Open khteh opened 2 years ago
@khteh Can I hear more about what you're trying to accomplish? Are you trying to create different versions of the model in different folders?
Hi @beccamc regardless of the versioning of the model, I was looking for the option to customize the path/location to put the model .zip
from the Model Builder in order for other projects in the solution (.sln
) to cosume it. It seems that the current behaviour always puts the .zip
together with the model builder config file....
I know that this is possible using the ML.Net API and I have tried it. This leads me to the following questions:
(1) What's the use case for the Model Builder? How can I use it for the end-to-end application development which involves ML/AI?
(2) Is the Model Builder only used in an isolated ML/AI step only for the purpose of preparing the data, ingest data, trying out different AI algorithms, train, evaluate and test the model?
(3) The Model Builder produces 3 artefacts:
(i) <foo>_training.cs
: It only contains 2 functions: RetrainPipeline
and BuildPipeline
. Where is the code for data preparation and ingestion (train data
, test/validation data
, etc)?
(ii) <foo>_consumption.cs
How can I easily resue the work from the Model Builder auto-generated code in the API use case?
We just released a VS Notebook Editor extension. If you have it installed we will generate a Notebook file, which gives a better step by step re-training flow. - https://marketplace.visualstudio.com/items?itemName=MLNET.notebook
Give it a try and let me know what you think!
Any tutorial for me to follow?
I am using VS 2022. Why is there no option to customize model training auto-generated files? For example I would like specific path to the auto-generated model files to be updated automatically during retrain.
I have tried to move
.mbconfig
to a different folder but to no avail. The.mbconfig
copies itself back to top-level directory of the project together with all the auto-generated files!