Awni Hannun
Awni Hannun
Check out the [getting started guide](https://ml-explore.github.io/mlx/build/html/usage/distributed.html#getting-started) for mx.distributed. Make sure you can run that simple example. If it doesn't work there are some tips on setting up MPI there that...
Could you share more details about the issue you are seeing?
Maybe try updating MLX? That line should work in a recent version.
Could you try running `python -c "import mlx_lm; print(mlx_lm.__version__)"` If that doesn't work `mlx_lm` didn't install where you think it installed and it's likely an environment issue.
We have Mamba in MLX LM already and there is a PR for Mamba 2 (#1009 ). As for the others, it would be helpful if you could point to...
Can you provide the full code (or some code that we can run) that reproduces the issue you are seeing? Without that it's pretty hard to say if this is...
I don't know of anyone working on it.. but I also haven't seen much demand for this fwiw. If you are not sure you could leave the issue open and...
Yes I think it's better this way. I'm debating if we should bother with the `compute_eigenvectors` flag as opposed to just having the primitive always populate both outputs and then...
That's interesting. It seems like an issue on the Metal side.. I wonder if there is a limit to how many command queues you can make 🤔
I'm going to leave this open for now. There is definitely a couple metal issues at play here: - The strange log: `Context leak detected, msgtracer returned -1` is a...