https://github.com/nixtla logo
#neural-forecast
Specify device
# neural-forecast
g

GoostValley

09/20/2023, 11:59 AM
Thanks for the great framework you implemented. Here's a small issue. I want to run the program on one GPU while the program calls all GPUs. I tried to set
accelerator="cuda:0"
but errors were raised. Any solutions except setting "CUDA_VISIBLE_DIVICE" in the command line?
c

Cristian (Nixtla)

09/21/2023, 4:02 PM
Hi! We also use the CUDA_VISIBLE_DIVICE flag before the python script, is the easiest option
3 Views