JetBrains/Mellum2-12B-A2.5B-Thinking
Text Generation • 12B • Updated • 18.3k • 283
Rod
Viac
AI & ML interests
playing around with a tool
Recent Activity
liked a model 1 day ago
JetBrains/Mellum2-12B-A2.5B-Thinking liked a model 13 days ago
openbmb/MiniCPM5-1B liked a model about 1 month ago
Vikhrmodels/Borealis-5b-itOrganizations
None yet
liked a
model 1 day ago
liked a
model 13 days ago
liked 2
models about 1 month ago
liked 8
models about 2 months ago
Qwen/Qwen3.6-27B
Image-Text-to-Text • 28B • Updated • 5.28M • • 1.68k
moonshotai/Kimi-K2.6
Image-Text-to-Text • 1.1T • Updated • 2.76M • • 1.45k
lordx64/Qwen3.6-35B-A3B-Claude-4.7-Opus-Reasoning-Distilled-IQ4_XS-GGUF
Text Generation • 35B • Updated • 6.69k • 31
Qwen/Qwen3.6-35B-A3B-FP8
Image-Text-to-Text • 36B • Updated • 5.04M • 263
nvidia/Lyra-2.0
Image-to-3D • Updated • 753 • 327
varjosoft/GLM-4.7-Flash-TQ3
13B • Updated • 13 • 1
Youssofal/MiniMax-M2.7-Abliterated-Heretic-GGUF
Text Generation • 229B • Updated • 1.36k • 46
MiniMaxAI/MiniMax-M2.7
Text Generation • 229B • Updated • 2.52M • • 1.21k
liked 2
models 2 months ago
liked a
dataset 2 months ago
reacted to reaperdoesntknow's post with 👍 2 months ago
Post
3293
We present a methodology for training small language models on CPU at FP32 precision
that achieves capability-per-dollar efficiency orders of magnitude beyond GPU-based training.
Across15modelsspanningfournovelarchitecturefamilies—MixtureofAttentions(MoA),cross-
architecture fusion (Qemma), swarm intelligence (SAGI), and metric-space causal language
models (DiscoverLM)—total compute cost was $24 on a single AMD EPYC 9454P proces-
sor. We introduce seven methodological pillars: (1) FP32 precision preservation, with exper-
iments demonstrating 5,810×single-operation error and 23,225×compounding error ratio for
FP16 at network depth; (2) sparse cognitive architectures where 0.02–7% of parameters activate
per token, matching CPU branching rather than GPU SIMD; (3) developmental curriculum
training progressing from language to logic to transfer to depth; (4) continuous belt-fed data
ingestion eliminating truncation waste; (5) hardware-native optimization for AMD Zen 4 via
AOCL/OpenMP/NUMA-aware allocation; (6) self-regulating thermodynamic governance with
emergent temperature measurement grounded in L2-star discrepancy; and (7) open-standard
compute (AVX2 SIMD at FP32) free of proprietary vendor dependency. We argue that transformers were designed for GPU hardware rather than mathematical optimality, and that architecture designed for geometric correctness—metric-space attention, triangle inequality enforcement, sparse expert routing—naturally favor CPU execution. For sub-2B parameter models, CPU training produces more capable models at a fraction of the cost.
that achieves capability-per-dollar efficiency orders of magnitude beyond GPU-based training.
Across15modelsspanningfournovelarchitecturefamilies—MixtureofAttentions(MoA),cross-
architecture fusion (Qemma), swarm intelligence (SAGI), and metric-space causal language
models (DiscoverLM)—total compute cost was $24 on a single AMD EPYC 9454P proces-
sor. We introduce seven methodological pillars: (1) FP32 precision preservation, with exper-
iments demonstrating 5,810×single-operation error and 23,225×compounding error ratio for
FP16 at network depth; (2) sparse cognitive architectures where 0.02–7% of parameters activate
per token, matching CPU branching rather than GPU SIMD; (3) developmental curriculum
training progressing from language to logic to transfer to depth; (4) continuous belt-fed data
ingestion eliminating truncation waste; (5) hardware-native optimization for AMD Zen 4 via
AOCL/OpenMP/NUMA-aware allocation; (6) self-regulating thermodynamic governance with
emergent temperature measurement grounded in L2-star discrepancy; and (7) open-standard
compute (AVX2 SIMD at FP32) free of proprietary vendor dependency. We argue that transformers were designed for GPU hardware rather than mathematical optimality, and that architecture designed for geometric correctness—metric-space attention, triangle inequality enforcement, sparse expert routing—naturally favor CPU execution. For sub-2B parameter models, CPU training produces more capable models at a fraction of the cost.
- 6 replies
liked 2
models 3 months ago
reacted to abdurrahmanbutler's post with 🤗 3 months ago
Post
2591
🚀 𝗜𝗻𝘁𝗿𝗼𝗱𝘂𝗰𝗶𝗻𝗴 𝗞𝗮𝗻𝗼𝗻 𝟮 𝗘𝗻𝗿𝗶𝗰𝗵𝗲𝗿: 𝘁𝗵𝗲 𝘄𝗼𝗿𝗹𝗱’𝘀 𝗳𝗶𝗿𝘀𝘁 𝗵𝗶𝗲𝗿𝗮𝗿𝗰𝗵𝗶𝗰𝗮𝗹 𝗴𝗿𝗮𝗽𝗵𝗶𝘁𝗶𝘇𝗮𝘁𝗶𝗼𝗻 𝗺𝗼𝗱𝗲𝗹
Today we’re publicly releasing Kanon 2 Enricher, and with it, an entirely new class of AI model that we’re calling a hierarchical graphitization model.
This is fundamentally different from both universal extraction models and generative models.
As a hierarchical graphitization model, Kanon 2 Enricher natively outputs a 𝗸𝗻𝗼𝘄𝗹𝗲𝗱𝗴𝗲 𝗴𝗿𝗮𝗽𝗵 rather than tokens, which makes it architecturally incapable of hallucinating or inventing text that wasn’t present in the input.
What that enables in practice is unlike any other model or ML architecture on the market:
• 𝗡𝗼 𝗵𝗮𝗹𝗹𝘂𝗰𝗶𝗻𝗮𝘁𝗶𝗼𝗻𝘀 🤖
It cannot hallucinate. All references and links are stored as spans, meaning exact character offsets anchored to the original text.
• 𝗛𝗶𝗲𝗿𝗮𝗿𝗰𝗵𝗶𝗰𝗮𝗹 𝘀𝗲𝗴𝗺𝗲𝗻𝘁𝗮𝘁𝗶𝗼𝗻, 𝗻𝗼𝘁 𝗷𝘂𝘀𝘁 𝗲𝘅𝘁𝗿𝗮𝗰𝘁𝗶𝗼𝗻 📑
It deconstructs a document’s full nested hierarchy, down to chapters, sections, clauses, schedules, signatures, and even singular sentences, and classifies each span with dozens of contextual features.
• 𝗘𝗻𝘁𝗶𝘁𝘆 𝗲𝘅𝘁𝗿𝗮𝗰𝘁𝗶𝗼𝗻, 𝗱𝗶𝘀𝗮𝗺𝗯𝗶𝗴𝘂𝗮𝘁𝗶𝗼𝗻, 𝗮𝗻𝗱 𝗹𝗶𝗻𝗸𝗶𝗻𝗴 🔗
It resolves what references actually point to, then links entities, citations, and cross-references into a single coherent graph.
• 𝗚𝗿𝗮𝗽𝗵-𝗳𝗶𝗿𝘀𝘁 𝗲𝗳𝗳𝗶𝗰𝗶𝗲𝗻𝗰𝘆 🏃➡️
Small enough to run locally on a consumer PC with sub-second latency, and it stays reliable on long documents where front
To read more about our new model, check out our latest Hugging Face article:
https://huggingface.co/blog/isaacus/introducing-kanon-2-enricher
Today we’re publicly releasing Kanon 2 Enricher, and with it, an entirely new class of AI model that we’re calling a hierarchical graphitization model.
This is fundamentally different from both universal extraction models and generative models.
As a hierarchical graphitization model, Kanon 2 Enricher natively outputs a 𝗸𝗻𝗼𝘄𝗹𝗲𝗱𝗴𝗲 𝗴𝗿𝗮𝗽𝗵 rather than tokens, which makes it architecturally incapable of hallucinating or inventing text that wasn’t present in the input.
What that enables in practice is unlike any other model or ML architecture on the market:
• 𝗡𝗼 𝗵𝗮𝗹𝗹𝘂𝗰𝗶𝗻𝗮𝘁𝗶𝗼𝗻𝘀 🤖
It cannot hallucinate. All references and links are stored as spans, meaning exact character offsets anchored to the original text.
• 𝗛𝗶𝗲𝗿𝗮𝗿𝗰𝗵𝗶𝗰𝗮𝗹 𝘀𝗲𝗴𝗺𝗲𝗻𝘁𝗮𝘁𝗶𝗼𝗻, 𝗻𝗼𝘁 𝗷𝘂𝘀𝘁 𝗲𝘅𝘁𝗿𝗮𝗰𝘁𝗶𝗼𝗻 📑
It deconstructs a document’s full nested hierarchy, down to chapters, sections, clauses, schedules, signatures, and even singular sentences, and classifies each span with dozens of contextual features.
• 𝗘𝗻𝘁𝗶𝘁𝘆 𝗲𝘅𝘁𝗿𝗮𝗰𝘁𝗶𝗼𝗻, 𝗱𝗶𝘀𝗮𝗺𝗯𝗶𝗴𝘂𝗮𝘁𝗶𝗼𝗻, 𝗮𝗻𝗱 𝗹𝗶𝗻𝗸𝗶𝗻𝗴 🔗
It resolves what references actually point to, then links entities, citations, and cross-references into a single coherent graph.
• 𝗚𝗿𝗮𝗽𝗵-𝗳𝗶𝗿𝘀𝘁 𝗲𝗳𝗳𝗶𝗰𝗶𝗲𝗻𝗰𝘆 🏃➡️
Small enough to run locally on a consumer PC with sub-second latency, and it stays reliable on long documents where front
To read more about our new model, check out our latest Hugging Face article:
https://huggingface.co/blog/isaacus/introducing-kanon-2-enricher
liked a
model 3 months ago
