未验证 提交 e9cfb187 编写于 作者: S shangliang Xu 提交者: GitHub

[DETR] fix find_unused_parameters in yaml (#5291)

上级 7dccc8f6
...@@ -20,7 +20,7 @@ Deformable DETR is an object detection model based on DETR. We reproduced the mo ...@@ -20,7 +20,7 @@ Deformable DETR is an object detection model based on DETR. We reproduced the mo
GPU multi-card training GPU multi-card training
```bash ```bash
export CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 export CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7
python -m paddle.distributed.launch --gpus 0,1,2,3,4,5,6,7 tools/train.py -c configs/deformable_detr/deformable_detr_r50_1x_coco.yml --fleet -o find_unused_parameters=True python -m paddle.distributed.launch --gpus 0,1,2,3,4,5,6,7 tools/train.py -c configs/deformable_detr/deformable_detr_r50_1x_coco.yml --fleet
``` ```
## Citations ## Citations
......
...@@ -6,3 +6,4 @@ _BASE_: [ ...@@ -6,3 +6,4 @@ _BASE_: [
'_base_/deformable_detr_reader.yml', '_base_/deformable_detr_reader.yml',
] ]
weights: output/deformable_detr_r50_1x_coco/model_final weights: output/deformable_detr_r50_1x_coco/model_final
find_unused_parameters: True
...@@ -20,7 +20,7 @@ DETR is an object detection model based on transformer. We reproduced the model ...@@ -20,7 +20,7 @@ DETR is an object detection model based on transformer. We reproduced the model
GPU multi-card training GPU multi-card training
```bash ```bash
export CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 export CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7
python -m paddle.distributed.launch --gpus 0,1,2,3,4,5,6,7 tools/train.py -c configs/detr/detr_r50_1x_coco.yml --fleet -o find_unused_parameters=True python -m paddle.distributed.launch --gpus 0,1,2,3,4,5,6,7 tools/train.py -c configs/detr/detr_r50_1x_coco.yml --fleet
``` ```
## Citations ## Citations
......
...@@ -6,3 +6,4 @@ _BASE_: [ ...@@ -6,3 +6,4 @@ _BASE_: [
'_base_/detr_reader.yml', '_base_/detr_reader.yml',
] ]
weights: output/detr_r50_1x_coco/model_final weights: output/detr_r50_1x_coco/model_final
find_unused_parameters: True
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册