RuntimeError: Found 0 images in subfolders of /tmp_data/classification/train.

[alphaProgrammingNow]

This is my command:

./distributed_train.sh 1 --data-dir /tmp_data/classification/ --train-split train --val-split val --model efficientnet_b2 -b 128 --sched step --epochs 450 --decay-epochs 2.4 --decay-rate .97 --opt rmsproptf --opt-eps .001 -j 8 --warmup-lr 1e-6 --weight-decay 1e-5 --drop 0.3 --drop-path 0.2 --model-ema --model-ema-decay 0.9999 --aa rand-m9-mstd0.5 --remode pixel --reprob 0.2 --amp --lr .016

The command is from https://huggingface.co/docs/timm/training_script.
I am sure there are five hundred imgs in /tmp_data/classification/train. But It just can't find.

[rwightman]

@alphaProgrammingNow they need to be in sub-folders that match the class of each sample in order to use the default folder dataset

[rwightman]

ie /tmp_data/classification/class_0/img0.jpg, /tmp_data/classification/class_1/img0.jpg .. etc

[alphaProgrammingNow]

Thank you very much! I solve the problem according to your method.

[ClayShoaf]

@rwightman So, if I am not using --class-map, should train.py and inference.py automatically create labels based on the directory names?

In your example, should it make the classes: ["class_0", "class_1", ...etc]?