object_detection/stm32ai_main.py

# /*---------------------------------------------------------------------------------------------
#  * Copyright (c) 2022-2023 STMicroelectronics.
#  * All rights reserved.
#  *
#  * This software is licensed under terms that can be found in the LICENSE file in
#  * the root directory of this software component.
#  * If no LICENSE file comes with this software, it is provided AS-IS.
#  *--------------------------------------------------------------------------------------------*/
import os
import sys
import hydra
import warnings
warnings.filterwarnings("ignore")
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3'
os.environ['TF_ENABLE_ONEDNN_OPTS'] = '0'

import argparse
from pathlib import Path
from omegaconf import DictConfig
from hydra.core.hydra_config import HydraConfig
import mlflow
import tensorflow as tf
from clearml import Task
from clearml.backend_config.defs import get_active_config_file
import torch
import torch.backends.cudnn as cudnn
import random

SCRIPT_DIR = os.path.dirname(os.path.abspath(__file__))
sys.path.append(os.path.dirname(SCRIPT_DIR))

from api.api import get_model, get_dataloaders, get_quantizer, get_predictor, get_evaluator, get_trainer
from common.utils import mlflow_ini, set_gpu_memory_limit, get_random_seed, log_to_file
from common.benchmarking import benchmark, cloud_connect
from common.evaluation import gen_load_val
from common.prediction import gen_load_val_predict
from object_detection.tf.src.utils import get_config
from object_detection.tf.src.deployment import deploy, deploy_mpu
from common.onnx_utils.onnx_model_convertor import torch_model_export_static
from object_detection.pt.src.utils.yolod import (configure_module,
                                                 configure_nccl, configure_omp,
                                                 init_distributed_mode)


# This function turns Tensorflow's eager mode on and off.
# Eager mode is for debugging the Model Zoo code and is slower.
# Do not set argument to True to avoid runtime penalties.
tf.config.run_functions_eagerly(False)


def process_mode(cfg: DictConfig):
    """
    Execution of the various services

    Args:
        cfg: Configuration dictionary.

    Returns:
        None
    """

    mode = cfg.operation_mode
    mlflow.log_param("model_path", cfg.model.model_path)
    # logging the operation_mode in the output_dir/stm32ai_main.log file
    log_to_file(cfg.output_dir, f'operation_mode: {mode}')

    # it makes class_names important for now, untill we work on the dataloader
    num_classes = len(cfg.dataset.class_names) if cfg.dataset.class_names else None
    dropout = cfg.training.dropout if cfg.training and 'dropout' in cfg.training else None
    num_anchors = len(cfg.postprocessing.yolo_anchors) if cfg.postprocessing and 'yolo_anchors' in cfg.postprocessing else None
    saved_model_dir = os.path.join(cfg.output_dir, cfg.general.saved_models_dir)
    os.makedirs(saved_model_dir, exist_ok=True)

    # create model 
    model = get_model(cfg=cfg)  
    if cfg.model.framework == 'torch' and isinstance(model, torch.nn.Module) and cfg.operation_mode not in ['training', 'chain_tb', 'chain_tqe', 'chain_tqeb', 'chain_tbqeb']:
        # Export Torch models in onnx format for all services but training 
        # (export to onnx is also handled at the end of the trainer.train() method)
        
        model = torch_model_export_static(cfg=cfg, 
                                            model_dir=saved_model_dir, 
                                            model=model)

    # Creates dataloaders
    if mode not in ['benchmarking', 'deployment']:
      dataloaders = get_dataloaders(cfg=cfg)    

    if mode == "training":
        trainer = get_trainer(cfg=cfg,
                            model=model,
                            dataloaders=dataloaders)
        trainer.train()
        print("[INFO] training complete")

    elif mode == "evaluation":
        
        # Generates the model to be loaded on the stm32n6 device using stedgeai core,
        # then loads it and validates in on the device if required.
        
        gen_load_val(cfg=cfg, model=model)
        # Launches evaluation on the target through the model zoo evaluation service
        os.chdir(os.path.dirname(os.path.realpath(__file__)))
        evaluator = get_evaluator(cfg=cfg,
                                model=model,
                                dataloaders=dataloaders)
        metrics = evaluator.evaluate()    
        print("[INFO] evaluation complete")

    elif mode == "quantization":
        quantizer = get_quantizer(cfg=cfg,
                                model=model,
                                dataloaders=dataloaders)
        quantized_model = quantizer.quantize()
        print("[INFO] quantization complete")

    elif mode == "prediction":
        # Generates the model to be loaded on the stm32n6 device using stedgeai core,
        # then loads it and validates in on the device if required.
        gen_load_val_predict(cfg=cfg, model=model)
        # Launches prediction on the target through the model zoo prediction service
        os.chdir(os.path.dirname(os.path.realpath(__file__)))
        predictor = get_predictor(cfg=cfg,
                                model=model,
                                dataloaders=dataloaders)
        predictor.predict()
        print("[INFO] prediction complete")

    elif mode == 'benchmarking':
        benchmark(cfg, model_path_to_benchmark=model.model_path)
        print("[INFO] benchmarking complete")

    elif mode == 'deployment':
        if cfg.hardware_type == "MPU":
            deploy_mpu(cfg, model_path_to_deploy=model.model_path, credentials=credentials)
        else:
            deploy(cfg=cfg, model_path_to_deploy=model.model_path)
        print("[INFO] deployment complete")
        if cfg.deployment.hardware_setup.board == "STM32N6570-DK":
            print('[INFO] : Please on STM32N6570-DK toggle the boot switches to the left and power cycle the board.')

    elif mode == 'chain_tqe':
        trainer = get_trainer(cfg=cfg,
                            model=model,
                            dataloaders=dataloaders)
        trained_model = trainer.train()
        quantizer = get_quantizer(cfg=cfg,
                                model=trained_model,
                                dataloaders=dataloaders)
        quantized_model = quantizer.quantize()
        evaluator = get_evaluator(cfg=cfg,
                                model=quantized_model,
                                dataloaders=dataloaders)
        metrics = evaluator.evaluate()
        print("Trained model path:", trained_model.model_path)
        print("Quantized model path:", quantized_model.model_path)
        print("[INFO] chain_tqe complete")

    elif mode == 'chain_tqeb':
        credentials = None
        if cfg.tools.stedgeai.on_cloud:
            _, _, credentials = cloud_connect(stedgeai_core_version=cfg.tools.stedgeai.version)
        trainer = get_trainer(cfg=cfg,
                            model=model,
                            dataloaders=dataloaders)
        trained_model = trainer.train()
        quantizer = get_quantizer(cfg=cfg,
                                model=trained_model,
                                dataloaders=dataloaders)
        quantized_model = quantizer.quantize()
        evaluator = get_evaluator(cfg=cfg,
                                model=quantized_model,
                                dataloaders=dataloaders)
        metrics = evaluator.evaluate()
        benchmark(cfg, model_path_to_benchmark=quantized_model.model_path, credentials=credentials)
        print("Trained model path:", trained_model.model_path)
        print("Quantized model path:", quantized_model.model_path)
        print("[INFO] chain_tqeb complete")

    elif mode == 'chain_eqe':
        evaluator = get_evaluator(cfg=cfg,
                                model=model,
                                dataloaders=dataloaders)
        metrics = evaluator.evaluate()
        quantizer = get_quantizer(cfg=cfg,
                                model=model,
                                dataloaders=dataloaders)
        quantized_model = quantizer.quantize()
        evaluator = get_evaluator(cfg=cfg,
                                model=quantized_model,
                                dataloaders=dataloaders)
        metrics = evaluator.evaluate()
        print("Quantized model path:", quantized_model.model_path)
        print("[INFO] chain_eqe complete")

    elif mode == 'chain_eqeb':
        credentials = None
        if cfg.tools.stedgeai.on_cloud:
            _, _, credentials = cloud_connect(stedgeai_core_version=cfg.tools.stedgeai.version)
        evaluator = get_evaluator(cfg=cfg,
                                model=model,
                                dataloaders=dataloaders)
        metrics = evaluator.evaluate()
        quantizer = get_quantizer(cfg=cfg,
                                model=model,
                                dataloaders=dataloaders)
        quantized_model = quantizer.quantize()
        evaluator = get_evaluator(cfg=cfg,
                                model=quantized_model,
                                dataloaders=dataloaders)
        metrics = evaluator.evaluate()
        benchmark(cfg, model_path_to_benchmark=quantized_model.model_path, credentials=credentials)
        print("Quantized model path:", quantized_model.model_path)
        print("[INFO] chain_eqeb complete")

    elif mode == 'chain_qb':
        credentials = None
        if cfg.tools.stedgeai.on_cloud:
            _, _, credentials = cloud_connect(stedgeai_core_version=cfg.tools.stedgeai.version)
        quantizer = get_quantizer(cfg=cfg,
                                model=model,
                                dataloaders=dataloaders)
        quantized_model = quantizer.quantize()
        benchmark(cfg, model_path_to_benchmark=quantized_model.model_path, credentials=credentials)
        print("Quantized model path:", quantized_model.model_path)
        print("[INFO] chain_qb complete")

    elif mode == 'chain_qd':
        quantizer = get_quantizer(cfg=cfg,
                                model=model,
                                dataloaders=dataloaders)
        quantized_model = quantizer.quantize()
        if cfg.hardware_type == "MPU":
            deploy_mpu(cfg, model_path_to_deploy=quantized_model.model_path)
        else:
            deploy(cfg, model_path_to_deploy=quantized_model.model_path)
        print("Quantized model path:", quantized_model.model_path)
        print("[INFO] chain_qd complete")

    else:
        raise RuntimeError(f"Internal error: invalid operation mode: {mode}")

    if mode in ['benchmarking', 'chain_tbqeb', 'chain_qb', 'chain_eqeb']:
        mlflow.log_param("stedgeai_core_version", cfg.tools.stedgeai.version)
        mlflow.log_param("target", cfg.benchmarking.board)

    # logging the completion of the chain
    log_to_file(cfg.output_dir, f'operation finished: {mode}')

    # ClearML - Example how to get task's context anywhere in the file.
    # Checks if there's a valid ClearML configuration file
    if get_active_config_file() is not None:
        print(f"[INFO] : ClearML task connection")
        task = Task.current_task()
        task.connect(cfg)

def _torch_specific_initializations(cfg: DictConfig = None) -> None:
    """
    PyTorch-specific initializations.

    This function is used for PyTorch-specific initializations,
    - multi GPU set up 
    - Setting a random seed for reproducibility.

    Args:
        cfg (DictConfig): Configuration object.
    """
    
    if cfg.general.global_seed is not None:
        random.seed(cfg.general.global_seed)
        torch.manual_seed(cfg.general.global_seed)
        cudnn.deterministic = True
        warnings.warn(
            "You have chosen to seed training. This will turn on the CUDNN deterministic setting, "
            "which can slow down your training considerably! You may see unexpected behavior "
            "when restarting from checkpoints."
        )

    device = 'cuda' if torch.cuda.is_available() else 'cpu'

    # Torch specific initializations for Multi GPU 
    configure_module()
    configure_nccl()
    configure_omp()
    init_distributed_mode()
    cudnn.benchmark = True

    cfg.device = device
    
    
@hydra.main(version_base=None, config_path="", config_name="user_config")
def main(cfg: DictConfig) -> None:
    """
    Main entry point of the script.

    Args:
        cfg: Configuration dictionary.

    Returns:
        None
    """

    # Configure the GPU (the 'general' section may be missing) 
    if "general" in cfg and cfg.general:
        # Set upper limit on usable GPU memory
        if "gpu_memory_limit" in cfg.general and cfg.general.gpu_memory_limit:
            set_gpu_memory_limit(cfg.general.gpu_memory_limit)
        else:
            print("[WARNING] The usable GPU memory is unlimited.\n"
                "Please consider setting the 'gpu_memory_limit' attribute "
                "in the 'general' section of your configuration file.")

    # Parse the configuration file
    cfg = get_config(cfg)
    cfg.output_dir = HydraConfig.get().runtime.output_dir
    mlflow_ini(cfg)

    if cfg.model.framework == 'torch': 
        print (f'[Info]: Torch specific initializations')
        _torch_specific_initializations(cfg)

    # Checks if there's a valid ClearML configuration file
    print(f"[INFO] : ClearML config check")
    if get_active_config_file() is not None:
        print(f"[INFO] : ClearML initialization and configuration")
        # ClearML - Initializing ClearML's Task object.
        task = Task.init(project_name=cfg.general.project_name,
                        task_name='od_modelzoo_task')
        # ClearML - Optional yaml logging
        task.connect_configuration(name=cfg.operation_mode,
                                    configuration=cfg)

    # Seed global seed for random generators
    seed = get_random_seed(cfg)
    print(f'[INFO] : The random seed for this simulation is {seed}')
    if seed is not None:
        tf.keras.utils.set_random_seed(seed)

    # The default hardware type is "MCU".
    process_mode(cfg)


if __name__ == "__main__":
    parser = argparse.ArgumentParser()
    parser.add_argument('--config-path', type=str, default='', help='Path to folder containing configuration file')
    parser.add_argument('--config-name', type=str, default='user_config', help='name of the configuration file')

    # Add arguments to the parser
    parser.add_argument('params', nargs='*',
                        help='List of parameters to over-ride in config.yaml')
    args = parser.parse_args()

    # Call the main function
    main()

    # log the config_path and config_name parameters
    mlflow.log_param('config_path', args.config_path)
    mlflow.log_param('config_name', args.config_name)
    mlflow.end_run()