Chris Taylor

https://github.com/andralex/MedianOfNinthers/blob/48f696fbbbe56ad0205acbe452359d9728d84071/src/common.h#L357

From here: https://github.com/catid/cm256/pull/3

Investigate Frobenius Additive FFT

3

I believe this can be a drop-in replacement for the FFT that is currently being used: https://arxiv.org/pdf/1802.03932.pdf More background: http://www.texmacs.org/joris/ffft/ffft.pdf

I had one contact who may have found this useful on cellphones so I should port it to ARM NEON

Looks like if you put the headset down for a while and then put it back on, the recording is stuck on the last screen.

This app should really have built-in recording. After setting up calibration, I had to move the cellphone to access the control panel to enable recording via iOS. And then had...

Multiple GPU performance worse than single GPU

2

``` ns-train nerfacto --machine.num-gpus 2 --vis viewer ``` The output indicates that training is slower with more GPUs. Using two GPUs, I see about 60% CPU utilization, and three GPUs...

Incompatible with latest Halide

3

I'd love to use this project but it seems to be incompatible with the latest Halide source code

Torchrun distributed running does not work

3

Running in a distributed manner either returns an error, or with the simplest example, produce obviously incorrect output. The following is the result of running 13B model across two nodes....

Add support for latest gymnasium

1

The examples now work properly.