if that happens, a Pirate Bay for AI models will rise from the ashes of HugginFace in record time
1
3
317
Esto es lo que hay que hacer en México y no sueños guajiros y gastos sin sentido. Los de la compañía de TI municipal del gobierno de la ciudad de Río de Janeiro han creado un modelo a partir de Qwen 3.5 Y han logrado buenos números en rendimiento. Está publicado en Hugginface. huggingface.co/prefeitura-ri…
1
1
139
توی hugginface زیاده نسخه قفل شکسته شده س. من دارم و جالبه.
معروف ترین ش سری abliterated هست.
gemma 4 abliterated
هست که نسخه ی e2b از مود q4 روی لپتاپ معمولی که 8گیگ vram gpu داره راحت جواب میده.
قبلا dolphin3 استفاده میکردم که اونم بد نبود اما gemma4 مدیا رو هم پردازش میکنه
2
1
169
Jun 13
43
Jun 11
this is another simple one by fable 5 in @threejs but take a look at how it was made: (especially side fins animated separately)
1- we created an image of a fish with grok imagine
2- then used trellis 2 at hugginface to create a 3d model and download its glb
3- fable then put the 3d model into the scene which was a single fused mesh
4- then it used some masking or something to animate the side fins separately
i don't know but it felt pretty AGI :)
below you can find the link to download this and 190 others from my patreon, and some other useful and cool links like a fable playground and my get amplified series
Jun 11
it is simple but fable did this one nicely using @threejs
download link and some other stuff in the comments like fable interactive simulation playground
2
2
68
10,756
Jun 11
So I actually cooked up a video editing workspace that uses ffmpeg hyperframes to do video editing. I was actually amazed at what it could do. It can do a lot of the stuff with just local models downloaded from hugginface basically. So maybe not as expensive as I thought...
1
1
55
You can use hugginface api for the data. That is what I did with LlamaStudio which i it's just a wrapper on llama.cpp. Or fork this and start there.
github.com/gnulnx/LlamaStudi…
1
3
330
Jun 10
Maturity is realizing you dont need to purchase any claude openai subscription, hugginface, Llama, openrouter is enough❤️❤️❤️
1
38
Jun 5
do you have any of your models on hugginface? I will like to try one.
1
170
Jun 4
Erdi bey öyle resimlerle biz tongaya düşmeyiz. Alican kardeşimizi de kimseye yedirmeyiz. ;)) Sizi hugginface model paylaşımlarına bekliyoruz.
187
Jun 4
Tech report: research.nvidia.com/labs/nem…
@Hugginface base model: huggingface.co/nvidia/NVIDIA…
HF BF16 post-trained: huggingface.co/nvidia/NVIDIA…
HF NVFP4 post-trained: huggingface.co/nvidia/NVIDIA…
82
Jun 2
With a dgx spark, hermes and an account on hugginface, Imagine what you can do in 2026.
It's crazy !
Add another computer with proxmox and some localhost apps like forgejo and you're good.
A frontier model for planning and building, and localhost models for tasks & workflows.
22
May 31
Afortunadamente, encontré uno que sí funcionaba, aunque lo descargué de páginas oficiales como Hugginface y otros creados por autores de CivitAI, algunos funcionaba y otros no. Ahora estoy probando los demás modelos y tomando notas para evitar problemas en el futuro.
1
133
🚀Output:
🔑This same pattern works for Groq, OpenAI, Anthropic and Open source models like HugginFace.
✨just swap the importn and initialization of the model
#LangChain #GoogleGemini #Python #GenAI #LearningInPublic
32
May 26
ASML, Carl Zeiss, Siemens, Schaefer und viele andere.
Direkt AI. Hugginface und Mistral. Deepmind in London darf man nicht vergessen. Gehoert aber Google.
Leider keine richtige grosse Startupszene im Bereich AI. Eine Menge Kleinkram .. ausser N8N kein richtig grosser Player
1
113
May 23
Broo! You wont believe this. Follow mradermacher on hugginface!
1
1
34
May 19
Are PRISM-PRO BF16 weights for 27B available? I've been waiting for it, didn't see it on your hugginface yet.
I'll see if I can get the drafter running on my llama-cpp fork 👍
1
210
May 13
Celui là est déjà inutile : les nœuds provisoires sont disponibles sur le nightbuild. Et Kijai a quantizé les poids sur le HugginFace Comfy.
Le workflow fonctionne, mais c'est idem que le HF space : très, très moche de près (même en BF16).
34