[QUESTION] Adding trees manually on txt file about lightgbm HOT 1 OPEN

Thanks for using LightGBM.

I've reformatted your post a bit to make the difference between code, output from code, and your own words a bit clearer. If you're new to GitHub and/or writing markdown, please see https://docs.github.com/en/get-started/writing-on-github/getting-started-with-writing-and-formatting-on-github/basic-writing-and-formatting-syntax#quoting-code to learn about how I did that.

I don't immediately see what's wrong with your model file but in general... don't manually merge files like this. There are just too many ways for it to go wrong (mixed line endings, accidental inclusion of special characters, failure to update all parameters correctly, etc.).

If it's acceptable to not be training both models at the same time, do this:

1. train the first model
2. save it to a text file with .save_model()
3. move that file over to the other machine
4. train another model on that machine, with new data, passing the path to the first model file to init_model

That will cause LightGBM to begin boosting from the end of the first model, using your new dataset. It's not exactly the same as the process you described (where you train two models from scratch and then combine them), but it's a good way to incorporate learnings from new data into an already-training model.

Like this pattern:

There are some more details on how to do this at https://stackoverflow.com/questions/73664093/lightgbm-train-vs-update-vs-refit/73669068#73669068

There is an API in LightGBM's C API, LGBM_BoosterMerge(), that I believe can be used to take 2 models and combine them together without training, but it's not currently exposed in the Python package.

We could consider adding that in the Python package.

from lightgbm.