v0.3.1

skypilot-org/skypilotv0.3.1Jun 4, 2023by concretevitamin

AI Summary

This is a patch release (v0.3.1) of SkyPilot that adds support for on-demand H100 GPUs from Lambda and includes several bug fixes across provisioner, logging, managed spot, storage, and on-prem GPU detection components.

Key Highlights

On-demand H100 GPU support from Lambda (requires removing previous Lambda catalog)
Fixed provision failover encountering FileNotFoundError (#2005)
Fixed user-level ray cluster causing SkyPilot cluster to be in INIT state (#2020)
Fixed logs of multi-node jobs not being streamed due to Ray 2.4 log deduplication (#2026)
Fixed `sky spot launch --retry-until-up` to actually retry until up (#2004)

New Features

On-demand H100 GPU support from Lambda (accessible via `sky launch --gpus h100`)

Full Release Notes

This is a patch release to ship **several important enhancements and bug fixes**:

Enhancements
- **On-demand H100 GPU from Lambda is supported!** `sky launch --gpus h100`
  - To use it, remove any previous Lambda catalog: `rm -rf ~/.sky/catalogs/v5/lambda`
- Managed spot: make job cancellation during failover more robust to mitigate a rare `FAILED_SETUP` error (#1998) 

Fixes
- Provisioner / Backend
  - Fix provision failover encountering FileNotFoundError (#2005) 
  - Fix user-level ray cluster causing SkyPilot cluster to be in INIT state (#2020) 
- Logging 
  - Fix certain logs of multi-node jobs not being streamed due to Ray 2.4 log dedup (#2026) 
  - Fix logs being created in current pwd `$PWD/~/sky_logs` in some cases (#2009)
- Managed spot
  - Fix `sky spot launch --retry-until-up` to make it actually retry until up (#2004) 
- Storage 
  - Fix a rare storage cloud check error if `sky check` has never been called (#2017)
- On-prem
  - Fix detecting A5000 and A6000 GPUs (#2023)
  
**Full Changelog**: https://github.com/skypilot-org/skypilot/compare/v0.3.0...v0.3.1