Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

from agents.navigation.global_route_planner import GlobalRoutePlanner #52

Closed
scuizhibin opened this issue Aug 13, 2024 · 11 comments
Closed

Comments

@scuizhibin
Copy link

scuizhibin commented Aug 13, 2024

我按照readme上的要求进行安装carla,运行bash leaderboard/scripts/run_evaluation_debug.sh,出现问题

@AvinsWang
Copy link

AvinsWang commented Aug 13, 2024

@scuizhibin 我之前也遇到过, 添加下面两行解决了

import sys
sys.path.append('/path_to_carla/PythonAPI/carla')

@scuizhibin
Copy link
Author

@scuizhibin 我之前也遇到过, 添加下面两行解决了

import sys
sys.path.append('/path_to_carla/PythonAPI/carla')

还有一个问题 如果把VAD跑起来,run_evaluation_debug.sh 中TEAM_CONFIG如何设置?

@AvinsWang
Copy link

AvinsWang commented Aug 13, 2024

@scuizhibin 和你使用ckpt对应即可, 我使用的是 Bench2DriveZoo/ckpts/uniad_tiny_b2d.pth

TEAM_CONFIG=Bench2DriveZoo/adzoo/uniad/configs/stage2_e2e/tiny_e2e_b2d.py

这样设置加载checkpoint可能出现问题, 可以手动修改下Bench2DriveZoo/team_code/uniad_b2d_agent.py

self.ckpt_path = 'Bench2DriveZoo/ckpts/uniad_tiny_b2d.pth'
checkpoint = load_checkpoint(self.model, self.ckpt_path, map_location='cpu', strict=True)

@scuizhibin
Copy link
Author

@scuizhibin 和你使用ckpt对应即可, 我使用的是 Bench2DriveZoo/ckpts/uniad_tiny_b2d.pth

TEAM_CONFIG=Bench2DriveZoo/adzoo/uniad/configs/stage2_e2e/tiny_e2e_b2d.py

这样设置加载checkpoint可能出现问题, 可以手动修改下Bench2DriveZoo/team_code/vad_b2d_agent.py

self.ckpt_path = 'Bench2DriveZoo/ckpts/uniad_tiny_b2d.pth'
checkpoint = load_checkpoint(self.model, self.ckpt_path, map_location='cpu', strict=True)

最后一个问题 我的torch版本是2.4,torchvision 版本0.19.0,但是运行bash命令出现,RuntimeError: operator torchvision::nms does not exist

@AvinsWang
Copy link

AvinsWang commented Aug 13, 2024

@scuizhibin 这是我使用的版本, 按照INSTALL.md安装能成功运行;你的问题我没遇到过, 不知道怎么解, 可以重装下torchvision或者按照官方重新配置环境

torch 2.4.0+cu118
torchaudio 2.4.0+cu118
torchmetrics 0.11.4
torchvision 0.19.0+cu118

@scuizhibin
Copy link
Author

@scuizhibin 这是我使用的版本, 按照INSTALL.md安装能成功运行;你的问题我没遇到过, 不知道怎么解, 可以重装下torchvision或者按照官方重新配置环境

torch 2.4.0+cu118
torchaudio 2.4.0+cu118
torchmetrics 0.11.4
torchvision 0.19.0+cu118

ATen/native/TensorShape.cpp:3587.)
return _VF.meshgrid(tensors, **kwargs) # type: ignore[attr-defined]
--- anchor_info_path /home/zzx/CZB/Bench2DriveZoo/data/others/b2d_motion_anchor_infos_mode6.pkl
load checkpoint from path: /media/zzx/data/workspace/Bench2Drive-base/uniad_tiny_b2d.pth

Running the route
=== [Agent] -- Wallclock = 2024-08-14 09:21:34.430 -- System time = 0.000 -- Game time = 0.050 -- Ratio = 0.000x
/home/zzx/.local/lib/python3.8/site-packages/scipy/optimize/_minpack_py.py:178: RuntimeWarning: The iteration is not making good progress, as measured by the
improvement from the last ten iterations.
warnings.warn(msg, RuntimeWarning)
leaderboard/scripts/run_evaluation.sh:行 44: 21984 段错误 (核心已转储) CUDA_VISIBLE_DEVICES=${GPU_RANK} python ${LEADERBOARD_ROOT}/leaderboard/leaderboard_evaluator.py --routes=${ROUTES} --repetitions=${REPETITI=${CHALLENGE_TRACK_CODENAME} --checkpoint=${CHECKPOINT_ENDPOINT} --agent=${TEAM_AGENT} --agent-config=${TEAM_CONFIG} --debug=${DEBUG_CHALLENGE} --record=${RECORD_PATH} --resume=${RESUME} --port=${PORT} --traffic-manager-port=${TM_PORT} --gpu-rank=${GPU_RANK}这个问题是显存不足吗?还是其他原因

@AvinsWang
Copy link

@scuizhibin 从上面信息不确定是哪一步出错了, 你可以调试代码看下具体发生在哪. 3060(notebook) 8G报OOM错误(uniad_tiny), 换成 A100 实际占用显存约9.4G(VAD);

@scuizhibin
Copy link
Author

scuizhibin commented Aug 14, 2024 via email

@happytianhao
Copy link

我按照readme上的要求进行安装carla,运行bash leaderboard/scripts/run_evaluation_debug.sh,出现问题

在Bench2Drive/leaderboard/scripts/run_evaluation.sh中设置一下CARLA_ROOT~

@JesseTowne
Copy link

@scuizhibin 这是我使用的版本, 按照INSTALL.md安装能成功运行;你的问题我没遇到过, 不知道怎么解, 可以重装下torchvision或者按照官方重新配置环境

torch 2.4.0+cu118
torchaudio 2.4.0+cu118
torchmetrics 0.11.4
torchvision 0.19.0+cu118

@AvinsWang @scuizhibin 请问在CUDA 11.8的条件下,如何配置VAD的虚拟环境?我发现INSTALL.md并未提及VAD所需的MMCV、MMDetection、MMDetection3D等依赖库的版本号

@jayyoung0802
Copy link
Member

Hi, @JesseTowne. You do not need another repo(MMCV、MMDetection、MMDetection3D) anymore, just follow step 6, pip install -v -e .

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

6 participants