KNN weight modeling on Omnisci

• Launch Omnisci using: https://github.com/cga-harvard/GIS_Apps_on_HPC/wiki/Omnisci-on-FASRC or https://www.youtube.com/watch?v=TvqqikT_V58

Location to map omnisci-storage: /n/holyscratch01/$lab/$USER/$SLURM_JOB_ID
Number of core: 2
Number of GPUs: 1
Memory allocated: 256 GB
Time: 168:00:00 (8 hours)
Partition: fas_gpu 

• Connect to Immerse by providing in the terminal of your PC the tunnel parameters which appear on the connection screen for e.g:

ssh -NL 7301:holygpu2c0701.rc.fas.harvard.edu:7301 [email protected]

Provide license to activate Omnisci Immerse and provide password: HyperInteractive

• Copy KNN_modelling script from /n/holyscratch01/enos_lab/dkakkar/partisan/knn_model.py
• Edit the script to change Omnisci port and location of your results files (.gz). The results file should be in following schema:

'source_id','neigbor_id','dist','dpost','rpost'

The omnisci port here is the backend port by clicking on your Session ID and then output.log file and find BACKEND TCP PORT.

• ssh to omnisci node (find node from your vdi connection string):

ssh holygpu2c0701.rc.fas.harvard.edu 

• Run the script

module load Anaconda3/5.0.1-fasrc02

source activate omnisci

python3 knn_model.py

• The output CSV file is in the same folder as the script, the name is same as input file

Refer here: https://github.com/cga-harvard/GIS_Apps_on_HPC/blob/master/Using%20FASRC%20Geospatial%20Resources.pdf