Refinedet Tensorrt prediction fails about deepdetect

[beniz]

Hi, can you try outside docker ? Thanks.

[YaYaB]

Honestly, it is kind of pain to install everything in a host as:

• it takes time
• it adds a lot of unecessary things in my host

I would rather avoid this except if it is absolutely needed.
Note that I gave you everything to test it in on your side.

[fantes]

Hi there. I cannot reproduce using fresh compiled master. @YaYaB could you try lower "maxWorkspaceSize" to 2000 ?

[YaYaB]

@fantes Yes I get a core dump as well with "maxWorkspaceSize" to 2000

[fantes]

thanks @YaYaB . Unfortunately, I cannot reproduce either using docker images, using your command lines (thanks for giving them , it helps a lot) . Maybe your BASE_PATH contains bad TRTengine_bs1 or bad net_tensorRT.proto ?

[YaYaB]

I make sure the directory does not contain any of those before creating the serving and launching the prediction. I'll try it in another machine to see If I get the same result

[YaYaB]

@fantes Ok my bad. After testing it in an another machine I found that I was using it with a Titan X (a bit older than 1080Ti). When I use a 1080Ti it works correctly. However it seems weird because I use classification models (and even ssd) with trt backend with Titan X. it may be the tensorrt_oss that makes the refinedet core dump. Is that possible?