Friedrich Marty

Smorty100

https://gitlab.com/users/Marty_Friedrich/projects

AI & ML interests

I'm most interested in content rerouting between LLM and VLLM agens for automation possibilities. Using templates for each agent which is then filled in by another agents inputs seems really useful.

Recent Activity

replied to MonsterMMORPG's post 2 days ago

It is now possible to generate 16 Megapixel (4096x4096) raw images with SANA 4K model using under 8GB VRAM, 4 Megapixel (2048x2048) images using under 6GB VRAM, and 1 Megapixel (1024x1024) images using under 4GB VRAM thanks to new optimizations 13 January 2024 Update Installers : https://www.patreon.com/posts/from-nvidia-labs-116474081 New 4K Tutorial Video : https://youtu.be/GjENQfHF4W8 Now the APP will use Diffusers Pipeline and it has huge VRAM optimizations You need to reinstall The models will be downloaded into your Hugging Face cache folder when you first time generate something How to Get Installation Logs and How to Change Hugging Face Cache Folder : https://www.patreon.com/posts/108419878 Please make a fresh install When you enable all 4 optimizations the VRAM usages are like below Make sure shared VRAM is enabled because initial loading of the model need more VRAM Enable VAE Tiling + Enable VAE Slicing + Enable Model CPU Offload + Enable Sequential CPU Offload 1K (1024x1024) : 4 GB GPUs 2K (2048x2048) : 6 GB GPUs 4K (4096x4096) : 8 GB GPUs Still in any case may work on your GPU test it Just Enable VAE Tiling + Enable Model CPU Offload works great in many cases All below attached images are generated via SANA 4K model, they are RAW and their resolution is 5376x3072 Official repo page : https://github.com/NVlabs/Sana

liked a model 4 days ago

PRIME-RL/Eurus-2-7B-PRIME

reacted to Severian's post with 👍 5 days ago

Interesting Solution to the Problem of Misguided Attention So I've been fascinated by the problem of Misguided Attention for a few weeks. I am trying to build an inference algorithm to help LLMs address that issue; but in the process, I found a cool short-term fix I call "Mindful Attention" using just prompt-engineering. Have you ever thought about how our brains filter reality through layers of past experiences, concepts, and mental images? For example, when you look at an oak tree, are you truly seeing that oak tree in all its unique details, or are you overlaying it with a generalized idea of "oak tree"? This phenomenon inspired the new approach. LLMs often fall into a similar trap, hence the Misguided Attention problem. They process input not as it’s uniquely presented but through patterns and templates they’ve seen before. This leads to responses that can feel "off," like missing the point of a carefully crafted prompt or defaulting to familiar but irrelevant solutions. I wanted to address this head-on by encouraging LLMs to slow down, focus, and engage directly with the input—free of assumptions. This is the core of the Mindful Attention Directive, a prompt designed to steer models away from over-generalization and back into the moment. You can read more about the broader issue here: https://github.com/cpldcpu/MisguidedAttention And if you want to try this mindful approach in action, check out the LLM I’ve set up for testing: https://hf.co/chat/assistant/677e7ebcb0f26b87340f032e. It works about 80% of the time to counteract these issues, and the results are pretty cool. I'll add the Gist with the full prompt. I admit, it is quite verbose but it's the most effective one I have landed on yet. I am working on a smaller version that can be appended to any System Prompt to harness the Mindful Attention. Feel free to experiment to find a better version for the community! Here is the Gist: https://gist.github.com/severian42/6dd96a94e546a38642278aeb4537cfb3

View all activity

Organizations

None yet

Smorty100's activity

replied to MonsterMMORPG's post 2 days ago

woah, some of these demo images look actually gud... i kinda lost hope for image diffusers there for a minute, but this is impressive. the one with the leaves cought me offguard.

And on such smol GPUs now too? that is super cool!

liked a model 4 days ago

PRIME-RL/Eurus-2-7B-PRIME

Text Generation • Updated 1 day ago • 995 • 53

reacted to Severian's post with 👍👀 5 days ago

Post

3750

Interesting Solution to the Problem of Misguided Attention

So I've been fascinated by the problem of Misguided Attention for a few weeks. I am trying to build an inference algorithm to help LLMs address that issue; but in the process, I found a cool short-term fix I call "Mindful Attention" using just prompt-engineering.

Have you ever thought about how our brains filter reality through layers of past experiences, concepts, and mental images? For example, when you look at an oak tree, are you truly seeing that oak tree in all its unique details, or are you overlaying it with a generalized idea of "oak tree"? This phenomenon inspired the new approach.

LLMs often fall into a similar trap, hence the Misguided Attention problem. They process input not as it’s uniquely presented but through patterns and templates they’ve seen before. This leads to responses that can feel "off," like missing the point of a carefully crafted prompt or defaulting to familiar but irrelevant solutions.

I wanted to address this head-on by encouraging LLMs to slow down, focus, and engage directly with the input—free of assumptions. This is the core of the Mindful Attention Directive, a prompt designed to steer models away from over-generalization and back into the moment.

You can read more about the broader issue here: https://github.com/cpldcpu/MisguidedAttention

And if you want to try this mindful approach in action, check out the LLM I’ve set up for testing: https://hf.co/chat/assistant/677e7ebcb0f26b87340f032e. It works about 80% of the time to counteract these issues, and the results are pretty cool.

I'll add the Gist with the full prompt. I admit, it is quite verbose but it's the most effective one I have landed on yet. I am working on a smaller version that can be appended to any System Prompt to harness the Mindful Attention. Feel free to experiment to find a better version for the community!

Here is the Gist: https://gist.github.com/severian42/6dd96a94e546a38642278aeb4537cfb3

reacted to Nitral-AI's post with 😔 6 days ago

Post

3286

That moment when you spend 5 days up babysitting trains, only for colab pro + to randomly disconnect the environment at every chance with 0 error indication of any kind (it just disconnects without an error). Nuke the session from the interface, but continue to eat my colab credits while it reports to wandb. 0 way of saving the models when this happens since it nukes the code preset up to auto-execute. And since the sessions 'exist' but also at the same time doesn't exist i cant close it. And have to wait till they auto timeout after 24hrs. Guess, i won't be using colab for 'quick' test trains anymore. Thanks google for scheming the very little model training budget i had for the month.