From bb4889e82b1e847d1cd664a0af1c6114f5c57ea8 Mon Sep 17 00:00:00 2001 From: wangyukai Date: Thu, 11 Sep 2025 12:34:22 +0000 Subject: [PATCH] fix fabric camera delay issue --- .../eval/configs/h1_internvla_n1_async_cfg.py | 40 +++++++++---------- scripts/eval/configs/h1_internvla_n1_cfg.py | 35 ++++++++-------- 2 files changed, 38 insertions(+), 37 deletions(-) diff --git a/scripts/eval/configs/h1_internvla_n1_async_cfg.py b/scripts/eval/configs/h1_internvla_n1_async_cfg.py index 6ce3e18b..15c101eb 100644 --- a/scripts/eval/configs/h1_internvla_n1_async_cfg.py +++ b/scripts/eval/configs/h1_internvla_n1_async_cfg.py @@ -14,33 +14,33 @@ model_name='internvla_n1', ckpt_path='', model_settings={ - 'env_num': 1, 'sim_num': 1, + 'env_num': 1, + 'sim_num': 1, 'model_path': "checkpoints/InternVLA-N1", - 'camera_intrinsic': [ - [585.0, 0.0, 320.0], [0.0, 585.0, 240.0], [0.0, 0.0, 1.0] - ], - - 'width': 640, 'height': 480, 'hfov': 79, - 'resize_w': 384, 'resize_h': 384, + 'camera_intrinsic': [[585.0, 0.0, 320.0], [0.0, 585.0, 240.0], [0.0, 0.0, 1.0]], + 'width': 640, + 'height': 480, + 'hfov': 79, + 'resize_w': 384, + 'resize_h': 384, 'max_new_tokens': 1024, 'num_frames': 32, 'num_history': 8, 'num_future_steps': 4, - 'device': 'cuda:1', 'predict_step_nums': 32, - 'continuous_traj': True, - - 'infer_mode': 'partial_async', # You can choose "sync" or "partial_async", but for this model, "partial_async" is better. + 'continuous_traj': True, + 'infer_mode': 'partial_async', # You can choose "sync" or "partial_async", but for this model, "partial_async" is better. # debug - 'vis_debug': True, # If vis_debug=True, you can get visualization results - 'vis_debug_path': './logs/test/vis_debug' + 'vis_debug': True, # If vis_debug=True, you can get visualization results + 'vis_debug_path': './logs/test/vis_debug', }, ), env=EnvCfg( env_type='vln_multi', env_settings={ - 'use_fabric': False, # Please set use_fabric=False due to the render delay; + 'rendering_interval': 0, # You can set rendering_interval=0 to avoid the render delay. + 'use_fabric': True, 'headless': True, }, ), @@ -48,27 +48,27 @@ task_name='test', task_settings={ 'env_num': 1, - 'use_distributed': False, # If the others setting in task_settings, please set use_distributed = False. + 'use_distributed': False, # If the others setting in task_settings, please set use_distributed = False. 'proc_num': 1, - # 'max_step': 1000, #If use flash modeļ¼Œdefault 1000; descrete mode, set 50000 + # 'max_step': 1000, #If use flash modeļ¼Œdefault 1000; descrete mode, set 50000 }, scene=SceneCfg( scene_type='mp3d', scene_data_dir='data/scene_data/mp3d_pe', ), robot_name='h1', - robot_flash=True, # If robot_flash is True, the mode is flash (set world_pose directly); else you choose physical mode. + robot_flash=True, # If robot_flash is True, the mode is flash (set world_pose directly); else you choose physical mode. robot_usd_path='data/Embodiments/vln-pe/h1/h1_internvla.usd', - camera_resolution=[640, 480], # (W,H) + camera_resolution=[640, 480], # (W,H) camera_prim_path='torso_link/h1_1_25_down_30', - one_step_stand_still = True, #For dual-system, please keep this param True. + one_step_stand_still=True, # For dual-system, please keep this param True. ), dataset=EvalDatasetCfg( dataset_type="mp3d", dataset_settings={ 'base_data_dir': 'data/vln_pe/raw_data/r2r', 'split_data_types': ['val_seen', 'val_unseen'], # 'val_seen' - 'filter_stairs': False, # For iros challenge, this is False; For results in the paper, this is True. + 'filter_stairs': False, # For iros challenge, this is False; For results in the paper, this is True. # 'selected_scans': ['zsNo4HB9uLZ'], # 'selected_scans': ['8194nk5LbLH', 'pLe4wQe7qrG'], }, diff --git a/scripts/eval/configs/h1_internvla_n1_cfg.py b/scripts/eval/configs/h1_internvla_n1_cfg.py index ac649b04..963d8e8d 100644 --- a/scripts/eval/configs/h1_internvla_n1_cfg.py +++ b/scripts/eval/configs/h1_internvla_n1_cfg.py @@ -14,31 +14,32 @@ model_name='internvla_n1', ckpt_path='', model_settings={ - 'env_num': 1, 'sim_num': 1, + 'env_num': 1, + 'sim_num': 1, 'model_path': "checkpoints/InternVLA-N1-Preview", - 'camera_intrinsic': [ - [585.0, 0.0, 320.0], [0.0, 585.0, 240.0], [0.0, 0.0, 1.0] - ], - - 'width': 640, 'height': 480, 'hfov': 79, - 'resize_w': 384, 'resize_h': 384, + 'camera_intrinsic': [[585.0, 0.0, 320.0], [0.0, 585.0, 240.0], [0.0, 0.0, 1.0]], + 'width': 640, + 'height': 480, + 'hfov': 79, + 'resize_w': 384, + 'resize_h': 384, 'max_new_tokens': 1024, 'num_frames': 32, 'num_history': 8, 'num_future_steps': 4, - 'device': 'cuda:1', 'predict_step_nums': 32, - 'continuous_traj': True, + 'continuous_traj': True, # debug - 'vis_debug': True, # If vis_debug=True, you can get visualization results - 'vis_debug_path': './logs/test/vis_debug' + 'vis_debug': True, # If vis_debug=True, you can get visualization results + 'vis_debug_path': './logs/test/vis_debug', }, ), env=EnvCfg( env_type='vln_multi', env_settings={ - 'use_fabric': False, # Please set use_fabric=False due to the render delay; + 'rendering_interval': 0, # You can set rendering_interval=0 to avoid the render delay. + 'use_fabric': True, 'headless': True, }, ), @@ -46,7 +47,7 @@ task_name='test', task_settings={ 'env_num': 1, - 'use_distributed': False, # If the others setting in task_settings, please set use_distributed = False. + 'use_distributed': False, # If the others setting in task_settings, please set use_distributed = False. 'proc_num': 1, }, scene=SceneCfg( @@ -54,18 +55,18 @@ scene_data_dir='data/scene_data/mp3d_pe', ), robot_name='h1', - robot_flash=True, # If robot_flash is True, the mode is flash (set world_pose directly); else you choose physical mode. + robot_flash=True, # If robot_flash is True, the mode is flash (set world_pose directly); else you choose physical mode. robot_usd_path='data/Embodiments/vln-pe/h1/h1_internvla.usd', - camera_resolution=[640, 480], # (W,H) + camera_resolution=[640, 480], # (W,H) camera_prim_path='torso_link/h1_1_25_down_30', - one_step_stand_still = True, #For dual-system, please keep this param True. + one_step_stand_still=True, # For dual-system, please keep this param True. ), dataset=EvalDatasetCfg( dataset_type="mp3d", dataset_settings={ 'base_data_dir': 'data/vln_pe/raw_data/r2r', 'split_data_types': ['val_unseen'], # 'val_seen' - 'filter_stairs': False, + 'filter_stairs': False, # 'selected_scans': ['zsNo4HB9uLZ'], # 'selected_scans': ['8194nk5LbLH', 'pLe4wQe7qrG'], },