From f76a6720e5244f4a9675506867c29e3827e4b7fa Mon Sep 17 00:00:00 2001 From: Ryan Date: Sat, 9 Apr 2022 14:31:10 +0800 Subject: [PATCH 1/2] =?UTF-8?q?VoxelPose=E6=A8=A1=E5=9E=8B=E5=88=A0?= =?UTF-8?q?=E9=99=A4=E5=8D=95=E5=8D=A1=E8=AE=AD=E7=BB=83=E9=83=A8=E5=88=86?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- PyTorch/contrib/cv/pose_estimation/VoxelPose/README.md | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) diff --git a/PyTorch/contrib/cv/pose_estimation/VoxelPose/README.md b/PyTorch/contrib/cv/pose_estimation/VoxelPose/README.md index 4974cf07a4..d6a3cc7c23 100644 --- a/PyTorch/contrib/cv/pose_estimation/VoxelPose/README.md +++ b/PyTorch/contrib/cv/pose_estimation/VoxelPose/README.md @@ -18,14 +18,15 @@ Set AscendProject = /usr/local/Ascend #use real path ### Training To train a model, run the following scripts with the real path of dataset: +PS：There is a large jitter on single card training, do not perform single card trainning. ```bash # real_data_path = data/shelf # training 1p accuracy -bash ./test/train_full_1p.sh --data_path=real_data_path --device_id=xxx +# bash ./test/train_full_1p.sh --data_path=real_data_path --device_id=xxx # training 1p performance -bash ./test/train_performance_1p.sh --data_path=real_data_path --device_id=xxx +# bash ./test/train_performance_1p.sh --data_path=real_data_path --device_id=xxx # training 8p accuracy bash ./test/train_full_8p.sh --data_path=real_data_path @@ -44,7 +45,8 @@ Log path: ### VoxelPose training result +说明：由于模型单卡训练（bs=1）抖动太大无法训练，暂不持支单卡训练。 + | 名称 | 精度 | FPS | AMP_Type | | :------: | :------: | :------: | :------: | -| NPU-1p | 96.62 | 0.289 | O1 | | NPU-8p | 97.10 | 1.267 | O1 | -- Gitee From e4a1d2abca9ba27a4d52c4c79c51d819876f8c39 Mon Sep 17 00:00:00 2001 From: Ryan Date: Sat, 9 Apr 2022 14:32:34 +0800 Subject: [PATCH 2/2] =?UTF-8?q?VoxelPose=E6=A8=A1=E5=9E=8B=E5=88=A0?= =?UTF-8?q?=E9=99=A4=E5=8D=95=E5=8D=A1=E8=AE=AD=E7=BB=83=E9=83=A8=E5=88=86?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- PyTorch/contrib/cv/pose_estimation/VoxelPose/README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/PyTorch/contrib/cv/pose_estimation/VoxelPose/README.md b/PyTorch/contrib/cv/pose_estimation/VoxelPose/README.md index d6a3cc7c23..72e63f8533 100644 --- a/PyTorch/contrib/cv/pose_estimation/VoxelPose/README.md +++ b/PyTorch/contrib/cv/pose_estimation/VoxelPose/README.md @@ -18,7 +18,7 @@ Set AscendProject = /usr/local/Ascend #use real path ### Training To train a model, run the following scripts with the real path of dataset: -PS：There is a large jitter on single card training, do not perform single card trainning. +PS：There is a large jitter on single card training with bs=1, do not perform single card trainning. ```bash # real_data_path = data/shelf -- Gitee