resnet50_trainer.py

# Module caffe2.python.examples.resnet50_trainer
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
from __future__ import unicode_literals

import argparse
import logging
import numpy as np
import time
import os

from caffe2.python import core, workspace, experiment_util, data_parallel_model
from caffe2.python import dyndep, optimizer
from caffe2.python import timeout_guard, model_helper, brew
from caffe2.proto import caffe2_pb2

import caffe2.python.models.resnet as resnet
from caffe2.python.modeling.initializers import Initializer, PseudoFP16Initializer
import caffe2.python.predictor.predictor_exporter as pred_exp
import caffe2.python.predictor.predictor_py_utils as pred_utils
from caffe2.python.predictor_constants import predictor_constants as predictor_constants

'''
Parallelized multi-GPU distributed trainer for Resne(X)t.
Can be used to train on imagenet data, for example.
The default parameters can train a standard Resnet-50 (1x64d), and parameters
can be provided to train ResNe(X)t models (e.g., ResNeXt-101 32x4d).

To run the trainer in single-machine multi-gpu mode by setting num_shards = 1.

To run the trainer in multi-machine multi-gpu mode with M machines,
run the same program on all machines, specifying num_shards = M, and
shard_id = a unique integer in the set [0, M-1].

For rendezvous (the trainer processes have to know about each other),
you can either use a directory path that is visible to all processes
(e.g. NFS directory), or use a Redis instance. Use the former by
passing the `file_store_path` argument. Use the latter by passing the
`redis_host` and `redis_port` arguments.
'''

logging.basicConfig()
log = logging.getLogger("ResNe(X)t_trainer")
log.setLevel(logging.DEBUG)

dyndep.InitOpsLibrary('@/caffe2/caffe2/distributed:file_store_handler_ops')
dyndep.InitOpsLibrary('@/caffe2/caffe2/distributed:redis_store_handler_ops')


def AddImageInput(
    model,
    reader,
    batch_size,
    img_size,
    dtype,
    is_test,
    mean_per_channel=None,
    std_per_channel=None,
):
    '''
    The image input operator loads image and label data from the reader and
    applies transformations to the images (random cropping, mirroring, ...).
    '''
    data, label = brew.image_input(
        model,
        reader, ["data", "label"],
        batch_size=batch_size,
        output_type=dtype,
        use_gpu_transform=True if core.IsGPUDeviceType(model._device_type) else False,
        use_caffe_datum=True,
        mean_per_channel=mean_per_channel,
        std_per_channel=std_per_channel,
        # mean_per_channel takes precedence over mean
        mean=128.,
        std=128.,
        scale=256,
        crop=img_size,
        mirror=1,
        is_test=is_test,
    )

    data = model.StopGradient(data, data)


def AddNullInput(model, reader, batch_size, img_size, dtype):
    '''
    The null input function uses a gaussian fill operator to emulate real image
    input. A label blob is hardcoded to a single value. This is useful if you
    want to test compute throughput or don't have a dataset available.
    '''
    suffix = "_fp16" if dtype == "float16" else ""
    model.param_init_net.GaussianFill(
        [],
        ["data" + suffix],
        shape=[batch_size, 3, img_size, img_size],
    )
    if dtype == "float16":
        model.param_init_net.FloatToHalf("data" + suffix, "data")

    model.param_init_net.ConstantFill(
        [],
        ["label"],
        shape=[batch_size],
        value=1,
        dtype=core.DataType.INT32,
    )


def SaveModel(args, train_model, epoch, use_ideep):
    prefix = "[]_{}".format(train_model._device_prefix, train_model._devices[0])
    predictor_export_meta = pred_exp.PredictorExportMeta(
        predict_net=train_model.net.Proto(),
        parameters=data_parallel_model.GetCheckpointParams(train_model),
        inputs=[prefix + "/data"],
        outputs=[prefix + "/softmax"],
        shapes={
            prefix + "/softmax": (1, args.num_labels),
            prefix + "/data": (args.num_channels, args.image_size, args.image_size)
        }
    )

    # save the train_model for the current epoch
    model_path = "%s/%s_%d.mdl" % (
        args.file_store_path,
        args.save_model_name,
        epoch,
    )

    # set db_type to be "minidb" instead of "log_file_db", which breaks
    # the serialization in save_to_db. Need to switch back to log_file_db
    # after migration
    pred_exp.save_to_db(
        db_type="minidb",
        db_destination=model_path,
        predictor_export_meta=predictor_export_meta,
        use_ideep = use_ideep
    )


def LoadModel(path, model, use_ideep):
    '''
    Load pretrained model from file
    '''
    log.info("Loading path: {}".format(path))
    meta_net_def = pred_exp.load_from_db(path, 'minidb')
    init_net = core.Net(pred_utils.GetNet(
        meta_net_def, predictor_constants.GLOBAL_INIT_NET_TYPE))
    predict_init_net = core.Net(pred_utils.GetNet(
        meta_net_def, predictor_constants.PREDICT_INIT_NET_TYPE))

    if use_ideep:
        predict_init_net.RunAllOnIDEEP()
    else:
        predict_init_net.RunAllOnGPU()
    if use_ideep:
        init_net.RunAllOnIDEEP()
    else:
        init_net.RunAllOnGPU()

    assert workspace.RunNetOnce(predict_init_net)
    assert workspace.RunNetOnce(init_net)

    # Hack: fix iteration counter which is in CUDA context after load model
    itercnt = workspace.FetchBlob("optimizer_iteration")
    workspace.FeedBlob(
        "optimizer_iteration",
        itercnt,
        device_option=core.DeviceOption(caffe2_pb2.CPU, 0)
    )


def RunEpoch(
    args,
    epoch,
    train_model,
    test_model,
    total_batch_size,
    num_shards,
    expname,
    explog,
):
    '''
    Run one epoch of the trainer.
    TODO: add checkpointing here.
    '''
    # TODO: add loading from checkpoint
    log.info("Starting epoch {}/{}".format(epoch, args.num_epochs))
    epoch_iters = int(args.epoch_size / total_batch_size / num_shards)
    test_epoch_iters = int(args.test_epoch_size / total_batch_size / num_shards)
    for i in range(epoch_iters):
        # This timeout is required (temporarily) since CUDA-NCCL
        # operators might deadlock when synchronizing between GPUs.
        timeout = args.first_iter_timeout if i == 0 else args.timeout
        with timeout_guard.CompleteInTimeOrDie(timeout):
            t1 = time.time()
            workspace.RunNet(train_model.net.Proto().name)
            t2 = time.time()
            dt = t2 - t1

        fmt = "Finished iteration {}/{} of epoch {} ({:.2f} images/sec)"
        log.info(fmt.format(i + 1, epoch_iters, epoch, total_batch_size / dt))
        prefix = "{}_{}".format(
            train_model._device_prefix,
            train_model._devices[0])
        accuracy = workspace.FetchBlob(prefix + '/accuracy')
        loss = workspace.FetchBlob(prefix + '/loss')
        train_fmt = "Training loss: {}, accuracy: {}"
        log.info(train_fmt.format(loss, accuracy))

    num_images = epoch * epoch_iters * total_batch_size
    prefix = "{}_{}".format(train_model._device_prefix, train_model._devices[0])
    accuracy = workspace.FetchBlob(prefix + '/accuracy')
    loss = workspace.FetchBlob(prefix + '/loss')
    learning_rate = workspace.FetchBlob(
        data_parallel_model.GetLearningRateBlobNames(train_model)[0]
    )
    test_accuracy = 0
    test_accuracy_top5 = 0
    if test_model is not None:
        # Run 100 iters of testing
        ntests = 0
        for _ in range(test_epoch_iters):
            workspace.RunNet(test_model.net.Proto().name)
            for g in test_model._devices:
                test_accuracy += np.asscalar(workspace.FetchBlob(
                    "{}_{}".format(test_model._device_prefix, g) + '/accuracy'
                ))
                test_accuracy_top5 += np.asscalar(workspace.FetchBlob(
                    "{}_{}".format(test_model._device_prefix, g) + '/accuracy_top5'
                ))
                ntests += 1
        test_accuracy /= ntests
        test_accuracy_top5 /= ntests
    else:
        test_accuracy = (-1)
        test_accuracy_top5 = (-1)

    explog.log(
        input_count=num_images,
        batch_count=(i + epoch * epoch_iters),
        additional_values={
            'accuracy': accuracy,
            'loss': loss,
            'learning_rate': learning_rate,
            'epoch': epoch,
            'top1_test_accuracy': test_accuracy,
            'top5_test_accuracy': test_accuracy_top5,
        }
    )
    assert loss < 40, "Exploded gradients :("

    # TODO: add checkpointing
    return epoch + 1


def Train(args):
    # Either use specified device list or generate one
    if args.gpus is not None:
        gpus = [int(x) for x in args.gpus.split(',')]
        num_gpus = len(gpus)
    else:
        gpus = list(range(args.num_gpus))
        num_gpus = args.num_gpus

    log.info("Running on GPUs: {}".format(gpus))

    # Verify valid batch size
    total_batch_size = args.batch_size
    batch_per_device = total_batch_size // num_gpus
    assert \
        total_batch_size % num_gpus == 0, \
        "Number of GPUs must divide batch size"

    # Verify valid image mean/std per channel
    if args.image_mean_per_channel:
        assert \
            len(args.image_mean_per_channel) == args.num_channels, \
            "The number of channels of image mean doesn't match input"

    if args.image_std_per_channel:
        assert \
            len(args.image_std_per_channel) == args.num_channels, \
            "The number of channels of image std doesn't match input"

    # Round down epoch size to closest multiple of batch size across machines
    global_batch_size = total_batch_size * args.num_shards
    epoch_iters = int(args.epoch_size / global_batch_size)

    assert \
        epoch_iters > 0, \
        "Epoch size must be larger than batch size times shard count"

    args.epoch_size = epoch_iters * global_batch_size
    log.info("Using epoch size: {}".format(args.epoch_size))

    # Create ModelHelper object
    if args.use_ideep:
        train_arg_scope = {
            'use_cudnn': False,
            'cudnn_exhaustive_search': False,
            'training_mode': 1
        }
    else:
        train_arg_scope = {
            'order': 'NCHW',
            'use_cudnn': True,
            'cudnn_exhaustive_search': True,
            'ws_nbytes_limit': (args.cudnn_workspace_limit_mb * 1024 * 1024),
        }
    train_model = model_helper.ModelHelper(
        name='resnext' + str(args.num_layers), arg_scope=train_arg_scope
    )

    num_shards = args.num_shards
    shard_id = args.shard_id

    # Expect interfaces to be comma separated.
    # Use of multiple network interfaces is not yet complete,
    # so simply use the first one in the list.
    interfaces = args.distributed_interfaces.split(",")

    # Rendezvous using MPI when run with mpirun
    if os.getenv("OMPI_COMM_WORLD_SIZE") is not None:
        num_shards = int(os.getenv("OMPI_COMM_WORLD_SIZE", 1))
        shard_id = int(os.getenv("OMPI_COMM_WORLD_RANK", 0))
        if num_shards > 1:
            rendezvous = dict(
                kv_handler=None,
                num_shards=num_shards,
                shard_id=shard_id,
                engine="GLOO",
                transport=args.distributed_transport,
                interface=interfaces[0],
                mpi_rendezvous=True,
                exit_nets=None)

    elif num_shards > 1:
        # Create rendezvous for distributed computation
        store_handler = "store_handler"
        if args.redis_host is not None:
            # Use Redis for rendezvous if Redis host is specified
            workspace.RunOperatorOnce(
                core.CreateOperator(
                    "RedisStoreHandlerCreate", [], [store_handler],
                    host=args.redis_host,
                    port=args.redis_port,
                    prefix=args.run_id,
                )
            )
        else:
            # Use filesystem for rendezvous otherwise
            workspace.RunOperatorOnce(
                core.CreateOperator(
                    "FileStoreHandlerCreate", [], [store_handler],
                    path=args.file_store_path,
                    prefix=args.run_id,
                )
            )

        rendezvous = dict(
            kv_handler=store_handler,
            shard_id=shard_id,
            num_shards=num_shards,
            engine="GLOO",
            transport=args.distributed_transport,
            interface=interfaces[0],
            exit_nets=None)

    else:
        rendezvous = None

    # Model building functions
    def create_resnext_model_ops(model, loss_scale):
        initializer = (PseudoFP16Initializer if args.dtype == 'float16'
                       else Initializer)

        with brew.arg_scope([brew.conv, brew.fc],
                            WeightInitializer=initializer,
                            BiasInitializer=initializer,
                            enable_tensor_core=args.enable_tensor_core,
                            float16_compute=args.float16_compute):
            pred = resnet.create_resnext(
                model,
                "data",
                num_input_channels=args.num_channels,
                num_labels=args.num_labels,
                num_layers=args.num_layers,
                num_groups=args.resnext_num_groups,
                num_width_per_group=args.resnext_width_per_group,
                no_bias=True,
                no_loss=True,
            )

        if args.dtype == 'float16':
            pred = model.net.HalfToFloat(pred, pred + '_fp32')

        softmax, loss = model.SoftmaxWithLoss([pred, 'label'],
                                              ['softmax', 'loss'])
        loss = model.Scale(loss, scale=loss_scale)
        brew.accuracy(model, [softmax, "label"], "accuracy", top_k=1)
        brew.accuracy(model, [softmax, "label"], "accuracy_top5", top_k=5)
        return [loss]

    def add_optimizer(model):
        stepsz = int(30 * args.epoch_size / total_batch_size / num_shards)

        if args.float16_compute:
            # TODO: merge with multi-prceision optimizer
            opt = optimizer.build_fp16_sgd(
                model,
                args.base_learning_rate,
                momentum=0.9,
                nesterov=1,
                weight_decay=args.weight_decay,   # weight decay included
                policy="step",
                stepsize=stepsz,
                gamma=0.1
            )
        else:
            optimizer.add_weight_decay(model, args.weight_decay)
            opt = optimizer.build_multi_precision_sgd(
                model,
                args.base_learning_rate,
                momentum=0.9,
                nesterov=1,
                policy="step",
                stepsize=stepsz,
                gamma=0.1
            )
        return opt

    # Define add_image_input function.
    # Depends on the "train_data" argument.
    # Note that the reader will be shared with between all GPUS.
    if args.train_data == "null":
        def add_image_input(model):
            AddNullInput(
                model,
                None,
                batch_size=batch_per_device,
                img_size=args.image_size,
                dtype=args.dtype,
            )
    else:
        reader = train_model.CreateDB(
            "reader",
            db=args.train_data,
            db_type=args.db_type,
            num_shards=num_shards,
            shard_id=shard_id,
        )

        def add_image_input(model):
            AddImageInput(
                model,
                reader,
                batch_size=batch_per_device,
                img_size=args.image_size,
                dtype=args.dtype,
                is_test=False,
                mean_per_channel=args.image_mean_per_channel,
                std_per_channel=args.image_std_per_channel,
            )

    def add_post_sync_ops(model):
        """Add ops applied after initial parameter sync."""
        for param_info in model.GetOptimizationParamInfo(model.GetParams()):
            if param_info.blob_copy is not None:
                model.param_init_net.HalfToFloat(
                    param_info.blob,
                    param_info.blob_copy[core.DataType.FLOAT]
                )

    # Create parallelized model
    data_parallel_model.Parallelize(
        train_model,
        input_builder_fun=add_image_input,
        forward_pass_builder_fun=create_resnext_model_ops,
        optimizer_builder_fun=add_optimizer,
        post_sync_builder_fun=add_post_sync_ops,
        devices=gpus,
        rendezvous=rendezvous,
        optimize_gradient_memory=False,
        cpu_device=args.use_cpu,
        ideep=args.use_ideep,
        shared_model=args.use_cpu,
        combine_spatial_bn=args.use_cpu,
    )

    data_parallel_model.OptimizeGradientMemory(train_model, {}, set(), False)

    workspace.RunNetOnce(train_model.param_init_net)
    workspace.CreateNet(train_model.net)

    # Add test model, if specified
    test_model = None
    if (args.test_data is not None):
        log.info("----- Create test net ----")
        if args.use_ideep:
            test_arg_scope = {
                'use_cudnn': False,
                'cudnn_exhaustive_search': False,
            }
        else:
            test_arg_scope = {
                'order': "NCHW",
                'use_cudnn': True,
                'cudnn_exhaustive_search': True,
            }
        test_model = model_helper.ModelHelper(
            name='resnext' + str(args.num_layers) + "_test",
            arg_scope=test_arg_scope,
            init_params=False,
        )

        test_reader = test_model.CreateDB(
            "test_reader",
            db=args.test_data,
            db_type=args.db_type,
        )

        def test_input_fn(model):
            AddImageInput(
                model,
                test_reader,
                batch_size=batch_per_device,
                img_size=args.image_size,
                dtype=args.dtype,
                is_test=True,
                mean_per_channel=args.image_mean_per_channel,
                std_per_channel=args.image_std_per_channel,
            )

        data_parallel_model.Parallelize(
            test_model,
            input_builder_fun=test_input_fn,
            forward_pass_builder_fun=create_resnext_model_ops,
            post_sync_builder_fun=add_post_sync_ops,
            param_update_builder_fun=None,
            devices=gpus,
            cpu_device=args.use_cpu,
        )
        workspace.RunNetOnce(test_model.param_init_net)
        workspace.CreateNet(test_model.net)

    epoch = 0
    # load the pre-trained model and reset epoch
    if args.load_model_path is not None:
        LoadModel(args.load_model_path, train_model, args.use_ideep)

        # Sync the model params
        data_parallel_model.FinalizeAfterCheckpoint(train_model)

        # reset epoch. load_model_path should end with *_X.mdl,
        # where X is the epoch number
        last_str = args.load_model_path.split('_')[-1]
        if last_str.endswith('.mdl'):
            epoch = int(last_str[:-4])
            log.info("Reset epoch to {}".format(epoch))
        else:
            log.warning("The format of load_model_path doesn't match!")

    expname = "resnext_%d_gpu%d_b%d_L%d_lr%.2f_v2" % (
        args.num_layers,
        args.num_gpus,
        total_batch_size,
        args.num_labels,
        args.base_learning_rate,
    )

    explog = experiment_util.ModelTrainerLog(expname, args)

    # Run the training one epoch a time
    while epoch < args.num_epochs:
        epoch = RunEpoch(
            args,
            epoch,
            train_model,
            test_model,
            total_batch_size,
            num_shards,
            expname,
            explog
        )

        # Save the model for each epoch
        SaveModel(args, train_model, epoch, args.use_ideep)

        model_path = "%s/%s_" % (
            args.file_store_path,
            args.save_model_name
        )
        # remove the saved model from the previous epoch if it exists
        if os.path.isfile(model_path + str(epoch - 1) + ".mdl"):
            os.remove(model_path + str(epoch - 1) + ".mdl")


def main():
    # TODO: use argv
    parser = argparse.ArgumentParser(
        description="Caffe2: ResNe(X)t training"
    )
    parser.add_argument("--train_data", type=str, default=None, required=True,
                        help="Path to training data (or 'null' to simulate)")
    parser.add_argument("--num_layers", type=int, default=50,
                        help="The number of layers in ResNe(X)t model")
    parser.add_argument("--resnext_num_groups", type=int, default=1,
                        help="The cardinality of resnext")
    parser.add_argument("--resnext_width_per_group", type=int, default=64,
                        help="The cardinality of resnext")
    parser.add_argument("--test_data", type=str, default=None,
                        help="Path to test data")
    parser.add_argument("--image_mean_per_channel", type=float, nargs='+',
                        help="The per channel mean for the images")
    parser.add_argument("--image_std_per_channel", type=float, nargs='+',
                        help="The per channel standard deviation for the images")
    parser.add_argument("--test_epoch_size", type=int, default=50000,
                        help="Number of test images")
    parser.add_argument("--db_type", type=str, default="lmdb",
                        help="Database type (such as lmdb or leveldb)")
    parser.add_argument("--gpus", type=str,
                        help="Comma separated list of GPU devices to use")
    parser.add_argument("--num_gpus", type=int, default=1,
                        help="Number of GPU devices (instead of --gpus)")
    parser.add_argument("--num_channels", type=int, default=3,
                        help="Number of color channels")
    parser.add_argument("--image_size", type=int, default=224,
                        help="Input image size (to crop to)")
    parser.add_argument("--num_labels", type=int, default=1000,
                        help="Number of labels")
    parser.add_argument("--batch_size", type=int, default=32,
                        help="Batch size, total over all GPUs")
    parser.add_argument("--epoch_size", type=int, default=1500000,
                        help="Number of images/epoch, total over all machines")
    parser.add_argument("--num_epochs", type=int, default=1000,
                        help="Num epochs.")
    parser.add_argument("--base_learning_rate", type=float, default=0.1,
                        help="Initial learning rate.")
    parser.add_argument("--weight_decay", type=float, default=1e-4,
                        help="Weight decay (L2 regularization)")
    parser.add_argument("--cudnn_workspace_limit_mb", type=int, default=64,
                        help="CuDNN workspace limit in MBs")
    parser.add_argument("--num_shards", type=int, default=1,
                        help="Number of machines in distributed run")
    parser.add_argument("--shard_id", type=int, default=0,
                        help="Shard id.")
    parser.add_argument("--run_id", type=str,
                        help="Unique run identifier (e.g. uuid)")
    parser.add_argument("--redis_host", type=str,
                        help="Host of Redis server (for rendezvous)")
    parser.add_argument("--redis_port", type=int, default=6379,
                        help="Port of Redis server (for rendezvous)")
    parser.add_argument("--file_store_path", type=str, default="/tmp",
                        help="Path to directory to use for rendezvous")
    parser.add_argument("--save_model_name", type=str, default="resnext_model",
                        help="Save the trained model to a given name")
    parser.add_argument("--load_model_path", type=str, default=None,
                        help="Load previously saved model to continue training")
    parser.add_argument("--use_cpu", type=bool, default=False,
                        help="Use CPU instead of GPU")
    parser.add_argument("--use_ideep", type=bool, default=False,
                        help="Use ideep")
    parser.add_argument('--dtype', default='float',
                        choices=['float', 'float16'],
                        help='Data type used for training')
    parser.add_argument('--float16_compute', action='store_true',
                        help="Use float 16 compute, if available")
    parser.add_argument('--enable_tensor_core', action='store_true',
                        help='Enable Tensor Core math for Conv and FC ops')
    parser.add_argument("--distributed_transport", type=str, default="tcp",
                        help="Transport to use for distributed run [tcp|ibverbs]")
    parser.add_argument("--distributed_interfaces", type=str, default="",
                        help="Network interfaces to use for distributed run")

    parser.add_argument("--first_iter_timeout", type=int, default=600,
                        help="Timeout (secs) of the first iteration "
                        "(default: %(default)s)")
    parser.add_argument("--timeout", type=int, default=60,
                        help="Timeout (secs) of each (except the first) iteration "
                        "(default: %(default)s)")

    args = parser.parse_args()

    Train(args)


if __name__ == '__main__':
    workspace.GlobalInit(['caffe2', '--caffe2_log_level=2'])
    main()