v0.20.0

What's Changed

• [docs] Update javadoc link by @xyang16 in #254
• Adds restart feature to management console by @c007456 in #255
• [docs] Update links by @xyang16 in #258
• [Docker][DeepSpeed] use accelerate to reduce memory cost by @lanking520 in #261
• [Docker] bump up torch version by @lanking520 in #264
• [docker] Fix pytorch-cu113 docker file by @frankfliu in #265
• [Docker] support integration test on release images by @lanking520 in #266
• [Docker] make aarch defaults in pytorch by @lanking520 in #267
• Add management console document by @c007456 in #262
• [Docker] fix arg in code by @lanking520 in #268
• [aarch64] Remove unsupported engine by @frankfliu in #269
• [Docker] mkdirs for Inferentia for bug fixing by @lanking520 in #270
• [docker] Fixed pytorch-cu113 docker build by @frankfliu in #271
• [Docker] change copy script location by @lanking520 in #272
• [Docker] upgrade deepspeed and transformers by @lanking520 in #273
• Fix the problem that the returned image of input text cannot be parsed by @c007456 in #276
• update huggingface version by @lanking520 in #277
• [serving] clean cache directory on model unload by @frankfliu in #278
• [python] Fixes logging issue by @frankfliu in #279
• [Docker] switch base image to devel by @lanking520 in #280
• [Docker] Add INT8 support on Large model by @lanking520 in #281
• apply CVE patches into the docker by @lanking520 in #284
• [Docker] minor fix by @lanking520 in #285
• Clean up ModelServerTest with request helper function by @zachgk in #283
• [Docker] add label patching for DLC by @lanking520 in #286
• [serving] Fixes install XGBoost engine bug by @frankfliu in #287
• [serving] Improve logging message by @frankfliu in #289
• [serving] Loads model from root of model_store directory by @frankfliu in #288
• [python] Adds large model inference support with MPI mode by @frankfliu in #291
• [python] Adds built-in DeepSpeed handler by @frankfliu in #292
• [python] Remove DsEngineProvider alias by @frankfliu in #296
• [Docker] add tagging labels by @lanking520 in #297
• [doc] Add some DJL Serving docs by @xyang16 in #299
• Reorganize docs by @zachgk in #298
• [doc] Update DJL Serving doc by @xyang16 in #300
• [Docker] rename parallelformers engine name to transformers by @lanking520 in #301
• [Integration] make 6 min timeout per model by @lanking520 in #304
• [Docker] Fix Accelerate version by @lanking520 in #306
• [serving] Sets default ONNXRuntime OMP threads to 1 by @frankfliu in #303
• start supporting multi-gpu in python mode by @lanking520 in #302
• Create an ensemble workflow by @zachgk in #282
• [Docker][DLC] upgrade for next release by @lanking520 in #307
• [ci] Upgrade deprecated github actions by @frankfliu in #309
• [benchmark] Adds HuggingFace model zoo to djl-bench by @frankfliu in #308
• [doc] Add DJL Serving packaging doc by @xyang16 in #310
• [serving] Adds hugginface tokenizer as default dependency by @frankfliu in #313
• [djl-bench] Update README to upgrade Java version to 11 by @frankfliu in #314
• fix typo in Pymodel mpi log by @siddvenk in #315
• [Docker][DLC] add deepspeed 0.7.5 by @lanking520 in #316
• [DLC] update docker with s5cmd by @lanking520 in #317
• fix shell script by @lanking520 in #319
• fix tar file unzip location by @lanking520 in #320
• [doc] Update serving docs by @xyang16 in #312
• [central] workaround webpack-cli 5.0.0 build issue by @frankfliu in #324
• [benchmark] Make warmup iteration configurable by @frankfliu in #323
• [benchmar] Update benchmark README by @frankfliu in #325
• support download model from s3 by @lanking520 in #322
• [doc] Update configurations document by @frankfliu in #327
• [Docker] build dlc telemetry by @lanking520 in #326
• fix s5cmd by @lanking520 in #328
• Upgrade dependencies version by @frankfliu in #331
• [serving] Uses Engine.getDjlVersion() for consistency by @frankfliu in #330
• [serving] Refactor NeuronUtils by @frankfliu in #333
• add telemetry collection testing by @lanking520 in #332
• [Docker][G5] add test artifacts by @lanking520 in #329
• fix the telemetry inaccessible issues by @lanking520 in #335
• fall back to imds v1 by @lanking520 in #338
• G5 test patch fixes by @lanking520 in #336
• [G5][Docker] add gptj model by @lanking520 in #339
• [Docker] upgrade cu117 by @lanking520 in #340
• [Docker][G5] add bloom 7b1 support by @lanking520 in #341
• update telemetry to follow DLC standard by @lanking520 in #343
• [Docker] add paddlepaddle docker build script by @lanking520 in #342
• fix on the tag by @lanking520 in #344
• [Docker] fix regex by @lanking520 in #345
• final regex fix by @lanking520 in #346
• [HF] add more information to HF Accelerate by @lanking520 in #318
• [ci] Upgrade dependencies version by @frankfliu in #347
• [docs] Update serving configurations document by @frankfliu in #348
• [Docker][G5] add huggingface tests by @lanking520 in #349
• [Handler] fix python grammer by @lanking520 in #350
• [Python] fix potential None on TP degree by @lanking520 in #351
• [serving] Fixes tensor_parallel_degree parsing bug by @frankfliu in #352
• fix telemetry by @lanking520 in #353
• fix mapping issues by @lanking520 in #354
• [G5] final fixes by @lanking520 in #355
• [serving] Support tensor_parallel_degree for commandline by @frankfliu in #356
• use default serve command by @lanking520 in #357
• [Docker] upgrade inferentia docker image by @lanking520 in #337
• [python] Fixes assert warning by @frankfliu in #360
• [Docker] unset min length by @lanking520 in #358
• feat: update deepspeed version to hosted deepspeed wheel by @tosterberg in #361
• [Inf] update model by @lanking520 in #362
• [serving] Avoid using special config.properties for DeepSpeed by @frankfliu in #363
• [serving] Remove unused setLoadOnDevice() by @frankfliu in #364
• fix on the test failure by @lanking520 in #365
• [WIP] Updating deepspeed handler for more models by @siddvenk in #359
• [DeepSpeed] Fix datatype and allow checkpoint loading by @lanking520 in #366
• [DeepSpeed] fix handler usage by @lanking520 in #367
• fix on pipeline by @lanking520 in #368
• [DeepSpeed] fix the get class error by @lanking520 in #369
• [CI] support faster failing by @lanking520 in #370
• Switch order of ds init_inference and pipeline construction to save m… by @siddvenk in #373
• [Docker] fix a few things by @lanking520 in #372
• [Docker] add opt 13b option by @lanking520 in #374
• Update SD handler to work with custom wheel by @siddvenk in #375
• [python] Add dynamic batching feature to python engine by @frankfliu in #371
• [python] Fixes python 3.8 error by @frankfliu in #376
• [python] use jitscript model if available by @frankfliu in #379
• [docker] Adds non-root user to docker by @frankfliu in #378
• Update sd handler to return image in request, fix some default values… by @siddvenk in #377
• [docker] Update PyTorch/TensorFlow version by @frankfliu in #380
• [docker] Fixes aarch64 docker file by @frankfliu in #381
• fix action test for stable-diffusion by @siddvenk in #382
• [python] Adds .npz input support for Python engine by @frankfliu in #383
• [serving] Adds a few more integration tests by @frankfliu in #384
• add status code field by @lanking520 in #385
• remove bracket by @lanking520 in #386
• fix issue in sd handler with data type by @siddvenk in #387
• [ci] Avoid run integration test as root by @frankfliu in #388
• [tests] Reformat python code by @frankfliu in #389
• Fix fp32 issues with DS fork wheel for stable diffusion, fix llm tests by @siddvenk in #390
• Import some dependencies only for stable diffusion test by @siddvenk in #391
• [serving] Make yaml support optional by @frankfliu in #393
• [ci] Rename resent18 model resource url by @frankfliu in #395
• [Backport] use torch wheel that has patched fixes by @lanking520 in #396

New Contributors

• @tosterberg made their first contribution in #361

Full Changelog: v0.19.0...v0.20.0