Hi, thanks for sharing the great work! I tried to reproduce the training process for agents with Habitat on the HM3D dataset. I changed habitat_baselines.evaluate=True to False in config/experiments/vlfm_objectnav_hm3d.yaml (is that correct? do I need to change anything else?). However, I ran into the following error. Does anyone have any idea what might be wrong? Thank you!
....
[05:52:20:041191]:[Assets] ResourceManager.cpp(2210)::loadMaterials : Idx 48:Flat.
[05:52:20:041203]:[Assets] ResourceManager.cpp(2210)::loadMaterials : Idx 49:Flat.
[05:52:20:041216]:[Assets] ResourceManager.cpp(2210)::loadMaterials : Idx 50:Flat.
[05:52:20:041228]:[Assets] ResourceManager.cpp(2210)::loadMaterials : Idx 51:Flat.
[05:52:20:041240]:[Assets] ResourceManager.cpp(2210)::loadMaterials : Idx 52:Flat.
[05:52:20:129431]:[Sim] Simulator.cpp(446)::instanceStageForSceneAttributes : Successfully loaded stage named : ./data/scene_datasets/hm3d/train/00657-TSJmdttd2GV/TSJmdttd2GV.basis.glb
[05:52:20:129535]:[Sim] Simulator.cpp(203)::reconfigure : CreateSceneInstance success == true for active scene name : data/scene_datasets/hm3d/train/00657-TSJmdttd2GV/TSJmdttd2GV.basis.glb with renderer.
[05:52:20:175509]:[Nav] PathFinder.cpp(571)::build : Building navmesh with 279 x 231 cells
[05:52:20:320075]:[Nav] PathFinder.cpp(845)::build : Created navmesh with 571 vertices 275 polygons
[05:52:20:320151]:[Sim] Simulator.cpp(908)::recomputeNavMesh : reconstruct navmesh successful
2025-08-21 05:52:20,324 Initializing task ObjectNav-v1
2025-08-21 05:52:25,739 agent number of parameters: 0
BLIP2ITMClient.cosine: (480, 640, 3), Seems like there is a potted plant ahead.
Step: 0 | Mode: initialize | Action: 2
BLIP2ITMClient.cosine: (480, 640, 3), Seems like there is a potted plant ahead.
Step: 1 | Mode: initialize | Action: 2
BLIP2ITMClient.cosine: (480, 640, 3), Seems like there is a potted plant ahead.
Step: 2 | Mode: initialize | Action: 2
BLIP2ITMClient.cosine: (480, 640, 3), Seems like there is a potted plant ahead.
Step: 3 | Mode: initialize | Action: 2
BLIP2ITMClient.cosine: (480, 640, 3), Seems like there is a potted plant ahead.
Step: 4 | Mode: initialize | Action: 2
BLIP2ITMClient.cosine: (480, 640, 3), Seems like there is a potted plant ahead.
Step: 5 | Mode: initialize | Action: 2
BLIP2ITMClient.cosine: (480, 640, 3), Seems like there is a potted plant ahead.
Step: 6 | Mode: initialize | Action: 2
Error executing job with overrides: []
Traceback (most recent call last):
File "/vlfm/run.py", line 55, in main
execute_exp(cfg, "eval" if cfg.habitat_baselines.evaluate else "train")
File "~/mambaforge/envs/vlfm2/lib/python3.9/site-packages/habitat_baselines/run.py", line 60, in execute_exp
trainer.train()
File "~/mambaforge/envs/vlfm2/lib/python3.9/contextlib.py", line 79, in inner
return func(*args, **kwds)
File "~/mambaforge/envs/vlfm2/lib/python3.9/site-packages/habitat_baselines/rl/ppo/ppo_trainer.py", line 710, in train
count_steps_delta += self._collect_environment_result(
File "~/mambaforge/envs/vlfm2/lib/python3.9/site-packages/habitat_baselines/rl/ppo/ppo_trainer.py", line 442, in _collect_environment_result
self._agent.rollouts.insert(
File "~/mambaforge/envs/vlfm2/lib/python3.9/site-packages/habitat_baselines/common/rollout_storage.py", line 150, in insert
self.buffers.set(
File "~/mambaforge/envs/vlfm2/lib/python3.9/site-packages/habitat_baselines/common/tensor_dict.py", line 238, in set
dst.set(index, v, strict=strict)
File "~/mambaforge/envs/vlfm2/lib/python3.9/site-packages/habitat_baselines/common/tensor_dict.py", line 241, in set
dst[index] = self._to_instance(v)
RuntimeError: The expanded size of the tensor (1) must match the existing size (2) at non-singleton dimension 1. Target sizes: [1, 1, 2]. Tensor sizes: [2, 2]
Set the environment variable HYDRA_FULL_ERROR=1 for a complete stack trace.
Exception ignored in: <function VectorEnv.__del__ at 0x14db308479d0>
Traceback (most recent call last):
File "~/mambaforge/envs/vlfm2/lib/python3.9/site-packages/habitat/core/vector_env.py", line 589, in __del__
self.close()
File "~/mambaforge/envs/vlfm2/lib/python3.9/site-packages/habitat/core/vector_env.py", line 460, in close
write_fn((CLOSE_COMMAND, None))
File "~/mambaforge/envs/vlfm2/lib/python3.9/site-packages/habitat/core/vector_env.py", line 129, in __call__
self.write_fn(data)
File "~/mambaforge/envs/vlfm2/lib/python3.9/site-packages/habitat/utils/pickle5_multiprocessing.py", line 63, in send
self.send_bytes(buf.getvalue())
File "~/mambaforge/envs/vlfm2/lib/python3.9/multiprocessing/connection.py", line 200, in send_bytes
self._send_bytes(m[offset:offset + size])
File "~/mambaforge/envs/vlfm2/lib/python3.9/multiprocessing/connection.py", line 411, in _send_bytes
self._send(header + buf)
File "~/mambaforge/envs/vlfm2/lib/python3.9/multiprocessing/connection.py", line 368, in _send
n = write(self._handle, buf)
BrokenPipeError: [Errno 32] Broken pipe
Hi, thanks for sharing the great work! I tried to reproduce the training process for agents with Habitat on the HM3D dataset. I changed
habitat_baselines.evaluate=TruetoFalseinconfig/experiments/vlfm_objectnav_hm3d.yaml(is that correct? do I need to change anything else?). However, I ran into the following error. Does anyone have any idea what might be wrong? Thank you!