HackerTrans
TopNewTrendsCommentsPastAskShowJobs

covi

no profile record

Submissions

VRAM Ghost Busting: Who You Gonna Close()?

hcompany.ai
1 points·by covi·vor 18 Tagen·0 comments

SkyPilot Sandboxes: Run Agent Code on Your Own Kubernetes, at Scale

blog.skypilot.co
6 points·by covi·letzten Monat·0 comments

RL Doesn't Work on Slurm

blog.skypilot.co
5 points·by covi·vor 2 Monaten·0 comments

SkyPilot: One system to use and manage all AI compute (K8s, 20 clouds, Slurm)

github.com
15 points·by covi·vor 6 Monaten·2 comments

The Evolution of AI Job Orchestration

blog.skypilot.co
2 points·by covi·vor 12 Monaten·0 comments

Large-Scale AI Batch Inference: 9x Faster Embedding Generation

blog.skypilot.co
1 points·by covi·letztes Jahr·0 comments

Abusing SQLite to Handle Concurrency

blog.skypilot.co
2 points·by covi·letztes Jahr·0 comments

Using DeepSeek R1 for RAG: Do's and Don'ts

blog.skypilot.co
13 points·by covi·letztes Jahr·0 comments

Building Large-Scale Image Search Using VectorDB and OpenAI Clip

blog.skypilot.co
3 points·by covi·letztes Jahr·2 comments

Fungible cloud compute and storage: AI training on any cloud

tigrisdata.com
2 points·by covi·vor 2 Jahren·0 comments

Getting $1M cloud credits for AI startups – and using them wisely

blog.skypilot.co
1 points·by covi·vor 2 Jahren·0 comments

You Ran the Operational Database on What? Testing Spot Instances

cockroachlabs.com
3 points·by covi·vor 2 Jahren·1 comments

Can Multi-Modal LLMs "See" Images? A Deep Dive with ASCII Art

blog.skypilot.co
3 points·by covi·vor 2 Jahren·0 comments

Embarrassingly parallel batch jobs on multiple clouds or regions

skypilot.readthedocs.io
1 points·by covi·vor 2 Jahren·0 comments

Llama 3.1 finetuning on your own data and infra, without vendor lock-in

blog.skypilot.co
2 points·by covi·vor 2 Jahren·0 comments

Guide: Finetune Llama 3.1 on your infra

github.com
5 points·by covi·vor 2 Jahren·0 comments

AI on Kubernetes Without the Pain

blog.skypilot.co
2 points·by covi·vor 2 Jahren·0 comments

comments

covi
·vor 4 Monaten·discuss
This feels like the chimpanzee with a power drill. An agent is honestly just brute-force search, but guided.
covi
·vor 5 Monaten·discuss
The post says Slurm supports gang scheduling, k8s doesn't (out of the box).
covi
·letztes Jahr·discuss
Take a look at SkyPilot. Good for running these batch workloads. You can use spot instances to save costs.
covi
·letztes Jahr·discuss
To massively increase the reliability to get GPUs, you can use something like SkyPilot (https://github.com/skypilot-org/skypilot) to fall back across regions, clouds, or GPU choices. E.g.,

$ sky launch --gpus H100

will fall back across GCP regions, AWS, your clusters, etc. There are options to say try either H100 or H200 or A100 or <insert>.

Essentially the way you deal with it is to increase the infra search space.
covi
·letztes Jahr·discuss
Related: https://skyplane.org/en/latest/ (mentioned in OP)

From what I know this idea underpins a few FAANG-level companies' data transfer systems. OP's value = a simple implementation of the idea that's OSS and applied to AI.
covi
·letztes Jahr·discuss
Congrats on the API launch (from SkyPilot)!
covi
·vor 2 Jahren·discuss
If you want to use your own GPUs or cloud accounts but with a great dev experience, see SkyPilot.