|
- import mindspore as ms
- import mindspore.nn as nn
- from mindspore import Model
-
- from mindcv.models import create_model
- from mindcv.data import create_dataset, create_transforms, create_loader
- from mindcv.loss import create_loss
- from config import parse_args
- from mindcv.utils.utils import check_batch_size
- from mindcv.utils.callbacks import ValCallback
-
- import os
- import json
- import time
- import moxing as mox
- from mindspore.communication import init, get_rank, get_group_size
- from mindspore import Model, load_checkpoint, load_param_into_net
- # "---Qi Zhi---"
- def ObsToEnv(obs_data_url, data_dir):
- try:
- mox.file.copy_parallel(obs_data_url, data_dir)
- print("Successfully Download {} to {}".format(obs_data_url, data_dir))
- except Exception as e:
- print('moxing download {} to {} failed: '.format(obs_data_url, data_dir) + str(e))
- # Set a cache file to determine whether the data has been copied to obs.
- # If this file exists during multi-card training, there is no need to copy the dataset multiple times.
- f = open("/cache/download_input.txt", 'w')
- f.close()
- try:
- if os.path.exists("/cache/download_input.txt"):
- print("download_input succeed")
- except Exception as e:
- print("download_input failed")
- return
-
-
- # Copy the output to obs###
- def EnvToObs(train_dir, obs_train_url):
- try:
- mox.file.copy_parallel(train_dir, obs_train_url)
- print("Successfully Upload {} to {}".format(train_dir, obs_train_url))
- except Exception as e:
- print('moxing upload {} to {} failed: '.format(train_dir, obs_train_url) + str(e))
- return
-
-
- def DownloadFromQizhi(obs_data_url, data_dir):
- device_num = int(os.getenv('RANK_SIZE'))
- if device_num == 1:
- ObsToEnv(obs_data_url, data_dir)
- ms.set_context(mode=ms.GRAPH_MODE, device_target=args.device_target)
- if device_num > 1:
- # set device_id and init for multi-card training
- ms.set_context(mode=ms.GRAPH_MODE, device_target=args.device_target,
- device_id=int(os.getenv('ASCEND_DEVICE_ID')))
- ms.reset_auto_parallel_context()
- ms.set_auto_parallel_context(device_num=device_num, parallel_mode='data_parallel', gradients_mean=True,
- parameter_broadcast=True)
- init()
- # Copying obs data does not need to be executed multiple times, just let the 0th card copy the data
- local_rank = int(os.getenv('RANK_ID'))
- if local_rank % 8 == 0:
- ObsToEnv(obs_data_url, data_dir)
- # If the cache file does not exist, it means that the copy data has not been completed,
- # and Wait for 0th card to finish copying data
- while not os.path.exists("/cache/download_input.txt"):
- time.sleep(1)
- return
-
-
- def UploadToQizhi(train_dir, obs_train_url):
- device_num = int(os.getenv('RANK_SIZE'))
- local_rank = int(os.getenv('RANK_ID'))
- if device_num == 1:
- EnvToObs(train_dir, obs_train_url)
- if device_num > 1:
- if local_rank % 8 == 0:
- EnvToObs(train_dir, obs_train_url)
- return
-
- def ObsUrlToEnv(obs_ckpt_url, ckpt_url):
- try:
- mox.file.copy(obs_ckpt_url, ckpt_url)
- print("Successfully Download {} to {}".format(obs_ckpt_url,ckpt_url))
- except Exception as e:
- print('moxing download {} to {} failed: '.format(obs_ckpt_url, ckpt_url) + str(e))
- return
- # "---Qi Zhi---"
- def validate(args):
- ms.set_context(mode=args.mode)
-
- # create dataset
- dataset_eval = create_dataset(
- name=args.dataset,
- root=args.data_dir,
- split=args.val_split,
- num_parallel_workers=args.num_parallel_workers,
- download=args.dataset_download)
-
-
- # create transform
- transform_list = create_transforms(
- dataset_name=args.dataset,
- is_training=False,
- image_resize=args.image_resize,
- crop_pct=args.crop_pct,
- interpolation=args.interpolation,
- mean=args.mean,
- std=args.std
- )
-
- # read num clases
- num_classes = dataset_eval.num_classes() if args.num_classes==None else args.num_classes
-
- # check batch size
- batch_size = check_batch_size(dataset_eval.get_dataset_size(), args.batch_size)
-
- # load dataset
- loader_eval = create_loader(
- dataset=dataset_eval,
- batch_size=batch_size,
- drop_remainder=False,
- is_training=False,
- transform=transform_list,
- num_parallel_workers=args.num_parallel_workers,
- )
-
- # create model
- network = create_model(model_name=args.model,
- num_classes=num_classes,
- drop_rate=args.drop_rate,
- drop_path_rate=args.drop_path_rate,
- pretrained=args.pretrained,
- checkpoint_path=args.ckpt_path,
- use_ema=args.use_ema)
- network.set_train(False)
-
- # create loss
- loss = create_loss(name=args.loss,
- reduction=args.reduction,
- label_smoothing=args.label_smoothing,
- aux_factor=args.aux_factor)
-
- # Define eval metrics.
- if num_classes >= 5:
- eval_metrics = {'Top_1_Accuracy': nn.Top1CategoricalAccuracy(),
- 'Top_5_Accuracy': nn.Top5CategoricalAccuracy(),
- 'loss': nn.metrics.Loss()
- }
- else:
- eval_metrics = {'Top_1_Accuracy': nn.Top1CategoricalAccuracy(),
- 'loss': nn.metrics.Loss()}
-
- # init model
- ckpt_url = '/cache/checkpoint.ckpt'
- ObsUrlToEnv(args.ckpt_url, ckpt_url)
- load_param_into_net(network, load_checkpoint(ckpt_url))
- model = Model(network, loss_fn=loss, metrics=eval_metrics)
-
- # log
- num_batches = loader_eval.get_dataset_size()
- print(f"Model: {args.model}")
- print(f"Num batches: {num_batches}")
- print(f"Start validating...")
-
- # validate
- result = model.eval(loader_eval,
- dataset_sink_mode=False,
- callbacks=[ValCallback(args.log_interval)])
- print(result)
-
-
- if __name__ == '__main__':
- args = parse_args()
- # "---Qi Zhi---"
- data_dir = '/cache/data'
- train_dir = '/cache/output'
- if not os.path.exists(data_dir):
- os.makedirs(data_dir, exist_ok=True)
- if not os.path.exists(train_dir):
- os.makedirs(train_dir, exist_ok=True)
-
- # Initialize and copy data to training image
- # DownloadFromQizhi is much slower than sync_data but sync_data usually abort;
- DownloadFromQizhi(args.data_url, data_dir)
- # data_url = args.data_url
- # local_data_path = '/cache/dataset'
- # os.makedirs(local_data_path, exist_ok=True)
- # sync_data(data_url, local_data_path, threads=256)
- print(f"data_dir:{os.listdir(data_dir)}")
- if "imagenet" in os.listdir(data_dir):
- data_dir = os.path.join(data_dir, "imagenet")
- args.data_dir = data_dir
- device_num = int(os.getenv('RANK_SIZE'))
- if device_num == 1:
- args.ckpt_save_dir = train_dir + "/"
- if device_num > 1:
- args.ckpt_save_dir = train_dir + "/" + str(get_rank()) + "/"
- # "---Qi Zhi---"
- validate(args)
-
- UploadToQizhi(args.ckpt_save_dir, args.train_url)
|