sagemaker-train 1.0__py3-none-any.whl

sagemaker/train/__init__.py

@@ -0,0 +1,34 @@
+# Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License"). You
+# may not use this file except in compliance with the License. A copy of
+# the License is located at
+#
+#     http://aws.amazon.com/apache2.0/
+#
+# or in the "license" file accompanying this file. This file is
+# distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF
+# ANY KIND, either express or implied. See the License for the specific
+# language governing permissions and limitations under the License.
+"""SageMaker Python SDK Train Module."""
+from __future__ import absolute_import
+
+# Lazy imports to avoid circular dependencies
+# Session and get_execution_role are available from sagemaker.core.helper.session_helper
+# Import them directly from there if needed, or use lazy import pattern
+
+def __getattr__(name):
+    """Lazy import to avoid circular dependencies."""
+    if name == "Session":
+        from sagemaker.core.helper.session_helper import Session
+        return Session
+    elif name == "get_execution_role":
+        from sagemaker.core.helper.session_helper import get_execution_role
+        return get_execution_role
+    elif name == "ModelTrainer":
+        from sagemaker.train.model_trainer import ModelTrainer
+        return ModelTrainer
+    elif name == "logger":
+        from sagemaker.core.utils.utils import logger
+        return logger
+    raise AttributeError(f"module '{__name__}' has no attribute '{name}'")

sagemaker/train/configs.py

@@ -0,0 +1,31 @@
+# Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License"). You
+# may not use this file except in compliance with the License. A copy of
+# the License is located at
+#
+#     http://aws.amazon.com/apache2.0/
+#
+# or in the "license" file accompanying this file. This file is
+# distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF
+# ANY KIND, either express or implied. See the License for the specific
+# language governing permissions and limitations under the License.
+"""
+DEPRECATED: This module has been moved to sagemaker.core.training.configs
+
+This is a backward compatibility shim. Please update your imports to:
+    from sagemaker.core.training.configs import ...
+"""
+from __future__ import absolute_import
+
+import warnings
+
+# Backward compatibility: re-export from core
+from sagemaker.core.training.configs import *  # noqa: F401, F403
+
+warnings.warn(
+    "sagemaker.train.configs has been moved to sagemaker.core.training.configs. "
+    "Please update your imports. This shim will be removed in a future version.",
+    DeprecationWarning,
+    stacklevel=2
+)

sagemaker/train/constants.py

@@ -0,0 +1,41 @@
+# Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License"). You
+# may not use this file except in compliance with the License. A copy of
+# the License is located at
+#
+#     http://aws.amazon.com/apache2.0/
+#
+# or in the "license" file accompanying this file. This file is
+# distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF
+# ANY KIND, either express or implied. See the License for the specific
+# language governing permissions and limitations under the License.
+"""
+DEPRECATED: This module has been moved to sagemaker.core.training.constants
+
+This is a backward compatibility shim. Please update your imports to:
+    from sagemaker.core.training.constants import ...
+"""
+from __future__ import absolute_import
+
+import os
+
+SM_CODE = "code"
+SM_CODE_CONTAINER_PATH = "/opt/ml/input/data/code"
+
+SM_DRIVERS = "sm_drivers"
+SM_DRIVERS_CONTAINER_PATH = "/opt/ml/input/data/sm_drivers"
+SM_DRIVERS_LOCAL_PATH = os.path.join(
+    os.path.dirname(os.path.abspath(__file__)), "container_drivers"
+)
+
+SOURCE_CODE_JSON = "sourcecode.json"
+DISTRIBUTED_JSON = "distributed.json"
+TRAIN_SCRIPT = "sm_train.sh"
+
+DEFAULT_CONTAINER_ENTRYPOINT = ["/bin/bash"]
+DEFAULT_CONTAINER_ARGUMENTS = [
+    "-c",
+    f"chmod +x {SM_DRIVERS_CONTAINER_PATH}/{TRAIN_SCRIPT} "
+    + f"&& {SM_DRIVERS_CONTAINER_PATH}/{TRAIN_SCRIPT}",
+]

sagemaker/train/container_drivers/__init__.py

@@ -0,0 +1,14 @@
+# Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License"). You
+# may not use this file except in compliance with the License. A copy of
+# the License is located at
+#
+#     http://aws.amazon.com/apache2.0/
+#
+# or in the "license" file accompanying this file. This file is
+# distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF
+# ANY KIND, either express or implied. See the License for the specific
+# language governing permissions and limitations under the License.
+"""Sagemaker modules container drivers directory."""
+from __future__ import absolute_import

sagemaker/train/container_drivers/common/__init__.py

@@ -0,0 +1,14 @@
+# Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License"). You
+# may not use this file except in compliance with the License. A copy of
+# the License is located at
+#
+#     http://aws.amazon.com/apache2.0/
+#
+# or in the "license" file accompanying this file. This file is
+# distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF
+# ANY KIND, either express or implied. See the License for the specific
+# language governing permissions and limitations under the License.
+"""Sagemaker modules container drivers - common directory."""
+from __future__ import absolute_import

sagemaker/train/container_drivers/common/utils.py

@@ -0,0 +1,205 @@
+# Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License"). You
+# may not use this file except in compliance with the License. A copy of
+# the License is located at
+#
+#     http://aws.amazon.com/apache2.0/
+#
+# or in the "license" file accompanying this file. This file is
+# distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF
+# ANY KIND, either express or implied. See the License for the specific
+# language governing permissions and limitations under the License.
+"""This module provides utility functions for the container drivers."""
+from __future__ import absolute_import
+
+import os
+import logging
+import sys
+import subprocess
+import traceback
+import json
+
+from typing import List, Dict, Any, Tuple, IO, Optional
+
+# Initialize logger
+SM_LOG_LEVEL = os.environ.get("SM_LOG_LEVEL", 20)
+logger = logging.getLogger(__name__)
+console_handler = logging.StreamHandler(sys.stdout)
+logger.addHandler(console_handler)
+logger.setLevel(int(SM_LOG_LEVEL))
+
+FAILURE_FILE = "/opt/ml/output/failure"
+DEFAULT_FAILURE_MESSAGE = """
+Training Execution failed.
+For more details, see CloudWatch logs at 'aws/sagemaker/TrainingJobs'.
+TrainingJob - {training_job_name}
+"""
+
+USER_CODE_PATH = "/opt/ml/input/data/code"
+SOURCE_CODE_JSON = "/opt/ml/input/data/sm_drivers/sourcecode.json"
+DISTRIBUTED_JSON = "/opt/ml/input/data/sm_drivers/distributed.json"
+
+HYPERPARAMETERS_JSON = "/opt/ml/input/config/hyperparameters.json"
+
+SM_EFA_NCCL_INSTANCES = [
+    "ml.g4dn.8xlarge",
+    "ml.g4dn.12xlarge",
+    "ml.g5.48xlarge",
+    "ml.p3dn.24xlarge",
+    "ml.p4d.24xlarge",
+    "ml.p4de.24xlarge",
+    "ml.p5.48xlarge",
+    "ml.trn1.32xlarge",
+]
+
+SM_EFA_RDMA_INSTANCES = [
+    "ml.p4d.24xlarge",
+    "ml.p4de.24xlarge",
+    "ml.trn1.32xlarge",
+]
+
+
+def write_failure_file(message: Optional[str] = None):
+    """Write a failure file with the message."""
+    if message is None:
+        message = DEFAULT_FAILURE_MESSAGE.format(training_job_name=os.environ["TRAINING_JOB_NAME"])
+    if not os.path.exists(FAILURE_FILE):
+        with open(FAILURE_FILE, "w") as f:
+            f.write(message)
+
+
+def read_source_code_json(source_code_json: Dict[str, Any] = SOURCE_CODE_JSON):
+    """Read the source code config json file."""
+    try:
+        with open(source_code_json, "r") as f:
+            source_code_dict = json.load(f) or {}
+    except FileNotFoundError:
+        source_code_dict = {}
+    return source_code_dict
+
+
+def read_distributed_json(distributed_json: Dict[str, Any] = DISTRIBUTED_JSON):
+    """Read the distribution config json file."""
+    try:
+        with open(distributed_json, "r") as f:
+            distributed_dict = json.load(f) or {}
+    except FileNotFoundError:
+        distributed_dict = {}
+    return distributed_dict
+
+
+def read_hyperparameters_json(hyperparameters_json: Dict[str, Any] = HYPERPARAMETERS_JSON):
+    """Read the hyperparameters config json file."""
+    try:
+        with open(hyperparameters_json, "r") as f:
+            hyperparameters_dict = json.load(f) or {}
+    except FileNotFoundError:
+        hyperparameters_dict = {}
+    return hyperparameters_dict
+
+
+def get_process_count(process_count: Optional[int] = None) -> int:
+    """Get the number of processes to run on each node in the training job."""
+    return (
+        process_count
+        or int(os.environ.get("SM_NUM_GPUS", 0))
+        or int(os.environ.get("SM_NUM_NEURONS", 0))
+        or 1
+    )
+
+
+def hyperparameters_to_cli_args(hyperparameters: Dict[str, Any]) -> List[str]:
+    """Convert the hyperparameters to CLI arguments."""
+    cli_args = []
+    for key, value in hyperparameters.items():
+        value = safe_deserialize(value)
+        cli_args.extend([f"--{key}", safe_serialize(value)])
+
+    return cli_args
+
+
+def safe_deserialize(data: Any) -> Any:
+    """Safely deserialize data from a JSON string.
+
+    This function handles the following cases:
+    1. If `data` is not a string, it returns the input as-is.
+    2. If `data` is a JSON-encoded string, it attempts to deserialize it using `json.loads()`.
+    3. If `data` is a string but cannot be decoded as JSON, it returns the original string.
+
+    Returns:
+        Any: The deserialized data, or the original input if it cannot be JSON-decoded.
+    """
+    if not isinstance(data, str):
+        return data
+
+    try:
+        return json.loads(data)
+    except json.JSONDecodeError:
+        return data
+
+
+def safe_serialize(data):
+    """Serialize the data without wrapping strings in quotes.
+
+    This function handles the following cases:
+    1. If `data` is a string, it returns the string as-is without wrapping in quotes.
+    2. If `data` is serializable (e.g., a dictionary, list, int, float), it returns
+       the JSON-encoded string using `json.dumps()`.
+    3. If `data` cannot be serialized (e.g., a custom object), it returns the string
+       representation of the data using `str(data)`.
+
+    Args:
+        data (Any): The data to serialize.
+
+    Returns:
+        str: The serialized JSON-compatible string or the string representation of the input.
+    """
+    if isinstance(data, str):
+        return data
+    try:
+        return json.dumps(data)
+    except TypeError:
+        return str(data)
+
+
+def get_python_executable() -> str:
+    """Get the python executable path."""
+    return sys.executable
+
+
+def log_subprocess_output(pipe: IO[bytes]):
+    """Log the output from the subprocess."""
+    for line in iter(pipe.readline, b""):
+        logger.info(line.decode("utf-8").strip())
+
+
+def execute_commands(commands: List[str]) -> Tuple[int, str]:
+    """Execute the provided commands and return exit code with failure traceback if any."""
+    try:
+        process = subprocess.Popen(
+            commands,
+            stdout=subprocess.PIPE,
+            stderr=subprocess.STDOUT,
+        )
+        with process.stdout:
+            log_subprocess_output(process.stdout)
+        exitcode = process.wait()
+        if exitcode != 0:
+            raise subprocess.CalledProcessError(exitcode, commands)
+        return exitcode, ""
+    except subprocess.CalledProcessError as e:
+        # Capture the traceback in case of failure
+        error_traceback = traceback.format_exc()
+        print(f"Command failed with exit code {e.returncode}. Traceback: {error_traceback}")
+        return e.returncode, error_traceback
+
+
+def is_worker_node() -> bool:
+    """Check if the current node is a worker node."""
+    return os.environ.get("SM_CURRENT_HOST") != os.environ.get("SM_MASTER_ADDR")
+
+
+def is_master_node() -> bool:
+    """Check if the current node is the master node."""
+    return os.environ.get("SM_CURRENT_HOST") == os.environ.get("SM_MASTER_ADDR")

sagemaker/train/container_drivers/distributed_drivers/__init__.py

@@ -0,0 +1,14 @@
+# Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License"). You
+# may not use this file except in compliance with the License. A copy of
+# the License is located at
+#
+#     http://aws.amazon.com/apache2.0/
+#
+# or in the "license" file accompanying this file. This file is
+# distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF
+# ANY KIND, either express or implied. See the License for the specific
+# language governing permissions and limitations under the License.
+"""Sagemaker modules container drivers - drivers directory."""
+from __future__ import absolute_import

sagemaker/train/container_drivers/distributed_drivers/basic_script_driver.py

@@ -0,0 +1,81 @@
+# Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License"). You
+# may not use this file except in compliance with the License. A copy of
+# the License is located at
+#
+#     http://aws.amazon.com/apache2.0/
+#
+# or in the "license" file accompanying this file. This file is
+# distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF
+# ANY KIND, either express or implied. See the License for the specific
+# language governing permissions and limitations under the License.
+"""This module is the entry point for the Basic Script Driver."""
+from __future__ import absolute_import
+
+import os
+import sys
+import json
+import shlex
+
+from pathlib import Path
+from typing import List
+
+sys.path.insert(0, str(Path(__file__).parent.parent))
+
+from common.utils import (  # noqa: E402 # pylint: disable=C0413,E0611
+    logger,
+    get_python_executable,
+    write_failure_file,
+    hyperparameters_to_cli_args,
+    execute_commands,
+)
+
+
+def create_commands() -> List[str]:
+    """Create the commands to execute."""
+    entry_script = os.environ["SM_ENTRY_SCRIPT"]
+    hyperparameters = json.loads(os.environ["SM_HPS"])
+    python_executable = get_python_executable()
+
+    args = hyperparameters_to_cli_args(hyperparameters)
+    if entry_script.endswith(".py"):
+        commands = [python_executable, entry_script]
+        commands += args
+    elif entry_script.endswith(".sh"):
+        args_str = " ".join(shlex.quote(arg) for arg in args)
+        commands = [
+            "/bin/sh",
+            "-c",
+            f"chmod +x {entry_script} && ./{entry_script} {args_str}",
+        ]
+    else:
+        raise ValueError(
+            f"Unsupported entry script type: {entry_script}. Only .py and .sh are supported."
+        )
+    return commands
+
+
+def main():
+    """Main function for the Basic Script Driver.
+
+    This function is the entry point for the Basic Script Driver.
+
+    Execution Lifecycle:
+    1. Read the source code and hyperparameters JSON files.
+    2. Set hyperparameters as command line arguments.
+    3. Create the commands to execute.
+    4. Execute the commands.
+    """
+
+    cmd = create_commands()
+
+    logger.info(f"Executing command: {' '.join(cmd)}")
+    exit_code, traceback = execute_commands(cmd)
+    if exit_code != 0:
+        write_failure_file(traceback)
+        sys.exit(exit_code)
+
+
+if __name__ == "__main__":
+    main()

sagemaker/train/container_drivers/distributed_drivers/mpi_driver.py

@@ -0,0 +1,105 @@
+# Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License"). You
+# may not use this file except in compliance with the License. A copy of
+# the License is located at
+#
+#     http://aws.amazon.com/apache2.0/
+#
+# or in the "license" file accompanying this file. This file is
+# distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF
+# ANY KIND, either express or implied. See the License for the specific
+# language governing permissions and limitations under the License.
+"""This module is the entry point for the MPI driver script."""
+from __future__ import absolute_import
+
+import os
+import sys
+import json
+
+from sagemaker.train.container_drivers.distributed_drivers.mpi_utils import (
+    start_sshd_daemon,
+    bootstrap_master_node,
+    bootstrap_worker_node,
+    get_mpirun_command,
+    write_status_file_to_workers,
+    write_env_vars_to_file,
+)
+
+
+from sagemaker.train.container_drivers.common.utils import (
+    logger,
+    hyperparameters_to_cli_args,
+    get_process_count,
+    execute_commands,
+    write_failure_file,
+)
+
+
+def main():
+    """Main function for the MPI driver script.
+
+    The MPI Dirver is responsible for setting up the MPI environment,
+    generating the correct mpi commands, and launching the MPI job.
+
+    Execution Lifecycle:
+    1. Setup General Environment Variables at /etc/environment
+    2. Start SSHD Daemon
+    3. Bootstrap Worker Nodes
+       a. Wait to establish connection with Master Node
+       b. Wait for Master Node to write status file
+    4. Bootstrap Master Node
+        a. Wait to establish connection with Worker Nodes
+        b. Generate MPI Command
+        c. Execute MPI Command with user script provided in `entry_script`
+        d. Write status file to Worker Nodes
+    5. Exit
+
+    """
+    entry_script = os.environ["SM_ENTRY_SCRIPT"]
+    distributed_config = json.loads(os.environ["SM_DISTRIBUTED_CONFIG"])
+    hyperparameters = json.loads(os.environ["SM_HPS"])
+
+    sm_current_host = os.environ["SM_CURRENT_HOST"]
+    sm_hosts = json.loads(os.environ["SM_HOSTS"])
+    sm_master_addr = os.environ["SM_MASTER_ADDR"]
+
+    write_env_vars_to_file()
+    start_sshd_daemon()
+
+    if sm_current_host != sm_master_addr:
+        bootstrap_worker_node(sm_master_addr)
+    else:
+        worker_hosts = [host for host in sm_hosts if host != sm_master_addr]
+        bootstrap_master_node(worker_hosts)
+
+        host_list = json.loads(os.environ["SM_HOSTS"])
+        host_count = int(os.environ["SM_HOST_COUNT"])
+        process_count = int(distributed_config["process_count_per_node"] or 0)
+        process_count = get_process_count(process_count)
+
+        if process_count > 1:
+            host_list = ["{}:{}".format(host, process_count) for host in host_list]
+
+        mpi_command = get_mpirun_command(
+            host_count=host_count,
+            host_list=host_list,
+            num_processes=process_count,
+            additional_options=distributed_config["mpi_additional_options"] or [],
+            entry_script_path=entry_script,
+        )
+
+        args = hyperparameters_to_cli_args(hyperparameters)
+        mpi_command += args
+
+        logger.info(f"Executing command: {' '.join(mpi_command)}")
+        exit_code, error_traceback = execute_commands(mpi_command)
+        write_status_file_to_workers(worker_hosts)
+
+        if exit_code != 0:
+            write_failure_file(error_traceback)
+            sys.exit(exit_code)
+
+
+if __name__ == "__main__":
+    main()