Loading vagen/env/primitive_skill/maniskill/env/align_two_cube.py +2 −2 Original line number Diff line number Diff line Loading @@ -45,12 +45,12 @@ class AlignTwoCubeEnv(BaseEnv): @property def _default_sensor_configs(self): pose = sapien_utils.look_at(eye=[0.3, 0, 0.6], target=[-0.1, 0, 0.1]) return [CameraConfig("base_camera", pose, 128, 128, np.pi / 2, 0.01, 100)] return [CameraConfig("base_camera", pose, 300, 300, np.pi / 2, 0.01, 100)] @property def _default_human_render_camera_configs(self): pose = sapien_utils.look_at([0.6, 0.7, 0.6], [0.0, 0.0, 0.35]) return CameraConfig("render_camera", pose, 512, 512, 1, 0.01, 100) return CameraConfig("render_camera", pose, 300,300, 1, 0.01, 100) def _load_scene(self, options: dict): self.cube_half_size = common.to_tensor([0.02] * 3) Loading vagen/env/primitive_skill/maniskill/env/place_two_cube.py +2 −2 Original line number Diff line number Diff line Loading @@ -46,12 +46,12 @@ class PlaceTwoCubeEnv(BaseEnv): @property def _default_sensor_configs(self): pose = sapien_utils.look_at(eye=[0.3, 0, 0.6], target=[-0.1, 0, 0.1]) return [CameraConfig("base_camera", pose, 128, 128, np.pi / 2, 0.01, 100)] return [CameraConfig("base_camera", pose, 300, 300, np.pi / 2, 0.01, 100)] @property def _default_human_render_camera_configs(self): pose = sapien_utils.look_at([0.6, 0.7, 0.6], [0.0, 0.0, 0.35]) return CameraConfig("render_camera", pose, 512, 512, 1, 0.01, 100) return CameraConfig("render_camera", pose, 300,300, 1, 0.01, 100) def _load_scene(self, options: dict): self.cube_half_size = common.to_tensor([0.02] * 3) Loading vagen/env/primitive_skill/maniskill/env/put_apple_in_drawer.py +2 −2 Original line number Diff line number Diff line Loading @@ -57,7 +57,7 @@ class PutAppleInDrawerEnv(BaseEnv): @property def _default_human_render_camera_configs(self): pose = sapien_utils.look_at([1.0, 0.3, 1.0], [-0.5, -0.2, 0.35]) return CameraConfig("render_camera", pose, 2048, 2048, 1, 0.01, 100) return CameraConfig("render_camera", pose, 300, 300, 1, 0.01, 100) def instruction(self): return "Please put the apple in the drawer and close the drawer." Loading Loading @@ -113,7 +113,7 @@ class PutAppleInDrawerEnv(BaseEnv): xyz = torch.zeros((b, 3), device=self.device) xyz[:, 2] = apple_height # Use apple's actual height sampler = randomization.UniformPlacementSampler(bounds=[ [self.workspace_x[0]+0.35, self.workspace_y[0]+0.5], [self.workspace_x[0]+0.35, self.workspace_y[0]+0.7], [self.workspace_x[1], self.workspace_y[1]] ], batch_size=b) apple_xy = sampler.sample(0.04, 100) # Adjust radius based on apple size if needed Loading vagen/env/primitive_skill/maniskill/env/stack_three_cube.py +2 −2 Original line number Diff line number Diff line Loading @@ -46,13 +46,13 @@ class StackThreeCubeEnv(BaseEnv): @property def _default_sensor_configs(self): pose = sapien_utils.look_at(eye=[0.3, 0, 0.6], target=[-0.1, 0, 0.1]) return [CameraConfig("base_camera", pose, 128, 128, np.pi / 2, 0.01, 100)] return [CameraConfig("base_camera", pose, 300, 300, np.pi / 2, 0.01, 100)] @property def _default_human_render_camera_configs(self): pose = sapien_utils.look_at([0.8, 0.8, 0.6], [0.14, +0.08, 0.12]) return CameraConfig( "render_camera", pose=pose, width=2048, height=2048, fov=0.63, near=0.01, far=100 "render_camera", pose=pose, width=300, height=300, fov=0.63, near=0.01, far=100 ) def instruction(self): Loading vagen/examples/navigation_aico/run.sh +5 −4 Original line number Diff line number Diff line Loading @@ -21,13 +21,13 @@ python3 -m vagen.trainer.main_ppo \ algorithm.high_level_gamma=0.95 \ data.train_files=data/navigation-vision-debug/train.parquet \ data.val_files=data/navigation-vision-debug/test.parquet \ data.train_batch_size=32 \ data.train_batch_size=64 \ data.val_batch_size=16 \ data.max_prompt_length=1024 \ data.max_response_length=128 \ data.max_trajectory_length=2400 \ data.image_key=images \ data.truncation=left \ data.truncation=error \ actor_rollout_ref.model.path=Qwen/Qwen2.5-VL-3B-Instruct \ actor_rollout_ref.actor.optim.lr=1e-6 \ actor_rollout_ref.model.use_remove_padding=True \ Loading @@ -42,7 +42,7 @@ python3 -m vagen.trainer.main_ppo \ actor_rollout_ref.rollout.log_prob_micro_batch_size_per_gpu=1 \ actor_rollout_ref.rollout.tensor_model_parallel_size=2 \ actor_rollout_ref.rollout.name=vllm \ actor_rollout_ref.rollout.gpu_memory_utilization=0.2 \ actor_rollout_ref.rollout.gpu_memory_utilization=0.1 \ actor_rollout_ref.rollout.enable_chunked_prefill=False \ actor_rollout_ref.rollout.enforce_eager=False \ actor_rollout_ref.rollout.free_cache_engine=False \ Loading Loading @@ -75,7 +75,8 @@ python3 -m vagen.trainer.main_ppo \ rollout_manager.use_gae_mask=True \ trainer.val_before_train=True \ trainer.val_generations_to_log_to_wandb=4 \ rollout_manager.n_trajectory=2 \ rollout_manager.n_trajectory=1 \ rollout_manager.use_service=True \ rollout_manager.timeout=240 \ rollout_manager.base_url="http://localhost:5001" \ 2>&1 | tee aico_navigation_vision.log No newline at end of file Loading
vagen/env/primitive_skill/maniskill/env/align_two_cube.py +2 −2 Original line number Diff line number Diff line Loading @@ -45,12 +45,12 @@ class AlignTwoCubeEnv(BaseEnv): @property def _default_sensor_configs(self): pose = sapien_utils.look_at(eye=[0.3, 0, 0.6], target=[-0.1, 0, 0.1]) return [CameraConfig("base_camera", pose, 128, 128, np.pi / 2, 0.01, 100)] return [CameraConfig("base_camera", pose, 300, 300, np.pi / 2, 0.01, 100)] @property def _default_human_render_camera_configs(self): pose = sapien_utils.look_at([0.6, 0.7, 0.6], [0.0, 0.0, 0.35]) return CameraConfig("render_camera", pose, 512, 512, 1, 0.01, 100) return CameraConfig("render_camera", pose, 300,300, 1, 0.01, 100) def _load_scene(self, options: dict): self.cube_half_size = common.to_tensor([0.02] * 3) Loading
vagen/env/primitive_skill/maniskill/env/place_two_cube.py +2 −2 Original line number Diff line number Diff line Loading @@ -46,12 +46,12 @@ class PlaceTwoCubeEnv(BaseEnv): @property def _default_sensor_configs(self): pose = sapien_utils.look_at(eye=[0.3, 0, 0.6], target=[-0.1, 0, 0.1]) return [CameraConfig("base_camera", pose, 128, 128, np.pi / 2, 0.01, 100)] return [CameraConfig("base_camera", pose, 300, 300, np.pi / 2, 0.01, 100)] @property def _default_human_render_camera_configs(self): pose = sapien_utils.look_at([0.6, 0.7, 0.6], [0.0, 0.0, 0.35]) return CameraConfig("render_camera", pose, 512, 512, 1, 0.01, 100) return CameraConfig("render_camera", pose, 300,300, 1, 0.01, 100) def _load_scene(self, options: dict): self.cube_half_size = common.to_tensor([0.02] * 3) Loading
vagen/env/primitive_skill/maniskill/env/put_apple_in_drawer.py +2 −2 Original line number Diff line number Diff line Loading @@ -57,7 +57,7 @@ class PutAppleInDrawerEnv(BaseEnv): @property def _default_human_render_camera_configs(self): pose = sapien_utils.look_at([1.0, 0.3, 1.0], [-0.5, -0.2, 0.35]) return CameraConfig("render_camera", pose, 2048, 2048, 1, 0.01, 100) return CameraConfig("render_camera", pose, 300, 300, 1, 0.01, 100) def instruction(self): return "Please put the apple in the drawer and close the drawer." Loading Loading @@ -113,7 +113,7 @@ class PutAppleInDrawerEnv(BaseEnv): xyz = torch.zeros((b, 3), device=self.device) xyz[:, 2] = apple_height # Use apple's actual height sampler = randomization.UniformPlacementSampler(bounds=[ [self.workspace_x[0]+0.35, self.workspace_y[0]+0.5], [self.workspace_x[0]+0.35, self.workspace_y[0]+0.7], [self.workspace_x[1], self.workspace_y[1]] ], batch_size=b) apple_xy = sampler.sample(0.04, 100) # Adjust radius based on apple size if needed Loading
vagen/env/primitive_skill/maniskill/env/stack_three_cube.py +2 −2 Original line number Diff line number Diff line Loading @@ -46,13 +46,13 @@ class StackThreeCubeEnv(BaseEnv): @property def _default_sensor_configs(self): pose = sapien_utils.look_at(eye=[0.3, 0, 0.6], target=[-0.1, 0, 0.1]) return [CameraConfig("base_camera", pose, 128, 128, np.pi / 2, 0.01, 100)] return [CameraConfig("base_camera", pose, 300, 300, np.pi / 2, 0.01, 100)] @property def _default_human_render_camera_configs(self): pose = sapien_utils.look_at([0.8, 0.8, 0.6], [0.14, +0.08, 0.12]) return CameraConfig( "render_camera", pose=pose, width=2048, height=2048, fov=0.63, near=0.01, far=100 "render_camera", pose=pose, width=300, height=300, fov=0.63, near=0.01, far=100 ) def instruction(self): Loading
vagen/examples/navigation_aico/run.sh +5 −4 Original line number Diff line number Diff line Loading @@ -21,13 +21,13 @@ python3 -m vagen.trainer.main_ppo \ algorithm.high_level_gamma=0.95 \ data.train_files=data/navigation-vision-debug/train.parquet \ data.val_files=data/navigation-vision-debug/test.parquet \ data.train_batch_size=32 \ data.train_batch_size=64 \ data.val_batch_size=16 \ data.max_prompt_length=1024 \ data.max_response_length=128 \ data.max_trajectory_length=2400 \ data.image_key=images \ data.truncation=left \ data.truncation=error \ actor_rollout_ref.model.path=Qwen/Qwen2.5-VL-3B-Instruct \ actor_rollout_ref.actor.optim.lr=1e-6 \ actor_rollout_ref.model.use_remove_padding=True \ Loading @@ -42,7 +42,7 @@ python3 -m vagen.trainer.main_ppo \ actor_rollout_ref.rollout.log_prob_micro_batch_size_per_gpu=1 \ actor_rollout_ref.rollout.tensor_model_parallel_size=2 \ actor_rollout_ref.rollout.name=vllm \ actor_rollout_ref.rollout.gpu_memory_utilization=0.2 \ actor_rollout_ref.rollout.gpu_memory_utilization=0.1 \ actor_rollout_ref.rollout.enable_chunked_prefill=False \ actor_rollout_ref.rollout.enforce_eager=False \ actor_rollout_ref.rollout.free_cache_engine=False \ Loading Loading @@ -75,7 +75,8 @@ python3 -m vagen.trainer.main_ppo \ rollout_manager.use_gae_mask=True \ trainer.val_before_train=True \ trainer.val_generations_to_log_to_wandb=4 \ rollout_manager.n_trajectory=2 \ rollout_manager.n_trajectory=1 \ rollout_manager.use_service=True \ rollout_manager.timeout=240 \ rollout_manager.base_url="http://localhost:5001" \ 2>&1 | tee aico_navigation_vision.log No newline at end of file
