Releases: leptonai/leptonai
Releases · leptonai/leptonai
0.20.4
0.20.3
What's Changed
- fix(sdk): add f-string to properly print error info by @Yangqing in #390
- fix(cli/api): make the Mount spec up to date with backend by @Yangqing in #391
- fix: add by_alias when creating job by @eahydra in #393
- fix: lint by @bddppq in #394
- feat: add some missing fields of job by @eahydra in #395
- release: 0.20.3 by @bddppq in #392
New Contributors
Full Changelog: 0.20.2...0.20.3
0.20.2
0.20.1
0.20.0
What's Changed
- fix(sdk): fix a bug in get_file_content that forgot to honor return_file by @Yangqing in #371
- chore(sdk): record pydantic version and double check its consistency by @Yangqing in #372
- chore(sdk): minor bugfix in config parameters by @Yangqing in #373
- feat(sdk): added functionality to force install pydantic and cloudpickle to match version at photon creation time. by @Yangqing in #374
- feat: update benchmark script to support fractional qps by @bddppq in #377
- fix(cli): fix return json version difference by @Yangqing in #380
- feat(benchmark): add criteria for no-ttft intertoken latency by @Yangqing in #381
- fix(cli): allow dep update to honor public/private photon namespace. by @Yangqing in #382
- fix: pod creation by @bddppq in #383
- release: 0.20.0 by @bddppq in #384
Full Changelog: 0.19.0...0.20.0
0.19.0
What's Changed
- chore(config): add proper resource shapes by @Yangqing in #365
- feat(cli): update deployment spec to match recent system updates by @Yangqing in #366
- feat(sdk): add http method to openapi schema, and added client support multiple methods for a path. by @Yangqing in #367
- fix: should read deployment name from metadata by @ccding in #368
- release: 0.19.0 by @bddppq in #369
New Contributors
Full Changelog: 0.18.3...0.19.0
0.18.3
What's Changed
- feat(whisperx): add align_only option with predefined text by @Yangqing in #339
- Add separate image with pre-installed sd webui by @bddppq in #340
- Split hf photons runtime requirements to extra dependencies by @bddppq in #341
- feat(photon): use full url for deployment endpoints, instead of the earlier header field by @Yangqing in #347
- release: 0.18.0 by @bddppq in #348
- doc(worker): add doc for worker setup by @bobmayuze in #349
- feat: update vllm photon to work with new version of vllm (0.3.3) by @bddppq in #350
- release: 0.18.1 by @bddppq in #351
- feat(sdk): add universal tokenizer readme by @Yangqing in #352
- feat(hf): support token-classification by @Yangqing in #355
- fix(hf): numpy only exists in runtime. by @Yangqing in #357
- release: 0.18.2 by @bddppq in #356
- feat(cli): add last modified time in object store by @Yangqing in #359
- fix: run on_task in parallel with worker_max_concurrency by @bddppq in #360
- feat(hf): automatically infer dependency from huggingface repo by @Yangqing in #361
- fix(cli): fix job port bug by @Yangqing in #362
- feat(cli): add a functionality to allow specifying additional dependencies by @Yangqing in #363
- release: 0.18.3 by @bddppq in #364
Full Changelog: 0.17.1...0.18.3
0.17.1
What's Changed
- chore(docs): update llm by lepton var with optional labels by @bobmayuze in #308
- fix(docs): add doc for benchmark test by @bobmayuze in #296
- Doc update rsync by @bobmayuze in #307
- Update hf_dependencies.py by @axissun1 in #311
- release: 0.17.0 by @bddppq in #310
- Update README: python -> Python by @AtomicVar in #315
- docs: update templates doc by @vthinkxie in #316
- chore: storrage to storage by @xudong963 in #317
- Support specifying prompt length in benchmark by @bddppq in #322
- fix(sdk): fix KV api json access by @Yangqing in #326
- feat(sdk): make graceful timeout configurable by package-level defaults by @Yangqing in #323
- fix(sdk): public photon deletion now supported. by @Yangqing in #324
- feat(sdk): set graceful timeoout default for on-platform deployments by @Yangqing in #327
- feat(job): complete the job arguments per the new spec. by @Yangqing in #328
- fix(job): change cli entry to /bin/bash by @bobmayuze in #329
- fix(whisperx): allow passing in empty string for language auto-detection, and keep default to English. by @Yangqing in #330
- Revert "pre-install hf-transfer for faster download" by @Yangqing in #331
- remove limits from python dep by @bddppq in #333
- minor tuning of grace period param by @Yangqing in #335
- release: 0.17.1 by @bddppq in #334
New Contributors
- @axissun1 made their first contribution in #311
- @AtomicVar made their first contribution in #315
- @xudong963 made their first contribution in #317
Full Changelog: 0.16.0...0.17.1