Building on HF

676 1232 4696

Victor Mustar PRO

victor

victormustar

AI & ML interests

Building the UX of this website

Recent Activity

reacted to danielhanchen's post with 🚀 22 minutes ago

You can now run Kimi K2.5 locally! 🔥 We shrank the 1T model to 240GB (-60%) via Dynamic 1-bit. Get >40 tok/s on 242GB or 622GB VRAM/RAM for near full precision. GGUF: https://huggingface.co/unsloth/Kimi-K2.5-GGUF Guide: https://unsloth.ai/docs/models/kimi-k2.5

liked a model about 2 hours ago

unsloth/Kimi-K2.5-GGUF

liked a Space about 4 hours ago

SammyLim/VideoMaMa

View all activity

Organizations

reacted to danielhanchen's post with 🚀 22 minutes ago

Post

You can now run Kimi K2.5 locally! 🔥

We shrank the 1T model to 240GB (-60%) via Dynamic 1-bit.
Get >40 tok/s on 242GB or 622GB VRAM/RAM for near full precision.

GGUF: unsloth/Kimi-K2.5-GGUF

Guide: https://unsloth.ai/docs/models/kimi-k2.5

liked a model about 2 hours ago

unsloth/Kimi-K2.5-GGUF

1T • Updated about 2 hours ago • 56

liked a Space about 4 hours ago

VideoMaMa

⚡

Remove video backgrounds and generate matte videos

upvoted a collection about 5 hours ago

Trinity-Large

Collection

4 items • Updated about 17 hours ago • 26

liked a model about 7 hours ago

zai-org/GLM-4.5-Air-FP8

Text Generation • 111B • Updated Aug 12, 2025 • 142k • • 77

reacted to tegridydev's post with ❤️ about 18 hours ago

Post

1016

Introducing OpenMALx

openmalx

Repository for Infosec and Machine Learning Resources

OpenMALx is an organization focused on the development of datasets and models for security analysis. The project objective is to provide structured data for training and evaluating large language models in a security context.

---

Technical Focus

**Dataset Formatting:** Processing raw security tool logs into instruction/response pairs for model training.
**Local Execution:** Optimizing models for local hardware to ensure data remains on-premises.
**Response Logic:** Developing structured formats for explaining security vulnerabilities and remediation steps.

Active Projects

**infosec-tool-output:** A dataset mapping static and dynamic analysis tool outputs to technical summaries.
openmalx/infosec-tool-output

**open-malsec:** A collection of text-based security threats, including phishing and social engineering samples, for classification tasks.
openmalx/open-malsec

replied to kelsend's post about 18 hours ago

will it be open sourced?

replied to schuler's post about 19 hours ago

Cool! can I use from my CLI or must do some python scripting?

reacted to danieldk's post with 🔥 about 19 hours ago

Post

1064

kernels 0.12 is out! 🎉

Changes:

* Support for kernel version branches to gracefully roll out kernel API changes.
* Support for PyTorch 2.10.
* kernel-builder is now merged into the kernels repo.
* Initial support for standardized kernel benchmarks.

https://github.com/huggingface/kernels/releases/tag/v0.12.0

upvoted an article about 19 hours ago

Article

Geometric Manifold Walking: Stable High-Accuracy Multi-Encoder Fusion Without Backbone Training

Dec 25, 2025

•

liked a model about 19 hours ago

AbstractPhil/sd15-flow-matching

Text-to-Image • Updated about 17 hours ago • 455 • 4

reacted to AbstractPhil's post with 👀 about 19 hours ago

Post

752

AbstractPhil/tinyflux-experts
Introducing the "blot" expert, sd15-flow-sol. The twin sister flow-matching experts for tinyflux-lailah; sd15-flow-lune AND sd15-flow-sol will be used in tandem to train tinyflux-Lailah. sd15-flow-sol never managed to reach full flow-matching prediction, so epsilon vpred conversion is required. All experts will exist within the tinyflux-experts repo, including all the critical checkpoint sets.
Lune was heavily finetuned in the sd3-style and adapted shift timestep system after David's interpolation converted sd15 into geometric basis.
Sol was left abandoned after 50 epochs with David and was considered overcooked and rigid, until I noticed the geometric structure today. Lune doesn't produce geometric structure as solid as Sol, not even close. Lune produces improved fidelity and detail, but Sol produces something very very different, aligned to sd15's behavior, and fully representative of the 5point 4simplex structure that David brought to the table.

Sol is essentially a nearly perfect blob-forming geometric blotter. Sol is SD15, and yet SOL was trained using a specific pattern recognizing and timestep aligned David model. David was tasked with classifying timesteps and patterns using complex deep-recognition structural analysis layer-by-layer, determining full-scale opinions after watching the entirety of sd15's structure during training.

Even though the sd15-flow-sol was left abandoned, the structure of Sol is HIGHLY effective at understanding TIMESTEP blotting interpolation. I didn't realize how crucially important this was until Lailah started to show rigidity and compartmentalized behavior with sequence - which likely happens to ALL flow matching models.

AbstractPhil/sd15-flow-matching

AbstractPhil/geo-david-collective-sd15-distilled
AbstractPhil/geo-david-collective-sd15-base-e40