dodox/dppo
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

support varying img size

Browse Source
This commit is contained in:
allenzren
2024-09-16 17:55:31 -04:00
parent 64595baca9
commit 1aaa6c2302
18 changed files with 131 additions and 81 deletions
Download Patch File Download Diff File Expand all files Collapse all files
cfg/robomimic/finetune/can/ft_ppo_diffusion_mlp_img.yaml
+6 -2
View File
@@ -122,7 +122,9 @@ model:
backbone:
_target_: model.common.vit.VitEncoder
obs_shape: ${shape_meta.obs.rgb.shape}
num_channel: ${eval:'${shape_meta.obs.rgb.shape[0]} * ${img_cond_steps}'} # each image patch is history concatenated
num_channel: ${eval:'3 * ${img_cond_steps}'} # each image patch is history concatenated
img_h: ${shape_meta.obs.rgb.shape[1]}
img_w: ${shape_meta.obs.rgb.shape[2]}
cfg:
patch_size: 8
depth: 1
@@ -146,7 +148,9 @@ model:
backbone:
_target_: model.common.vit.VitEncoder
obs_shape: ${shape_meta.obs.rgb.shape}
num_channel: ${eval:'${shape_meta.obs.rgb.shape[0]} * ${img_cond_steps}'} # each image patch is history concatenated
num_channel: ${eval:'3 * ${img_cond_steps}'} # each image patch is history concatenated
img_h: ${shape_meta.obs.rgb.shape[1]}
img_w: ${shape_meta.obs.rgb.shape[2]}
cfg:
patch_size: 8
depth: 1
cfg/robomimic/finetune/can/ft_ppo_gaussian_mlp_img.yaml
+6 -2
View File
@@ -101,7 +101,9 @@ model:
backbone:
_target_: model.common.vit.VitEncoder
obs_shape: ${shape_meta.obs.rgb.shape}
num_channel: ${eval:'${shape_meta.obs.rgb.shape[0]} * ${img_cond_steps}'} # each image patch is history concatenated
num_channel: ${eval:'3 * ${img_cond_steps}'} # each image patch is history concatenated
img_h: ${shape_meta.obs.rgb.shape[1]}
img_w: ${shape_meta.obs.rgb.shape[2]}
cfg:
patch_size: 8
depth: 1
@@ -128,7 +130,9 @@ model:
backbone:
_target_: model.common.vit.VitEncoder
obs_shape: ${shape_meta.obs.rgb.shape}
num_channel: ${eval:'${shape_meta.obs.rgb.shape[0]} * ${img_cond_steps}'} # each image patch is history concatenated
num_channel: ${eval:'3 * ${img_cond_steps}'} # each image patch is history concatenated
img_h: ${shape_meta.obs.rgb.shape[1]}
img_w: ${shape_meta.obs.rgb.shape[2]}
cfg:
patch_size: 8
depth: 1
cfg/robomimic/finetune/lift/ft_ppo_diffusion_mlp_img.yaml
+6 -2
View File
@@ -122,7 +122,9 @@ model:
backbone:
_target_: model.common.vit.VitEncoder
obs_shape: ${shape_meta.obs.rgb.shape}
num_channel: ${eval:'${shape_meta.obs.rgb.shape[0]} * ${img_cond_steps}'} # each image patch is history concatenated
num_channel: ${eval:'3 * ${img_cond_steps}'} # each image patch is history concatenated
img_h: ${shape_meta.obs.rgb.shape[1]}
img_w: ${shape_meta.obs.rgb.shape[2]}
cfg:
patch_size: 8
depth: 1
@@ -146,7 +148,9 @@ model:
backbone:
_target_: model.common.vit.VitEncoder
obs_shape: ${shape_meta.obs.rgb.shape}
num_channel: ${eval:'${shape_meta.obs.rgb.shape[0]} * ${img_cond_steps}'} # each image patch is history concatenated
num_channel: ${eval:'3 * ${img_cond_steps}'} # each image patch is history concatenated
img_h: ${shape_meta.obs.rgb.shape[1]}
img_w: ${shape_meta.obs.rgb.shape[2]}
cfg:
patch_size: 8
depth: 1
cfg/robomimic/finetune/lift/ft_ppo_gaussian_mlp_img.yaml
+6 -2
View File
@@ -101,7 +101,9 @@ model:
backbone:
_target_: model.common.vit.VitEncoder
obs_shape: ${shape_meta.obs.rgb.shape}
num_channel: ${eval:'${shape_meta.obs.rgb.shape[0]} * ${img_cond_steps}'} # each image patch is history concatenated
num_channel: ${eval:'3 * ${img_cond_steps}'} # each image patch is history concatenated
img_h: ${shape_meta.obs.rgb.shape[1]}
img_w: ${shape_meta.obs.rgb.shape[2]}
cfg:
patch_size: 8
depth: 1
@@ -128,7 +130,9 @@ model:
backbone:
_target_: model.common.vit.VitEncoder
obs_shape: ${shape_meta.obs.rgb.shape}
num_channel: ${eval:'${shape_meta.obs.rgb.shape[0]} * ${img_cond_steps}'} # each image patch is history concatenated
num_channel: ${eval:'3 * ${img_cond_steps}'} # each image patch is history concatenated
img_h: ${shape_meta.obs.rgb.shape[1]}
img_w: ${shape_meta.obs.rgb.shape[2]}
cfg:
patch_size: 8
depth: 1
cfg/robomimic/finetune/square/ft_ppo_diffusion_mlp_img.yaml
+6 -2
View File
@@ -122,7 +122,9 @@ model:
backbone:
_target_: model.common.vit.VitEncoder
obs_shape: ${shape_meta.obs.rgb.shape}
num_channel: ${eval:'${shape_meta.obs.rgb.shape[0]} * ${img_cond_steps}'} # each image patch is history concatenated
num_channel: ${eval:'3 * ${img_cond_steps}'} # each image patch is history concatenated
img_h: ${shape_meta.obs.rgb.shape[1]}
img_w: ${shape_meta.obs.rgb.shape[2]}
cfg:
patch_size: 8
depth: 1
@@ -146,7 +148,9 @@ model:
backbone:
_target_: model.common.vit.VitEncoder
obs_shape: ${shape_meta.obs.rgb.shape}
num_channel: ${eval:'${shape_meta.obs.rgb.shape[0]} * ${img_cond_steps}'} # each image patch is history concatenated
num_channel: ${eval:'3 * ${img_cond_steps}'} # each image patch is history concatenated
img_h: ${shape_meta.obs.rgb.shape[1]}
img_w: ${shape_meta.obs.rgb.shape[2]}
cfg:
patch_size: 8
depth: 1
cfg/robomimic/finetune/square/ft_ppo_gaussian_mlp_img.yaml
+6 -2
View File
@@ -101,7 +101,9 @@ model:
backbone:
_target_: model.common.vit.VitEncoder
obs_shape: ${shape_meta.obs.rgb.shape}
num_channel: ${eval:'${shape_meta.obs.rgb.shape[0]} * ${img_cond_steps}'} # each image patch is history concatenated
num_channel: ${eval:'3 * ${img_cond_steps}'} # each image patch is history concatenated
img_h: ${shape_meta.obs.rgb.shape[1]}
img_w: ${shape_meta.obs.rgb.shape[2]}
cfg:
patch_size: 8
depth: 1
@@ -128,7 +130,9 @@ model:
backbone:
_target_: model.common.vit.VitEncoder
obs_shape: ${shape_meta.obs.rgb.shape}
num_channel: ${eval:'${shape_meta.obs.rgb.shape[0]} * ${img_cond_steps}'} # each image patch is history concatenated
num_channel: ${eval:'3 * ${img_cond_steps}'} # each image patch is history concatenated
img_h: ${shape_meta.obs.rgb.shape[1]}
img_w: ${shape_meta.obs.rgb.shape[2]}
cfg:
patch_size: 8
depth: 1
cfg/robomimic/finetune/transport/ft_ppo_diffusion_mlp_img.yaml
+6 -2
View File
@@ -126,7 +126,9 @@ model:
backbone:
_target_: model.common.vit.VitEncoder
obs_shape: ${shape_meta.obs.rgb.shape}
num_channel: ${eval:'${shape_meta.obs.rgb.shape[0]} * ${img_cond_steps}'} # each image patch is history concatenated
num_channel: ${eval:'3 * ${img_cond_steps}'} # each image patch is history concatenated
img_h: ${shape_meta.obs.rgb.shape[1]}
img_w: ${shape_meta.obs.rgb.shape[2]}
cfg:
patch_size: 8
depth: 1
@@ -152,7 +154,9 @@ model:
backbone:
_target_: model.common.vit.VitEncoder
obs_shape: ${shape_meta.obs.rgb.shape}
num_channel: ${eval:'${shape_meta.obs.rgb.shape[0]} * ${img_cond_steps}'} # each image patch is history concatenated
num_channel: ${eval:'3 * ${img_cond_steps}'} # each image patch is history concatenated
img_h: ${shape_meta.obs.rgb.shape[1]}
img_w: ${shape_meta.obs.rgb.shape[2]}
cfg:
patch_size: 8
depth: 1
cfg/robomimic/finetune/transport/ft_ppo_gaussian_mlp_img.yaml
+6 -2
View File
@@ -105,7 +105,9 @@ model:
backbone:
_target_: model.common.vit.VitEncoder
obs_shape: ${shape_meta.obs.rgb.shape}
num_channel: ${eval:'${shape_meta.obs.rgb.shape[0]} * ${img_cond_steps}'} # each image patch is history concatenated
num_channel: ${eval:'3 * ${img_cond_steps}'} # each image patch is history concatenated
img_h: ${shape_meta.obs.rgb.shape[1]}
img_w: ${shape_meta.obs.rgb.shape[2]}
cfg:
patch_size: 8
depth: 1
@@ -134,7 +136,9 @@ model:
backbone:
_target_: model.common.vit.VitEncoder
obs_shape: ${shape_meta.obs.rgb.shape}
num_channel: ${eval:'${shape_meta.obs.rgb.shape[0]} * ${img_cond_steps}'} # each image patch is history concatenated
num_channel: ${eval:'3 * ${img_cond_steps}'} # each image patch is history concatenated
img_h: ${shape_meta.obs.rgb.shape[1]}
img_w: ${shape_meta.obs.rgb.shape[2]}
cfg:
patch_size: 8
depth: 1