search

SthPhoenix / InsightFace-REST

InsightFace REST API for easy deployment of face recognition services with TensorRT in Docker.
Apache License 2.0
501 stars 117 forks source link

Need for config.pbtxt models for triton server #94

Closed zerodwide closed 2 years ago

zerodwide commented 2 years ago

Hi I followed guide on TRT model structure at #60 exact comment

but when running triton server with this command: docker run --gpus=1 --rm -p8000:8000 -p8001:8001 -p8002:8002 -v/model/triton:/models nvcr.io/nvidia/tritonserver:22.03-py3 tritonserver --model-repository=/models

this error come out:

... model_repository_manager.cc:1927] Poll failed for model directory 'arcface_r100_v1': failed to open text file for read /models/arcface_r100_v1/config.pbtxt: No such file or directory

Can you provide config.pbtxt for arcface_r100_v1, scrfd_10g_gnkps or retinaface_r50_v1 ?

Thanks in Advance.

SthPhoenix commented 2 years ago

Hi you can run Triton server with --strict-model-config=false argument, it should generate config automatically.

Though I'm not sure if this project still works with Triton, I haven't checked it for a while since current version works much faster with TRT backend.

zerodwide commented 2 years ago

@SthPhoenix Thanks for quick response.

I have check it out. still it works but as you mentioned TRT is faster.

image: tritonserver:22.03-py3 retinaface_r50_v1 (OK) scrfd_10g_gnkps (FAILED) arcface_r100_v1 (OK)