|
| 1 | +# Changelog |
| 2 | + |
| 3 | +## v0.1.0 (2019-05-22) |
| 4 | + |
| 5 | +### Bug fixes and other changes |
| 6 | + |
| 7 | + * skip setup on second remote run |
| 8 | + * add setup file back |
| 9 | + * add branch name to remote gpu test run command |
| 10 | + * remove setup file in release build gpu test |
| 11 | + * ignore coverage in release build tests |
| 12 | + * use tar file name as framework_support_installable in build_all.py |
| 13 | + * Add release build |
| 14 | + * Explicitly set lower-bound for botocore version |
| 15 | + * Pull request to test codebuild trigger on TensorFlow script mode |
| 16 | + * Update integ test for checking Python version |
| 17 | + * Upgrade to TensorFlow 1.13.1 |
| 18 | + * Add mpi4py to pip installs |
| 19 | + * Add SageMaker integ test for hyperparameter tuning model_dir logic |
| 20 | + * Add Horovod benchmark |
| 21 | + * Fix model_dir adjustment for hyperparameter tuning jobs |
| 22 | + * change model_dir to training job name if it is for tuning. |
| 23 | + * Tune test_s3_plugin test |
| 24 | + * Skip the s3_plugin test before new binary released |
| 25 | + * Add model saving warning at end of training |
| 26 | + * Specify region when creating S3 resource in integ tests |
| 27 | + * Fix instance_type fixture setup for tests |
| 28 | + * Read framework version from Python SDK for integ test default |
| 29 | + * Fix SageMaker Session handling in Horovod test |
| 30 | + * Configure encoding to be utf-8 |
| 31 | + * Use the test argement framework_version in all tests |
| 32 | + * Fix broken test test_distributed_mnist_no_ps |
| 33 | + * Add S3 plugin tests |
| 34 | + * Skip horovod local CPU test in GPU instances |
| 35 | + * Add Horovod tests |
| 36 | + * Skip horovod integration tests |
| 37 | + * TensorFlow 1.12 and Horovod support |
| 38 | + * Deprecate get_marker. Use get_closest_marker instead |
| 39 | + * Force parameter server to run on CPU |
| 40 | + * Add python-dev and build-essential to Dockerfiles |
| 41 | + * Update script_mode_train_any_tf_script_in_sage_maker.ipynb |
| 42 | + * Skip keras local mode test on gpu and use random port for serving in the test |
| 43 | + * Fix Keras test |
| 44 | + * Create parameter server in different thread |
| 45 | + * Add Keras support |
| 46 | + * Fix broken unit tests |
| 47 | + * Unset CUDA_VISIBLE_DEVICES for worker processes |
| 48 | + * Disable GPU for parameter process |
| 49 | + * Set parameter process waiting to False |
| 50 | + * Update sagemaker containers |
| 51 | + * GPU fix |
| 52 | + * Set S3 environment variables |
| 53 | + * Add CI configuration files |
| 54 | + * Add distributed training support |
| 55 | + * Edited the tf script mode notebook |
| 56 | + * Add benchmarking script |
| 57 | + * Add Script Mode example |
| 58 | + * Add integration tests to run training jobs with sagemaker |
| 59 | + * Add tox.ini and configure coverage and flake runs |
| 60 | + * Scriptmode single machine training implementation |
| 61 | + * Update region in s3 boto client in serve |
| 62 | + * Update readme with instructions for 1.9.0 and above |
| 63 | + * Fix deserialization of dicts for json predict requests |
| 64 | + * Add dockerfile and update test for tensorflow 1.10.0 |
| 65 | + * Support tensorflow 1.9.0 |
| 66 | + * Add integ tests to verify that tensorflow in gpu-image can access gpu-devices. |
| 67 | + * train on 3 epochs for pipe mode test |
| 68 | + * Change error classes used by _default_input_fn() and _default_output_fn() |
| 69 | + * Changing assertion to check only existence |
| 70 | + * Install sagemaker-tensorflow from pypi. Add MKL environment variables for TF 1.8 |
| 71 | + * get most recent saved model to export |
| 72 | + * pip install tensorflow 1.8 in 1.8 cpu image |
| 73 | + * install tensorflow extensions |
| 74 | + * upgrade cpu binaries in docker build |
| 75 | + * Force upgrade of the framework binaries to make sure the right binaries are installed. |
| 76 | + * Add Pillow to pip install list |
| 77 | + * Increase train steps for cifar distributed test to mitigate race condition |
| 78 | + * Add TensorFlow 1.8 dockerfiles |
| 79 | + * Add TensorFlow 1.7 dockerfiles |
| 80 | + * Explain how to download tf binaries from PyPI |
| 81 | + * Allow training without S3 |
| 82 | + * Fix hyperparameter name for detecting a tuning job |
| 83 | + * Checkout v1.4.1 tag instead of r1.4 branch |
| 84 | + * Move processing of requirements file in. |
| 85 | + * Generate checkpoint path using TRAINING_JOB_NAME environment variable if needed |
| 86 | + * Wrap user-provided model_fn to pass arguments positionally (maintains compatibility with existing behavior) |
| 87 | + * Add more unit tests for trainer, fix __all__ and rename train.py to avoid import conflict |
| 88 | + * Use regional endpoint for S3 client |
| 89 | + * Update README.rst |
| 90 | + * Pass input_channels to eval_input_fn if defined |
| 91 | + * Fix setup.py to refer to renamed README |
| 92 | + * Add test and build instructions |
| 93 | + * Fix year in license headers |
| 94 | + * Add TensorFlow 1.6 |
| 95 | + * Add test instructions in README |
| 96 | + * Add container support to install_requires |
| 97 | + * Add Apache license headers |
| 98 | + * Use wget to install tensorflow-model-server |
| 99 | + * Fix file path for integ test |
| 100 | + * Fix s3_prefix path in integ test |
| 101 | + * Fix typo in path for integ test |
| 102 | + * Add input_channels to train_input_fn interface. |
| 103 | + * Update logging and make serving_input_fn optional. |
| 104 | + * remove pip install in tensorflow training |
| 105 | + * Modify integration tests to run nvidia-docker for gpu |
| 106 | + * add h5py for keras models |
| 107 | + * Add local integ tests & resources |
| 108 | + * Restructure repo to use a directory per TF version for dockerfiles |
| 109 | + * Rename "feature_map" variables to "feature_dict" to avoid overloading it with the ML term "feature map" |
| 110 | + * Copying in changes from internal repo: |
| 111 | + * Add functional test |
| 112 | + * Fix FROM image names for final build dockerfiles |
| 113 | + * Add dockerfiles for building our production images (TF 1.4) |
| 114 | + * GPU Dockerfile and setup.py fixes |
| 115 | + * Add base image Dockerfiles for 1.4 |
| 116 | + * Merge pull request #1 from aws/mvs-first-commit |
| 117 | + * first commit |
| 118 | + * Updating initial README.md from template |
| 119 | + * Creating initial file from template |
| 120 | + * Creating initial file from template |
| 121 | + * Creating initial file from template |
| 122 | + * Creating initial file from template |
| 123 | + * Creating initial file from template |
| 124 | + * Initial commit |
| 125 | + |
0 commit comments