I want to profile my pytorch1.7 / tensorRT code in docker, to check GPU, DLA, TensorCore usage with DLProf. However, the DLProf installation fails.
The installation steps are as in: DLProf User Guide :: NVIDIA Deep Learning Frameworks Documentation
1.
pip install nvidia-pyindex
2.
pip install nvidia-dlprof
But step 2 fails.
By the way, I run my pytorch / tensorRT application in docker, should I install DLProf inside or outside docker?
my CUDA version is 10.2
The error message is:
ERROR: Cannot install nvidia-dlprof==0.15.0, nvidia-dlprof==0.16.0, nvidia-dlprof==0.17.0, nvidia-dlprof==0.18.0, nvidia-dlprof==0.19.0, nvidia-dlprof==1.0.0, nvidia-dlprof==1.1.0 and nvidia-dlprof==1.2.0 because these package versions have conflicting dependencies.
The conflict is caused by:
nvidia-dlprof 1.2.0 depends on nvidia-nsys-cli>=2020.4.1.117
nvidia-dlprof 1.1.0 depends on nvidia-nsys-cli==2021.2.1.58
nvidia-dlprof 1.0.0 depends on nvidia-nsys-cli>=2020.4.1.117
nvidia-dlprof 0.19.0 depends on nvidia-nsys-cli>=2020.4.1.117
nvidia-dlprof 0.18.0 depends on nvidia-nsys-cli>=2020.4.1.117
nvidia-dlprof 0.17.0 depends on nvidia-nsys-cli>=2020.4.1.117
nvidia-dlprof 0.16.0 depends on nvidia-nsys-cli==2020.4.1.117
nvidia-dlprof 0.15.0 depends on nvidia-nsys-cli==2020.4.1.117
To fix this you could try to:
- loosen the range of package versions you’ve specified
- remove package versions to allow pip attempt to solve the dependency conflict
ERROR: ResolutionImpossible: for help visit User Guide - pip documentation v22.3.dev0
Thanks in advance.