4.11.9

4.11.9 (2023-05-01)

Bug Fixes

• wait for ray workers prior to server-side job submit (0d8731f)

4.11.8

4.11.8 (2023-04-28)

Bug Fixes

• add websocat to custodian to avoid having to wget it every time (2334b1d)
• increase resilience to network disconnects, restore helm delete in custodian (5a47eb7)

4.11.7

4.11.7 (2023-04-27)

Bug Fixes

• avoid helm delete in custodian for now (6a12f44)
• increase memory for ray head logs container (85385f1)
• increase memory for runtime-env custodian pod (c7f2f19)

4.11.6

4.11.6 (2023-04-27)

Bug Fixes

• add h5py to pyarrow base image (dd9875f)
• torchx volume mount paths may have extra quotes (5db0410)

4.11.5

4.11.5 (2023-04-27)

Bug Fixes

• improve custodian memory requests for larger jobs (0201c00)
• remove reliance on wget in ray head container (6018e7f)

4.11.4

4.11.4 (2023-04-25)

Bug Fixes

• another fix for loading yoga.wasm (codeflare top exits silently) (3a10932)

4.11.3

4.11.3 (2023-04-24)

Bug Fixes

• add fsspec[s3] to lightning image, and conda clean -afy (3f8f4ed)
• fix for torchx+s3 and lightning+s3 (77637d4)
• improve support for pytorch lightning's fsspec[s3] support (6cdbdd5)

4.11.2

4.11.2 (2023-04-21)

Bug Fixes

• shorter uuids resulted in uppercase kubernetes resource names (f11c0c8)

4.11.1

4.11.1 (2023-04-21)

Bug Fixes

• use shorter uuids, custodian trackers for runtime-env and worker status (ba6eaeb)

4.11.0

4.11.0 (2023-04-20)

Bug Fixes

• force yoga.wasm to be loaded as a file resource (214250a)
• update kind tests to delete new name of cleaner -> logs (1d26251)

Features

• update custodian to track cpu, mem, and gpu utilization (dfbf0a9)