Upgrade XgBoost to 1.7.x (#362)

* Upgrade XgBoost version to 1.7.3

* Restructure checkpointing.py and fix Unit tests

* Resolve Flake8 style errors

* Address Comments and improvements

* Add comments and improvements

* Upgrade xgboost to 1.7.4

* remove grow_local_histmaker and Single precision histogram

---------

Co-authored-by: Malav Shastri <[email protected]>

Author: malav-shastri

README.rst

@@ -253,4 +253,4 @@ SageMaker XGboost Framework Container is licensed under the Apache 2.0 License.
 .com, Inc. or its affiliates. All Rights Reserved. The license is available at:
 http://aws.amazon.com/apache2.0/
 
-.. |XGBoostLatestVersion| replace:: 1.5-1
+.. |XGBoostLatestVersion| replace:: 1.7-1

docker/1.5-1/base/Dockerfile.cpu renamed to docker/1.7-1/base/Dockerfile.cpu

@@ -11,7 +11,7 @@ ARG CONDA_PKG_VERSION=4.10.1
 ARG PYTHON_VERSION=3.8.13
 ARG PYARROW_VERSION=1.0
 ARG MLIO_VERSION=0.7.0
-ARG XGBOOST_VERSION=1.5.2
+ARG XGBOOST_VERSION=1.7.4
 
 ENV DEBIAN_FRONTEND=noninteractive
 ENV LANG=C.UTF-8

docker/1.5-1/base/Dockerfile_arm64.cpu renamed to docker/1.7-1/base/Dockerfile_arm64.cpu

@@ -1,4 +1,4 @@
-ARG SAGEMAKER_XGBOOST_VERSION=1.5-1
+ARG SAGEMAKER_XGBOOST_VERSION=1.7-1
 ARG PYTHON_VERSION=3.8
 
 FROM xgboost-container-base:${SAGEMAKER_XGBOOST_VERSION}-cpu-py3

docker/1.5-1/final/Dockerfile.cpu renamed to docker/1.7-1/final/Dockerfile.cpu

@@ -3,7 +3,7 @@ PyYAML==5.4.1
 Pillow==9.1.1
 boto3==1.17.52
 botocore==1.20.52
-cryptography==35.0.0
+cryptography==39.0.1
 dask==2022.11.1
 dask-cuda==22.12.0
 gunicorn==19.10.0

docker/1.5-1/final/Dockerfile_arm64.cpu renamed to docker/1.7-1/final/Dockerfile_arm64.cpu

@@ -29,7 +29,6 @@ def updater_validator(value, dependencies):
             "grow_colmaker",
             "distcol",
             "grow_histmaker",
-            "grow_local_histmaker",
             "grow_skmaker",
             "sync",
             "refresh",
@@ -40,7 +39,6 @@ def updater_validator(value, dependencies):
             "grow_colmaker",
             "distcol",
             "grow_histmaker",
-            "grow_local_histmaker",
             "grow_colmaker",
             "grow_quantile_histmaker",
         ]
@@ -62,7 +60,7 @@ def updater_validator(value, dependencies):
             if not all(x in valid_tree_plugins for x in value):
                 raise exc.UserError(
                     "Tree updater should be selected from these options: 'grow_colmaker', 'distcol', 'grow_histmaker', "
-                    "'grow_local_histmaker', 'grow_skmaker', 'grow_quantile_histmaker', 'sync', 'refresh', 'prune', "
+                    "'grow_skmaker', 'grow_quantile_histmaker', 'sync', 'refresh', 'prune', "
                     "'shortgun', 'coord_descent'."
                 )
             # validate only one tree updater is selected
@@ -74,7 +72,7 @@ def updater_validator(value, dependencies):
                 raise exc.UserError(
                     "Only one tree grow plugin can be selected. Choose one from the"
                     "following: 'grow_colmaker', 'distcol', 'grow_histmaker', "
-                    "'grow_local_histmaker', 'grow_skmaker'"
+                    "'grow_skmaker'"
                 )
 
     @hpv.range_validator(["auto", "cpu_predictor", "gpu_predictor"])
@@ -239,15 +237,13 @@ def interaction_constraints_validator(value, dependencies):
                 "grow_colmaker",
                 "distcol",
                 "grow_histmaker",
-                "grow_local_histmaker",
                 "grow_skmaker",
                 "sync",
                 "refresh",
                 "prune",
                 "grow_colmaker",
                 "distcol",
                 "grow_histmaker",
-                "grow_local_histmaker",
                 "grow_colmaker",
                 "shotgun",
                 "coord_descent",
@@ -334,7 +330,6 @@ def interaction_constraints_validator(value, dependencies):
         hpv.ContinuousHyperparameter(
             name="aft_loss_distribution_scale", range=hpv.Interval(min_closed=0), required=False
         ),
-        hpv.CategoricalHyperparameter(name="single_precision_histogram", range=["true", "false"], required=False),
         hpv.CategoricalHyperparameter(name="deterministic_histogram", range=["true", "false"], required=False),
         hpv.CategoricalHyperparameter(name="sampling_method", range=["uniform", "gradient_based"], required=False),
         hpv.IntegerHyperparameter(name="prob_buffer_row", range=hpv.Interval(min_open=1.0), required=False),