6
playing with Expert expansion in MOE Qwen 3.6 35B mmlu-pro intermediate results : dyn 20 vs static 8 experts 0.820 vs 0.805 (the same 6bit quant gguf) LESS tokens -9% , more latency +1%
40
Pelican without bicycle found on the beach of #milazzo github.com/vagrillo/ds4/blob… ready for inferencing #qwen #dwarfstar
43
Replying to @VAgrillo @antirez
over 12 tok/sec on my MBP M4p 48GB with DeepSeek V4 Flash... is like magic
52
local Kimi K3?
34
what is : "Drafted / verify call 0.00tok/call 0 drafted · 1,071 verifies" in my M4Pro is always zero, but i can guarantee that old fused kernels drive Qwen 3.6 35B A3B so well that has never been so fast (without losing smartness , in comparison to unsloth 6b XL )
1
1
50
More than 10 tok/sec with a near SOTA model likes Deepseek v4 flash on my MacBook pro with M4 pro and only 48GB of RAM using the great DwarfStar tool from @antirez and it's SSD streaming feature! github.com/antirez/ds4
1
99
Dopo porta a porta la meloni da Gerry Scotti alla ruota della fortuna, basta no?
106
Non hanno un amico ha conquistato il primo posto. E c'è di più nel mio #SpotifyWrapped. spotify.com/wrapped-share/8d…
25
They are surely stochastic parrots!
30
Replying to @VAgrillo @karpathy
It's a colour contrast prob
5
Replying to @karpathy
It struggles on italian menu? :-D Great idea pricing LLM payxuse
1
92
Rock'n'roll never dies!
1
1
12
"io non voglio restare seduto in dispare, ne arte ne parte" #Mentana , ricominciamo?
1
215
Replying to @engineers_feed
Who used Winplay3 instead ?
12
Ottobrata a Milazzo
1
30
33° alle 20:30 in Autunno !
32
Salvini, Meloni, mangiate tranquilli Il M5S non fa alleanze con quelli che da decenni sono complici della distruzione del Paese