DJL Monthly Update - January 2022

[zachgk]

Also including content from December 2021

Released DJL 0.15.0 (Release Notes)

Key Features:

• Introduces Huggingface tokenizers extension which allows user to leverage high performance fast tokenizer in Java
• Upgrades PyTorch engine to 1.10.0 with CUDA 11.3 support
• Upgrades TensorFlow to 2.7.0 with CUDA 11.3 support
• Upgrades ONNXRuntime engine to 1.10.0
• Upgrades TensorFlowLite to 2.6.2
• Provides better PyTorch engine backward compatibility support
• Adds load model from InputStream support
• Adds Windows support for SentencePiece
• Removes -auto packages to simplify DJL dependencies
• Fixes log4j CVEs

Community

The Amazon DJL team is hiring new developers for DJL. Apply on the Amazon jobs site.

Read the blog post Simplified MLOps with Deep Java Library by hypefactors.

Feature Enhancements

DJL

• api Allows dynamic add model zoo #1397
• add sentencepiece load from bytes method #1399
• api Adds load model from Inpustream to public API #1400
• Adds load model from InputStream for OnnxRuntime and TFLite #1402
• Upgrade log4j to 1.15.0 #1413
• sentencepiece Adds Windows support #1412
• Upgrade log4j to 2.16.0 #1418
• Upgrade log4j to 2.17.0 #1424
• Add djl-admin as codeowners #1425
• Create IndexEvaluator and IndexLoss #1414
• huggingface Adds huggingface tokenizer extension #1406
• Update djl-bench version to 0.14.1 #1426
• Support for Tuples as input values for PyTorch models #1436
• Upgrade log4j to 2.17.1 #1444
• Update TrainableWordEmbedding.java #1450
• onnx Upgrade OnnxRuntime to 1.10.0 #1455
• Upgrade testng to 7.5 #1457
• bom Use release version of PyTorch and TFLite #1458
• tensorflow Upgrade tensorflow to 2.7.0 #1462
• hadoop Exclude log4j dependency from hadoop extension #1463
• pytorch Publish 1.8.1 and 1.9.1 JNI in nightly build #1465
• bom Update tensorflow cu113 package in bom #1468
• Upgrade windows docker to use opendjk11 #1472
• Upgrade DJL version from 0.14.0 to 0.15.0 #1473
• djl-bench Upgrade djl-bench to 0.15.0 #1476
• Bump up version to 0.16.0 #1474
• onnx Add SessionOptions support for OnnxRuntime #1479
• android Update to 0.15.0 #1483
• tokenizers Adding huggingface tokenizer extension to BOM #1487

DJL Serving

• serving Install engine dependency ondemand #45
• Upgrade log4j to 2.15.0 #46
• serving Refactor debug flag #47
• Upgrade log4j to 2.16.0 #48
• Add djl-admin as codeowners #50
• Upgrade log4j to 2.17.0 #49
• Upgrade log4j to 2.17.1 #51
• Bump follow-redirects from 1.14.3 to 1.14.7 in /central #53
• Update nodejs package version #54
• Upgrade netty to 4.1.72.Final #55
• Upgrade to DJL 0.15.0 #56
• Load all model by default when model store is specified #57
• Workflow file parsing #52
• Combine WorkflowInfo into Workflow #58

Documentation & Examples

DJL

• Update demo blockrunner example code #1435
• pytorch Update pytorch native library README #1449
• docs Update engine documents to remove -auto package #1459
• pytorch Update PyTorch engine dependency document #1466
• docs Update package model document #1471
• Update README for 0.15.0 release #1477
• android Update README to use 0.16.0-SNAPSHOT version #1486

DJL Demos

• Bump log4j-api from 2.13.1 to 2.15.0 in /camel-djl #190
• Upgrade live object detection to DJL 0.14.0 #189
• Upgrade log4j to 2.15.0 #191
• Bump log4j-api from 2.15.0 to 2.16.0 in /camel-djl #192
• Upgrade log4j to 2.16.0 #193
• Bump log4j-api from 2.16.0 to 2.17.0 in /camel-djl #194
• Add djl-admin as codeowners #197
• Upgrade log4j to 2.17.0 #196
• Upgrade logback to 1.2.9 #195
• Upgrade web-demo to 0.14.0 #199
• Bump log4j-api from 2.17.0 to 2.17.1 in /camel-djl #201
• Upgrade log4j to 2.17.1 #202
• Bump protobuf-java from 3.8.0 to 3.16.1 in /quarkus/extension-example #205
• Bump protobuf-java from 3.8.0 to 3.16.1 in /quarkus/example #204
• Upgrade to DJL 0.15.0 #207
• android Upgrade to DJL 0.16.0-SNAPSHOT #208

D2l-Java

• Upgrade log4j to 2.15.0 #178
• Upgrade log4j to 2.16.0 #179
• Upgrade log4j to 2.17.0 #180
• Upgrade log4j to 2.17.1 #181
• Add djl-admin as codeowner #182
• Fix images display issue #183

D2l-Java zh

• Upgrade log4j to 2.15.0 #4
• Upgrade log4j to 2.16.0 #5
• Translation chapter_recurrent-neural-networks folder #3
• Upgrade log4j to 2.17.0 #6
• Translation chapter_recurrent-neural-networks #7
• Upgrade log4j to 2.17.1 #8
• translated chapter 11 optimization #11
• Add djl-admin as code owner #13
• translated chapter 12 computational performance #12
• Fix images display issue #14
• Fixes wrong cell type for multiple-gpus-concise #16
• translated chapter 9 recurrent-modern #9
• translate chapter 14 nlp-pretrained #15
• translate chapter 14 nlp-pretrained index page #10

Bug Fixes

DJL

• ci Invalidate cloudfront cache after sentencepiece build #1404
• benchmark Fixes benchmark README version #1410
• [pytorch Fixes Core dump when trying to create empty IValue list for pytorch engine #1437, check empty IValue
• pytorch Fixes unit test failure on GPU. #1430
• api Fix NPE bug in getContextClassLoader(). #1445
• api Fix model name detection in loading model from jar #1446
• Add mavenCentral() to buildscript repository #1456
• tensorflow Fix tensorflow native library cuda version #1464
• pytorch workaround libtorch.dylob 1.8.1 loading issue #1467
• pytorch Fix pytorch engine version string #1469
• test Fix flaky test for tensorflow 2.7.0 on GPU #1475
• api Fixes topK items for DetectedObjects and make it configurable to Classifications #1478
• ci Remove unecessary github actions workflow #1484
• pytorch Fix load native library failure on Android #1485
• ci Use JDK 11 for github actions build. #1489