First of all thank you who those who read it
I have two machines. And both machines have DIGITS localhost. But right now i want to create a DIGITS dataset backup and model to another machine without retraining them as i have alot that i need to backup. Is it to possible to do it? Or is there any other way to do it. Hope to receive help. Thanks.
On the machine where you already have trained a model that you would like to save, on the DIGITS model creation page, after training is finished, towards the bottom of the page you should see a green button labelled “Make Pretrained Model”.
Click this button to start the process of saving a pretrained model. Each time you click this button, it will spin up a new job (which will normally finish quite quickly) to create the pretrained model. This new model will appear in the digits/digts/jobs directory in a subdirectory identified by the jobname. Since this process is usually pretty quick, you can probably identify the new job directory just be see what new one appeared after you click the pretrained model, or else use the file creation dates to determine.
As an alternative to the above, you can just click on the blue button “Download Model” next to the green button. This will download the model corresponding to the currently selected epoch of training to your local/client machine. From here you can upload this saved file to the other machine, extract the archive contents, and use that.
On the machine where you would like to load the pretrained model, you can select a pretrained model to load during the model creation process. Load the model you had previously saved on the other machine. You will need to transfer or otherwise make that model file available to the machine where you would like to load it.