If the benchmark.py --model argument is used to benchmark a single model, and the --results-file argument is also used, then the results file will not be created. There is no warning.
I expected the results file to be created for the model, just as it would if I used --mode-list instead of --model.
https://github.com/rwightman/pytorch-image-models/blob/e8ddc6865c2c74871e6953e8d42873829768ba14/benchmark.py#L647