Skip to content

GitLab

Commit dee0625f authored by sunxx1's avatar sunxx1
Browse files

Merge branch 'sun_22.10' into 'main' 

Update openmmlab_test/mmclassification-speed-benchmark/train.md,...

See merge request dcutoolkit/deeplearing/dlexamples_new!44
parents 6d8087d4 0d449bca
Show whitespace changes
Inline Side-by-side
Showing with 2 additions and 36 deletions
openmmlab_test/mmclassification-speed-benchmark/train.md
View file @ dee0625f
...@@ -38,7 +38,6 @@ pip3 install -e . ...@@ -38,7 +38,6 @@ pip3 install -e .
configs/speed_test/datasets/imagenet_bs32.py 中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time configs/speed_test/datasets/imagenet_bs32.py 中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time
![1](image/train/1659061854685.png) ![1](image/train/1659061854685.png)
#### 性能关注：time #### 性能关注：time
![1659062180839](image/train/1659062180839.png)
### 多卡测试（单精度） ### 多卡测试（单精度）
```python ```python
./multi_test.sh configs/speed_test/resnet18_b32x8_imagenet.py ./multi_test.sh configs/speed_test/resnet18_b32x8_imagenet.py
...@@ -60,7 +59,6 @@ configs/speed_test/datasets/imagenet_bs32.py 中batch_size=samples_per_gpu*卡 ...@@ -60,7 +59,6 @@ configs/speed_test/datasets/imagenet_bs32.py 中batch_size=samples_per_gpu*卡
configs/speed_test/datasets/imagenet_bs32.py 中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time configs/speed_test/datasets/imagenet_bs32.py 中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time
![1659064427635](image/train/1659064427635.png) ![1659064427635](image/train/1659064427635.png)
#### 性能关注：time #### 性能关注：time
![1659064222206](image/train/1659064222206.png)
### 多卡测试（单精度） ### 多卡测试（单精度）
```python ```python
./multi_test.sh configs/speed_test/resnet34_b32x8_imagenet.py ./multi_test.sh configs/speed_test/resnet34_b32x8_imagenet.py
...@@ -82,7 +80,6 @@ configs/speed_test/datasets/imagenet_bs32.py 中batch_size=samples_per_gpu*卡 ...@@ -82,7 +80,6 @@ configs/speed_test/datasets/imagenet_bs32.py 中batch_size=samples_per_gpu*卡
configs/speed_test/datasets/imagenet_bs32.py 中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time configs/speed_test/datasets/imagenet_bs32.py 中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time
![1659064905610](image/train/1659064905610.png) ![1659064905610](image/train/1659064905610.png)
#### 性能关注：time #### 性能关注：time
![1659064925468](image/train/1659064925468.png)
### 多卡测试（单精度） ### 多卡测试（单精度）
```python ```python
./multi_test.sh configs/speed_test/resnet50_b32x8_imagenet.py ./multi_test.sh configs/speed_test/resnet50_b32x8_imagenet.py
...@@ -105,7 +102,6 @@ configs/speed_test/datasets/imagenet_bs32.py 中batch_size=samples_per_gpu*卡 ...@@ -105,7 +102,6 @@ configs/speed_test/datasets/imagenet_bs32.py 中batch_size=samples_per_gpu*卡
configs/speed_test/datasets/imagenet_bs32.py 中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time configs/speed_test/datasets/imagenet_bs32.py 中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time
![1659064905610](image/train/1659064905610.png) ![1659064905610](image/train/1659064905610.png)
#### 性能关注：time #### 性能关注：time
![1659065333529](image/train/1659065333529.png)
### 多卡测试（单精度） ### 多卡测试（单精度）
```python ```python
./multi_test.sh configs/speed_test/resnet152_b32x8_imagenet.py ./multi_test.sh configs/speed_test/resnet152_b32x8_imagenet.py
...@@ -128,7 +124,6 @@ configs/speed_test/datasets/imagenet_bs32.py 中batch_size=samples_per_gpu*卡 ...@@ -128,7 +124,6 @@ configs/speed_test/datasets/imagenet_bs32.py 中batch_size=samples_per_gpu*卡
configs/speed_test/datasets/imagenet_bs32.py 中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time configs/speed_test/datasets/imagenet_bs32.py 中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time
![1659064905610](image/train/1659064905610.png) ![1659064905610](image/train/1659064905610.png)
#### 性能关注：time #### 性能关注：time
![1659065333529](image/train/1659065333529.png)
### 多卡测试（单精度） ### 多卡测试（单精度）
```python ```python
./multi_test.sh configs/speed_test/vgg11_b32x8_imagenet.py ./multi_test.sh configs/speed_test/vgg11_b32x8_imagenet.py
...@@ -150,7 +145,6 @@ configs/speed_test/datasets/imagenet_bs32.py 中batch_size=samples_per_gpu*卡 ...@@ -150,7 +145,6 @@ configs/speed_test/datasets/imagenet_bs32.py 中batch_size=samples_per_gpu*卡
configs/speed_test/datasets/imagenet_bs32.py 中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time configs/speed_test/datasets/imagenet_bs32.py 中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time
![1659064905610](image/train/1659064905610.png) ![1659064905610](image/train/1659064905610.png)
#### 性能关注：time #### 性能关注：time
![1659065659769](image/train/1659065659769.png)
### 多卡测试（单精度） ### 多卡测试（单精度）
```python ```python
./multi_test.sh configs/speed_test/seresnet50_b32x8_imagenet.py ./multi_test.sh configs/speed_test/seresnet50_b32x8_imagenet.py
...@@ -173,7 +167,6 @@ configs/speed_test/datasets/imagenet_bs32.py 中batch_size=samples_per_gpu*卡 ...@@ -173,7 +167,6 @@ configs/speed_test/datasets/imagenet_bs32.py 中batch_size=samples_per_gpu*卡
configs/speed_test/datasets/imagenet_bs32.py 中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time configs/speed_test/datasets/imagenet_bs32.py 中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time
![1659064905610](image/train/1659064905610.png) ![1659064905610](image/train/1659064905610.png)
#### 性能关注：time #### 性能关注：time
![1659065746317](image/train/1659065746317.png)
### 多卡测试（单精度） ### 多卡测试（单精度）
```python ```python
./multi_test.sh configs/speed_test/resnext50_32x4d_b32x8_imagenet.py ./multi_test.sh configs/speed_test/resnext50_32x4d_b32x8_imagenet.py
...@@ -195,7 +188,6 @@ configs/speed_test/datasets/imagenet_bs32.py 中batch_size=samples_per_gpu*卡 ...@@ -195,7 +188,6 @@ configs/speed_test/datasets/imagenet_bs32.py 中batch_size=samples_per_gpu*卡
configs/speed_test/datasets/imagenet_bs32.py 中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time configs/speed_test/datasets/imagenet_bs32.py 中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time
![1659064905610](image/train/1659064905610.png) ![1659064905610](image/train/1659064905610.png)
#### 性能关注：time #### 性能关注：time
![1659065746317](image/train/1659065746317.png)
### 多卡测试（单精度） ### 多卡测试（单精度）
```python ```python
./multi_test.sh configs/speed_test/mobilenet_v2_b32x8_imagenet.py ./multi_test.sh configs/speed_test/mobilenet_v2_b32x8_imagenet.py
...@@ -217,7 +209,6 @@ configs/speed_test/datasets/imagenet_bs32.py 中batch_size=samples_per_gpu*卡 ...@@ -217,7 +209,6 @@ configs/speed_test/datasets/imagenet_bs32.py 中batch_size=samples_per_gpu*卡
configs/speed_test/datasets/imagenet_bs64.py 中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time configs/speed_test/datasets/imagenet_bs64.py 中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time
![1659064905610](image/train/1659064905610.png) ![1659064905610](image/train/1659064905610.png)
#### 性能关注：time #### 性能关注：time
![1659066120939](image/train/1659066120939.png)
### 多卡测试（单精度） ### 多卡测试（单精度）
```python ```python
./multi_test.sh configs/speed_test/shufflenet_v1_1x_b64x16_linearlr_bn_nowd_imagenet.py ./multi_test.sh configs/speed_test/shufflenet_v1_1x_b64x16_linearlr_bn_nowd_imagenet.py
...@@ -240,7 +231,6 @@ configs/speed_test/datasets/imagenet_bs64.py 中batch_size=samples_per_gpu*卡 ...@@ -240,7 +231,6 @@ configs/speed_test/datasets/imagenet_bs64.py 中batch_size=samples_per_gpu*卡
configs/speed_test/datasets/imagenet_bs64.py 中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time configs/speed_test/datasets/imagenet_bs64.py 中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time
![1659064905610](image/train/1659064905610.png) ![1659064905610](image/train/1659064905610.png)
#### 性能关注：time #### 性能关注：time
![1659066120939](image/train/1659066120939.png)
### 多卡测试（单精度） ### 多卡测试（单精度）
```python ```python
./multi_test.sh configs/speed_test/shufflenet_v2_1x_b64x16_linearlr_bn_nowd_imagenet.py ./multi_test.sh configs/speed_test/shufflenet_v2_1x_b64x16_linearlr_bn_nowd_imagenet.py
...@@ -262,7 +252,6 @@ configs/speed_test/datasets/imagenet_bs64.py 中batch_size=samples_per_gpu*卡 ...@@ -262,7 +252,6 @@ configs/speed_test/datasets/imagenet_bs64.py 中batch_size=samples_per_gpu*卡
configs/speed_test/datasets/imagenet_bs32.py 中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time configs/speed_test/datasets/imagenet_bs32.py 中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time
![1659064905610](image/train/1659064905610.png) ![1659064905610](image/train/1659064905610.png)
#### 性能关注：time #### 性能关注：time
![1659067079718](image/train/1659067079718.png)
### 多卡测试（单精度） ### 多卡测试（单精度）
```python ```python
./multi_test.sh configs/vgg/vgg16_b32x8_imagenet.py ./multi_test.sh configs/vgg/vgg16_b32x8_imagenet.py
......
openmmlab_test/mmdetection-speed_xinpian/train.md
View file @ dee0625f
...@@ -38,7 +38,6 @@ pip3 install -e ...@@ -38,7 +38,6 @@ pip3 install -e
configs/_base_/datasets/coco_detection.py 中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time configs/_base_/datasets/coco_detection.py 中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time
![1659163369453](image/train/1659163369453.png) ![1659163369453](image/train/1659163369453.png)
#### 性能关注：time #### 性能关注：time
![1659163379020](image/train/1659163379020.png)
### 多卡测试（单精度） ### 多卡测试（单精度）
```python ```python
./multi_test.sh configs/faster_rcnn/faster_rcnn_r50_fpn_1x_coco.py ./multi_test.sh configs/faster_rcnn/faster_rcnn_r50_fpn_1x_coco.py
...@@ -60,7 +59,6 @@ configs/_base_/datasets/coco_detection.py 中batch_size=samples_per_gpu*卡数, ...@@ -60,7 +59,6 @@ configs/_base_/datasets/coco_detection.py 中batch_size=samples_per_gpu*卡数,
configs/_base_/datasets/coco_instance.py 中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time configs/_base_/datasets/coco_instance.py 中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time
![1659164315340](image/train/1659164315340.png) ![1659164315340](image/train/1659164315340.png)
#### 性能关注：time #### 性能关注：time
![1659164326948](image/train/1659164326948.png)
### 多卡测试（单精度） ### 多卡测试（单精度）
```python ```python
./multi_test.sh configs/mask_rcnn/mask_rcnn_r50_caffe_fpn_mstrain-poly_1x_coco.py ./multi_test.sh configs/mask_rcnn/mask_rcnn_r50_caffe_fpn_mstrain-poly_1x_coco.py
...@@ -82,7 +80,6 @@ configs/_base_/datasets/coco_instance.py 中batch_size=samples_per_gpu*卡数, ...@@ -82,7 +80,6 @@ configs/_base_/datasets/coco_instance.py 中batch_size=samples_per_gpu*卡数,
configs/_base_/datasets/coco_detection.py 中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time configs/_base_/datasets/coco_detection.py 中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time
![1659164497462](image/train/1659164497462.png) ![1659164497462](image/train/1659164497462.png)
#### 性能关注：time #### 性能关注：time
![1659164507155](image/train/1659164507155.png)
### 多卡测试（单精度） ### 多卡测试（单精度）
```python ```python
./multi_test.sh configs/double_heads/dh_faster_rcnn_r50_fpn_1x_coco.py ./multi_test.sh configs/double_heads/dh_faster_rcnn_r50_fpn_1x_coco.py
...@@ -105,7 +102,6 @@ configs/_base_/datasets/coco_detection.py 中batch_size=samples_per_gpu*卡数, ...@@ -105,7 +102,6 @@ configs/_base_/datasets/coco_detection.py 中batch_size=samples_per_gpu*卡数,
configs/_base_/datasets/coco_instance.py 中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time configs/_base_/datasets/coco_instance.py 中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time
![1659165101880](image/train/1659165101880.png) ![1659165101880](image/train/1659165101880.png)
#### 性能关注：time #### 性能关注：time
![1659165112953](image/train/1659165112953.png)
### 多卡测试（单精度） ### 多卡测试（单精度）
```python ```python
./multi_test.sh configs/cascade_rcnn/cascade_mask_rcnn_r50_fpn_1x_coco.py ./multi_test.sh configs/cascade_rcnn/cascade_mask_rcnn_r50_fpn_1x_coco.py
...@@ -130,7 +126,6 @@ configs/_base_/datasets/coco_instance.py 中batch_size=samples_per_gpu*卡数, ...@@ -130,7 +126,6 @@ configs/_base_/datasets/coco_instance.py 中batch_size=samples_per_gpu*卡数,
单卡测试需将type='SyncBN'改为BN： 单卡测试需将type='SyncBN'改为BN：
![1659165231678](image/train/1659165231678.png) ![1659165231678](image/train/1659165231678.png)
#### 性能关注：time #### 性能关注：time
![1659165240225](image/train/1659165240225.png)
### 多卡测试（单精度） ### 多卡测试（单精度）
```python ```python
./multi_test.sh configs/resnest/mask_rcnn_s50_fpn_syncbn-backbone+head_mstrain_1x_coco.py ./multi_test.sh configs/resnest/mask_rcnn_s50_fpn_syncbn-backbone+head_mstrain_1x_coco.py
...@@ -152,7 +147,6 @@ configs/_base_/datasets/coco_instance.py 中batch_size=samples_per_gpu*卡数, ...@@ -152,7 +147,6 @@ configs/_base_/datasets/coco_instance.py 中batch_size=samples_per_gpu*卡数,
configs/_base_/datasets/coco_detection.py 中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time configs/_base_/datasets/coco_detection.py 中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time
![1659165385359](image/train/1659165385359.png) ![1659165385359](image/train/1659165385359.png)
#### 性能关注：time #### 性能关注：time
![1659165394619](image/train/1659165394619.png)
### 多卡测试（单精度） ### 多卡测试（单精度）
```python ```python
./multi_test.sh configs/dcn/faster_rcnn_r50_fpn_mdpool_1x_coco.py ./multi_test.sh configs/dcn/faster_rcnn_r50_fpn_mdpool_1x_coco.py
...@@ -175,7 +169,6 @@ configs/_base_/datasets/coco_detection.py 中batch_size=samples_per_gpu*卡数, ...@@ -175,7 +169,6 @@ configs/_base_/datasets/coco_detection.py 中batch_size=samples_per_gpu*卡数,
configs/_base_/datasets/coco_detection.py 中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time configs/_base_/datasets/coco_detection.py 中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time
![1659165489693](image/train/1659165489693.png) ![1659165489693](image/train/1659165489693.png)
#### 性能关注：time #### 性能关注：time
![1659165500683](image/train/1659165500683.png)
### 多卡测试（单精度） ### 多卡测试（单精度）
```python ```python
./multi_test.sh configs/retinanet/retinanet_r50_caffe_fpn_1x_coco.py ./multi_test.sh configs/retinanet/retinanet_r50_caffe_fpn_1x_coco.py
...@@ -197,7 +190,6 @@ configs/_base_/datasets/coco_detection.py 中batch_size=samples_per_gpu*卡数, ...@@ -197,7 +190,6 @@ configs/_base_/datasets/coco_detection.py 中batch_size=samples_per_gpu*卡数,
configs/vfnet/vfnet_r50_fpn_1x_coco.py中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time configs/vfnet/vfnet_r50_fpn_1x_coco.py中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time
![1659165606806](image/train/1659165606806.png) ![1659165606806](image/train/1659165606806.png)
#### 性能关注：time #### 性能关注：time
![1659165616259](image/train/1659165616259.png)
### 多卡测试（单精度） ### 多卡测试（单精度）
```python ```python
./multi_test.sh configs/vfnet/vfnet_r50_fpn_1x_coco.p ./multi_test.sh configs/vfnet/vfnet_r50_fpn_1x_coco.p
...@@ -219,7 +211,6 @@ configs/vfnet/vfnet_r50_fpn_1x_coco.py中batch_size=samples_per_gpu*卡数,性 ...@@ -219,7 +211,6 @@ configs/vfnet/vfnet_r50_fpn_1x_coco.py中batch_size=samples_per_gpu*卡数,性
configs/ssd/ssd300_coco.py中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time configs/ssd/ssd300_coco.py中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time
![1659165721407](image/train/1659165721407.png) ![1659165721407](image/train/1659165721407.png)
#### 性能关注：time #### 性能关注：time
### 多卡测试（单精度） ### 多卡测试（单精度）
```python ```python
./multi_test.sh configs/ssd/ssd300_coco.py ./multi_test.sh configs/ssd/ssd300_coco.py
...@@ -241,7 +232,6 @@ configs/ssd/ssd300_coco.py中batch_size=samples_per_gpu*卡数,性能计算方 ...@@ -241,7 +232,6 @@ configs/ssd/ssd300_coco.py中batch_size=samples_per_gpu*卡数,性能计算方
configs/yolo/yolov3_d53_mstrain-608_273e_coco.py中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time configs/yolo/yolov3_d53_mstrain-608_273e_coco.py中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time
![1659166259835](image/train/1659166259835.png) ![1659166259835](image/train/1659166259835.png)
#### 性能关注：time #### 性能关注：time
![1659165730006](image/train/1659165730006.png)
### 多卡测试（单精度） ### 多卡测试（单精度）
```python ```python
./multi_test.sh configs/yolo/yolov3_d53_mstrain-416_273e_coco.py ./multi_test.sh configs/yolo/yolov3_d53_mstrain-416_273e_coco.py
......
openmmlab_test/mmpose-speed_test/train.md
View file @ dee0625f
...@@ -41,7 +41,6 @@ pip3 install -e . ...@@ -41,7 +41,6 @@ pip3 install -e .
configs/speed_test/bottomup_hrnet_w32_coco_512x512_dummy.py 中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time configs/speed_test/bottomup_hrnet_w32_coco_512x512_dummy.py 中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time
![1659168500714](image/train/1659168500714.png) ![1659168500714](image/train/1659168500714.png)
#### 性能关注：time #### 性能关注：time
![1659168511464](image/train/1659168511464.png)
### 多卡测试（单精度） ### 多卡测试（单精度）
```python ```python
./multi_test.sh configs/speed_test/bottomup_hrnet_w32_coco_512x512_dummy.py ./multi_test.sh configs/speed_test/bottomup_hrnet_w32_coco_512x512_dummy.py
...@@ -63,7 +62,6 @@ configs/speed_test/bottomup_hrnet_w32_coco_512x512_dummy.py 中batch_size=sample ...@@ -63,7 +62,6 @@ configs/speed_test/bottomup_hrnet_w32_coco_512x512_dummy.py 中batch_size=sample
configs/speed_test/res50_coco_256x192_dummy.py 中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time configs/speed_test/res50_coco_256x192_dummy.py 中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time
![1659169127556](image/train/1659169127556.png) ![1659169127556](image/train/1659169127556.png)
#### 性能关注：time #### 性能关注：time
![1659169137095](image/train/1659169137095.png)
### 多卡测试（单精度） ### 多卡测试（单精度）
```python ```python
./multi_test.sh configs/speed_test/res50_coco_256x192_dummy.py ./multi_test.sh configs/speed_test/res50_coco_256x192_dummy.py
...@@ -85,7 +83,6 @@ configs/speed_test/res50_coco_256x192_dummy.py 中batch_size=samples_per_gpu*卡 ...@@ -85,7 +83,6 @@ configs/speed_test/res50_coco_256x192_dummy.py 中batch_size=samples_per_gpu*卡
configs/speed_test/hrnet_w32_coco_256x192_dummy.py 中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time configs/speed_test/hrnet_w32_coco_256x192_dummy.py 中batch_size=samples_per_gpu*卡数,性能计算方法：batch_size/time
![1659169253165](image/train/1659169253165.png) ![1659169253165](image/train/1659169253165.png)
#### 性能关注：time #### 性能关注：time
![1659169263508](image/train/1659169263508.png)
### 多卡测试（单精度） ### 多卡测试（单精度）
```python ```python
./multi_test.sh configs/speed_test/hrnet_w32_coco_256x192_dummy.py ./multi_test.sh configs/speed_test/hrnet_w32_coco_256x192_dummy.py
......
openmmlab_test/mmsegmentation/train.md
View file @ dee0625f
...@@ -50,8 +50,6 @@ configs/_base_/datasets/cityscapes.py中batch_size=samples_per_gpu*卡数,性能 ...@@ -50,8 +50,6 @@ configs/_base_/datasets/cityscapes.py中batch_size=samples_per_gpu*卡数,性能
#### 性能关注：time #### 性能关注：time
![1659172396166](image/train/1659172396166.png)
### 多卡测试（单精度） ### 多卡测试（单精度）
```python ```python
...@@ -73,8 +71,6 @@ configs/_base_/datasets/cityscapes.py中batch_size=samples_per_gpu*卡数,性能 ...@@ -73,8 +71,6 @@ configs/_base_/datasets/cityscapes.py中batch_size=samples_per_gpu*卡数,性能
#### 性能关注：time #### 性能关注：time
![1659172478200](image/train/1659172478200.png)
### 多卡测试（单精度） ### 多卡测试（单精度）
```python ```python
...@@ -96,8 +92,6 @@ configs/_base_/datasets/cityscapes.py中batch_size=samples_per_gpu*卡数,性能 ...@@ -96,8 +92,6 @@ configs/_base_/datasets/cityscapes.py中batch_size=samples_per_gpu*卡数,性能
#### 性能关注：time #### 性能关注：time
![1659172540450](image/train/1659172540450.png)
### 多卡测试（单精度） ### 多卡测试（单精度）
```python ```python
...@@ -119,8 +113,6 @@ configs/_base_/datasets/cityscapes.py中batch_size=samples_per_gpu*卡数,性能 ...@@ -119,8 +113,6 @@ configs/_base_/datasets/cityscapes.py中batch_size=samples_per_gpu*卡数,性能
#### 性能关注：time #### 性能关注：time
![1659172640432](image/train/1659172640432.png)
### 多卡测试（单精度） ### 多卡测试（单精度）
```python ```python
...@@ -142,8 +134,6 @@ configs/_base_/datasets/cityscapes.py中batch_size=samples_per_gpu*卡数,性能 ...@@ -142,8 +134,6 @@ configs/_base_/datasets/cityscapes.py中batch_size=samples_per_gpu*卡数,性能
#### 性能关注：time #### 性能关注：time
![1659172709859](image/train/1659172709859.png)
### 多卡测试（单精度） ### 多卡测试（单精度）
```python ```python
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment