Search results for Prune
Prune community
One keyword maps to one global community path.
People
Not Found
Tweets including Prune
Prune Juice Was Served On A Packed Flight With Just Two Lavatories. What Could Go Wrong?
benchmarks of a 50% pruned Qwen3.6-35b-a3b and expert-specific quantization technique (made by me)
7.3gb model preforming => 51gb model, exiting to see where I can bring this technique to. I have some more things lined up too.
I need a DGX spark😭
Show more
Kijai's 😃
Minimax_h3_ref2va_pruned_w6a8_g32
its experimental, highly quantized work-in-progress (WIP) base model for MiniMax H3 video generation. It is a massive 16.8 GB file designed to drastically reduce the VRAM required to run this model locally.
👇
Show more
GM. today's agenda: prune, measure, repeat. 🪴
Kijai's 😃
Minimax H3 4step_lora_flashgen_v1.0 768p_fl2va_pruned_avg_rank_13_bf16
This Lora is a Turbo/Flash acceleration adapter. using only 4 steps, delivering a roughly 5x speedup in generation times
👇
Show more
Updated papercuts and DCE to 0.3.0
Papercuts: you can now prune the papercuts list after you resolve
DCE: the agent asks you at the end of a session if any of the tools that were promoted should be alwaysActive.
pi install npm:pi-papercuts
pi install npm:pi-deferred-context-engine
Show more
Qwen3.8-2.4T-A95B, compressed two ways at once: 25% of the experts pruned with REAP, and the rest quantized to NVFP4.
Even with a quarter of the experts gone and 4-bit weights, GPQA Diamond holds at 91.5 vs 92.6 for the full-precision base. ~99% recovery.
Serve on @vllm_project:
Show more
H3 🐰💋👠 N$FW LoRA
-PinkFluffyBunny lora that truly brings the style and pose
-Maximum results achieved at 0.5 str on pruned int8 model.
-Use the h3_fl2va_pruned_int8_convrot.safetensors model. Alpha quality so temper expectations
👇😉
Show more
MiniMax H3 💋💄👠Naughty Times LoRA
Tip-
Run it with a lora loader at value 0.5 using the model
(minimax_h3_fl2va_pruned_int8_convrot.safetensors)
👇
OpenAI’s Navier–Stokes result reportedly used ~10,000 autonomous agents, but only 2.7M messages and 130B tokens for the final proof search.
That’s not “lots of agents = lots of chatter.” It’s tight coordination: orchestration logic aggressively prunes, routes, and reuses partial work.
Show more
