vooom/Fast_Growth-storage
358 kB
FAJU85 PRO
vooom
AI & ML interests
None yet
Recent Activity
updated a bucket 2 days ago
vooom/Fast_Growth-storage updated a bucket 3 days ago
vooom/home reacted to Banaxi-Tech's post with 👍 4 days ago
We're excited to release BananaMind 2 MoE, a new addition to the BananaMind 2 series!
BananaMind 2 MoE is a sparse mixture-of-experts model with 25M total parameters but only 2M active per token, Like the rest of the series, it uses our custom digit-aware BPE tokenizer that keeps every digit isolated, fixing the core arithmetic weakness of our earlier models. It's trained on 30B tokens from FineWeb-Edu, DCLM, Cosmopedia-v2 and FineMath-4+, and outperforms Pythia-31M on average despite activating just 2M parameters.
Check it out at
https://huggingface.co/BananaMind/BananaMind-2-MoE
BananaMind 2 Nano is Coming Next. Already training. Apache 2.0.Organizations
None yet
updated a
bucket 2 days ago
updated a
bucket 3 days ago
reacted to Banaxi-Tech's post with 👍 4 days ago
Post
137
We're excited to release BananaMind 2 MoE, a new addition to the BananaMind 2 series!
BananaMind 2 MoE is a sparse mixture-of-experts model with 25M total parameters but only 2M active per token, Like the rest of the series, it uses our custom digit-aware BPE tokenizer that keeps every digit isolated, fixing the core arithmetic weakness of our earlier models. It's trained on 30B tokens from FineWeb-Edu, DCLM, Cosmopedia-v2 and FineMath-4+, and outperforms Pythia-31M on average despite activating just 2M parameters.
Check it out at
BananaMind/BananaMind-2-MoE
BananaMind 2 Nano is Coming Next. Already training. Apache 2.0.
BananaMind 2 MoE is a sparse mixture-of-experts model with 25M total parameters but only 2M active per token, Like the rest of the series, it uses our custom digit-aware BPE tokenizer that keeps every digit isolated, fixing the core arithmetic weakness of our earlier models. It's trained on 30B tokens from FineWeb-Edu, DCLM, Cosmopedia-v2 and FineMath-4+, and outperforms Pythia-31M on average despite activating just 2M parameters.
Check it out at
BananaMind/BananaMind-2-MoE
BananaMind 2 Nano is Coming Next. Already training. Apache 2.0.
reacted to Smooke's post with 👍 4 days ago
Post
100
Say it with me, AI is moving beyond the phone, AI is moving beyond the phone, AI is moving beyond the phone. The weight Apple is putting on this says it all. They're fighting for the next hardware market before it hits the market. Harder to say what it means for OpenAI's consumer trust.
Re
apple vs. 
openai : https://hackernoon.com/apple-v-openai-trade-secrets-lawsuit-the-42-most-explosive-allegations-and-the-internets-reactions
Re
reacted to satgeze's post with 👍 4 days ago
Post
112
Two abliterated Ornith-1.0 builds, now public. Refusal removed with a single rank-1 weight edit on the MLP down-projections, direction extracted from matched harmful/harmless pairs. Measured, not claimed: 35B at **0% refusal, KL 0.036, no capability loss** (GSM8K 96.0 vs 94.0); 9B at **2% residual, KL 0.070**, an honest 2.5-point tradeoff. Verified on a real-world gate (no language drift, output integrity, real engagement), not just a refusal count. GGUF, Q8_0/Q4_K_M.
satgeze/Ornith-1.0-35B-abliterated-SatGeZe
satgeze/Ornith-1.0-9B-abliterated-SatGeZe
satgeze/Ornith-1.0-35B-abliterated-SatGeZe
satgeze/Ornith-1.0-9B-abliterated-SatGeZe
reacted to ajibawa-2023's post with 🚀 4 days ago
Post
3470
Technical-Architectures-Large
Dataset: ajibawa-2023/Technical-Architectures-Large
This dataset provides over 210,000 distinct enterprise software architectures generated using two open source models: GPT-OSS-120B and Qwen3-Coder-Next-FP8.
These architectures model realistic enterprise systems complete with client layers, edge security, API gateways, service meshes, compliance boundaries, and multi-cloud infrastructure topologies.
Dataset: ajibawa-2023/Technical-Architectures-Large
This dataset provides over 210,000 distinct enterprise software architectures generated using two open source models: GPT-OSS-120B and Qwen3-Coder-Next-FP8.
These architectures model realistic enterprise systems complete with client layers, edge security, API gateways, service meshes, compliance boundaries, and multi-cloud infrastructure topologies.
reacted to salma-remyx's post with 🚀 4 days ago
Post
4049
It's conference season, so you'll find an uptick in chatter around the research reproducibility crisis. Consider this a PSA on where the real challenges in working with research actually live.
After all, AI has made it way easier to release code and model artifacts alongside the preprints. And how many times do you really need to replicate the authors' exact configuration?
Downstream of that, as engineers evaluate candidate methods for improving THEIR systems, they rarely find a drop-in solution. More often, they're making tough tradeoffs in fidelity to the documented technique and the constraints of their deployment scenario.
They're swapping models or data indexing strategies. They have their own benchmarks to measure changes against. They're making principled reductions of a technique to respect some resource limit not considered in the source paper.
AI coding has made replication cheap when a paper provides starting point for your own experiments. But the work of adoption requires validation grounded in real-world outcomes.
So put these techniques to the test in your own system, and you'll understand a method's impact well before the survey paper drops in six months.
At Remyx AI, we're helping teams discover, implement, and validate what's next for their systems.
Get Outrider: https://github.com/remyxai/outrider
After all, AI has made it way easier to release code and model artifacts alongside the preprints. And how many times do you really need to replicate the authors' exact configuration?
Downstream of that, as engineers evaluate candidate methods for improving THEIR systems, they rarely find a drop-in solution. More often, they're making tough tradeoffs in fidelity to the documented technique and the constraints of their deployment scenario.
They're swapping models or data indexing strategies. They have their own benchmarks to measure changes against. They're making principled reductions of a technique to respect some resource limit not considered in the source paper.
AI coding has made replication cheap when a paper provides starting point for your own experiments. But the work of adoption requires validation grounded in real-world outcomes.
So put these techniques to the test in your own system, and you'll understand a method's impact well before the survey paper drops in six months.
At Remyx AI, we're helping teams discover, implement, and validate what's next for their systems.
Get Outrider: https://github.com/remyxai/outrider
- 8 replies
reacted to scthornton's post with ❤️ 7 days ago
Post
2644
SecureCode update: we went back and fact-checked our own security dataset and corrected what didn't hold up.
The original claim was "complete incident grounding, every example ties to a documented CVE." An adversarial re-audit found that it was overstated: many CVEs were misattributed, and many "incidents" were representative scenarios carrying invented statistics. So we fixed it.
- Grounding: re-verified every reference. Removed 802 misattributed CVEs on the web side, corrected or honestly relabeled the incident narratives, and confirmed the AI/ML conversation CVEs are real (EchoLeak CVE-2025-32711, EmailGPT CVE-2024-5184, and others).
- Fix-correctness: reviewed whether each "secure" example actually eliminates the vulnerability. Removed 28 that did not (a "secure" secret scanner whose entropy check always returned zero, an Angular example still using bypassSecurityTrustHtml, and more).
- Leakage: re-split so near-duplicates stay on one side. Test contamination went from 11.6% to zero.
- Viewer, schema, and metadata: rebuilt as parquet under a shared schema. All three viewers are live.
- Models: retrained the whole family on the corrected data so the fix reaches the weights, not just the cards. Now ten open models (3B to 26B), including two new Gemma 4 variants, refreshed locally on a DGX Spark GB10. The paper (arXiv:2512.18542) was revised to match.
Counts moved from 2,185 to 2,372 unified (web 1,625 + AI/ML 747). A slightly smaller, fully-checked dataset beats a larger one you have to take on faith. Full writeup and links in the article.
Datasets: scthornton/securecode, scthornton/securecode-web, scthornton/securecode-aiml
The original claim was "complete incident grounding, every example ties to a documented CVE." An adversarial re-audit found that it was overstated: many CVEs were misattributed, and many "incidents" were representative scenarios carrying invented statistics. So we fixed it.
- Grounding: re-verified every reference. Removed 802 misattributed CVEs on the web side, corrected or honestly relabeled the incident narratives, and confirmed the AI/ML conversation CVEs are real (EchoLeak CVE-2025-32711, EmailGPT CVE-2024-5184, and others).
- Fix-correctness: reviewed whether each "secure" example actually eliminates the vulnerability. Removed 28 that did not (a "secure" secret scanner whose entropy check always returned zero, an Angular example still using bypassSecurityTrustHtml, and more).
- Leakage: re-split so near-duplicates stay on one side. Test contamination went from 11.6% to zero.
- Viewer, schema, and metadata: rebuilt as parquet under a shared schema. All three viewers are live.
- Models: retrained the whole family on the corrected data so the fix reaches the weights, not just the cards. Now ten open models (3B to 26B), including two new Gemma 4 variants, refreshed locally on a DGX Spark GB10. The paper (arXiv:2512.18542) was revised to match.
Counts moved from 2,185 to 2,372 unified (web 1,625 + AI/ML 747). A slightly smaller, fully-checked dataset beats a larger one you have to take on faith. Full writeup and links in the article.
Datasets: scthornton/securecode, scthornton/securecode-web, scthornton/securecode-aiml
- 8 replies
reacted to PsamSahil's post with 👍 7 days ago
Post
126
I dont know why dataset is not visible and download speed is terrible today
#huggingfacebug
#huggingfacebug
- 2 replies
reacted to kanaria007's post with 👍 7 days ago
Post
122
✅ Article highlight: Constitutions, Amendments, and Emergency Powers for Simulated Polities (art-60-243, v0.1)
TL;DR:
This article asks a practical design question for persistent simulated worlds:
Once NPC societies form recognized polities, how do those polities survive leadership change, crisis, amendment, and emergency rule without collapsing into arbitrary operator control?
243 argues that recognition is not constitutional continuity. Durable simulated institutions need bounded amendment, emergency, succession, review, suspension, revocation, and supersession paths.
Read:
kanaria007/agi-structural-intelligence-protocols
Why it matters:
• separates constitution from ordinary policy
• distinguishes amendment from coup, patch, or lore rewrite
• prevents emergency powers from becoming permanent rule
• treats succession as continuity of offices, archives, duties, and legitimacy
• gives constitutions explicit lifecycle states
What’s inside:
• polity constitution objects
• amendment proposals with ratification paths
• emergency activations with scope, expiry, review, and forbidden actions
• elective, hereditary, appointive, rotating, federated, and mixed succession modes
• constitutional review reports
• states: ACTIVE, SUSPENDED, REVOKED, SUPERSEDED, and ARCHIVED
Key idea:
Do not say:
“the ruler changed the rules during the crisis, so the constitution evolved.”
Say:
“this simulated polity activated bounded emergency powers under this constitutional trigger, preserved these forbidden surfaces, and reviewed whether the frame remained active, required amendment, became suspended, or was superseded.”
A virtual state may survive by force.
A constitution shows whether its authority can continue without pretending every rupture was lawful.
TL;DR:
This article asks a practical design question for persistent simulated worlds:
Once NPC societies form recognized polities, how do those polities survive leadership change, crisis, amendment, and emergency rule without collapsing into arbitrary operator control?
243 argues that recognition is not constitutional continuity. Durable simulated institutions need bounded amendment, emergency, succession, review, suspension, revocation, and supersession paths.
Read:
kanaria007/agi-structural-intelligence-protocols
Why it matters:
• separates constitution from ordinary policy
• distinguishes amendment from coup, patch, or lore rewrite
• prevents emergency powers from becoming permanent rule
• treats succession as continuity of offices, archives, duties, and legitimacy
• gives constitutions explicit lifecycle states
What’s inside:
• polity constitution objects
• amendment proposals with ratification paths
• emergency activations with scope, expiry, review, and forbidden actions
• elective, hereditary, appointive, rotating, federated, and mixed succession modes
• constitutional review reports
• states: ACTIVE, SUSPENDED, REVOKED, SUPERSEDED, and ARCHIVED
Key idea:
Do not say:
“the ruler changed the rules during the crisis, so the constitution evolved.”
Say:
“this simulated polity activated bounded emergency powers under this constitutional trigger, preserved these forbidden surfaces, and reviewed whether the frame remained active, required amendment, became suspended, or was superseded.”
A virtual state may survive by force.
A constitution shows whether its authority can continue without pretending every rupture was lawful.
reacted to Banaxi-Tech's post with 👍 7 days ago
Post
3786
We are announcing 3 more models in our BananaMind 2 Family of models!
BananaMind 2 Nano, a small 10M parameter model, fits on your Pentium 4
BananaMind 2 Medium, our medium model, 50M parameters
BananaMind 2 MoE, 25M parameters, 2M active per tokens as fast as a 2M.
Because of this our release dates have changed a bit our currently estimates are:
BananaMind 2 MoE July 16-18
BananaMind 2 Nano July 18-20
BananaMind 2 Medium July 24-28
BananaMind 2 Pro August 10-16
Keep in mind these dates are estimates and we don't have a speed number currently, we will post for details going forward!
BananaMind 2 Nano, a small 10M parameter model, fits on your Pentium 4
BananaMind 2 Medium, our medium model, 50M parameters
BananaMind 2 MoE, 25M parameters, 2M active per tokens as fast as a 2M.
Because of this our release dates have changed a bit our currently estimates are:
BananaMind 2 MoE July 16-18
BananaMind 2 Nano July 18-20
BananaMind 2 Medium July 24-28
BananaMind 2 Pro August 10-16
Keep in mind these dates are estimates and we don't have a speed number currently, we will post for details going forward!
reacted to mrs83's post with 🔥 7 days ago
Post
116
A new classifier model for the OpenAIRE AI Hackathon 2026! 🇪🇺 it’s a 98M parameter Hybrid RNN that classifies research paper intents (Methodology, Dataset, Review, etc.) running entirely on CPU, currently classifying data from the OpenAIRE Graph API.
We are using a multi-model LLM-as-a-Judge pipeline to validate classifier’s predictions, but we need human experts to catch the edge cases.
Drop into the Collab Hub, review the classifications, and help us curate the open dataset for the upcoming v0.1.4 fine-tune!
Model card: ethicalabs/Echo-DSRN-v0.1.3-Research-Intent-CLF
Collab Hub: https://openaire-2026.ethicalabs.ai/
Made in 🇪🇺 with ❤️ for Open Science by
ethicalabs 🤗
We are using a multi-model LLM-as-a-Judge pipeline to validate classifier’s predictions, but we need human experts to catch the edge cases.
Drop into the Collab Hub, review the classifications, and help us curate the open dataset for the upcoming v0.1.4 fine-tune!
Model card: ethicalabs/Echo-DSRN-v0.1.3-Research-Intent-CLF
Collab Hub: https://openaire-2026.ethicalabs.ai/
Made in 🇪🇺 with ❤️ for Open Science by
reacted to danielhanchen's post with 🔥 7 days ago
Post
4407
We’re releasing Gemma 4 NVFP4 quants that run 1.5× faster on your GPU.
Gemma-4-12B NVFP4 works on 11GB VRAM.
26B-A4B hits 13K tok/s (B200).
Unsloth NVFP4 enables faster, more accurate 4-bit Blackwell inference.
Blog: https://unsloth.ai/docs/basics/nvfp4
Gemma NVFP4: https://huggingface.co/collections/unsloth/nvfp4
Gemma-4-12B NVFP4 works on 11GB VRAM.
26B-A4B hits 13K tok/s (B200).
Unsloth NVFP4 enables faster, more accurate 4-bit Blackwell inference.
Blog: https://unsloth.ai/docs/basics/nvfp4
Gemma NVFP4: https://huggingface.co/collections/unsloth/nvfp4
- 2 replies
reacted to AbstractPhil's post with 🔥 7 days ago
Post
4148
Massive AlephLM success. The task collective is producing powerful MOE shared knowledge adapters. A serious success and a massive first step towards the next stage. The current family collective results are present here; AbstractPhil/geolip-aleph-qwen
This is akin to a stackable non-intrusive lora that enables increased shared collective behavior.
This includes the three mentioned json tasks, a math task, a tinystories task, and a diffusion task for cifar10. Each adapter anchored to the knowledge within model that already exists while enhancing the knowledge through anchored lookup systems and decision-driven hierarchical access trees.
All tasks activate independently upon manual override, all tasks handle direct shared knowledge when left to greedy decoding, each task issued multiple tests alongside to determine fidelity and accuracy throughout the process.
The results show the gating is more than willing to hop from sector to sector, using alternating weight shifts from the cooperative anchored systems - even systems never trained for the tasks contributing to the accuracy of the results for other tasks due to the lookup accuracy to the heuristic chains, never having seen the tasks before. Each structure is independently trained and the collective cooperates together through a dense activation network.
Full writeup and article https://huggingface.co/blog/AbstractPhil/aleph-autoregression-differentiation-ft2.
This is akin to a stackable non-intrusive lora that enables increased shared collective behavior.
This includes the three mentioned json tasks, a math task, a tinystories task, and a diffusion task for cifar10. Each adapter anchored to the knowledge within model that already exists while enhancing the knowledge through anchored lookup systems and decision-driven hierarchical access trees.
All tasks activate independently upon manual override, all tasks handle direct shared knowledge when left to greedy decoding, each task issued multiple tests alongside to determine fidelity and accuracy throughout the process.
The results show the gating is more than willing to hop from sector to sector, using alternating weight shifts from the cooperative anchored systems - even systems never trained for the tasks contributing to the accuracy of the results for other tasks due to the lookup accuracy to the heuristic chains, never having seen the tasks before. Each structure is independently trained and the collective cooperates together through a dense activation network.
Full writeup and article https://huggingface.co/blog/AbstractPhil/aleph-autoregression-differentiation-ft2.
- 4 replies
reacted to Quazim0t0's post with 🤯🔥 7 days ago
Post
4259
I never really posted about my DaisyChain project because it's still work in progress. I decided to post a small bit about it and the demo.
DaisyChain Genomics: four small DNA/RNA specialists chained behind a learned router that behave like one big genomics model, at ~7× less active compute.
I built a modular genomics model chasing a 500M-parameter foundation model, then caught myself measuring it wrong. Here's the honest version.
DaisyChain is a different bet: instead of one monolithic DNA model, it's four ~74M specialists (eukaryote, prokaryote, mRNA, splice) chained behind a learned router, each distilled per-domain from HuggingFaceBio's Carbon-500M. Every specialist reports how surprised it is (bits/base) and the router hands each sequence to the link most at home with it. In lineage it's a cluster Branch-Train-Merge mixture of experts, so you can chain on a new domain without retraining the others.
The pitch: ~295M total params (under Carbon-500M), but only one ~74M specialist runs per query, so ~7× cheaper per token, routing at 100% held-out.
The mistake: Carbon works in 6-mers, and I'd been scoring likelihood as 6-mer cross-entropy. By that number I was +0.043 bits/base behind, splice even "beating" Carbon. But Carbon scores at the base-pair level, which is harder and more honest. Re-run their way:
Real gap: 1.862 vs 1.787 bits/base, +0.089 behind, not +0.043
No domain actually beats Carbon; the "splice win" was an artifact
Seq recovery: euk 31.5% vs 38.9%, bacteria 40.9% vs 54.1%
DaisyChain is still behind Carbon-500M (itself a draft model, not built to top benchmarks), but by a number I can defend, and the gap closes with every per-domain pass. 🌼
DaisyChainAI
DaisyChainAI/Daisychain-Genomics-Demo
DaisyChainAI/daisychain-genomics
DaisyChain Genomics: four small DNA/RNA specialists chained behind a learned router that behave like one big genomics model, at ~7× less active compute.
I built a modular genomics model chasing a 500M-parameter foundation model, then caught myself measuring it wrong. Here's the honest version.
DaisyChain is a different bet: instead of one monolithic DNA model, it's four ~74M specialists (eukaryote, prokaryote, mRNA, splice) chained behind a learned router, each distilled per-domain from HuggingFaceBio's Carbon-500M. Every specialist reports how surprised it is (bits/base) and the router hands each sequence to the link most at home with it. In lineage it's a cluster Branch-Train-Merge mixture of experts, so you can chain on a new domain without retraining the others.
The pitch: ~295M total params (under Carbon-500M), but only one ~74M specialist runs per query, so ~7× cheaper per token, routing at 100% held-out.
The mistake: Carbon works in 6-mers, and I'd been scoring likelihood as 6-mer cross-entropy. By that number I was +0.043 bits/base behind, splice even "beating" Carbon. But Carbon scores at the base-pair level, which is harder and more honest. Re-run their way:
Real gap: 1.862 vs 1.787 bits/base, +0.089 behind, not +0.043
No domain actually beats Carbon; the "splice win" was an artifact
Seq recovery: euk 31.5% vs 38.9%, bacteria 40.9% vs 54.1%
DaisyChain is still behind Carbon-500M (itself a draft model, not built to top benchmarks), but by a number I can defend, and the gap closes with every per-domain pass. 🌼
DaisyChainAI/Daisychain-Genomics-Demo
DaisyChainAI/daisychain-genomics
reacted to Quazim0t0's post with 🚀🤯 7 days ago
Post
5003
🌼 DaisyChain-Web: train a language model with friends or by yourself with multiple devices, in the browser, no install
Open a webpage, share a room link, and every device that joins becomes part of the training cluster. Phones, laptops, old PCs: they connect peer-to-peer over WebRTC and train one shared transformer together, entirely in the browser.
What's actually happening under the hood:
🧠 A mini transformer LM trains on FineWeb-Edu, streamed live from the HuggingFace Hub. Each device pulls its own slice (data parallelism), tokenized with our 16.5k-token Spikewhale tokenizer
⚡ Every single multiply runs through verified INT8 neural units, no float fallback. On WebGPU browsers it uses the GPU's DP4A integer dot-product hardware, admitted only after proving bit-identical results against the verified units, with a 3×INT8 fast-accurate scheme (CUTLASS's 3xTF32 trick, ported to 8-bit)
🔒 Devices average gradients every step under a sync guard: a per-step roster protocol plus weight-hash verification keeps every device's model bit-identical. If anything drifts, training stops instead of silently forking
📊 Live logs show exactly what every device contributes, step by step
💾 When you're done: test generations right on the page, download a checkpoint, or grab the inference kit, a single self-contained HTML file with the weights baked in that runs generations offline, anywhere
Works solo too. Every extra device just grows the effective batch.
👉 Try it: Quazim0t0/DaisyChain-Web
🛠 Training framework: DaisyChainAI/DaisyChain-Train
Updates:
- Block-scaled INT8 quantization
- Batched attention GEMM
- Fused dequant+ReLU epilogue
- Weight-tied unembedding
- WebSocket relay fallback
- Server keepalive ping/pong every 30s
- disconnected-state redial
- Visibility/network-change reconnect
Open a webpage, share a room link, and every device that joins becomes part of the training cluster. Phones, laptops, old PCs: they connect peer-to-peer over WebRTC and train one shared transformer together, entirely in the browser.
What's actually happening under the hood:
🧠 A mini transformer LM trains on FineWeb-Edu, streamed live from the HuggingFace Hub. Each device pulls its own slice (data parallelism), tokenized with our 16.5k-token Spikewhale tokenizer
⚡ Every single multiply runs through verified INT8 neural units, no float fallback. On WebGPU browsers it uses the GPU's DP4A integer dot-product hardware, admitted only after proving bit-identical results against the verified units, with a 3×INT8 fast-accurate scheme (CUTLASS's 3xTF32 trick, ported to 8-bit)
🔒 Devices average gradients every step under a sync guard: a per-step roster protocol plus weight-hash verification keeps every device's model bit-identical. If anything drifts, training stops instead of silently forking
📊 Live logs show exactly what every device contributes, step by step
💾 When you're done: test generations right on the page, download a checkpoint, or grab the inference kit, a single self-contained HTML file with the weights baked in that runs generations offline, anywhere
Works solo too. Every extra device just grows the effective batch.
👉 Try it: Quazim0t0/DaisyChain-Web
🛠 Training framework: DaisyChainAI/DaisyChain-Train
Updates:
- Block-scaled INT8 quantization
- Batched attention GEMM
- Fused dequant+ReLU epilogue
- Weight-tied unembedding
- WebSocket relay fallback
- Server keepalive ping/pong every 30s
- disconnected-state redial
- Visibility/network-change reconnect
- 11 replies