feat(vla): vla框架初始化

roboimi/vla/models/backbones/__init__.py

@@ -0,0 +1,6 @@
+# Backbone models
+from .siglip import SigLIPBackbone
+from .clip import CLIPBackbone
+from .dinov2 import DinoV2Backbone
+
+__all__ = ["SigLIPBackbone", "CLIPBackbone", "DinoV2Backbone"]

roboimi/vla/models/backbones/clip.py

@@ -0,0 +1 @@
+# CLIP Backbone 实现

roboimi/vla/models/backbones/dinov2.py

@@ -0,0 +1 @@
+# DinoV2 Backbone 实现

roboimi/vla/models/backbones/siglip.py

@@ -0,0 +1 @@
+# SigLIP Backbone 实现

roboimi/vla/models/heads/__init__.py

@@ -0,0 +1,5 @@
+# Action Head models
+from .diffusion import DiffusionActionHead
+from .act import ACTHead
+
+__all__ = ["DiffusionActionHead", "ACTHead"]

roboimi/vla/models/heads/act.py

@@ -0,0 +1 @@
+# ACT-VAE Action Head 实现

roboimi/vla/models/heads/diffusion.py

@@ -0,0 +1 @@
+# Diffusion Policy Action Head 实现

roboimi/vla/models/projectors/__init__.py

@@ -0,0 +1,5 @@
+# Projector models
+from .mlp import MLPProjector
+from .perceiver import PerceiverResampler
+
+__all__ = ["MLPProjector", "PerceiverResampler"]

roboimi/vla/models/projectors/mlp.py

@@ -0,0 +1 @@
+# MLP Projector 实现

roboimi/vla/models/projectors/perceiver.py

@@ -0,0 +1 @@
+# Perceiver Resampler 实现