mediapipe RET_CHECK failure display != EGL_NO_DISPLAYeglGetDisplay()

[mixiazhiyang]

I0000 00:00:1708958024.457143 2343 task_runner.cc:85] GPU suport is not available: INTERNAL: ; RET_CHECK failure (mediapipe/gpu/gl_context_egl.cc:77) display != EGL_NO_DISPLAYeglGetDisplay() returned error 0x300c
I0000 00:00:1708958025.256465 2343 task_runner.cc:85] GPU suport is not available: INTERNAL: ; RET_CHECK failure (mediapipe/gpu/gl_context_egl.cc:77) display != EGL_NO_DISPLAYeglGetDisplay() returned error 0x300c
W0000 00:00:1708958025.257329 2343 face_landmarker_graph.cc:174] Sets FaceBlendshapesGraph acceleration to xnnpack by default.
batch 0 | loss_lan: 0.40, loss_reg_id: 0.27,loss_reg_exp: 0.97
| To extract pose from static
{'src_image_name': 'infer_out/tmp/cropped_src_img.png', 'drv_audio_name': '/tmp/gradio/89f924df8f6ba2a90f4ff1ad4f4f3ec670ba241f/audio.wav', 'drv_pose_name': 'static', 'bg_image_name': '/tmp/gradio/77e40a84169eb20cd5632ed343840895826c8a11/bg.png', 'blink_mode': 'period', 'temperature': 0.2, 'mouth_amp': 0.45, 'out_mode': 'concat_debug', 'map_to_init_pose': True, 'low_memory_usage': False, 'hold_eye_opened': False, 'a2m_ckpt': '/hpc2hdd/home/yyang937/digital_human/video/Real3DPortrait/checkpoints/240210_real3dportrait_orig/audio2secc_vae/model_ckpt_steps_400000.ckpt', 'head_ckpt': None, 'torso_ckpt': '/hpc2hdd/home/yyang937/digital_human/video/Real3DPortrait/checkpoints/240210_real3dportrait_orig/secc2plane_torso_orig/model_ckpt_steps_100000.ckpt', 'min_face_area_percent': 0.2, 'out_name': '', 'seed': 42}
| audio-to-motion finished
rendering drv secc: 0%| | 0/5 [00:00<?, ?it/s]create rasterizer on device cuda:0
rendering drv secc: 0%| | 0/5 [00:00<?, ?it/s]
Inference ERROR: Not compiled with GPU support?

I installed everything correctly, including mediapipe. I tried bazel run --copt -DMESA_EGL_NO_X11_HEADERS --copt -DEGL_NO_X11 \ mediapipe/examples/desktop/hello_world:hello_world and it works under cloned mediapipe folder.
But when I run python inference/app_real3dportrait.py , it will raise errors above.

uname -a Linux 4c2a87d133e3 4.18.0-305.el8.x86_64 #1 SMP Thu Apr 29 08:54:30 EDT 2021 x86_64 x86_64 x86_64 GNU/Linux
I am using HPC, so I can only access the docker if I want to get the computation node. I cannot docker run to get nvidia docker.
I find some solutions here, but they are not working.
google-ai-edge/mediapipe#1651
google-ai-edge/mediapipe#5120
https://stanfordvl.github.io/iGibson/issues.html
google-ai-edge/mediapipe#4091
google-ai-edge/mediapipe#2458
google-ai-edge/mediapipe#2693

[DYJNG]

same question