Closed dwei-k closed 1 year ago
hi! @dwei-k. I'm running the eval of onjectnav on mp3d now, but I am blocked in the following result, I don't know if it is running or entering an endless loop of the process.
I0326 19:33:12.711207 2695857 ManagedContainerBase.cpp:19] ManagedContainerBase::convertFilenameToJSON : Filename : data/scene_datasets/mp3d/8194nk5LbLH/8194nk5LbLH.glb changed to proposed JSON configuration filename : data/scene_datasets/mp3d/8194nk5LbLH/8194nk5LbLH.stage_config.json
I0326 19:33:12.711282 2695857 AbstractObjectAttributesManagerBase.h:180] File (data/scene_datasets/mp3d/8194nk5LbLH/8194nk5LbLH.glb) exists but is not a recognized config filename extension, so new default Stage attributes created and registered.
I0326 19:33:12.711297 2695857 Simulator.cpp:156] Loading navmesh from data/scene_datasets/mp3d/8194nk5LbLH/8194nk5LbLH.navmesh
I0326 19:33:12.711426 2695857 Simulator.cpp:158] Loaded.
I0326 19:33:12.711445 2695857 SceneGraph.h:93] Created DrawableGroup:
Renderer: NVIDIA GeForce RTX 3090/PCIe/SSE2 by NVIDIA Corporation
OpenGL version: 4.6.0 NVIDIA 510.47.03
Using optional features:
GL_ARB_ES2_compatibility
GL_ARB_direct_state_access
GL_ARB_get_texture_sub_image
GL_ARB_invalidate_subdata
GL_ARB_multi_bind
GL_ARB_robustness
GL_ARB_separate_shader_objects
GL_ARB_texture_filter_anisotropic
GL_ARB_texture_storage
GL_ARB_texture_storage_multisample
GL_ARB_vertex_array_object
GL_KHR_debug
Using driver workarounds:
no-forward-compatible-core-context
nv-egl-incorrect-gl11-function-pointers
no-layout-qualifiers-on-old-glsl
nv-zero-context-profile-mask
nv-implementation-color-read-format-dsa-broken
nv-cubemap-inconsistent-compressed-image-size
nv-cubemap-broken-full-compressed-image-query
nv-compressed-block-size-in-bits
I0326 19:33:12.787256 2695857 ResourceManager.cpp:205] ResourceManager::loadStage : Loading Semantic Stage mesh : data/scene_datasets/mp3d/8194nk5LbLH/8194nk5LbLH_semantic.ply
I0326 19:33:12.787293 2695857 SceneGraph.h:93] Created DrawableGroup:
I0326 19:33:14.959846 2695857 ResourceManager.cpp:237] ResourceManager::loadStage : Semantic Stage mesh : data/scene_datasets/mp3d/8194nk5LbLH/8194nk5LbLH_semantic.ply loaded.
I0326 19:33:14.959901 2695857 ResourceManager.cpp:1146] Importing Basis files as BC3
I0326 19:33:16.179716 2695857 simulator.py:220] Loaded navmesh data/scene_datasets/mp3d/8194nk5LbLH/8194nk5LbLH.navmesh
I0326 19:33:16.180754 2695857 simulator.py:232] Recomputing navmesh for agent's height 0.88 and radius 0.18.
W0326 19:33:16.188374 2695857 PathFinder.cpp:716] Building naavmesh before << -1 -- [9.26109,5.3914,9.69957]
W0326 19:33:16.188406 2695857 PathFinder.cpp:721] Building naavmesh after << -1 -- [9.26109,5.3914,9.69957]
I0326 19:33:16.188418 2695857 PathFinder.cpp:407] Building navmesh with 403x274 cells
I0326 19:33:16.316104 2695857 PathFinder.cpp:675] Created navmesh with 365 vertices 183 polygons
I0326 19:33:16.316138 2695857 Simulator.cpp:767] reconstruct navmesh successful
2023-03-26 19:33:16,319 Initializing task ObjectNav-v1
2023-03-26 19:33:16,320 max object cat: 20
2023-03-26 19:33:16,321 cats: dict_values([0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20])
Have you ever been in a similar situation? And how long have you run the eval script?
Hi @dwei-k @me-no-money ,
The evaluation script runs for ~8 hours to complete evaluation on all 2k episodes. Are you still facing this issue?
@Ram81 thank you very much for replying. I recently found that it took a extremly long time because I set num_environments
by 1
Got it. Increasing num environments to > 1 should resolve this issue.
Closing the issue as it is resolved.
hi, @Ram81 thanks for your great work. after I followed the steps of this repo to start eval of onjectnav on mp3d, i got an extremly long time to finish it
I change the eval scripts:
!/bin/bash
SBATCH --job-name=onav_eval
SBATCH --gres gpu:1
SBATCH --nodes 1
SBATCH --cpus-per-task 6
SBATCH --ntasks-per-node 1
SBATCH --partition=long
SBATCH --constraint=rtx_6000
SBATCH --output=slurm_logs/eval/eval-%j.out
SBATCH --error=slurm_logs/eval/eval-%j.err
source /srv/share3/rramrakhya6/miniconda3/etc/profile.d/conda.sh
conda deactivate
conda activate habitat-3
export GLOG_minloglevel=2 export MAGNUM_LOG=quiet
MASTER_ADDR=$(srun --ntasks=1 hostname 2>&1 | tail -n1) export MASTER_ADDR
path=$1 val_dataset_path=$2 checkpoint=$3
set -x
echo "Evaluating..." echo "Hab-Sim: ${PYTHONPATH}"
python -u -m habitat_baselines.run \ --exp-config $path \ --run-type eval \ TASK_CONFIG.DATASET.DATA_PATH "$val_dataset_path/{split}/{split}.json.gz" \ TASK_CONFIG.TASK.SENSORS "['OBJECTGOAL_SENSOR', 'COMPASS_SENSOR', 'GPS_SENSOR']" \ EVAL_CKPT_PATH_DIR $checkpoint
and start it by this command:
job_scripts/run_objectnav_eval.sh habitat_baselines/config/objectnav/il_objectnav.yaml data/datasets/objectnav checkpoint/objectnav_semseg.ckpt
I wonder how does it take you to run the eval scripts, and whether I did something wrong. looking forward to your reply!