Zach Deane-Mayer

all the breaking tests are running on r-devel. They all seem to use `expect_warning`. It seems that predict.lm used to say `prediction from a rank-deficient fit may be misleading` but...

Closing in favor of https://github.com/zachmayer/caretEnsemble/pull/260

@mewo2 This commit looks good to me, you should merge it! @ihar in the meantime, maybe check out [caretEnsemble](https://github.com/zachmayer/caretEnsemble) which is my package for solving similar problems (using the [caret](https://github.com/topepo/caret)...

Good to hear! I'm not volunteering to do so yet, but was just curious. I'll make a fork if I have any ideas.

@mht-sharma @IlyasMoutawwakil the optimizer is definitely doing something to the model. I tested it on the [vikp/surya_layout](https://huggingface.co/vikp/surya_layout) segformer, where I exported the original model to onnx, optimized it, and then...

I tried 3 ways of handling the lists: - Sum - Max - Replace the list with 0 (and let microsoft's optimizer infer based on the graph) Sum/Max yield pretty...

@mht-sharma @IlyasMoutawwakil what do you think? The tests pass when I run them locally, and the optimizer seems to be able to reduce the size of the model a lot....

huh. I also tried using 0, which the docs said would infer the number of heads based on the model graph. I changed from `sum` to `max` and the values...

@IlyasMoutawwakil I changed the normalized config to use 0 instead of the max. How's it look now?

Awesome! No worries and thank you!