search

geodesymiami / MiNoPy

MIami NOn linear phase linking in PYthon
Other
19 stars 5 forks source link

Homogenization of MinoPy and MintPy compute options ? #35

Closed falkamelung closed 2 years ago

falkamelung commented 3 years ago

Currently we have different options. Can that be homogenized, either by changing MintPy or MinoPy? E.g. MintPy has auto for 4.

########## 1. parallel job setting ##################################
MINOPY.compute.num_workers              = auto         # If mpi is used: auto for 1
MINOPY.compute.num_nodes                = auto         # auto for 1
MINOPY.compute.job_memory               = auto         # auto for 20000 (20 G)
MINOPY.compute.job_walltime             = auto         # auto for 02:00:00
########## computing resource configuration
mintpy.compute.maxMemory = auto #[float > 0.0], auto for 4, max memory to allocate in GB
## parallel processing with dask
## currently apply to steps: invert_network, correct_topography
## cluster   = none to turn off the parallel computing
## numWorker = all  to use all locally available cores (for cluster = local only)
## config    = none to rollback to the default name (same as the cluster type; for cluster != local)
mintpy.compute.cluster   = local #[local / slurm / pbs / lsf / none], auto for none, cluster type
mintpy.compute.numWorker = 32 #[int > 1 / all], auto for 4 (local) or 40 (non-local), num of workers
mintpy.compute.config    = auto #[none / slurm / pbs / lsf ], auto for none (same as cluster), config name
mirzaees commented 3 years ago

@falkamelung, Minopy uses mpi for parallelization while mintpy is using dask I had named it minopy.parallel instead of minopy.compute to distinguish them but if it does not matter, we can use mintpy options for that except that some of them are of no use like cluster and config What is your opinion?