From f006d508143fc971958c2aa5b0ee36f798d76d09 Mon Sep 17 00:00:00 2001 From: gouhanke <12219217+gouhanke@user.noreply.gitee.com> Date: Fri, 6 Feb 2026 15:33:07 +0800 Subject: [PATCH] =?UTF-8?q?chore:=20=E8=87=AA=E5=8A=A8=E8=8E=B7=E5=8F=96ca?= =?UTF-8?q?meras=E7=9A=84=E9=95=BF=E5=BA=A6?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- roboimi/vla/agent.py | 2 +- roboimi/vla/conf/agent/resnet_diffusion.yaml | 5 +---- roboimi/vla/conf/data/resnet_dataset.yaml | 2 +- 3 files changed, 3 insertions(+), 6 deletions(-) diff --git a/roboimi/vla/agent.py b/roboimi/vla/agent.py index f29901c..ac1371e 100644 --- a/roboimi/vla/agent.py +++ b/roboimi/vla/agent.py @@ -17,7 +17,7 @@ class VLAAgent(nn.Module): pred_horizon=16, # 预测未来多少步动作 obs_horizon=4, # 使用多少步历史观测 diffusion_steps=100, - num_cams=2, # 视觉输入的摄像头数量 + num_cams=3, # 视觉输入的摄像头数量 ): super().__init__() # Store parameters diff --git a/roboimi/vla/conf/agent/resnet_diffusion.yaml b/roboimi/vla/conf/agent/resnet_diffusion.yaml index 4851b5f..b1b3d8f 100644 --- a/roboimi/vla/conf/agent/resnet_diffusion.yaml +++ b/roboimi/vla/conf/agent/resnet_diffusion.yaml @@ -19,7 +19,4 @@ obs_horizon: 2 diffusion_steps: 100 # Number of diffusion timesteps for training # Camera Configuration -# num_cams 应与 data.camera_names 列表长度一致 -# 可使用 Hydra OmegaConf resolver: ${oc.len:data.camera_names} -# 但部分版本不支持，这里手动保持同步 -num_cams: 3 # len(data.camera_names) = 3 +num_cams: ${oc.len:data.camera_names} # 自动从 data.camera_names 列表长度获取 \ No newline at end of file diff --git a/roboimi/vla/conf/data/resnet_dataset.yaml b/roboimi/vla/conf/data/resnet_dataset.yaml index 73b7435..b2822da 100644 --- a/roboimi/vla/conf/data/resnet_dataset.yaml +++ b/roboimi/vla/conf/data/resnet_dataset.yaml @@ -16,4 +16,4 @@ camera_names: - front # Normalization Type: 'gaussian' (mean/std) or 'min_max' ([-1, 1]) -normalization_type: gaussian +normalization_type: min_max