding your project, it's motivation, and results, that's a win. Present as if you're presenting for THEM, not as if you're presenting to the one expert. + ### Structuring a Talk (Beyond Lab Meeting) From 7215d6dc5d2bd9e133c4cd6c0d881b89b73bbc90 Mon Sep 17 00:00:00 2001 From: Andrea Cumpelik Date: Mon, 16 Mar 2026 17:12:39 -0400 Subject: [PATCH 3/3] Typo --- docs/Resources/presentations.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/docs/Resources/presentations.md b/docs/Resources/presentations.md index 41783c3..2e792eb 100644 --- a/docs/Resources/presentations.md +++ b/docs/Resources/presentations.md @@ -53,7 +53,7 @@ When presenting, your job is not to get through your slides. It's to make sure e **Watch good talks and PAY ATTENTION.** You'll learn more about presenting from actively watching great speakers than from any list of tips (including this one). Pay attention to how they structure their story, how they use their slides, how they handle transitions, how they explain figures. -**Present to the least knowledgable person in the room.** A common mistaken assumption most students have is that everyone else knows everything they know, and more. However, most of your audience (including professors!) will not know everything you know about your project or its background. If the students in the audience come away understanding your project, it's motivation, and results, that's a win. Present as if you're presenting for THEM, not as if you're presenting to the one expert. +**Present to the least knowledgable person in the room.** A common mistaken assumption most students have is that everyone else knows everything they know, and more. However, most of your audience (including professors!) will not know everything you know about your project or its background. If the students in the audience come away understanding your project, its motivation, and results, that's a win. Present as if you're presenting for THEM, not as if you're presenting to the one expert. ### Structuring a Talk (Beyond Lab Meeting)