alt Hacker NewsI’m confused too and the slop readme doesn’t help at all.
> Shrinks DeepSeek-V4-Flash from 76.87 GiB (which crashes in OOM on 64 GB Linux) down to 56.56 GiB
And later
> Standard IQ2_XXS (76.87 GiB, Unpatched)
Except unsloth’s IQ2_XXS GGUF is 91GB, even their smallest IQ1_S is 82.5GB. So I wonder which model/quant the number even refers to.
Also the mentions of “compression” and “MoE pruning” also make it sound lobotomized even more than unsloth’s IQ1 quants.
PS: only now noticed that 76.87GiB is 82.5GB, which happens to be size of the unsloth IQ1_S. The “technical” paper also mentions IQ1_S at some point. Sounds to me like the author’s model or the author mixed IQ2_XXS with IQ1_S at some point.
Also, just noticed that “moe pruning” likely means “the model is supposed to run 6(DS)/8(Qwen) experts per token per layer, but we only run 4 to save on time”.
Either way, the final gguf is 75% of the size of unsloth IQ1_S, which by itself should tell you a lot about its quality; they (somehow; I have no interest digging that info from that slop paper) shrink the (already IQ1) FFN weights, and shrink some attention weights from Q5 to Q2. I’m surprised it was even able to generate the text in the demo.