If you load a model once, then change the normalization of a template in the file without changing the name, and load the model again, the model will load the old template from pdf_cache. One way to solve this would be to hash the template file itself instead of just the file path.
If you load a model once, then change the normalization of a template in the file without changing the name, and load the model again, the model will load the old template from
pdf_cache. One way to solve this would be to hash the template file itself instead of just the file path.