Audio2Face crash after second launching

AI & Data Science Digital Humans (closed) Audio2Face (closed)
1

HW: a laptop with 3070.
SW: audio2face 2022.1.1
how to reproduce:
start a2f and then do audio player
exit a2f
relaunch a2f, then it crash
logs attached.

kit_20220908_112635.log (484.5 KB)
[omni.audio2face.common] [INFO] Loading network: fullface_male_hi_v1
[TRT INFERENCE] [INFO] TensorRT Processor: Using CUDA device: [0] NVIDIA GeForce RTX 3070 Laptop GPU
[omni.audio2face.common] [INFO] Setting track: [/World/audio2face/Player] → [/World/audio2face/CoreFullface]
[omni.audio2face.common] [INFO] Waiting for Audio2Face Instance to be ready: /World/audio2face/CoreFullface
[omni.audio2face.common] [INFO] Waiting for Instance to be ready: /World/audio2face/CoreFullface
[omni.audio2face.common] [INFO] Waiting for Instance to be ready: /World/audio2face/CoreFullface – DONE
[omni.audio2face.common] [INFO] OG Connection ADDED: [/World/audio2face/CoreFullface] → [/World/audio2face/l_eye_grp_hi_WritePrim]
[omni.audio2face.common] [INFO] OG Connection ADDED: [/World/audio2face/CoreFullface] → [/World/audio2face/r_eye_grp_hi_WritePrim]
[omni.audio2face.common] [INFO] OG Connection ADDED: [/World/audio2face/CoreFullface] → [/World/audio2face/c_bottomDenture_grp_hi_WritePrim]
2022-09-08 03:28:34 [118,786ms] [Warning] [omni.hydra] Mesh /World/male/char_male_model_hi/c_headWatertight_hi update topology/point without updating normal, fallback to smooth normal.
2022-09-08 03:28:34 [118,848ms] [Warning] [omni.audio2face.common.scripts.utils] Unable to read TRT info: c:\users\terry\appdata\local\ov\pkg\audio2face-2022.1.1\exts\omni.audio2face.core_deps\deps\audio2face-nets\nets_fullface\fullface_male_hi_v1\trt_info.json ; Details: [Errno 2] No such file or directory: ‘c:\users\terry\appdata\local\ov\pkg\audio2face-2022.1.1\exts\omni.audio2face.core_deps\deps\audio2face-nets\nets_fullface\fullface_male_hi_v1\trt_info.json’
[omni.audio2face.common] [INFO] Preparing and loading TensorRT model for: c:\users\terry\appdata\local\ov\pkg\audio2face-2022.1.1\exts\omni.audio2face.core_deps\deps\audio2face-nets\nets_fullface\fullface_male_hi_v1
[omni.audio2face.common] [INFO] TensorRT model is found: C:\Users\terry\AppData\Local\ov\cache\Kit\103.1\641148c1\audio2face-nets\fullface_male_hi_v1\model_device0_fp32.trt
[Audio2Face] [INFO] Reading data file: c:\users\terry\appdata\local\ov\pkg\audio2face-2022.1.1\exts\omni.audio2face.core_deps\deps\audio2face-nets\nets_fullface\fullface_male_hi_v1\emo_db.bin
[Audio2Face] [INFO] Emotion Database: Using CUDA device: [0] NVIDIA GeForce RTX 3070 Laptop GPU
[Audio2Face] [INFO] Setting default emotion: (g1a_neutral, 0)
[Audio2Face] [INFO] Getting emotion vector for (g1a_neutral, 0)
[Audio2Face] [INFO] Reading data file: C:\Users\terry\AppData\Local\ov\cache\Kit\103.1\641148c1\audio2face-nets\fullface_male_hi_v1\model_device0_fp32.trt
[Audio2Face] [INFO] Inference Engine: Using CUDA device: [0] NVIDIA GeForce RTX 3070 Laptop GPU
[Audio2Face] [INFO] [TensorRT] [MemUsageChange] Init CUDA: CPU +313, GPU +0, now: CPU 12411, GPU 6746 (MiB)
[Audio2Face] [INFO] [TensorRT] Loaded engine size: 6 MiB
[Audio2Face] [INFO] [TensorRT] [MemUsageChange] Init cuDNN: CPU +136, GPU +58, now: CPU 12582, GPU 6812 (MiB)
[Audio2Face] [INFO] [TensorRT] [MemUsageChange] TensorRT-managed allocation in engine deserialization: CPU +0, GPU +6, now: CPU 0, GPU 6 (MiB)
[Audio2Face] [INFO] [TensorRT] [MemUsageChange] Init cuDNN: CPU +1, GPU +8, now: CPU 12581, GPU 6812 (MiB)
Could not load library cudnn_cnn_infer64_8.dll. Error code 1455
crash_2022-09-08_10-56-47_11908.txt (4.1 KB)

2

Hello @terryzhv83! Sorry to hear you are crashing, I am sending this over to the dev team now.
I am trying to reproduce this issue on my computer and I’m not able to. When you say “then do audio player”, what exactly are you doing? Are you just playing the demo audio clip then closing A2F?

3

Hi Wendy, thanks for the reply. Actually, I was writng a extension to do voice tts and asr, and using the streaming client to drive the avatar. So it is not simply as I mentioned earlier. But the theory is smiliar as playing the audio clip. So that is why you can’t reproduce the issue.
And when this issue happen, there are two ways to recover: 1, reboot my laptop, 2 exit the omniverse launcher and monitor, then kill the omniverse process in the task manager.
Terry

4

Gotcha. I am sending this information over to the dev team for help.