0.1.0rc1

What's Changed

• Scaffolding for nos (pyproject.toml, Makefile, pre-commit) by @spillai in #2
• Added test data to tests/test_data #9 by @spillai in #12
• Adding Github Actions CI scaffolding #11 by @spillai in #13
• Moved to mamba forge for cached builds by @spillai in #15
• Added pypi project autonomi-nos with new makefile target by @spillai in #16
• New NOS CLI entrypoint for downloading, optimizing models by @spillai in #18
• Added new base dockerfile and makefile targets for py39 and trt by @spillai in #19
• Updated README with badges and key features by @spillai in #23
• Improved pre-commit with ruff linting support by @spillai in #24
• New nos cli subcommands and sample system info by @spillai in #25
• Add self-hosted gpu benchmark tests by @spillai in #27
• Trigger self-hosted benchmark manually with /benchmark comment in PR by @spillai in #28
• nos hub registry + example stable diffusion model registered by @spillai in #26
• Add nos logger and improved default error logging by @spillai in #29
• Add openai/clip-vit and laion/CLIP-ViT models with tests/benchmarks by @spillai in #30
• Persist conda env for mmdeploy flow by @outtanames in #31
• Rename mmdeploy conda environment to mmdeploy-cuda118 by @spillai in #32
• Experimental ray-serve based model deployment by @spillai in #48
• Experimental gRPC-based inference server by @spillai in #50
• Moved cuda base image to cuda:11.8.0-base-ubuntu22.04 by @spillai in #58
• Reduce docker image sizes for cpu/gpu (up to 2x) by @spillai in #59
• gRPC serve CLI with txt2img, txt2vec and img2vec prediction examples by @spillai in #60
• MacOS support for make develop-cpu by @spillai in #61
• Configs for mmdetection efficientdet by @spillai in #63
• Efficientdet endpoint by @outtanames in #65
• Add ray executor as a daemon process for background serving/inference by @spillai in #49
• Docker executor daemon for background execution of NOS jobs. by @spillai in #51
• Add endpoint for faster-rcnn by @outtanames in #73
• Torch FasterRCNN endpoint by @outtanames in #75
• Dockerized inference runtime with model-multiplexing by @spillai in #71
• New nos.client module for simplified gRPC client interface by @spillai in #76
• Build cpu docker container as part of CI actions to check conda env by @outtanames in #82
• TensorRT conversion flow for Faster-RCNN by @outtanames in #66
• Refactor nos/server and nos/client modules for better extensibility. by @spillai in #83
• nos CPU and GPU docker builds with tests by @spillai in #91
• Refactor requirements for cpu/gpu builds by @spillai in #96
• Client docs with mkdocs by @spillai in #95
• make update-conda by @outtanames in #84
• M1 Mac support for autonomi/nos:latest-cpu by @spillai in #97
• Exclude nos/experimental from wheel build by @outtanames in #90
• Use docker tmp volume for ray storage in cpu/gpu docker builds by @outtanames in #99
• docker compose command->entrypoint by @outtanames in #100
• Add a quickstart guide and docker-compose by @outtanames in #101
• Updates to README, ROADMAP and fixes to gRPC e2e tests by @spillai in #98
• Updates to readmes: QUICKSTART, ROADMAP and docs scaffolding by @spillai in #102
• Fix for ray-gcs server errors on multiple restarts by @spillai in #103
• Improved dockerfile builds with caching and fix for proto includes in… by @outtanames in #104
• Bump to 0.0.3 by @outtanames in #106
• Updated client-side requirements with full testing (client, e2e, … by @spillai in #107
• Client-side WaitForServer() and IsHealthy() by @spillai in #108
• [docs] Public mkdocs site for client API by @spillai in #109
• Add CLI tests for txt2vec (CPU), txt2img (GPU), img2bbox (GPU) by @outtanames in #114
• Give CPU tests a non-parametrized name to see if it shows up in gatin… by @outtanames in #122
• Rename cpu tests to allow it to be used for gating tests by @outtanames in #123
• [ci] Fix name indentation for CI by @spillai in #127
• Refactored prediction API with ModelSpec and TensorSpec signatures by @spillai in #125
• Update requirements.txt with pydantic for nos-client by @spillai in #121
• New type system for input/output specification by @spillai in #128
• Updates to docs and tests for 0.0.4a3 release by @spillai in #131
• Simplify type annotations for ImageT, TensorT and Batch by @spillai in #132
• Refactor InferenceServiceRuntime and DockerRuntime by @spillai in #133
• Release version 0.0.4 by @spillai in #139
• Simplify server-side task dispatch without switch-case by @spillai in #145
• Add segmentation service with SAM by @outtanames in #126
• Fully functional runtime environment for mmdet models with tests by @spillai in #149
• ModelSpec serialization to_proto bugfix by @spillai in #151
• [benchmark] Fix benchmark tests for 0.0.5a0 by @spillai in #153
• Segmentation test/fixes by @outtanames in #154
• Basic opencv-based video reader for I/O by @spillai in #152
• Refactored InferenceServiceImpl with new ModelManager by @spillai in #155
• Batched inference tests for all image-based models by @spillai in #157
• Updated various benchmark tests with consistent comments/fixtures by @spillai in #159
• Fixed FasterRCNN scores passthrough by @spillai in #160
• Version 0.0.5a1 bump / release by @spillai in #161
• [hub] New YOLOX model with unified object detection tests / benchmarks by @spillai in #163
• Fix allowable image resolutions forr yolox and faster rcnn by @spillai in #167
• Deregister SAM by @outtanames in #166
• NOS v0.0.5 by @outtanames in #168
• Makefile target for creating tags by @spillai in #169
• Update pyproject.toml to exclude build, experimental etc. by @outtanames in #170
• Set resource limits for nos container to avoid cached raylets by @spillai in #174
• Refactor stable diffusion with more models and ControlNet guides by @spillai in #173
• New makefile targets for benchmark tests (CPU/GPU) w/ docker variants by @spillai in #176
• Improved docker CLI with simplified nos docker stop and nos docker logs by @spillai in #178
• Torch FX TensorRT Conversion for YOLOX by @outtanames in #177
• OpenCV-based video writer by @spillai in #180
• Inference client demo with notebook by @spillai in #179
• Refactored model manager and cache model spec in inference service by @spillai in #184
• CLIP compilation workflow with Torch TensorRT with tests by @spillai in #189
• Updated notebooks with inference client and acceleration ...

0.0.10

What's Changed

• Add arch diagram to arch overview by @outtanames in #296
• Discord bot for image generation by @outtanames in #298
• New mmdetection interface with custom weights support by @spillai in #309
• New model spec metadata format with runtime and model resource information by @spillai in #268
• Updates to various docs for 0.0.10 release. by @spillai in #310
• New super-resolution models based on LDM and Swin2SR by @spillai in #311
• SDv2 Dreambooth LoRA fine-tuning API by @spillai in #312
• Discord bot with NOS fine-tuning API by @spillai in #314
• Support for nos.init(runtime="local") by @spillai in #318
• Fine-tuning API docs with YOLOX/Dreambooth support by @spillai in #319
• Whisper and whisperx models for speech-to-text and transcriptions by @spillai in #317
• Support manual weights path for civit models by @outtanames in #320
• Added benchmarks docs section by @spillai in #322
• Add Civit image generation to bot by @outtanames in #323
• SDXL model support by @outtanames in #331
• Fixes for 0.0.10 release by @spillai in #335

Full Changelog: 0.0.9...0.0.10

0.0.9

What's Changed

• Updates to API docs, and mkdocs config by @spillai in #280
• Hotfix for MacOS shared memory transport by @spillai in #282
• 0.0.9 docs updates v2 by @spillai in #283
• Add docs assets for 0.0.9 release by @spillai in #284
• Enable multi-processing for ray workers with OMP_NUM_THREADS>1 by @spillai in #291

Full Changelog: 0.0.8...0.0.9

0.0.8

What's Changed

• Enable Memray tracking by @outtanames in #229
• More granular/useful client exception types by @outtanames in #245
• Exclude example notebooks from pre-commit json formatting by @outtanames in #254
• Re-register SAM by @outtanames in #246
• Update pixeltable integration test by @outtanames in #253
• Update readme to use new server launch setup by @outtanames in #251
• Reduce nos.init() startup time if container is already running by @spillai in #264
• CI integrations for nos and pixeltable by @spillai in #263
• Add nos profiles for 0.0.7 and 0.0.7a1 (2080 and 2080Ti) by @spillai in #262
• [hub] Add MIDAS monodepth model by @spillai in #164
• Add SAM demo, mesh grid for autosegmentation by @outtanames in #257
• Raise NosInputValidationException on input count mismatch, Fix Pydantic version by @outtanames in #269
• Support user-defined models with custom runtime-environments by @spillai in #270
• Improved ModelHandle with multiple-replicas and task dispatch by @spillai in #241
• Fix NOS_MEMRAY_ENABLED flags with defaults by @spillai in #272
• Include image resolution in CLIP config by @outtanames in #271
• Move exception types to common by @outtanames in #274
• Recommend better batch sizes for all NOS models by @spillai in #275
• Spruce up main README with badges and links to docs and blog by @spillai in #277

Full Changelog: 0.0.7...0.0.8

0.0.7

What's Changed

• Update grpcio-tools>=1.51.3 by @spillai in #216
• Link nvidia docker instructions in quickstart by @outtanames in #222
• Add noop grpc call with image input for benchmarking by @outtanames in #221
• NOS profiler with notebook report by @spillai in #219
• Pixeltable integration tests with no-op/yolox benchmarks by @spillai in #228
• Improved nos.init(logging_level=...) with verbose debug logs by @spillai in #230
• 0.0.7a2 release by @spillai in #231
• Shared memory transport for gRPC by @spillai in #224
• Resolve issue with version import by @outtanames in #240
• Warn users about shm allocation rate if exceeds max-rate by @spillai in #242
• Re-initialize new object ids and namespace on RegisterSharedMemory() by @spillai in #244

Full Changelog: 0.0.6...0.0.7

0.0.6

What's Changed

• Makefile target for creating tags by @spillai in #169
• Update pyproject.toml to exclude build, experimental etc. by @outtanames in #170
• Set resource limits for nos container to avoid cached raylets by @spillai in #174
• Refactor stable diffusion with more models and ControlNet guides by @spillai in #173
• New makefile targets for benchmark tests (CPU/GPU) w/ docker variants by @spillai in #176
• Improved docker CLI with simplified nos docker stop and nos docker logs by @spillai in #178
• Torch FX TensorRT Conversion for YOLOX by @outtanames in #177
• OpenCV-based video writer by @spillai in #180
• Inference client demo with notebook by @spillai in #179
• Refactored model manager and cache model spec in inference service by @spillai in #184
• CLIP compilation workflow with Torch TensorRT with tests by @spillai in #189
• Updated notebooks with inference client and acceleration examples by @spillai in #190
• Accelerated SDv2 implementation with VAE, UNet, and CLIP by @spillai in #192
• YOLOX compilation workflow with new Torch TRT compilation backend by @spillai in #195
• CLIP compilation workflow with tests by @spillai in #196
• Overhaul system information and benchmarking instrumentation for prof… by @spillai in #201
• Adding MacOS testing to CI by @spillai in #203
• Add init inference service utility method by @spillai in #204
• Pull server docker image on nos.init with optional arg by @spillai in #205
• Version bump 0.0.6a0 -> 0.0.6a1 by @spillai in #206
• Mac M1 fixes for nos.init by @spillai in #207

Full Changelog: 0.0.5...0.0.6

0.0.5

What's Changed

• Simplify server-side task dispatch without switch-case by @spillai in #145
• Add segmentation service with SAM by @outtanames in #126
• Fully functional runtime environment for mmdet models with tests by @spillai in #149
• ModelSpec serialization to_proto bugfix by @spillai in #151
• [benchmark] Fix benchmark tests for 0.0.5a0 by @spillai in #153
• Segmentation test/fixes by @outtanames in #154
• Basic opencv-based video reader for I/O by @spillai in #152
• Refactored InferenceServiceImpl with new ModelManager by @spillai in #155
• Batched inference tests for all image-based models by @spillai in #157
• Updated various benchmark tests with consistent comments/fixtures by @spillai in #159
• Fixed FasterRCNN scores passthrough by @spillai in #160
• Version 0.0.5a1 bump / release by @spillai in #161
• [hub] New YOLOX model with unified object detection tests / benchmarks by @spillai in #163
• Fix allowable image resolutions forr yolox and faster rcnn by @spillai in #167
• Deregister SAM by @outtanames in #166

Full Changelog: 0.0.4...0.0.5

0.0.4

What's Changed

• Updated client-side requirements with full testing (client, e2e, … by @spillai in #107
• Client-side WaitForServer() and IsHealthy() by @spillai in #108
• [docs] Public mkdocs site for client API by @spillai in #109
• Add CLI tests for txt2vec (CPU), txt2img (GPU), img2bbox (GPU) by @outtanames in #114
• Give CPU tests a non-parametrized name to see if it shows up in gatin… by @outtanames in #122
• Rename cpu tests to allow it to be used for gating tests by @outtanames in #123
• [ci] Fix name indentation for CI by @spillai in #127
• Refactored prediction API with ModelSpec and TensorSpec signatures by @spillai in #125
• Update requirements.txt with pydantic for nos-client by @spillai in #121
• New type system for input/output specification by @spillai in #128
• Updates to docs and tests for 0.0.4a3 release by @spillai in #131
• Simplify type annotations for ImageT, TensorT and Batch by @spillai in #132
• Refactor InferenceServiceRuntime and DockerRuntime by @spillai in #133

Full Changelog: 0.0.3...0.0.4