Are you sure you want to delete this task? Once this task is deleted, it cannot be recovered.
|
|
4 years ago | |
|---|---|---|
| .. | ||
| scripts | 4 years ago | |
| src | 4 years ago | |
| README.md | 4 years ago | |
| eval.py | 5 years ago | |
| export.py | 4 years ago | |
| train.py | 4 years ago | |
Xception by Google is extreme version of Inception. With a modified depthwise separable convolution, it is even better than Inception-v3. This paper was published in 2017.
Paper Franois Chollet. Xception: Deep Learning with Depthwise Separable Convolutions. 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) IEEE, 2017.
The overall network architecture of Xception is show below:
Dataset used can refer to paper.
The mixed precision training method accelerates the deep learning neural network training process by using both the single-precision and half-precision data formats, and maintains the network precision achieved by the single-precision training at the same time. Mixed precision training can accelerate the computation process, reduce memory usage, and enable a larger model or batch size to be trained on specific hardware.
For FP16 operators, if the input data type is FP32, the backend of MindSpore will automatically handle it with reduced precision. Users could check the reduced-precision operators by enabling INFO log and then searching ‘reduce precision’.
.
└─Xception
├─README.md
├─scripts
├─run_standalone_train.sh # launch standalone training with ascend platform(1p)
├─run_distribute_train.sh # launch distributed training with ascend platform(8p)
└─run_eval.sh # launch evaluating with ascend platform
├─src
├─config.py # parameter configuration
├─dataset.py # data preprocessing
├─Xception.py # network definition
├─loss.py # Customized CrossEntropy loss function
└─lr_generator.py # learning rate generator
├─train.py # train net
├─export.py # export net
└─eval.py # eval net
Major parameters in train.py and config.py are:
'num_classes': 1000 # dataset class numbers
'batch_size': 128 # input batchsize
'loss_scale': 1024 # loss scale
'momentum': 0.9 # momentum
'weight_decay': 1e-4 # weight decay
'epoch_size': 250 # total epoch numbers
'save_checkpoint': True # save checkpoint
'save_checkpoint_epochs': 1 # save checkpoint epochs
'keep_checkpoint_max': 5 # max numbers to keep checkpoints
'save_checkpoint_path': "./" # save checkpoint path
'warmup_epochs': 1 # warmup epoch numbers
'lr_decay_mode': "liner" # lr decay mode
'use_label_smooth': True # use label smooth
'finish_epoch': 0 # finished epochs numbers
'label_smooth_factor': 0.1 # label smoothing factor
'lr_init': 0.00004 # initiate learning rate
'lr_max': 0.4 # max bound of learning rate
'lr_end': 0.00004 # min bound of learning rate
You can start training using python or shell scripts. The usage of shell scripts as follows:
# distribute training example(8p)
sh scripts/run_distribute_train.sh RANK_TABLE_FILE DATA_PATH
# standalone training
sh scripts/run_standalone_train.sh DEVICE_ID DATA_PATH
Notes: RANK_TABLE_FILE can refer to Link, and the device_ip can be got as Link.
# training example
python:
Ascend:
python train.py --device_target Ascend --dataset_path /dataset/train
shell:
Ascend:
# distribute training example(8p)
sh scripts/run_distribute_train.sh RANK_TABLE_FILE DATA_PATH
# standalone training
sh scripts/run_standalone_train.sh DEVICE_ID DATA_PATH
Training result will be stored in the example path. Checkpoints will be stored at . /ckpt_0 by default, and training log will be redirected to log.txt like following.
epoch: 1 step: 1251, loss is 4.8427444
epoch time: 701242.350 ms, per step time: 560.545 ms
epoch: 2 step: 1251, loss is 4.0637593
epoch time: 598591.422 ms, per step time: 478.490 ms
You can start training using python or shell scripts. The usage of shell scripts as follows:
sh scripts/run_eval.sh DEVICE_ID DATA_DIR PATH_CHECKPOINT
# eval example
python:
Ascend: python eval.py --device_target Ascend --checkpoint_path PATH_CHECKPOINT --dataset_path DATA_DIR
shell:
Ascend: sh scripts/run_eval.sh DEVICE_ID DATA_DIR PATH_CHECKPOINT
checkpoint can be produced in training process.
Evaluation result will be stored in the example path, you can find result like the following in eval.log.
result: {'Loss': 1.7797744848789312, 'Top_1_Acc': 0.7985777243589743, 'Top_5_Acc': 0.9485777243589744}
| Parameters | Ascend |
|---|---|
| Model Version | Xception |
| Resource | HUAWEI CLOUD Modelarts |
| uploaded Date | 12/10/2020 |
| MindSpore Version | 1.1.0 |
| Dataset | 1200k images |
| Batch_size | 128 |
| Training Parameters | src/config.py |
| Optimizer | Momentum |
| Loss Function | CrossEntropySmooth |
| Loss | 1.78 |
| Accuracy (8p) | Top1[79.8%] Top5[94.8%] |
| Per step time (8p) | 479 ms/step |
| Total time (8p) | 42h |
| Params (M) | 180M |
| Scripts | Xception script |
| Parameters | Ascend |
|---|---|
| Model Version | Xception |
| Resource | HUAWEI CLOUD Modelarts |
| Uploaded Date | 12/10/2020 |
| MindSpore Version | 1.1.0 |
| Dataset | 50k images |
| Batch_size | 128 |
| Accuracy | Top1[79.8%] Top5[94.8%] |
| Total time | 3mins |
In dataset.py, we set the seed inside create_dataset function. We also use random seed in train.py.
Please check the official homepage.
MindSpore is a new open source deep learning training/inference framework that could be used for mobile, edge and cloud scenarios.
C++ Python Text Unity3D Asset C other