From b9e30a0afea422b604752b009e3105d3f98acaad Mon Sep 17 00:00:00 2001 From: LielinJiang Date: Thu, 13 Feb 2020 14:03:09 +0800 Subject: [PATCH] update distill readme --- slim/distillation/README.md | 16 ++++++++++++---- 1 file changed, 12 insertions(+), 4 deletions(-) diff --git a/slim/distillation/README.md b/slim/distillation/README.md index 33897707..9d045995 100644 --- a/slim/distillation/README.md +++ b/slim/distillation/README.md @@ -70,7 +70,17 @@ distill_loss = l2_loss('teacher_bilinear_interp_2.tmp_0', 'bilinear_interp_0.tmp ### 执行示例 -下载teacher的预训练模型([deeplabv3p_xception65_bn_cityscapes.tgz](https://paddleseg.bj.bcebos.com/models/xception65_bn_cityscapes.tgz))和student的预训练模型([mobilenet_cityscapes.tgz](https://paddleseg.bj.bcebos.com/models/mobilenet_cityscapes.tgz)), 替换如下命令中的```your_tearcher_pretrained_model_path```和```your_student_pretrained_model``` +下载teacher的预训练模型([deeplabv3p_xception65_bn_cityscapes.tgz](https://paddleseg.bj.bcebos.com/models/xception65_bn_cityscapes.tgz))和student的预训练模型([mobilenet_cityscapes.tgz](https://paddleseg.bj.bcebos.com/models/mobilenet_cityscapes.tgz)), +修改student config file(./slim/distillation/cityscape.yaml)中预训练模型的路径: +``` +TRAIN: + PRETRAINED_MODEL_DIR: your_student_pretrained_model_dir +``` +修改teacher config file(./slim/distillation/cityscape_teacher.yaml)中预训练模型的路径: +``` +SLIM: + KNOWLEDGE_DISTILL_TEACHER_MODEL_DIR: your_teacher_pretrained_model_dir +``` 执行如下命令启动训练，每间隔```cfg.TRAIN.SNAPSHOT_EPOCH```会进行一次评估。 ```shell @@ -80,9 +90,7 @@ python -m paddle.distributed.launch ./slim/distillation/train_distill.py \ --teacher_cfg ./slim/distillation/cityscape_teacher.yaml \ --use_gpu \ --use_mpio \ ---do_eval \ -SLIM.KNOWLEDGE_DISTILL_TEACHER_MODEL_DIR your_tearcher_pretrained_model_path \ -TRAIN.PRETRAINED_MODEL_DIR your_student_pretrained_model +--do_eval ``` ## 评估预测 -- GitLab