Howard Liberty

I can now successfully merge the weights. Thanks!

Noticed that the weights are in pickle format not safetensors. I created this PR to default to safetensors https://github.com/huggingface/accelerate/pull/2853

The FA provided by https://github.com/Dao-AILab/flash-attention has only MI200 or MI300 GPUs. With Trition 3.0, the kernel can work on a much broad range of AMD GPUs. Tested kernels on AMD...

Unfortunately AnyMemo is using legacy API to access file system, which is not compatible with Android scoped storage anymore. I just don't have time to look into this issue now.

The zxing library does not support that. May need to seek a different QR recognization library.

Hmmm, very weird issue. Thanks for the repo steps. I will see if I can repo.

What is the triton version? I used Triton 3.0.0 and successfully run all tests on AMD 7900 on Rocm 6.2. @DocShotgun

The translation is quite incomplete though.

Thanks for the suggestion. Yes, I like weblate as an OSS translation platform. I did an experimental setup: https://hosted.weblate.org/projects/freeotpplus/freeotpplus-android-app/

Same. Rocm 3.5 broke davinci resolve on CentOS 7. The davinci resolve launches fine but all video previews are blank but audio works fine. Downgrading to 3.3 and everything works.