v0.10.0

0.10.0 (2024-06-05)

Features

• client, resources: adopt latest api spec (#145) (bafa292)
• dockerfile: separate config and weights into different layers (#156) (8d25fc3)
• ray: support cuda version and fix user root permission (#158) (cd1fa69)

Bug Fixes

• clients: fix message length (#147) (8083761)
• resources: fix nil response handling (#151) (c6df35c)

Documentation

• readme: update workflow name (ba36a91)

v0.9.0

0.9.0 (2024-04-30)

Features

• ray: implement standalone executable and optional system packages (#141) (c897080)

v0.8.1

0.8.1 (2024-04-29)

Bug Fixes

• ray: fix missing imports (#139) (02199b1)

v0.8.0

0.8.0 (2024-04-25)

Features

• deps: upgrade ray version (cc61b85)
• ray: adapt to native docker client instead of docker sdk (#138) (7d19ccb)
• ray: add accelerator and custom resource support (#118) (f974f98)
• ray: add llava 13b to predeploy list (3fd5914)
• ray: add metadata and infer constructor for llm tasks (#137) (be122d1)
• ray: generate sha256 as tag if not presented (#120) (6abb538)
• ray: inject accelerator type at runtime (#121) (f78a2d0)
• ray: support containerized model serving (#116) (ad0f250)
• ray: support custom accelerator type (#134) (ae6c139)
• ray: use env for resource and deprecate deploy/undeploy (#124) (a58bc50)
• ray: use tmp folder for image building (#122) (9512cec)

Bug Fixes

• deps: downgrade ray to avoid grpc servicer issue (#128) (9ead421)
• dockerfile: avoid build hang at ARG statement (#130) (f02a27c)
• ray: fix etrypoint module not found (#126) (f1ed83d)
• ray: fix missing default resource value (#129) (b2f564a)
• ray: fix multi-platform build stage (6f358fd)
• ray: support target platform for image building (#127) (f4825fc)

v0.7.1

0.7.1 (2024-02-22)

Bug Fixes

• ray: update default resources allocation for test (#102) (1aaa905)

Miscellaneous Chores

• release: release v0.7.1 (b7dec13)

v0.7.0

0.7.0 (2024-01-30)

Features

• ray: determine ram usage by file size (#89) (7a0023d)
• ray: determine vram usage by file size (#87) (71e84e6)

Bug Fixes

• ray: add vram ceiling and override list (#94) (5804e4a)
• ray: fix application name (f2bb563)
• ray: fix gpu resource > 1 (#91) (b121f56)
• ray: fix ray autoscaling (#95) (5bd8c2a)

v0.6.0

0.6.0 (2024-01-12)

Features

• connector,operator,component: adopt jsonscema validation and dataclass type hint (#74) (0c47e51)
• helpers: add wrapper func for protobuf message (#85) (279bd1e)
• resource: adopt dataclass as config in component for type hinting (#79) (8c25bd1)
• resources: support recipe update in pipeline resource (#83) (89431fa)

Bug Fixes

• resources: fix resource schema path (#76) (25f4418)

Documentation

• notebook: update notebook (e5da642)

v0.5.0

0.5.0 (2024-01-02)

⚠ BREAKING CHANGES

• ray: retire non-decorator deploy and update scaling config (#67)

Features

• model: Update Text-Generation Task Schema to Align with OpenAI Standards (#69) (703789a)

Bug Fixes

• model: fix image decoding issue (#71) (5c571b2)
• ray: allow cwd as runtime env (#66) (00c0497)
• ray: avoid agressive downscale and non upscale (#70) (f159314)

Documentation

• notebooks: add async example (49ca895)

Code Refactoring

• ray: retire non-decorator deploy and update scaling config (#67) (89ef078)

v0.4.0

0.4.0 (2023-12-11)

⚠ BREAKING CHANGES

• client: support asyncio and add better client type hint (#55)

Features

• client: support asyncio and add better client type hint (#55) (aa41246)
• org: adopt organization endpoints for VDP (#60) (0c313d6)
• ray: add io helpers for llm tasks (#59) (1876a20)
• ray: add nested decorators (#63) (900dda8)
• ray: add text to image io helper (#58) (0430977)

Bug Fixes

• clients,resources: fix resource creating will get None type (#57) (4516d46)
• makefile: fix wheel build missing submodule (9aa73eb)
• ray: fix missing ray init (#64) (db0b5c4)

Documentation

• notebooks: update notebooks to adopt latest ray decorators (52b90c6)

v0.3.2

0.3.2 (2023-11-29)

Bug Fixes

• client: fix async pipeline trigger and get user (#53) (c86274b)
• ray: fix mismatched grpcio version (024471a)
• ray: fix model weight file extension (4f3a40c)