more changes:

eval.sh

@@ -6,4 +6,3 @@ lerobot-eval \
   --eval.batch_size=1 \
   --eval.n_episodes=1 \
   --seed=142
-

src/lerobot/envs/factory.py

@@ -19,13 +19,13 @@ from typing import Any
 import gymnasium as gym
 from gymnasium.envs.registration import registry as gym_registry
 
+from lerobot.configs.policies import PreTrainedConfig
 from lerobot.envs.configs import AlohaEnv, EnvConfig, LiberoEnv, PushtEnv
 from lerobot.envs.utils import _call_make_env, _download_hub_file, _import_hub_module, _normalize_hub_result
+from lerobot.policies.xvla.configuration_xvla import XVLAConfig
 from lerobot.processor import ProcessorStep
 from lerobot.processor.env_processor import LiberoProcessorStep
 from lerobot.processor.pipeline import PolicyProcessorPipeline
-from lerobot.policies.xvla.configuration_xvla import XVLAConfig
-from lerobot.configs.policies import PreTrainedConfig
 
 
 def make_env_config(env_type: str, **kwargs) -> EnvConfig:
@@ -66,12 +66,12 @@ def make_env_pre_post_processors(
     postprocessor_steps: list[ProcessorStep] = []
     if isinstance(policy_cfg, XVLAConfig):
         from lerobot.policies.xvla.processor_xvla import make_xvla_libero_pre_post_processors
+
         return make_xvla_libero_pre_post_processors()
 
     # For LIBERO environments, add the LiberoProcessorStep to preprocessor
     if isinstance(env_cfg, LiberoEnv) or "libero" in env_cfg.type:
         preprocessor_steps.append(LiberoProcessorStep())
-    
 
     preprocessor = PolicyProcessorPipeline(steps=preprocessor_steps)
     postprocessor = PolicyProcessorPipeline(steps=postprocessor_steps)

src/lerobot/envs/libero.py

@@ -240,11 +240,11 @@ class LiberoEnv(gym.Env):
             image = raw_obs[camera_name]
             images[self.camera_name_mapping[camera_name]] = image
 
-        eef_pos = self._env.robots[0].controller.ee_pos #raw_obs.get("robot0_eef_pos")
+        eef_pos = raw_obs.get("robot0_eef_pos")
         eef_quat = raw_obs.get("robot0_eef_quat")
 
         # rotation matrix from controller
-        eef_mat = self._env.robots[0].controller.ee_ori_mat
+        eef_mat = self._env.robots[0].controller.ee_ori_mat if eef_pos is not None else None
         gripper_qpos = raw_obs.get("robot0_gripper_qpos")
         gripper_qvel = raw_obs.get("robot0_gripper_qvel")
         joint_pos = raw_obs.get("robot0_joint_pos")
@@ -297,7 +297,7 @@ class LiberoEnv(gym.Env):
         # Increasing this value can improve determinism and reproducibility across resets.
         for _ in range(self.num_steps_wait):
             raw_obs, _, _, _ = self._env.step(get_libero_dummy_action())
-        
+
         if self.control_mode == "absolute":
             for robot in self._env.robots:
                 robot.controller.use_delta = False
@@ -422,7 +422,6 @@ def create_libero_envs(
         suite = _get_suite(suite_name)
         total = len(suite.tasks)
         selected = _select_task_ids(total, task_ids_filter)
-        selected = [0]
         if not selected:
             raise ValueError(f"No tasks selected for suite '{suite_name}' (available: {total}).")
 

src/lerobot/policies/xvla/__init__.py

@@ -1,6 +1,2 @@
-from lerobot.policies.xvla.processor_xvla import (
-    XVLAAddDomainIdProcessorStep,
-    XVLAImageScaleProcessorStep,
-    XVLARotation6DToAxisAngleProcessorStep,
-    make_xvla_pre_post_processors,
-)
+# add domainid
+from lerobot.policies.xvla.processor_xvla import XVLAAddDomainIdProcessorStep

src/lerobot/policies/xvla/compare.py

@@ -1,43 +0,0 @@
-import numpy as np
-import os
-
-# --- CHANGE THESE TWO PATHS ---
-dir_a = "/home/jade_choghari/robot/lerobot"
-dir_b = "/home/jade_choghari/robot/robot2/lerobot"
-
-# keys = ["input_ids", "image_input", "image_mask", "domain_id", "proprio"]
-keys = ["domain_id"]
-
-def load_np(path, key):
-    file_path = os.path.join(path, f"{key}.npy")
-    if not os.path.exists(file_path):
-        raise FileNotFoundError(f"Missing file: {file_path}")
-    return np.load(file_path)
-
-for k in keys:
-    print(f"\n===== {k} =====")
-    a = load_np(dir_a, k)
-    b = load_np(dir_b, k)
-    print(a)
-    print(b)
-    print("Shapes:", a.shape, b.shape)
-    same_shape = a.shape == b.shape
-    print("Same shape:", same_shape)
-
-    if not same_shape:
-        continue
-
-    # Allclose for numeric, array_equal for boolean
-    if a.dtype == bool or b.dtype == bool:
-        equal = np.array_equal(a, b)
-        print("Equal:", equal)
-
-        # For boolean arrays, diff = xor
-        diff = np.sum(a ^ b)
-        print("Num differing elements:", diff)
-    else:
-        close = np.allclose(a, b, atol=1e-6, rtol=1e-6)
-        print("Allclose:", close)
-
-        diff = np.max(np.abs(a - b))
-        print("Max difference:", diff)

src/lerobot/policies/xvla/processor_xvla.py

@@ -20,15 +20,16 @@ from typing import Any
 import numpy as np
 import torch
 
+from lerobot.configs.types import PipelineFeatureType, PolicyFeature
 from lerobot.policies.xvla.configuration_xvla import XVLAConfig
 from lerobot.policies.xvla.utils import rotate6d_to_axis_angle
 from lerobot.processor import (
     AddBatchDimensionProcessorStep,
     DeviceProcessorStep,
     NormalizerProcessorStep,
+    ObservationProcessorStep,
     PolicyAction,
     PolicyProcessorPipeline,
-    ObservationProcessorStep,
     ProcessorStep,
     ProcessorStepRegistry,
     RenameObservationsProcessorStep,
@@ -37,8 +38,13 @@ from lerobot.processor import (
 )
 from lerobot.processor.converters import policy_action_to_transition, transition_to_policy_action
 from lerobot.processor.core import EnvTransition, TransitionKey
-from lerobot.utils.constants import POLICY_POSTPROCESSOR_DEFAULT_NAME, POLICY_PREPROCESSOR_DEFAULT_NAME, OBS_STATE, OBS_IMAGES
-from lerobot.configs.types import PipelineFeatureType, PolicyFeature
+from lerobot.utils.constants import (
+    OBS_IMAGES,
+    OBS_STATE,
+    POLICY_POSTPROCESSOR_DEFAULT_NAME,
+    POLICY_PREPROCESSOR_DEFAULT_NAME,
+)
+
 
 def make_xvla_pre_post_processors(
     config: XVLAConfig,
@@ -138,8 +144,8 @@ class LiberoProcessorStep(ObservationProcessorStep):
 
             extra = torch.zeros((eef_pos.shape[0], 1), dtype=torch.float32, device=eef_pos.device)
 
-            proprio_state = torch.cat((eef_pos, eef_rot6d, extra), dim=-1) # (B, 10)
-            state = torch.cat((proprio_state, torch.zeros_like(proprio_state)), dim=-1) # (B, 20)
+            proprio_state = torch.cat((eef_pos, eef_rot6d, extra), dim=-1)  # (B, 10)
+            state = torch.cat((proprio_state, torch.zeros_like(proprio_state)), dim=-1)  # (B, 20)
             # ensure float32
             state = state.float()
             if state.dim() == 1:
@@ -178,7 +184,7 @@ class LiberoProcessorStep(ObservationProcessorStep):
     def _mat_to_rotate6d(self, rot_mats: torch.Tensor) -> torch.Tensor:
         """
         Convert batched rotation matrices (B, 3, 3) into 6D rotation representation (B, 6).
-        
+
         Args:
             rot_mats (Tensor): Rotation matrices of shape (B, 3, 3)
 
@@ -191,21 +197,17 @@ class LiberoProcessorStep(ObservationProcessorStep):
         """
 
         if not isinstance(rot_mats, torch.Tensor):
-            raise TypeError(
-                f"mat_to_rot6d expects a torch.Tensor, got {type(rot_mats)}"
-            )
+            raise TypeError(f"mat_to_rot6d expects a torch.Tensor, got {type(rot_mats)}")
 
         if rot_mats.ndim != 3 or rot_mats.shape[1:] != (3, 3):
-            raise ValueError(
-                f"mat_to_rot6d expects shape (B, 3, 3), got {tuple(rot_mats.shape)}"
-            )
+            raise ValueError(f"mat_to_rot6d expects shape (B, 3, 3), got {tuple(rot_mats.shape)}")
 
         rot_mats = rot_mats.to(torch.float32)
 
         col1 = rot_mats[:, :3, 0]  # (B, 3)
-        col2 = rot_mats[:, :3, 1]   # (B, 3)
+        col2 = rot_mats[:, :3, 1]  # (B, 3)
 
-        rot6d = torch.cat([col1, col2], dim=-1) # (B, 6)
+        rot6d = torch.cat([col1, col2], dim=-1)  # (B, 6)
 
         return rot6d
 
@@ -213,7 +215,6 @@ class LiberoProcessorStep(ObservationProcessorStep):
         return self._process_observation(observation)
 
 
-
 @dataclass
 @ProcessorStepRegistry.register(name="xvla_image_scale")
 class XVLAImageScaleProcessorStep(ProcessorStep):
@@ -294,7 +295,7 @@ class XVLAAddDomainIdProcessorStep(ProcessorStep):
                 if isinstance(v, torch.Tensor):
                     batch_size = v.shape[0]
                     break
-    
+
         # Add domain_id tensor
         comp["domain_id"] = torch.tensor([int(self.domain_id)] * batch_size, dtype=torch.long).to(self.device)
 
@@ -378,8 +379,8 @@ class XVLARotation6DToAxisAngleProcessorStep(ProcessorStep):
             "expected_action_dim": self.expected_action_dim,
         }
 
-def make_xvla_libero_pre_post_processors(
-) -> tuple[
+
+def make_xvla_libero_pre_post_processors() -> tuple[
     PolicyProcessorPipeline[dict[str, Any], dict[str, Any]],
     PolicyProcessorPipeline[PolicyAction, PolicyAction],
 ]:
@@ -398,3 +399,7 @@ def make_xvla_libero_pre_post_processors(
             steps=post_processor_steps,
         ),
     )
+
+__all__ = [
+    "XVLAAddDomainIdProcessorStep",
+]