Closed sjg02122 closed 2 years ago
Can you try using our provided dockerfile, so it's the same configuration?
I use the conda environments Is it impossible on conda environments?
I guess it is, if you set up all required packages inside the conda environment.
The model works well on my server
and I try to run the model using multi-gpu (horovorun -np # ~~)
Even though I never change anything, I face the problem GPU out of memory (It was working well only single gpu)
How to can I fix this problem?