Antheas Kapenekakis

Now that my project is getting big not anymore. What's even worse, however, is having half of the nodes with the pretty print not working. It's one of the reasons...

Hi Luke, I think the reason it is this way is because COPR has license requirements that the firmware does not meet, so it can't be made as a COPR...

I am having the same issue. `gamescope -- sh -c "mangoapp& vkcube"` runs fine but mangohud is not launching on wayland or gamescope with steam using a similar script to...

The crash happens when attempting to create the atom after a failed initialization. For some reason, the initialization does not return false. https://github.com/flightlessmango/MangoHud/blob/1baecfc493ecc32c9c2b14c025769485eb30de20/src/app/main.cpp#L290-L305 ``` systemd-coredump[31893]: [🡕] Process 31874 (mangoapp) of...

I only have 40GB of RAM. So the default code did not work for me for 7B. By changing the first lines to this, RAM is limited to 17GB and...

with `device_map=auto` I get the following map. The last 4 layers don't fit. ```python model.hf_device_map = {'gpt_neox.embed_in': 0, 'gpt_neox.layers.0': 0, 'gpt_neox.layers.1': 0, 'gpt_neox.layers.2': 0, 'gpt_neox.layers.3': 0, 'gpt_neox.layers.4': 0, 'gpt_neox.layers.5': 0,...

So I came up with the following, to use 8 bit quantization @cduk: ```python import torch from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig, StoppingCriteria, StoppingCriteriaList quantization_config = BitsAndBytesConfig( load_in_8bit=True, llm_int8_threshold=6.0, )...

> > However, with 4 prompts run a bit faster (?) > > Isn't this expected? The lower the threshold, the more weights are converted to int8 (hence less compute...

I just commented out the check. As this part of the code does not have the reach to know what is the architecture of each package, it can not be...

Actually I take that back, part of the error is the nevra. So potentially a solution is refactoring the check to cut out the version and keep name, architecture. Just...