skypilot-org

skypilot-org/skypilot

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

10.2K stars 1.1K forks 10.2K watchers Python Apache License 2.0
cloud-computing cloud-management cost-optimization deep-learning distributed-training gpu hyperparameter-tuning job-queue job-scheduler llm-serving llm-training machine-learning ml-infrastructure ml-platform mlops multicloud slurm spot-instances tpu
View on GitHub Website
89 Open Issues Need Help Last updated: Jul 1, 2026

Open Issues Need Help

View All on GitHub
.skyignore file does not respect negate [!] 5 days ago
good first issue good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
New Azure cloud bucket (container) created in different region from cluster 6 days ago
good first issue Stale good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[GCP] Support g4 Instances 6 days ago
good first issue Stale good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[Core] nvcr dynamo image doesn't work on SkyPilot 22 days ago
good first issue Stale good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[GCP] Move from `gsutil cp/rsync` to `gcloud storage cp/rsync` 24 days ago
good first issue good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[Storage] CoreWeave/Nebius/R2/VastData file_mount sync fails on images that already provide awscli 24 days ago
good first issue good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[Kubernetes] GPU prefix matching should not allow non-existent GPU variants 25 days ago
good first issue good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[Azure] _create_storage_account passes a raw dict instead of a typed model, breaking on azure-mgmt-storage 25.0.0 28 days ago
good first issue
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[RunPod] FUSE mount fails inside Docker container - /dev/fuse: operation not permitted about 1 month ago
good first issue good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[UX] Time in `sky jobs queue` CLI is incorrect about 2 months ago
good first issue Stale good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[Dashboard] when an API request fails, the dashboard should show the error message, e.g., the cluster or jobs page about 2 months ago
good first issue Stale good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[UX] Add server side filter for user or request name for `sky api status` about 2 months ago
good first issue Stale good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
serve: autoscaler `latest_version` resets on controller restart, causing scale churn 2 months ago
good first issue good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[Dashboard] Delete cluster / jobs from the history 2 months ago
good first issue good idea
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[Dev] Migrate to `ty` from `mypy` 3 months ago
good first issue Stale good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[Jobs] Show jobs names in the confirmation message for `sky jobs cancel` 3 months ago
good first issue Stale good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[Core] SkyPilot base image is missing infiniband dependencies 3 months ago
good first issue good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
Runpod catalog mismatch? 3 months ago
good first issue good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
Mark a storage mount as read-only, even if user has write permissions to the backing store 3 months ago
good first issue Initial-User-Issue
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[AWS] Add tags to security groups for scoped IAM policies 3 months ago
good first issue good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[Dashboard] GCP VM instance links point to wrong VM names 4 months ago
good first issue good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[UX] Hint for misaligned config from client and server appear twice 4 months ago
good first issue Stale good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[Dashboard] Safari prompt for username and password even included in URL 4 months ago
good first issue Stale good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[UX] Support `--secret-file` and automatically populate the secret in `--env-file` to secrets defined in the task 5 months ago
good first issue good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
Duplicate p4de.24xlarge entries in AWS catalog cause AssertionError 5 months ago
bug good first issue good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
FAILED_PROVISION is bloating the sky serve status output 5 months ago
good first issue Stale serve
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
Passing `--network=host` in run options lead to error `docker: network "host" is specified multiple times` 5 months ago
good first issue Stale
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[Core] Add env var that is accessible by cluster jobs/managed jobs indicating the user launched the job 5 months ago
good first issue good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
not auto-recovering from terminated runpod pod 5 months ago
good first issue Stale good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[Docker] silently not failing to pass SSH command 6 months ago
good first issue good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[SSH] After `helm upgrade`, `sky ssh up` needs to be run again 6 months ago
good first issue P0 good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[Core] No way to specify candidate resources with different num_nodes 6 months ago
good first issue Stale good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[core] Resources specified as a dict work, but not as a list 6 months ago
good first issue Stale
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[Dashboard] Log download from the cluster/managed jobs should be compressed 6 months ago
good first issue good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
`sky down -p cluster` fails to terminate a cluster without the cloud credential 6 months ago
good first issue Stale good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[Serve] Regression by HTTPS Support: Double Protocol in Endpoint URL 6 months ago
bug good first issue Stale good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[AWS] `sky launch --infra aws -t p5e.48xlarge` fails with `Try specifying a different CPU count` 6 months ago
good first issue good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[Pool] Logs and dashboard show L4:1 instead of 0.5 when specifying L4:0.5 for jobs on a pool 6 months ago
good first issue good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[Docs] Update examples in docs to use `uv` and validate them 6 months ago
good first issue Stale good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[Vast] SSH not working on vast instances 6 months ago
good first issue Stale good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[UX] Show a user is a service account in `sky api status` 6 months ago
good first issue good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[SDK] Programmatic way to add docker credential to a Resource/Task with DockerLoginConfig 7 months ago
good first issue Stale good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[API server] sky api start --deploy does not honor host flag when api_server.endpoint is set 7 months ago
good first issue Stale good starter issues api server
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[Dashboard] Volume page should utilize caches, instead of always refresh 7 months ago
good first issue good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[Dashboard] Show how old the info are for the cluster / jobs are, i.e. how old the cache is, so people now if they need to refresh 7 months ago
good first issue good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[k8s] support `get_label_value` for GFDLabelFormatter 7 months ago
good first issue Stale good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[API server] Do not run as root in helm deployment by default 7 months ago
good first issue Stale good starter issues api server
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
AWS Dependenices pin Colorama to a lower version 7 months ago
good first issue
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[Dashboard] Do not round down < 1 CPU in resources column 7 months ago
good first issue good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[Examples] Move pip installs to uv for faster setup 7 months ago
documentation good first issue Stale
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
Progress bar of sky launch 7 months ago
good first issue Stale good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[API Server] Add a way to search for requests for a specific cluster 7 months ago
good first issue good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
Bypass port-forward if API server and cluster runs in same kubernetes cluster 7 months ago
enhancement good first issue feature-request Stale good starter issues api server
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
RunPod spot instances more expensive through SkyPilot 7 months ago
good first issue Stale good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[k8s] gpu labeler should be callable with remote API server 7 months ago
good first issue good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[UX] launching on existing cluster with the same `sky launch` command fails 8 months ago
good first issue Stale good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[k8s] Pod using latest image tag is launched with IfNotPresent imagePullPolicy 8 months ago
bug good first issue Stale good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[k8s] `sky start/launch` fails if kubeconfig is missing and k8s is enabled 8 months ago
good first issue Stale good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[API server] API server needs to be restarted to pick up credentials newly setup 8 months ago
good first issue Stale good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[Dependency] GCP package showing warning 8 months ago
good first issue good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[UX] More informative `ResourceMismatchError` 8 months ago
good first issue Stale good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[SSH Node Pool] Infra column in dashboard shows SSH context with a `ssh-` prefix 8 months ago
good first issue good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[Storage] Refactor storage (e.g., S3Store/R2Store) 8 months ago
help wanted Stale
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[SSH Node Pools] `lstrip` to strip `ssh-` is a bug 8 months ago
help wanted good first issue good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[Spot/Serve] Optimize the translation of filemounts 8 months ago
good first issue Stale spot
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[k8s] Unexpected error when relaunching an INIT cluster on k8s which failed due to capacity error 9 months ago
good first issue Stale good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[Jobs] Big overhead for triggerring `check` for every `sky jobs launch` when no storage cloud is enabled 9 months ago
good first issue Stale good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[UX] `sky show-gpus --infra k8s` fails during the kubernetes cluster GPU node pool is scaling up 9 months ago
good first issue good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[k8s] Deprecate `nodeport` networking mode 9 months ago
good first issue Stale
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
Ulimit low on MacOS, but incorrect way to update it 9 months ago
good first issue Stale interface/ux good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[Examples] Improve nemo rl examples 10 months ago
good first issue good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[Performance] Investigate `sky exec` speed on multi-node cluster 10 months ago
help wanted Stale core
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[Resources] Resource yaml cannot use `k8s` as a valid name 10 months ago
good first issue good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[ux] Sky down output stops if one cluster fails 10 months ago
good first issue good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[Core] Unblock user program for SIGINT 10 months ago
good first issue Stale triage good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[GCP] Fail to handle already terminated cluster correctly in `sky down` 10 months ago
good first issue Stale good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[k8s] Support non-debian custom images 10 months ago
good first issue Stale k8s good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
Support barrier for in the run section 10 months ago
good first issue feature-request Stale good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
`ordered` does not take `disk_size` into account 10 months ago
good first issue Stale good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[k8s] Prevent mounting of /dev/shm in pods 10 months ago
good first issue Stale k8s good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[UX] Auto completion in `sky check` for cloud name 10 months ago
good first issue Stale
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[Dashboard] 0cpu 0 mem display 10 months ago
good first issue good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[Core][Spot] Change Spot Dashboard from port forwarding to open ports 11 months ago
good first issue Stale
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[UX] Support completion of partially specified instance type 11 months ago
good first issue Stale
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[Catalog] Lambda catalog fetching failed due to recently added B200 11 months ago
good first issue good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[k8s] Add validation for `pod_config` 11 months ago
help wanted k8s
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[UX] shell complete fails if cookie is needed for the API requests 11 months ago
good first issue good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
sudo: yum: command not found with MOUNT_CACHED 11 months ago

AI Summary: The user is experiencing a `sudo: yum: command not found` error while using SkyPilot to mount an S3 bucket. The issue stems from the SkyPilot script attempting to use `yum` (a Red Hat package manager) on a system that uses `apt` (a Debian package manager). The solution involves modifying the SkyPilot mount script to use `apt` instead of `yum` for installing `rclone` on Debian-based systems, ensuring compatibility with the underlying OS of the launched instance.

Complexity: 3/5
good first issue good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu
[UX] unexpected output in sky status 11 months ago

AI Summary: The task is to fix a user experience issue in the SkyPilot `sky status` command. The command unexpectedly prints a message about adding a default admin role to a user, which should only be logged to the server log file. The solution involves modifying the code to redirect this message to the appropriate log instead of standard output.

Complexity: 3/5
bug good first issue interface/ux good starter issues
skypilot-org/skypilot
10.2K

Run, manage, and scale AI workloads on any AI infrastructure. Use one system to access & manage all AI compute (Kubernetes, Slurm, 20+ clouds, on-prem).

Python
#cloud-computing#cloud-management#cost-optimization#deep-learning#distributed-training#gpu#hyperparameter-tuning#job-queue#job-scheduler#llm-serving#llm-training#machine-learning#ml-infrastructure#ml-platform#mlops#multicloud#slurm#spot-instances#tpu