Oliver Simons

Same here. It starts to error for all releases `>=2.0.0` (to be fair I didn't test older than [1.26.7](https://hub.docker.com/layers/aklitzing/ausweisapp2/1.26.7/images/sha256-dde262fe9e08cc70e9d450b0c2b00e52e33bd71184468013ec28120be5b4d95d?context=explore). Specifically, 2.0.0 fails with ``` qt.qpa.plugin: Could not find the Qt...

Any progress on this? it currently prevents us from resuming training with PyTorch 2.6 and ~lightning 2.5~ lightning 2.3

> Wanted to check this issue as well. My existing workflow of trainer.fit(..., ckpt_path=) is also currently blocked because of the default behavior change in torch 2.6 Curious but what's...

I observe this issue now as well on a p8 pro. For me, it seems to have been introduced in a recent update

> Secondly fusion is a common source of bugs in the cuda backend, I don't want to add another layer of complexity on top of it. This points to a...

> Did you pass the env flag? Why of course not 🫠 Updated numbers ``` + ./scripts/compare-llama-bench.py -b b8595b16e69e3029e06be3b8f6635f9812b2bc3f -c d3a8d93a04d4cb8cc8daa62d12e6763a2f4c0221 --tool llama-bench -i llama-bench.sqlite ``` | Model | Test...

> My problem is that the buffer gets reused in this case. If my current understanding of ggml is correct, we should be able to get the same behavior (fusion...