So maybe you want to try and see inference time using evaluate.py codebase then, which is closer to a intensive-inference workload.
@sonofhotmale,
Hi.
any news about Tensor RT ??
any improvments in speed, like 3 to 5X (as said in nvidia tensor RT pages)
Working on a TX2, and soon, on a Xavier, it could help me to accelerate my researchs.
Thanks a lot.
Vincent
@elpimous_robot,
Hi, apologies for the late reply. I have converted the deepspeech model into a .pbmm file for TensorRT. However, I cannot run it on the demo. I can use it for the evaluation script if that is of any use for you. Happy to share the file.
@sonofhotmale,
Hi thanks for answer.
What is your gpu board ? (mine is TX2 for now)
you seems to create a TensorRt model, not really optimized…
Did you see speedup, regarding to standard *.pbmm model ?
Thanks.
Vincent