Unsloth Dynamic 2.0 Quants
Collection
New 2.0 version of our Dynamic GGUF + Quants. Dynamic 2.0 achieves superior accuracy & SOTA quantization performance. โข 111 items โข Updated โข 780
UygarUsta
RivianG
AI & ML interests
Computer Vision
Recent Activity
upvoted a collection 2 days ago
Unsloth Dynamic 2.0 Quants reacted to SeaWolf-AI's post with ๐ฅ 16 days ago
๐ต VKUE โ No GPU? Runs anyway.
"Frontier models need a datacenter GPU" rests on a hidden assumption: that the model reads ALL its parameters every token. Decode is memory-bandwidth bound โ sweep 34B params/token and an 8 GB card dies at 1โ2 tok/s.
So we ran ONE 34.7B reasoning model โ Ourbox-35B-JGOS, a sparse Mixture-of-Experts โ as the identical weights across the whole hardware spectrum. All measured:
โข B200: 18,057 tok/s (aggregate)
โข 1ร A10G: 126 tok/s
โข 8 GB laptop (RTX 5060): 20 tok/s
โข GPU-less CPU: 17 tok/s
Why it works: Ourbox holds 34.7B params but only ~3B are active per token (256 experts, top-8). Since decode is bandwidth-bound, a dense 34B moves ~16.7 GB/token while Ourbox moves ~1.45 GB โ ~11ร less traffic. Put the experts in system RAM, keep attention/router/shared on the GPU, and a 34.7B reasoner runs on an 8 GB laptop โ or no GPU at all.
Sparsity alone, proven (same laptop, same quant, ~same footprint): Ourbox-35B (A3B) 20.01 tok/s vs Qwen2.5-32B (dense) 5.36 โ 3.7ร from sparsity alone, ~2ร the best dense-32B on any 8 GB machine. Not a toy: GPQA Diamond 86.4% (maj@8).
Try it live (same prompt, GPU vs GPU-less CPU, live tok/s). Honest scope: one machine's measurements; the CPU path proves it RUNS without a GPU, not that it beats one.
๐ Article: https://huggingface.co/blog/FINAL-Bench/vkue
๐ต GPU vs CPU demo: https://final-bench-ourbox-35b-vkue-demo.hf.space/
๐ต CPU-only demo: https://final-bench-ourbox-35b-vkue-cpu.hf.space
๐ VKUE leaderboard: https://huggingface.co/spaces/FINAL-Bench/VKUE
๐ค Model: https://huggingface.co/FINAL-Bench/Ourbox-35B-JGOS-GGUF
โก VKAE (speed): https://huggingface.co/spaces/VIDraft/vkae
VKUE is the "runs anywhere" side of our serving line; VKAE the "fast on datacenter GPUs" side. VKAE is fast; VKUE is everywhere. reacted to MiniMax-AI's post with ๐ฅ 18 days ago
Huge news from MiniMax: weโve secured a $2B funding round, paired with a formal long-term commitment from our CEO IO to allocate 1% of total company equity from his personal holdings to support the global open-source AI community over the next four years.
This capital backs our continuous open model releases, community tooling and transparent frontier AI research. Weโre just getting started on our open-source roadmap toward accessible AGI.
If you build with open foundation models and want to push frontier AI together, come join us.
Intelligence with Everyone. ๐
https://huggingface.co/MiniMaxAIOrganizations
None yet
upvoted a collection 2 days ago
reacted to SeaWolf-AI's post with ๐ฅ 16 days ago
Post
5402
๐ต VKUE โ No GPU? Runs anyway.
"Frontier models need a datacenter GPU" rests on a hidden assumption: that the model reads ALL its parameters every token. Decode is memory-bandwidth bound โ sweep 34B params/token and an 8 GB card dies at 1โ2 tok/s.
So we ran ONE 34.7B reasoning model โ Ourbox-35B-JGOS, a sparse Mixture-of-Experts โ as the identical weights across the whole hardware spectrum. All measured:
โข B200: 18,057 tok/s (aggregate)
โข 1ร A10G: 126 tok/s
โข 8 GB laptop (RTX 5060): 20 tok/s
โข GPU-less CPU: 17 tok/s
Why it works: Ourbox holds 34.7B params but only ~3B are active per token (256 experts, top-8). Since decode is bandwidth-bound, a dense 34B moves ~16.7 GB/token while Ourbox moves ~1.45 GB โ ~11ร less traffic. Put the experts in system RAM, keep attention/router/shared on the GPU, and a 34.7B reasoner runs on an 8 GB laptop โ or no GPU at all.
Sparsity alone, proven (same laptop, same quant, ~same footprint): Ourbox-35B (A3B) 20.01 tok/s vs Qwen2.5-32B (dense) 5.36 โ 3.7ร from sparsity alone, ~2ร the best dense-32B on any 8 GB machine. Not a toy: GPQA Diamond 86.4% (maj@8).
Try it live (same prompt, GPU vs GPU-less CPU, live tok/s). Honest scope: one machine's measurements; the CPU path proves it RUNS without a GPU, not that it beats one.
๐ Article: https://huggingface.co/blog/FINAL-Bench/vkue
๐ต GPU vs CPU demo: https://final-bench-ourbox-35b-vkue-demo.hf.space/
๐ต CPU-only demo: https://final-bench-ourbox-35b-vkue-cpu.hf.space
๐ VKUE leaderboard: FINAL-Bench/VKUE
๐ค Model: FINAL-Bench/Ourbox-35B-JGOS-GGUF
โก VKAE (speed): VIDraft/vkae
VKUE is the "runs anywhere" side of our serving line; VKAE the "fast on datacenter GPUs" side. VKAE is fast; VKUE is everywhere.
"Frontier models need a datacenter GPU" rests on a hidden assumption: that the model reads ALL its parameters every token. Decode is memory-bandwidth bound โ sweep 34B params/token and an 8 GB card dies at 1โ2 tok/s.
So we ran ONE 34.7B reasoning model โ Ourbox-35B-JGOS, a sparse Mixture-of-Experts โ as the identical weights across the whole hardware spectrum. All measured:
โข B200: 18,057 tok/s (aggregate)
โข 1ร A10G: 126 tok/s
โข 8 GB laptop (RTX 5060): 20 tok/s
โข GPU-less CPU: 17 tok/s
Why it works: Ourbox holds 34.7B params but only ~3B are active per token (256 experts, top-8). Since decode is bandwidth-bound, a dense 34B moves ~16.7 GB/token while Ourbox moves ~1.45 GB โ ~11ร less traffic. Put the experts in system RAM, keep attention/router/shared on the GPU, and a 34.7B reasoner runs on an 8 GB laptop โ or no GPU at all.
Sparsity alone, proven (same laptop, same quant, ~same footprint): Ourbox-35B (A3B) 20.01 tok/s vs Qwen2.5-32B (dense) 5.36 โ 3.7ร from sparsity alone, ~2ร the best dense-32B on any 8 GB machine. Not a toy: GPQA Diamond 86.4% (maj@8).
Try it live (same prompt, GPU vs GPU-less CPU, live tok/s). Honest scope: one machine's measurements; the CPU path proves it RUNS without a GPU, not that it beats one.
๐ Article: https://huggingface.co/blog/FINAL-Bench/vkue
๐ต GPU vs CPU demo: https://final-bench-ourbox-35b-vkue-demo.hf.space/
๐ต CPU-only demo: https://final-bench-ourbox-35b-vkue-cpu.hf.space
๐ VKUE leaderboard: FINAL-Bench/VKUE
๐ค Model: FINAL-Bench/Ourbox-35B-JGOS-GGUF
โก VKAE (speed): VIDraft/vkae
VKUE is the "runs anywhere" side of our serving line; VKAE the "fast on datacenter GPUs" side. VKAE is fast; VKUE is everywhere.
- 7 replies
reacted to MiniMax-AI's post with ๐ฅ 18 days ago
Post
4660
Huge news from MiniMax: weโve secured a $2B funding round, paired with a formal long-term commitment from our CEO IO to allocate 1% of total company equity from his personal holdings to support the global open-source AI community over the next four years.
This capital backs our continuous open model releases, community tooling and transparent frontier AI research. Weโre just getting started on our open-source roadmap toward accessible AGI.
If you build with open foundation models and want to push frontier AI together, come join us.
Intelligence with Everyone. ๐
MiniMaxAI
This capital backs our continuous open model releases, community tooling and transparent frontier AI research. Weโre just getting started on our open-source roadmap toward accessible AGI.
If you build with open foundation models and want to push frontier AI together, come join us.
Intelligence with Everyone. ๐
- 3 replies
liked a
Space about 1 month ago
liked a
model about 2 months ago
liked a
model 2 months ago
upvoted an article 5 months ago
Article
PRX Part 3 โ Training a Text-to-Image Model in 24h!
Photoroom
โข โข 68 liked a
model 5 months ago
liked 3
models 6 months ago
liked a
model 7 months ago
reacted to hesamation's post with โค๏ธ 8 months ago
Post
5551
this is big... 50 AI researchers from Bytedance, Alibaba, Tencent, and other labs/universities just published a 300-page paper with surprising lessons about coding models and agents (data, pre and post-training, etc).
key highlights:
> small LLMs can beat proprietary giants
RL (RLVR specifically) gives small open-source models an edge over big models in reasoning. a 14B model trained with RLVR on high-quality verified problems can match the performance of OpenAI's o3.
> models have a hard time learning Python.
mixing language models during pre-training is good, but Python behaves different from statically typed languages. languages with similar syntax (Java and C#, or JavaScript and TypeScript) creates high positive synergy. mixing Python heavily into the training of statically typed languages can actually hurt because of Python's dynamic typing.
> not all languages are equal (coding scaling laws)
the amount of data required to specialize a model on a language drastically depends on the language. paper argues like C# and Java are easier to learn (less training data required). languages like Python and Javascript are actually more tricky to learn, ironically (you see AI most used for these languages :)
> MoE vs Dense (ability vs stability)
MoE models offer higher capacity, but are much more fragile during SFT than dense models. hyperparams in training have a more drastic effect in MoE models, while dense models are more stable. MoE models also require constant learning rate schedules to avoid routing instability.
> code models are "insecure" by default (duh)
training on public repos makes models learn years of accumulated insecure coding patterns. safety fine-tuning often fails to work much on code. a model might refuse to write a hate speech email but will happily generate a SQL-injection vulnerable function because it "works."
read the full paper:
From Code Foundation Models to Agents and Applications: A Practical Guide to Code Intelligence (2511.18538)
key highlights:
> small LLMs can beat proprietary giants
RL (RLVR specifically) gives small open-source models an edge over big models in reasoning. a 14B model trained with RLVR on high-quality verified problems can match the performance of OpenAI's o3.
> models have a hard time learning Python.
mixing language models during pre-training is good, but Python behaves different from statically typed languages. languages with similar syntax (Java and C#, or JavaScript and TypeScript) creates high positive synergy. mixing Python heavily into the training of statically typed languages can actually hurt because of Python's dynamic typing.
> not all languages are equal (coding scaling laws)
the amount of data required to specialize a model on a language drastically depends on the language. paper argues like C# and Java are easier to learn (less training data required). languages like Python and Javascript are actually more tricky to learn, ironically (you see AI most used for these languages :)
> MoE vs Dense (ability vs stability)
MoE models offer higher capacity, but are much more fragile during SFT than dense models. hyperparams in training have a more drastic effect in MoE models, while dense models are more stable. MoE models also require constant learning rate schedules to avoid routing instability.
> code models are "insecure" by default (duh)
training on public repos makes models learn years of accumulated insecure coding patterns. safety fine-tuning often fails to work much on code. a model might refuse to write a hate speech email but will happily generate a SQL-injection vulnerable function because it "works."
read the full paper:
From Code Foundation Models to Agents and Applications: A Practical Guide to Code Intelligence (2511.18538)
- 1 reply
liked a
model 8 months ago
liked a
Space 8 months ago
liked a
model 8 months ago
updated a
model 8 months ago
published a
model 8 months ago
reacted to nouamanetazi's post with ๐ค 9 months ago
Post
5053
After training ๐๐ฆ๐จ๐ฅ๐๐๐ on ๐๐๐ ๐๐๐๐๐ฌ for nearly a month, I've come to realize something most people overlook: ๐ข๐ง๐๐ซ๐๐ฌ๐ญ๐ซ๐ฎ๐๐ญ๐ฎ๐ซ๐ ๐ข๐ฌ ๐ญ๐ก๐ ๐ฆ๐๐ค๐-๐จ๐ซ-๐๐ซ๐๐๐ค ๐๐๐๐ญ๐จ๐ซ ๐ข๐ง ๐๐๐ ๐ญ๐ซ๐๐ข๐ง๐ข๐ง๐ . ๐ฅ
Everyone talks about model architecture and data quality. And yes, those matter immensely. But here's what nobody tells you: when your training run fails at 2 AM because of mysterious ๐๐๐๐ ๐๐ซ๐ซ๐จ๐ซ๐ฌ, or when your expensive GPU cluster is running at ๐๐% ๐๐๐๐ข๐๐ข๐๐ง๐๐ฒ, the problem isn't your model. It's most probably a ๐ฆ๐ข๐ฌ๐ฎ๐ฌ๐ ๐จ๐ ๐ญ๐ก๐ ๐ก๐๐ซ๐๐ฐ๐๐ซ๐. ๐ ๏ธ
Questions that seemed simple but had no clear answers: Why is ๐๐จ๐ ๐ญ๐ซ๐๐ข๐ง๐ข๐ง๐ ๐ฌ๐ฅ๐จ๐ฐ๐๐ซ ๐ญ๐ก๐๐ง ๐๐๐ง๐ฌ๐ ๐ฆ๐จ๐๐๐ฅ๐ฌ? Which ๐๐๐๐ ๐๐ฅ๐๐ ๐ฌ should we actually set? How often should we checkpoint without killing throughput?
That's why we built ๐๐ก๐ ๐๐ฆ๐จ๐ฅ ๐๐ซ๐๐ข๐ง๐ข๐ง๐ ๐๐ฅ๐๐ฒ๐๐จ๐จ๐ค ๐: a complete guide covering everything from model architecture and data curation to the SmolLM3 training marathon, post-training techniques, and crucially, the ๐ข๐ง๐๐ซ๐๐ฌ๐ญ๐ซ๐ฎ๐๐ญ๐ฎ๐ซ๐ ๐ฅ๐๐ฒ๐๐ซ that most teams get wrong.
We validated real vs theoretical bandwidth across the entire stack: ๐๐๐๐ ๐ก๐ข๐ญ๐ญ๐ข๐ง๐ ๐ ๐๐/๐ฌ, ๐๐๐๐ข๐ง๐ค ๐.๐ ๐ซ๐๐๐๐ก๐ข๐ง๐ ๐๐๐ ๐๐/๐ฌ, ๐๐๐๐ ๐๐๐ง๐ ๐๐ญ ๐๐.๐ ๐๐/๐ฌ. Then we ran collective operations across ๐๐๐ ๐๐๐๐ฌ (16 nodes, 8xH100s each) and measured how performance degrades at scale: all-reduce drops from ๐๐๐ ๐๐/๐ฌ on a single node to ๐๐๐-๐๐๐ ๐๐/๐ฌ across 16 nodes.
If you've ever wondered why your training runs are slower than they should be, or you're planning to scale up and want to avoid expensive mistakes, this guide might save you weeks of debugging.
๐๐ก๐ ๐๐ฆ๐จ๐ฅ ๐๐ซ๐๐ข๐ง๐ข๐ง๐ ๐๐ฅ๐๐ฒ๐๐จ๐จ๐ค: https://lnkd.in/e5MKXUHS
Shared with โค๏ธ by the HuggingFace team
Everyone talks about model architecture and data quality. And yes, those matter immensely. But here's what nobody tells you: when your training run fails at 2 AM because of mysterious ๐๐๐๐ ๐๐ซ๐ซ๐จ๐ซ๐ฌ, or when your expensive GPU cluster is running at ๐๐% ๐๐๐๐ข๐๐ข๐๐ง๐๐ฒ, the problem isn't your model. It's most probably a ๐ฆ๐ข๐ฌ๐ฎ๐ฌ๐ ๐จ๐ ๐ญ๐ก๐ ๐ก๐๐ซ๐๐ฐ๐๐ซ๐. ๐ ๏ธ
Questions that seemed simple but had no clear answers: Why is ๐๐จ๐ ๐ญ๐ซ๐๐ข๐ง๐ข๐ง๐ ๐ฌ๐ฅ๐จ๐ฐ๐๐ซ ๐ญ๐ก๐๐ง ๐๐๐ง๐ฌ๐ ๐ฆ๐จ๐๐๐ฅ๐ฌ? Which ๐๐๐๐ ๐๐ฅ๐๐ ๐ฌ should we actually set? How often should we checkpoint without killing throughput?
That's why we built ๐๐ก๐ ๐๐ฆ๐จ๐ฅ ๐๐ซ๐๐ข๐ง๐ข๐ง๐ ๐๐ฅ๐๐ฒ๐๐จ๐จ๐ค ๐: a complete guide covering everything from model architecture and data curation to the SmolLM3 training marathon, post-training techniques, and crucially, the ๐ข๐ง๐๐ซ๐๐ฌ๐ญ๐ซ๐ฎ๐๐ญ๐ฎ๐ซ๐ ๐ฅ๐๐ฒ๐๐ซ that most teams get wrong.
We validated real vs theoretical bandwidth across the entire stack: ๐๐๐๐ ๐ก๐ข๐ญ๐ญ๐ข๐ง๐ ๐ ๐๐/๐ฌ, ๐๐๐๐ข๐ง๐ค ๐.๐ ๐ซ๐๐๐๐ก๐ข๐ง๐ ๐๐๐ ๐๐/๐ฌ, ๐๐๐๐ ๐๐๐ง๐ ๐๐ญ ๐๐.๐ ๐๐/๐ฌ. Then we ran collective operations across ๐๐๐ ๐๐๐๐ฌ (16 nodes, 8xH100s each) and measured how performance degrades at scale: all-reduce drops from ๐๐๐ ๐๐/๐ฌ on a single node to ๐๐๐-๐๐๐ ๐๐/๐ฌ across 16 nodes.
If you've ever wondered why your training runs are slower than they should be, or you're planning to scale up and want to avoid expensive mistakes, this guide might save you weeks of debugging.
๐๐ก๐ ๐๐ฆ๐จ๐ฅ ๐๐ซ๐๐ข๐ง๐ข๐ง๐ ๐๐ฅ๐๐ฒ๐๐จ๐จ๐ค: https://lnkd.in/e5MKXUHS
Shared with โค๏ธ by the HuggingFace team
liked a
model 9 months ago
