My "silly" experiment.
ZeroWw 'SILLY' version. The original model has been quantized (fq8 version) and a percentage of it's tensors have been modified adding some noise.
Full colab: https://colab.research.google.com/drive/1a7seagBzu5l3k3FL4SFk0YJocl7nsDJw?usp=sharing
Fast colab: https://colab.research.google.com/drive/1SDD7ox21di_82Y9v68AUoy0PhkxwBVvN?usp=sharing
Original reddit post: https://www.reddit.com/r/LocalLLaMA/comments/1ec0s8p/i_made_a_silly_test/
I created a program to randomize the weights of a model. The program has 2 parameters: the percentage of weights to modify and the percentage of the original value to randmly apply to each weight.
At the end I check the resulting GGUF file for binary differences. In this example I set to modify 100% of the weights of Mistral 7b Instruct v0.3 by a maximum of 15% deviation.
Since the deviation is calculated on the F32 weights, when quantized to Q8_0 this changes. So, in the end I got a file that compared to the original has:
Bytes Difference percentage: 73.04%
Average value divergence: 2.98%
The cool thing is that chatting with the model I see no apparent difference and the model still works nicely as the original.
Since I am running everything on CPU, I could not run perplexity scores or anything computing intensive.
As a small test, I asked the model a few questions (like the history of the roman empire) and then fact check its answer using a big model. No errors were detected.
Update: all procedure tested and created on COLAB.
Example: https://huggingface.co/ZeroWw/Meta-Llama-3.1-8B-Instruct-abliterated-SILLY
Haha that's a funny experiment, thanks for sharing it. You should bruteforce evals with random noise
Haha that's a funny experiment, thanks for sharing it. You should bruteforce evals with random noise
I'd love to eval them at small percentages first then bigger ones, but with colab because I just have an old laptop to work with. I live abroad and I don't have any LLM worthy hardware.
I am updating this now after the llama.cpp update. Will be up in a few minutes.
Haha that's a funny experiment, thanks for sharing it. You should bruteforce evals with random noise
How can I contact you? Facebook/discord?
