1 1

No_Thanks

Matrix-Array

AI & ML interests

None yet

Recent Activity

new activity 10 days ago

unsloth/Qwen3-32B-GGUF:PPL vs model size - safe to assume larger size == better accuracy regardless of UD vs non-UD?

upvoted an article about 1 month ago

Comparing sub 50GB Llama 4 Scout quants (KLD/Top P)

commented on an article about 1 month ago

Comparing sub 50GB Llama 4 Scout quants (KLD/Top P)

View all activity

Organizations

Matrix-Array's activity

New activity in unsloth/Qwen3-32B-GGUF 10 days ago

PPL vs model size - safe to assume larger size == better accuracy regardless of UD vs non-UD?

#6 opened 10 days ago by

Thireus

upvoted an article about 1 month ago

Article

Comparing sub 50GB Llama 4 Scout quants (KLD/Top P)

•

Apr 9

• 39

commented on Comparing sub 50GB Llama 4 Scout quants (KLD/Top P) about 1 month ago

I took the liberty of trying to add more visualization of the data. You should be able to view it here:

[Lite]
https://page.genspark.site/page/toolu_013iF9H8mPpqYyNHfi3sc8KA/performance_metrics_analysis.html

[Advanced]
https://page.genspark.site/page/toolu_018LCdFVi8w6aXChyMqERMdx/revised_quantization_performance_analysis.html

[Notes]
Edit includes an updated revision with enhanced visuals, keeping the lite version for those who want to see it. Enhanced mode is not optimized for mobile. Landscape mode fixes the issue it seems, and desktop looks fine.