Anthony

[Original comment removed, see below]

The silent failing of OpenELM 3B seems to be caused by commit [https://github.com/ml-explore/mlx-swift-examples/commit/ff7610d522a57e403140e8431ebd6d032dcdb86c](https://github.com/ml-explore/mlx-swift-examples/commit/ff7610d522a57e403140e8431ebd6d032dcdb86c). It doesn't happen when checking out the commit before that.

I can test it on my iPad, if you let me know what package/commit I need to import.

Using the chat templates with Llama 3.2 will result in an error until `none` ~~and `tojson` are~~ is implemented in the Jinja package: https://github.com/maiqingqiang/Jinja/pull/4

> OK, should we hold off on taking this until that is merged (and we can pick it up here)? Yes, since LLMEval includes model configs for Llama 3.2. In...

~~We had a similar issue with an EOS token at the end of the output of Gemma 2: https://github.com/ml-explore/mlx-swift-examples/issues/89~~ ~~This shouldn't be appearing in the output if it's included in...

The problem with Mistral 7B is resolved here: https://github.com/huggingface/swift-transformers/pull/134

Llama 3.2 1B and 3B and Mistral 7B now work with chat templates, thanks to the changes in swift-transformers and Jinja.

Is there a reason the default version hasn't been updated? I've been fighting Xcode for an hour, just trying to get my app to build with the various dependencies.

Currently, when I try to run Phi 3.5 mini in the example app with [this branch](https://github.com/ml-explore/mlx-swift-examples/pull/107), I get the following error: ``` MLX error: Shapes (14) and (48) cannot be...