forked from mindspore-Ecosystem/mindspore
optimized readme and add per_step_time
This commit is contained in:
parent
00336b7df5
commit
eb80f555b7
|
|
@ -117,13 +117,13 @@ python train.py \
|
|||
--lr_scheduler=cosine_annealing > log.txt 2>&1 &
|
||||
|
||||
# standalone training example(1p) by shell script
|
||||
sh run_standalone_train.sh dataset/coco2014 darknet53_backbone.ckpt
|
||||
bash run_standalone_train.sh dataset/coco2014 darknet53_backbone.ckpt
|
||||
|
||||
# For Ascend device, distributed training example(8p) by shell script
|
||||
sh run_distribute_train.sh dataset/coco2014 darknet53_backbone.ckpt rank_table_8p.json
|
||||
bash run_distribute_train.sh dataset/coco2014 darknet53_backbone.ckpt rank_table_8p.json
|
||||
|
||||
# For GPU device, distributed training example(8p) by shell script
|
||||
sh run_distribute_train_gpu.sh dataset/coco2014 darknet53_backbone.ckpt
|
||||
bash run_distribute_train_gpu.sh dataset/coco2014 darknet53_backbone.ckpt
|
||||
|
||||
# run evaluation by python command
|
||||
python eval.py \
|
||||
|
|
@ -132,7 +132,7 @@ python eval.py \
|
|||
--testing_shape=416 > log.txt 2>&1 &
|
||||
|
||||
# run evaluation by shell script
|
||||
sh run_eval.sh dataset/coco2014/ checkpoint/0-319_102400.ckpt
|
||||
bash run_eval.sh dataset/coco2014/ checkpoint/0-319_102400.ckpt
|
||||
```
|
||||
|
||||
## [Script Description](#contents)
|
||||
|
|
@ -269,13 +269,13 @@ The model checkpoint will be saved in outputs directory.
|
|||
For Ascend device, distributed training example(8p) by shell script
|
||||
|
||||
```command
|
||||
sh run_distribute_train.sh dataset/coco2014 darknet53_backbone.ckpt rank_table_8p.json
|
||||
bash run_distribute_train.sh dataset/coco2014 darknet53_backbone.ckpt rank_table_8p.json
|
||||
```
|
||||
|
||||
For GPU device, distributed training example(8p) by shell script
|
||||
|
||||
```command
|
||||
sh run_distribute_train_gpu.sh dataset/coco2014 darknet53_backbone.ckpt
|
||||
bash run_distribute_train_gpu.sh dataset/coco2014 darknet53_backbone.ckpt
|
||||
```
|
||||
|
||||
The above shell script will run distribute training in the background. You can view the results through the file `train_parallel[X]/log.txt`. The loss value will be achieved as follows:
|
||||
|
|
@ -307,7 +307,7 @@ python eval.py \
|
|||
--pretrained=yolov3.ckpt \
|
||||
--testing_shape=416 > log.txt 2>&1 &
|
||||
OR
|
||||
sh run_eval.sh dataset/coco2014/ checkpoint/0-319_102400.ckpt
|
||||
bash run_eval.sh dataset/coco2014/ checkpoint/0-319_102400.ckpt
|
||||
```
|
||||
|
||||
The above python command will run in the background. You can view the results through the file "log.txt". The mAP of the test dataset will be as follows:
|
||||
|
|
|
|||
|
|
@ -90,7 +90,7 @@ YOLOv3使用DarkNet53执行特征提取，这是YOLOv2中的Darknet-19和残差
|
|||
可以从网站[下载](https://pjreddie.com/media/files/darknet53.conv.74) darknet53.conv.74文件。
|
||||
也可以在linux系统中使用指令下载该文件。
|
||||
|
||||
```command
|
||||
```command
|
||||
wget https://pjreddie.com/media/files/darknet53.conv.74
|
||||
```
|
||||
|
||||
|
|
@ -118,17 +118,17 @@ python train.py \
|
|||
|
||||
```shell script
|
||||
# shell脚本单机训练示例(1卡)
|
||||
sh run_standalone_train.sh dataset/coco2014 darknet53_backbone.ckpt
|
||||
bash run_standalone_train.sh dataset/coco2014 darknet53_backbone.ckpt
|
||||
```
|
||||
|
||||
```shell script
|
||||
# 对于Ascend设备，使用shell脚本分布式训练示例(8卡)
|
||||
sh run_distribute_train.sh dataset/coco2014 darknet53_backbone.ckpt rank_table_8p.json
|
||||
bash run_distribute_train.sh dataset/coco2014 darknet53_backbone.ckpt rank_table_8p.json
|
||||
```
|
||||
|
||||
```shell script
|
||||
# 对于GPU设备，使用shell脚本分布式训练示例(8卡)
|
||||
sh run_distribute_train_gpu.sh dataset/coco2014 darknet53_backbone.ckpt
|
||||
bash run_distribute_train_gpu.sh dataset/coco2014 darknet53_backbone.ckpt
|
||||
```
|
||||
|
||||
```python
|
||||
|
|
@ -141,7 +141,7 @@ python eval.py \
|
|||
|
||||
```shell script
|
||||
# 通过shell脚本运行评估
|
||||
sh run_eval.sh dataset/coco2014/ checkpoint/0-319_102400.ckpt
|
||||
bash run_eval.sh dataset/coco2014/ checkpoint/0-319_102400.ckpt
|
||||
```
|
||||
|
||||
# 脚本说明
|
||||
|
|
@ -270,13 +270,13 @@ python train.py \
|
|||
对于Ascend设备，使用shell脚本分布式训练示例(8卡)
|
||||
|
||||
```shell script
|
||||
sh run_distribute_train.sh dataset/coco2014 darknet53_backbone.ckpt rank_table_8p.json
|
||||
bash run_distribute_train.sh dataset/coco2014 darknet53_backbone.ckpt rank_table_8p.json
|
||||
```
|
||||
|
||||
对于GPU设备，使用shell脚本分布式训练示例(8卡)
|
||||
|
||||
```shell script
|
||||
sh run_distribute_train_gpu.sh dataset/coco2014 darknet53_backbone.ckpt
|
||||
bash run_distribute_train_gpu.sh dataset/coco2014 darknet53_backbone.ckpt
|
||||
```
|
||||
|
||||
上述shell脚本将在后台运行分布训练。您可以通过`train_parallel[X]/log.txt`文件查看结果。损失值的实现如下：
|
||||
|
|
@ -312,7 +312,7 @@ python eval.py \
|
|||
或者
|
||||
|
||||
```shell script
|
||||
sh run_eval.sh dataset/coco2014/ checkpoint/0-319_102400.ckpt
|
||||
bash run_eval.sh dataset/coco2014/ checkpoint/0-319_102400.ckpt
|
||||
```
|
||||
|
||||
上述python命令将在后台运行，您可以通过log.txt文件查看结果。测试数据集的mAP如下：
|
||||
|
|
|
|||
|
|
@ -274,10 +274,12 @@ def train():
|
|||
if i % args.log_interval == 0:
|
||||
time_used = time.time() - t_end
|
||||
epoch = int(i / args.steps_per_epoch)
|
||||
per_step_time = time_used/args.log_interval
|
||||
fps = args.per_batch_size * (i - old_progress) * args.group_size / time_used
|
||||
if args.rank == 0:
|
||||
args.logger.info(
|
||||
'epoch[{}], iter[{}], {}, {:.2f} imgs/sec, lr:{}'.format(epoch, i, loss_meter, fps, lr[i]))
|
||||
'epoch[{}], iter[{}], {}, {:.2f} imgs/sec, lr:{},'
|
||||
' per_step_time:{}'.format(epoch, i, loss_meter, fps, lr[i], per_step_time))
|
||||
t_end = time.time()
|
||||
loss_meter.reset()
|
||||
old_progress = i
|
||||
|
|
|
|||
Loading…
Reference in New Issue