m53297601
/
mindspore
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

modify readme and some confusion code, fix some bug

Browse Source
pull/6237/head
Payne 5 years ago
parent 7393c665e0
commit cd767f0f29
9 changed files with 97 additions and 128 deletions
Show Stats Download Patch File Download Diff File

118
model_zoo/official/cv/mobilenetv2/README.md
Unescape View File

@ -4,17 +4,16 @@
- [Model Architecture](#model-architecture)
- [Dataset](#dataset)
- [Features](#features)
- [Mixed Precision](#mixed-precision)
- [Mixed Precision](#mixed-precision(ascend))
- [Environment Requirements](#environment-requirements)
- [Script Description](#script-description)
- [Script and Sample Code](#script-and-sample-code)
- [Training Process](#training-process)
- [Evaluation Process](#evaluation-process)
- [Evaluation](#evaluation)
- [Evaluation Process](#eval-process)
- [Model Description](#model-description)
- [Performance](#performance)
- [Training Performance](#evaluation-performance)
- [Inference Performance](#evaluation-performance)
- [Training Performance](#training-performance)
- [Evaluation Performance](#evaluation-performance)
- [Description of Random Situation](#description-of-random-situation)
- [ModelZoo Homepage](#modelzoo-homepage)
@ -38,7 +37,7 @@ Dataset used: [imagenet](http://www.image-net.org/)
- Train: 120G, 1.2W images
- Test: 5G, 50000 images
- Data format: RGB images.
- Note: Data will be processed in src/dataset.py
- Note: Data will be processed in src/dataset.py
# [Features](#contents)
@ -92,84 +91,84 @@ You can start training using python or shell scripts. The usage of shell scripts
### Launch
```
```shell
# training example
python:
Ascend: python train.py --dataset_path ~/imagenet/train/ --platform Ascend --train_method train
GPU: python train.py --dataset_path ~/imagenet/train/ --platform GPU --train_method train
CPU: python train.py --dataset_path ~/imagenet/train/ --platform CPU --train_method train
Ascend: python train.py --platform Ascend --dataset_path [TRAIN_DATASET_PATH] --train_method train
GPU: python train.py --platform GPU --dataset_path [TRAIN_DATASET_PATH] --train_method train
CPU: python train.py --platform CPU --dataset_path [TRAIN_DATASET_PATH] --train_method train
shell:
Ascend: sh run_train.sh Ascend 8 0,1,2,3,4,5,6,7 hccl_config.json ~/imagenet/train/ train
GPU: sh run_train.sh GPU 8 0,1,2,3,4,5,6,7 ~/imagenet/train/ train
CPU: sh run_train.sh CPU ~/imagenet/train/ train
Ascend: sh run_train.sh Ascend 8 0,1,2,3,4,5,6,7 hccl_config.json [TRAIN_DATASET_PATH] train
GPU: sh run_train.sh GPU 8 0,1,2,3,4,5,6,7 [TRAIN_DATASET_PATH] train
CPU: sh run_train.sh CPU [TRAIN_DATASET_PATH] train
# fine tune example
python:
Ascend: python train.py --dataset_path ~/imagenet/train/ --platform Ascend --train_method fine_tune ./pretrain_checkpoint/mobilenetv2_199.ckpt
GPU: python train.py --dataset_path ~/imagenet/train/ --platform GPU --train_method fine_tune ./pretrain_checkpoint/mobilenetv2_199.ckpt
CPU: python train.py --dataset_path ~/imagenet/train/ --platform CPU --train_method fine_tune ./pretrain_checkpoint/mobilenetv2_199.ckpt
Ascend: python train.py --platform Ascend --dataset_path [TRAIN_DATASET_PATH] --train_method fine_tune ./pretrain_checkpoint/mobilenetv2.ckpt
GPU: python train.py --platform GPU --dataset_path [TRAIN_DATASET_PATH] --train_method fine_tune ./pretrain_checkpoint/mobilenetv2.ckpt
CPU: python train.py --platform CPU --dataset_path [TRAIN_DATASET_PATH] --train_method fine_tune ./pretrain_checkpoint/mobilenetv2.ckpt
shell:
Ascend: sh run_train.sh Ascend 8 0,1,2,3,4,5,6,7 hccl_config.json ~/imagenet/train/ fine_tune ./pretrain_checkpoint/mobilenetv2_199.ckpt
GPU: sh run_train.sh GPU 8 0,1,2,3,4,5,6,7 ~/imagenet/train/ fine_tune ./pretrain_checkpoint/mobilenetv2_199.ckpt
CPU: sh run_train.sh CPU ~/imagenet/train/ fine_tune ./pretrain_checkpoint/mobilenetv2_199.ckpt
Ascend: sh run_train.sh Ascend 8 0,1,2,3,4,5,6,7 hccl_config.json [TRAIN_DATASET_PATH] fine_tune ./pretrain_checkpoint/mobilenetv2.ckpt
GPU: sh run_train.sh GPU 8 0,1,2,3,4,5,6,7 [TRAIN_DATASET_PATH] fine_tune ./pretrain_checkpoint/mobilenetv2.ckpt
CPU: sh run_train.sh CPU [TRAIN_DATASET_PATH] fine_tune ./pretrain_checkpoint/mobilenetv2.ckpt
# incremental learn example
python:
Ascend: python train.py --dataset_path ~/imagenet/train/ --platform Ascend --train_method incremental_learn ./pretrain_checkpoint/mobilenetv2_199.ckpt
GPU: python train.py --dataset_path ~/imagenet/train/ --platform GPU --train_method incremental_learn ./pretrain_checkpoint/mobilenetv2_199.ckpt
CPU: python train.py --dataset_path ~/imagenet/train/ --platform CPU --train_method incremental_learn ./pretrain_checkpoint/mobilenetv2_199.ckpt
Ascend: python --platform Ascend train.py --dataset_path [TRAIN_DATASET_PATH] --train_method incremental_learn ./pretrain_checkpoint/mobilenetv2.ckpt ./checkpoint/mobilenetv2_head_15.ckpt
GPU: python --platform GPU train.py --dataset_path [TRAIN_DATASET_PATH] --train_method incremental_learn ./pretrain_checkpoint/mobilenetv2.ckpt ./checkpoint/mobilenetv2_head_15.ckpt
CPU: python --platform CPU train.py --dataset_path [TRAIN_DATASET_PATH] --train_method incremental_learn ./pretrain_checkpoint/mobilenetv2.ckpt ./checkpoint/mobilenetv2_head_15.ckpt
shell:
Ascend: sh run_train.sh Ascend 8 0,1,2,3,4,5,6,7 hccl_config.json ~/imagenet/train/ incremental_learn ./pretrain_checkpoint/mobilenetv2_199.ckpt
GPU: sh run_train.sh GPU 8 0,1,2,3,4,5,6,7 ~/imagenet/train/ incremental_learn ./pretrain_checkpoint/mobilenetv2_199.ckpt
CPU: sh run_train.sh CPU ~/imagenet/train/ incremental_learn ./pretrain_checkpoint/mobilenetv2_199.ckpt
Ascend: sh run_train.sh Ascend 8 0,1,2,3,4,5,6,7 hccl_config.json [TRAIN_DATASET_PATH] incremental_learn ./pretrain_checkpoint/mobilenetv2.ckpt ./checkpoint/mobilenetv2_head_15.ckpt
GPU: sh run_train.sh GPU 8 0,1,2,3,4,5,6,7 [TRAIN_DATASET_PATH] incremental_learn ./pretrain_checkpoint/mobilenetv2.ckpt ./checkpoint/mobilenetv2_head_15.ckpt
CPU: sh run_train.sh CPU [TRAIN_DATASET_PATH] incremental_learn ./pretrain_checkpoint/mobilenetv2_199.ckpt ./checkpoint/mobilenetv2_head_15.ckpt
```
### Result
Training result will be stored in the example path. Checkpoints will be stored at `. /checkpoint` by default, and training log will be redirected to `./train/train.log` like followings.
Training result will be stored in the example path. Checkpoints will be stored at `. /checkpoint` by default, and training log will be redirected to `./train.log` like followings with the platform CPU and GPU, will be wrote to `./train/rank*/log*.log` with the platform Ascend .
```
```shell
epoch: [ 0/200], step:[ 624/ 625], loss:[5.258/5.258], time:[140412.236], lr:[0.100]
epoch time: 140522.500, per step time: 224.836, avg loss: 5.258
epoch: [ 1/200], step:[ 624/ 625], loss:[3.917/3.917], time:[138221.250], lr:[0.200]
epoch time: 138331.250, per step time: 221.330, avg loss: 3.917
```
## [Eval process](#contents)
## [Evaluation process](#contents)
### Usage
You can start training using python or shell scripts. The usage of shell scripts as follows:
You can start training using python or shell scripts.If the train method is train or fine tune, should not input the `[CHECKPOINT_PATH]` The usage of shell scripts as follows:
- Ascend: sh run_infer.sh Ascend [DATASET_PATH] [CHECKPOINT_PATH] [HEAD_CKPT_PATH]
- GPU: sh run_infer.sh GPU [DATASET_PATH] [CHECKPOINT_PATH] [HEAD_CKPT_PATH]
- CPU: sh run_infer.sh CPU [DATASET_PATH] [BACKBONE_CKPT_PATH] [HEAD_CKPT_PATH]
- Ascend: sh run_eval.sh Ascend [DATASET_PATH] [CHECKPOINT_PATH] [HEAD_CKPT_PATH]
- GPU: sh run_eval.sh GPU [DATASET_PATH] [CHECKPOINT_PATH] [HEAD_CKPT_PATH]
- CPU: sh run_eval.sh CPU [DATASET_PATH] [BACKBONE_CKPT_PATH] [HEAD_CKPT_PATH]
### Launch
```
# infer example
```shell
# eval example
python:
Ascend: python eval.py --dataset_path ~/imagenet/val/ --pretrain_ckpt ~/train/mobilenet-200_625.ckpt --platform Ascend --head_ckpt ./checkpoint/mobilenetv2_199.ckpt
GPU: python eval.py --dataset_path ~/imagenet/val/ --pretrain_ckpt ~/train/mobilenet-200_625.ckpt --platform GPU --head_ckpt ./checkpoint/mobilenetv2_199.ckpt
CPU: python eval.py --dataset_path ~/imagenet/val/ --pretrain_ckpt ~/train/mobilenet-200_625.ckpt --platform CPU --head_ckpt ./checkpoint/mobilenetv2_199.ckpt
Ascend: python eval.py --platform Ascend --dataset_path [VAL_DATASET_PATH] --pretrain_ckpt ./pretrain_ckpt/mobilenetv2.ckpt --head_ckpt ./checkpoint/mobilenetv2_head_15.ckpt
GPU: python eval.py --platform GPU --dataset_path [VAL_DATASET_PATH] --pretrain_ckpt ./pretrain_ckpt/mobilenetv2.ckpt --head_ckpt ./checkpoint/mobilenetv2_head_15.ckpt
CPU: python eval.py --platform CPU --dataset_path [VAL_DATASET_PATH] --pretrain_ckpt ./pretrain_ckpt/mobilenetv2.ckpt --head_ckpt ./checkpoint/mobilenetv2_head_15.ckpt
shell:
Ascend: sh run_infer.sh Ascend ~/imagenet/val/ ~/train/mobilenet-200_625.ckpt ./checkpoint/mobilenetv2_199.ckpt
GPU: sh run_infer.sh GPU ~/imagenet/val/ ~/train/mobilenet-200_625.ckpt ./checkpoint/mobilenetv2_199.ckpt
CPU: sh run_infer.sh GPU ~/imagenet/val/ ~/train/mobilenet-200_625.ckpt ./checkpoint/mobilenetv2_199.ckpt
Ascend: sh run_eval.sh Ascend [VAL_DATASET_PATH] ./pretrain_ckpt/mobilenetv2.ckpt ./checkpoint/mobilenetv2_head_15.ckpt
GPU: sh run_eval.sh GPU [VAL_DATASET_PATH] ./pretrain_ckpt/mobilenetv2.ckpt ./checkpoint/mobilenetv2_head_15.ckpt
CPU: sh run_eval.sh CPU [VAL_DATASET_PATH] ./pretrain_ckpt/mobilenetv2.ckpt ./checkpoint/mobilenetv2_head_15.ckpt
```
> checkpoint can be produced in training process.
### Result
Inference result will be stored in the example path, you can find result like the followings in `val.log`.
Inference result will be stored in the example path, you can find result like the followings in `eval.log`.
```
```shell
result: {'acc': 0.71976314102564111} ckpt=/path/to/checkpoint/mobilenet-200_625.ckpt
```
@ -181,7 +180,7 @@ result: {'acc': 0.71976314102564111} ckpt=/path/to/checkpoint/mobilenet-200_625.
| Parameters | MobilenetV2 | |
| -------------------------- | ---------------------------------------------------------- | ------------------------- |
| Model Version | | large |
| Model Version | V1 | V1 |
| Resource | Ascend 910, cpu:2.60GHz 56cores, memory:314G | NV SMX2 V100-32G |
| uploaded Date | 05/06/2020 | 05/06/2020 |
| MindSpore Version | 0.3.0 | 0.3.0 |
@ -189,33 +188,18 @@ result: {'acc': 0.71976314102564111} ckpt=/path/to/checkpoint/mobilenet-200_625.
| Training Parameters | src/config.py | src/config.py |
| Optimizer | Momentum | Momentum |
| Loss Function | SoftmaxCrossEntropy | SoftmaxCrossEntropy |
| outputs | | |
| Loss | | 1.913 |
| Accuracy | | ACC1[77.09%] ACC5[92.57%] |
| Total time | | |
| Params (M) | | |
| Checkpoint for Fine tuning | | |
| Model for inference | | |
#### Inference Performance
| Parameters | | | |
| -------------------------- | ----------------------------- | ------------------------- | -------------------- |
| Model Version | V1 | | |
| Resource | Ascend 910 | NV SMX2 V100-32G | Ascend 310 |
| uploaded Date | 05/06/2020 | 05/22/2020 | |
| MindSpore Version | 0.2.0 | 0.2.0 | 0.2.0 |
| Dataset | ImageNet, 1.2W | ImageNet, 1.2W | ImageNet, 1.2W |
| batch_size | | 130(8P) | |
| outputs | | | |
| Accuracy | | ACC1[72.07%] ACC5[90.90%] | |
| Speed | | | |
| Total time | | | |
| Model for inference | | | |
| outputs | probability | probability |
| Loss | 1.908 | 1.913 |
| Accuracy | ACC1[71.78%] | ACC1[71.08%] |
| Total time | 753 min | 845 min |
| Params (M) | 3.3 M | 3.3 M |
| Checkpoint for Fine tuning | 27.3 M | 27.3 M |
| Scripts | [Link](https://gitee.com/mindspore/mindspore/tree/master/model_zoo/official/cv/mobilenetv2)|
# [Description of Random Situation](#contents)
In dataset.py, we set the seed inside “create_dataset" function. We also use random seed in train.py.
<!-- In dataset.py, we set the seed inside “create_dataset" function. We also use random seed in train.py. -->
In train.py, we set the seed which is used by numpy.random, mindspore.common.Initializer, mindspore.ops.composite.random_ops and mindspore.nn.probability.distribution.
# [ModelZoo Homepage](#contents)

2
model_zoo/official/cv/mobilenetv2/eval.py
Unescape View File

@ -51,7 +51,7 @@ if __name__ == '__main__':
loss = nn.SoftmaxCrossEntropyWithLogits(sparse=True, reduction='mean')
model = Model(net, loss_fn=loss, metrics={'acc'})
res = model.eval(dataset, dataset_sink_mode=False)
res = model.eval(dataset)
print(f"result:{res}\npretrain_ckpt={args_opt.pretrain_ckpt}")
if args_opt.head_ckpt:
print(f"head_ckpt={args_opt.head_ckpt}")

29
model_zoo/official/cv/mobilenetv2/scripts/run_infer.sh → model_zoo/official/cv/mobilenetv2/scripts/run_eval.sh
Unescape View File

@ -18,10 +18,10 @@
run_ascend()
{
# check checkpoint file
# check pretrain_ckpt file
if [ ! -f $3 ]
then
echo "error: CHECKPOINT_PATH=$3 is not a file"
echo "error: PRETRAIN_CKPT=$3 is not a file"
exit 1
fi
@ -44,15 +44,15 @@ run_ascend()
--dataset_path=$2 \
--pretrain_ckpt=$3 \
--head_ckpt=$4 \
&> ../infer.log & # dataset val folder path
&> ../eval.log & # dataset val folder path
}
run_gpu()
{
# check checkpoint file
# check pretrain_ckpt file
if [ ! -f $3 ]
then
echo "error: CHECKPOINT_PATH=$3 is not a file"
echo "error: PRETRAIN_CKPT=$3 is not a file"
exit 1
fi
@ -70,26 +70,18 @@ run_gpu()
--dataset_path=$2 \
--pretrain_ckpt=$3 \
--head_ckpt=$4 \
&> ../infer.log & # dataset train folder
&> ../eval.log & # dataset train folder
}
run_cpu()
{
# check checkpoint file
# check pretrain_ckpt file
if [ ! -f $3 ]
then
echo "error: BACKBONE_CKPT=$3 is not a file"
echo "error: PRETRAIN_CKPT=$3 is not a file"
exit 1
fi
# check checkpoint file
if [ ! -f $4 ]
then
echo "error: HEAD_CKPT=$4 is not a file"
exit 1
fi
BASEPATH=$(cd "`dirname $0`" || exit; pwd)
export PYTHONPATH=${BASEPATH}:$PYTHONPATH
if [ -d "../eval" ];
@ -104,13 +96,14 @@ run_cpu()
--dataset_path=$2 \
--pretrain_ckpt=$3 \
--head_ckpt=$4 \
&> ../infer.log & # dataset train folder
&> ../eval.log & # dataset train folder
}
if [ $# -gt 4 ] || [ $# -lt 3 ]
then
echo "Ascend: sh run_infer.sh [PLATFORM] [DATASET_PATH] [PRETRAIN_CKPT] \
echo "Usage:
Ascend: sh run_infer.sh [PLATFORM] [DATASET_PATH] [PRETRAIN_CKPT]
GPU: sh run_infer.sh [PLATFORM] [DATASET_PATH] [PRETRAIN_CKPT]
CPU: sh run_infer.sh [PLATFORM] [DATASET_PATH] [BACKBONE_CKPT] [HEAD_CKPT]"
exit 1

9
model_zoo/official/cv/mobilenetv2/scripts/run_train.sh
Unescape View File

@ -109,11 +109,10 @@ run_cpu()
if [ $# -gt 7 ] || [ $# -lt 4 ]
then
echo "Usage:\n \
Ascend: sh run_train.sh Ascend [DEVICE_NUM] [VISIABLE_DEVICES(0,1,2,3,4,5,6,7)] [RANK_TABLE_FILE] [DATASET_PATH] [TRAIN_METHOD] [CKPT_PATH] \n \
GPU: sh run_train.sh GPU [DEVICE_NUM] [VISIABLE_DEVICES(0,1,2,3,4,5,6,7)] [DATASET_PATH] [TRAIN_METHOD] [CKPT_PATH]\n \
CPU: sh run_train.sh CPU [DATASET_PATH] [TRAIN_METHOD] [CKPT_PATH]\n \
"
echo "Usage:
Ascend: sh run_train.sh Ascend [DEVICE_NUM] [VISIABLE_DEVICES(0,1,2,3,4,5,6,7)] [RANK_TABLE_FILE] [DATASET_PATH] [TRAIN_METHOD] [CKPT_PATH]
GPU: sh run_train.sh GPU [DEVICE_NUM] [VISIABLE_DEVICES(0,1,2,3,4,5,6,7)] [DATASET_PATH] [TRAIN_METHOD] [CKPT_PATH]
CPU: sh run_train.sh CPU [DATASET_PATH] [TRAIN_METHOD] [CKPT_PATH]"
exit 1
fi

19
model_zoo/official/cv/mobilenetv2/src/args.py
Unescape View File

@ -38,31 +38,28 @@ def launch_parse_args():
def train_parse_args():
train_parser = argparse.ArgumentParser(description='Image classification trian')
train_parser.add_argument('--dataset_path', type=str, required=True, help='Dataset path')
train_parser.add_argument('--platform', type=str, default="Ascend", choices=("CPU", "GPU", "Ascend"), \
help='run platform, only support CPU, GPU and Ascend')
train_parser.add_argument('--pretrain_ckpt', type=str, default=None, help='Pretrained checkpoint path \
for fine tune or incremental learning')
train_parser.add_argument('--run_distribute', type=ast.literal_eval, default=True, help='Run distribute')
train_parser.add_argument('--dataset_path', type=str, required=True, help='Dataset path')
train_parser.add_argument('--train_method', type=str, choices=("train", "fine_tune", "incremental_learn"), \
help="\"fine_tune\"or \"incremental_learn\" if to fine tune the net after loading the ckpt, \"train\" to \
train from initialization model")
train_parser.add_argument('--pretrain_ckpt', type=str, default=None, help='Pretrained checkpoint path \
for fine tune or incremental learning')
train_parser.add_argument('--run_distribute', type=ast.literal_eval, default=True, help='Run distribute')
train_args = train_parser.parse_args()
return train_args
def eval_parse_args():
eval_parser = argparse.ArgumentParser(description='Image classification eval')
eval_parser.add_argument('--dataset_path', type=str, required=True, help='Dataset path')
eval_parser.add_argument('--platform', type=str, default="Ascend", choices=("Ascend", "GPU", "CPU"), \
help='run platform, only support GPU, CPU and Ascend')
eval_parser.add_argument('--pretrain_ckpt', type=str, default=None, help='Pretrained checkpoint path \
eval_parser.add_argument('--dataset_path', type=str, required=True, help='Dataset path')
eval_parser.add_argument('--pretrain_ckpt', type=str, required=True, help='Pretrained checkpoint path \
for fine tune or incremental learning')
eval_parser.add_argument('--head_ckpt', type=str, default=None, help='Pretrained checkpoint path \
for fine tune or incremental learning')
eval_parser.add_argument('--run_distribute', type=ast.literal_eval, default=False, help='Run distribute')
for incremental learning')
eval_parser.add_argument('--run_distribute', type=ast.literal_eval, default=False, help='If run distribute in GPU.')
eval_args = eval_parser.parse_args()
return eval_args

18
model_zoo/official/cv/mobilenetv2/src/dataset.py
Unescape View File

@ -16,7 +16,6 @@
create train or eval dataset.
"""
import os
from tqdm import tqdm
import numpy as np
from mindspore import Tensor
@ -97,7 +96,11 @@ def create_dataset(dataset_path, do_train, config, repeat_num=1):
# apply dataset repeat operation
ds = ds.repeat(repeat_num)
return ds
step_size = ds.get_dataset_size()
if step_size == 0:
raise ValueError("The step_size of dataset is zero. Check if the images of train dataset is more than batch_\
size in config.py")
return ds, step_size
def extract_features(net, dataset_path, config):
@ -109,19 +112,16 @@ def extract_features(net, dataset_path, config):
config=config,
repeat_num=1)
step_size = dataset.get_dataset_size()
pbar = tqdm(list(dataset.create_dict_iterator(output_numpy=True)))
model = Model(net)
i = 0
for data in pbar:
for i, data in enumerate(dataset.create_dict_iterator(output_numpy=True)):
features_path = os.path.join(features_folder, f"feature_{i}.npy")
label_path = os.path.join(features_folder, f"label_{i}.npy")
if not (os.path.exists(features_path) and os.path.exists(label_path)):
if not os.path.exists(features_path or not os.path.exists(label_path)):
image = data["image"]
label = data["label"]
features = model.predict(Tensor(image))
np.save(features_path, features.asnumpy())
np.save(label_path, label)
pbar.set_description("Process dataset batch: %d" % (i + 1))
i += 1
print(f"Complete the batch {i}/{step_size}")
return step_size

12
model_zoo/official/cv/mobilenetv2/src/launch.py
Unescape View File

@ -38,17 +38,17 @@ def main():
for rank_id in range(0, args.nproc_per_node):
os.chdir(cur_path)
device_id = visible_devices[rank_id]
device_dir = os.path.join(cur_path, 'device{}'.format(rank_id))
rank_dir = os.path.join(cur_path, 'rank{}'.format(rank_id))
env['RANK_ID'] = str(rank_id)
env['DEVICE_ID'] = str(device_id)
if os.path.exists(device_dir):
shutil.rmtree(device_dir)
os.mkdir(device_dir)
os.chdir(device_dir)
if os.path.exists(rank_dir):
shutil.rmtree(rank_dir)
os.mkdir(rank_dir)
os.chdir(rank_dir)
cmd = [sys.executable, '-u']
cmd.append(args.training_script)
cmd.extend(args.training_script_args)
log_file = open('{dir}/log{id}.log'.format(dir=device_dir, id=rank_id), 'w')
log_file = open(f'{rank_dir}/log{rank_id}.log', 'w')
process = subprocess.Popen(cmd, stdout=log_file, stderr=log_file, env=env)
processes.append(process)
cmds.append(cmd)

10
model_zoo/official/cv/mobilenetv2/src/utils.py
Unescape View File

@ -76,14 +76,12 @@ def config_ckpoint(config, lr, step_size):
if config.save_checkpoint:
config_ck = CheckpointConfig(save_checkpoint_steps=config.save_checkpoint_epochs * step_size,
keep_checkpoint_max=config.keep_checkpoint_max)
ckpt_save_dir = config.save_checkpoint_path
if config.platform == "GPU":
if config.run_distribute:
ckpt_save_dir += "ckpt_" + str(get_rank()) + "/"
else:
ckpt_save_dir += "ckpt_" + "/"
rank = 0
if config.run_distribute:
rank = get_rank()
ckpt_save_dir = config.save_checkpoint_path + "ckpt_" + str(rank) + "/"
ckpt_cb = ModelCheckpoint(prefix="mobilenetV2", directory=ckpt_save_dir, config=config_ck)
cb += [ckpt_cb]
return cb

8
model_zoo/official/cv/mobilenetv2/train.py
Unescape View File

@ -60,8 +60,7 @@ if __name__ == '__main__':
elif args_opt.train_method in ("train", "fine_tune"):
if args_opt.platform == "CPU":
raise ValueError("Currently, CPU only support \"incremental_learn\", not \"fine_tune\" or \"train\".")
dataset = create_dataset(dataset_path=args_opt.dataset_path, do_train=True, config=config)
step_size = dataset.get_dataset_size()
dataset, step_size = create_dataset(dataset_path=args_opt.dataset_path, do_train=True, config=config)
# Currently, only Ascend support switch precision.
switch_precision(net, mstype.float16, config)
@ -108,9 +107,8 @@ if __name__ == '__main__':
losses.append(network(feature, label).asnumpy())
epoch_mseconds = (time.time()-epoch_start) * 1000
per_step_mseconds = epoch_mseconds / step_size
print("\r epoch[{}], iter[{}] cost: {:5.3f}, per step time: {:5.3f}, avg loss: {:5.3f}"\
.format(epoch + 1, step_size, epoch_mseconds, per_step_mseconds, np.mean(np.array(losses))), \
end="")
print("epoch[{}], iter[{}] cost: {:5.3f}, per step time: {:5.3f}, avg loss: {:5.3f}"\
.format(epoch + 1, step_size, epoch_mseconds, per_step_mseconds, np.mean(np.array(losses))))
if (epoch + 1) % config.save_checkpoint_epochs == 0:
save_checkpoint(network, os.path.join(config.save_checkpoint_path, \
f"mobilenetv2_head_{epoch+1}.ckpt"))

Write Preview
Loading…
Cancel
Save