alkinun

AtAndDev

22 150 767

alkinun

AI & ML interests

decentralize

Recent Activity

reacted to Bc-AI's post with 🔥 about 2 hours ago

@juiceb0xc0de @ProCreations @Keeby-smilyai @MUK-IS-GOAT @Sbui503 @Fishtiks Hello Beta testers! I will release our new model in our beta testers organisation as soon as I can! Thanks for being part of our beta testers community!

liked a model about 2 hours ago

SupraLabs/Supra2-Medium-Base

liked a model about 2 hours ago

SupraLabs/Supra2-Medium-Instruct

View all activity

Organizations

reacted to Bc-AI's post with 🔥 about 2 hours ago

Post

2313

6 replies

reacted to Banaxi-Tech's post with 🚀👀🔥 1 day ago

Post

3263

We're releasing BananaMind 2.1 Unified, a 35M three-tower relay model where the two output towers can only talk to each other through a silent middle tower that has no output head and no loss term.

Tower B trains entirely on indirect gradient. It was never told what to predict. It woke up anyway. Jacobian lens shows it carrying the correct answer ("Paris", "oxygen", "blue") at its deepest layer. Its bridge gates grew 4-47x from init. Feed it from only one side and the representations collapse to junk — it needs both outer towers to become semantic.

The PIQA result is the cleanest demonstration: Tower A alone scores 50.11 (chance). Tower C alone 52.12. Full system 61.75. All physical reasoning lives in the integration. The 35M three-tower beats the 50M single-tower BananaMind 2 Medium on PIQA.

Trained on 38B tokens in ~7.5 hours on 8x RTX PRO 6000. Ships with 7 ablation modes so you can surgically cut the model apart without retraining. Full training logs, J-lens fits, and eval outputs for every mode included.

This is a research model. It is interesting because you can take it apart.
BananaMind/BananaMind-2.1-Unified

Follow us for more:

BananaMind

@Banaxi-Tech

9 replies

reacted to Felladrin's post with 👀🔥👍 1 day ago

Post

4044

I've open-sourced the trainer I've been using to build tiny language models from scratch, together with the 95M base model I trained with it.

The trainer runs on Deno (https://deno.com, cross-platform), trains on WebGPU, and it writes GGUF directly. No Python/PyTorch. The weights live in a GGUF file from the first step to the last, so every checkpoint is already something llama.cpp can load.

The model is Felladrin/Minueza-3-95M-Base: 94.7M parameters, 1.95B tokens seen, 8192 context.

And here’s the repository on GitHub: https://github.com/felladrin/gguf-trainer

Here on Hugging Face, I published the optimizer state next to the weights, so you can continue the pretraining instead of starting over.

Or start your own from nothing: deno run -A cli.ts demo trains a tiny one end to end in under a minute.

And the docs are written for coding agents, so you can point your agent of choice at the GitHub repo and have it drive the whole pipeline.

36 replies

replied to Enderchef's post 1 day ago

ill be on this list soon!
hopefully

reacted to Enderchef's post with 👀🚀🤗🔥 1 day ago

Post

2721

Do you support the SLM(1M-150M parameter) community?
If so, join the SLM discord(https://discord.gg/BBYaERvvn), and give these orgs some follows!

BananaMind

SupraLabs

fromziro

AxiomicLabs

2 replies

reacted to ProCreations's post with 👍 2 days ago

Post

3346

I have hit 300 followers, and I think this calls for a bit of a giveaway 👀 a unique one, too.
I have had countless AI projects I have wanted to make but have been (brutally) blocked by compute. Now that I finally have just enough compute to sort of get around (i still don't have enough 😭) and for hitting 300 followers (tysm!) I will be funding three of the communities projects via HuggingFace jobs, giving them 150 dollars max worth of compute each. I will personally be picking the winners, I am looking for projects that genuinely hit the compute wall: great ideas, blocked by compute, just like the countless ideas I've had.
To join, head over to https://giveaway.ssh.codes
RULES:
- Final result must be open weight or open source
- Only one submission per person
- Have fun!

34 replies

reacted to etemiz's post with 🔥 4 days ago

Post

2915

made two new "surgeries" on Qwen 3.8

etemiz/Ostrich-27B-Qwen3.8-260815

etemiz/Ostrich-27B-Qwen3.8-260816-Abliterated

these are past loras, applied to new model. they work because they have similar architecture and 3.8 itself is a further training of previous versions.

reacted to JonathanColetti's post with 🔥 5 days ago

Post

5474

I just uncensored qwen 3.8 27B. You can check out the demo here: JonathanColetti/Qwen3.8-27B-Uncensored-Demo or the full repo: JonathanColetti/Qwen3.8-27B-Uncensored-GGUF

What I did was used an H100 NVL and use https://github.com/p-e-w/heretic to uncensor it. I measured the KL divergence and the refusal rate (Refusal rate is the count of refusals over 100 hold out prompts from mlabonne/harmful_behaviors). Check out the full repo and give it a like if you think its cool

Thanks!

1 reply

reacted to appvoid's post with 🔥 9 days ago

Post

854

- GLM 5.2
- Flux 3
- New Qwen model
- New small model leaderboards
- Lots of people finetuning smol models.
- Some even under 12 year olds clauders are here (was not on my bingo card this year)
- ChatGPT's Sol became a lot faster this week
- LFM2.5 2.6b
- Kimi K3 (though only a few will run it)
- New Ling 3.0 Tiny
- New video model that is making south park videos?
- Deepseek v4 flash being more honest than bigger models
- The new model from meta

Everything Everywhere All At Once

1 reply

reacted to wop's post with 🔥 10 days ago

Post

2163

bench-labs/GCTokenizer-v1 , a multilingual tokenizer which does not require a training corpus

bench-labs developed **GCTokenizer-v1**, which is a multi-lingual tokenizer
Available in four sizes: 32K, 65K, 131K and 262K tokens "S, M, L, XL"
It utilizes an encoding scheme which allows it to handle characters in any language around the world

General (multi lingual)
Consensus (from multiple model tokenizers consensus)
Tokenizer

We included an implementation script too,
built like BPE- it can encode arbitrary text, most of the time, efficiently

replied to SoulInPsyAbstract's post 14 days ago

holy chatgpt

reacted to JonnaMat's post with 🔥 14 days ago

Post

3152

🚗 The reasoning backbone quadruples from 8B to 32B , while the action expert remains at 2.3B!

👀 We took a closer look at the architectural evolution from nvidia/Alpamayo-1.5-10B to nvidia/Alpamayo2-Super .

Read the analysis here:
https://huggingface.co/blog/JonnaMat/alpamayo2-super

Our analysis explores some implications of this design choice, especially from a distillation perspective where keeping the expert compact could be key for efficient deployment. 🧠

3 replies

replied to SoulInPsyAbstract's post 14 days ago

holy chatgpt

alkinun

AI & ML interests

Recent Activity

Organizations

AtAndDev's activity