ffmpeg_trt Error Code 3 Parameter check failed

[efschu]

Hi,

built ffmpeg_trt with

compute_75,code=sm_75

which is for Turing.

Then, when I run

ffmpeg -hwaccel cuda -hwaccel_output_format cuda -i input.mkv -vf format_cuda=rgbpf32,tensorrt=model_RealESRGAN_x2plus_2080ti.engine,format_cuda=nv12,hwdownload,format=nv12 -c:v h264 out.mkv

I get following error:

[tensorrt @ 0x55e138eaddc0] Load trt engine
[tensorrt @ 0x55e138eaddc0] Loaded engine size: 115 MiB
[tensorrt @ 0x55e138eaddc0] [MemUsageChange] TensorRT-managed allocation in engine deserialization: CPU +0, GPU +114, now: CPU 0, GPU 114 (MiB)
[tensorrt @ 0x55e138eaddc0] 3: [runtime.cpp::~Runtime::346] Error Code 3: API Usage Error (Parameter check failed at: runtime/rt/runtime.cpp::~Runtime::346, condition: mEngineCounter.use_count() == 1. Destroying a runtime before destroying deserialized engines created by the runtime leads to undefined behavior.
)
[tensorrt @ 0x55e138eaddc0] [MemUsageChange] TensorRT-managed allocation in IExecutionContext creation: CPU +1, GPU +2026, now: CPU 1, GPU 2140 (MiB)

Same happens for

compute_61,code=sm_61

on Pascal.

Thought it may be related to the RealESRGAN_x2plus engine but same error happens on converted realesr-animevideov3 engine

With vapoursynth the engines runs "as expected"

[styler00dollar]

Have you tried building the engine in the docker itself? Not loading the engine from the vapoursynth docker.

[efschu]

Yes, I have tried that first. Makes no difference.