Running on cluster

[alib-l]

Hello,
I want to use deepmd on a cluster. It seems "dp train" uses all cores which is not possible in a HPC.
How can I specify the number of cores for training?

[njzjz]

Use environmental variables: OMP_NUM_THREADS, TF_INTRA_OP_PARALLELISM_THREADS and TF_INTER_OP_PARALLELISM_THREADS.
See https://github.com/deepmodeling/deepmd-kit/blob/v1.3.3/doc/use-deepmd-kit.md#training.

[alib-l]

Thank you so much