ZeroWw 'SILLY' version.
The original model has been quantized (fq8 version)
and a percentage of it's tensors have
been modified adding some noise.
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 the original model 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.
Created on: Sat Jul 27, 02:36:00
