Upload to Main

scripts/dapt/dp_few_shot.sh

@@ -0,0 +1,38 @@
+#!/bin/bash
+
+#cd ../..
+
+# custom config
+DATA="/home/ubuntu/Data_file/few_shot_data"   #Your data path
+TRAINER=MaPLe
+
+DATASET=$1    #10 dataset from, like "stanford_cars"
+CFG=vit_b16_t
+MODE=dapt-g
+
+#seed = ("0" "1" "2")
+#method "Uniform"    "Uncertainty" "Herding" "Submodular" "Glister" "GraNd" "Craig" "Cal"    "Forgetting"
+#sample_rate = 0.05 0.1 0.2 0.3 0.5 1.0  #"Uncertainty" "Herding"
+# Normally method == "Uniform" and sample_rate = "1.0", which means no data selection process
+
+for seed in 1 2 3; do
+  for rate in 1.0; do   #
+    for shot in 1 2 4 8 16; do
+      for method in "Uniform"; do
+        echo "Running with seed =${seed} and sample rate=${rate} and Method=${method}"
+        CUDA_VISIBLE_DEVICES=1 python train.py \
+        --root ${DATA} \
+        --seed ${seed} \
+        --trainer ${TRAINER} \
+        --dataset-config-file configs/datasets/${DATASET}.yaml \
+        --config-file configs/trainers/${TRAINER}/${CFG}.yaml \
+        --output-dir output_few \
+        --mode ${MODE} \
+        DATASET.NUM_SHOTS ${shot} \
+        DATASET.SUBSAMPLE_CLASSES all \
+        DATASET.SELECTION_RATIO ${rate} \
+        DATASET.SELECTION_METHOD ${method}
+        done
+    done
+  done
+done