#!/bin/bash PHASE2_DATA_DPATH=/flash/smart_data_dvc PHASE2_EXPT_DPATH=$HOME/data/dvc-repos/smart_expt_dvc WORKDIR=$PHASE2_EXPT_DPATH/training/$HOSTNAME/$USER # DATASET_CODE=Drop5 # DATASET_CODE=Aligned-Drop4-2022-08-08-TA1-S2-WV-PD-ACC DATASET_CODE=Aligned-Drop4-2022-08-08-TA1-S2-L8-ACC KWCOCO_BUNDLE_DPATH=$PHASE2_DATA_DPATH/$DATASET_CODE TRAIN_FPATH=$KWCOCO_BUNDLE_DPATH/data_train.kwcoco.json VALI_FPATH=$KWCOCO_BUNDLE_DPATH/data_vali.kwcoco.json TEST_FPATH=$KWCOCO_BUNDLE_DPATH/data_vali.kwcoco.json EXPERIMENT_NAME=Drop4_SC_Heterogeneous_Multi DEFAULT_ROOT_DIR=$WORKDIR/$DATASET_CODE/runs/$EXPERIMENT_NAME python -m geowatch.tasks.fusion fit \ --config=config_common.yaml \ --config=config_sc_native.yaml \ --model.init_args.name=$EXPERIMENT_NAME \ --data.train_dataset="$TRAIN_FPATH" \ --data.vali_dataset="$VALI_FPATH" \ --data.num_workers=4 \ --trainer.default_root_dir="$DEFAULT_ROOT_DIR" \ --trainer.accelerator="gpu" \ --trainer.devices=1 \ --trainer.precision=16 \ --trainer.max_steps=200000 # --trainer.detect_anomaly=true \