When you do a

bentoml containerize pytorch_mnist_demo:latest

, for example, how does BentoML know which model from the

~/bentoml/models

directory to copy into the container? For example, suppose the following models exist, how does

bentoml containerize pytorch_mnist_demo:latest

know to use the

pytorch_mnist

model, and how does it know which version to use, since in this example, there are two versions of the

pytorch_mnist

model?

$ bentoml models list
 Tag                             Module           Size      Creation Time       
 iris_clf:yrgcaphel2f4tlg6       bentoml.sklearn  5.98 KiB  2024-03-17 09:03:36 
 pytorch_mnist:7sm2kgxekkguplg6  bentoml.pytorch  1.30 MiB  2024-03-17 07:39:17 
 pytorch_mnist:aq7jcehd66tj3lg6  bentoml.pytorch  1.30 MiB  2024-03-16 20:40:56

it takes latest created ones based on the timestamp

Thank you, Jian. So it's strictly based on the timestamp and not what's contained in the

latest

file, correct?

$ cat ~/bentoml/models/pytorch_mnist/latest
7sm2kgxekkguplg6

Also, is that configurable? That is, can you specify a specific model to use in the

bentoml containerize

command? Just trying to understand how things work. Thanks.

if you want to specify a model you should use the

tag

!

Thank you.