Matricardi Fabio's picture

Matricardi Fabio

FM-1976

·

https://medium.com/@fabio.matricardi

AI & ML interests

control system engineering, AI, LLM with python. ThePoorGPUguy on substack

Recent Activity

liked a model about 20 hours ago

Tiiny/SmallThinker-3B-Preview

liked a model 5 days ago

LiquidAI/LFM2-2.6B-Exp

reacted to codelion's post with 🚀 5 days ago

Introducing Dhara-70M: A diffusion language model that achieves 3.8x higher throughput than autoregressive models! Key findings from our research on optimal architectures for small language models: → Depth beats width: 32 layers outperforms 12 layers at the same parameter count → Best-in-class factuality: 47.5% on TruthfulQA → 10x training efficiency using WSD (Warmup-Stable-Decay) conversion → Canon layers add only 0.13% parameters but improve reasoning We trained on 1B tokens using the optimal 50-30-20 dataset mix (PDFs + filtered web + educational content), then converted to diffusion with just 100M additional tokens. Blog: https://huggingface.co/blog/codelion/optimal-model-architecture Model: https://huggingface.co/codelion/dhara-70m

View all activity

Organizations

None yet

upvoted a paper 7 days ago

LoPA: Scaling dLLM Inference via Lookahead Parallel Decoding

Paper • 2512.16229 • Published 13 days ago • 15

upvoted an article 19 days ago

Article

Ellora: Enhancing LLMs with LoRA - Standardized Recipes for Capability Enhancement

28 days ago

•

12

upvoted a collection 23 days ago

RWKV7-G0a-G1a-GGUF

gguf of RWKV7 models • 6 items • Updated 7 days ago • 7

upvoted a collection 28 days ago

Ovis-U1

An unified model for multimodal understanding, text-to-image generation, and image editing. • 3 items • Updated Jul 2 • 6

upvoted a collection 3 months ago

Granite 4.0 Language Models

13 items • Updated Nov 17 • 198

upvoted 2 collections 5 months ago

GLiNER-decoder

A joint encoder-decoder GLiNER model for a scalable open-ontology entity recognition • 3 items • Updated Aug 15 • 16

SD-prompter

A series of models designed to output Stable-diffusion prompts when given a character appearance. Made for the CharGen Project. • 6 items • Updated Sep 28 • 3

upvoted a collection 7 months ago

MiniCPM4

MiniCPM4: Ultra-Efficient LLMs on End Devices • 29 items • Updated Sep 8 • 82

upvoted 2 papers 8 months ago

ReplaceMe: Network Simplification via Layer Pruning and Linear Transformations

Paper • 2505.02819 • Published May 5 • 26

Practical Efficiency of Muon for Pretraining

Paper • 2505.02222 • Published May 4 • 40

upvoted an article 9 months ago

Article

Bamba: Inference-Efficient Hybrid Mamba2 Model

+27

Dec 18, 2024

•

60

upvoted a collection 9 months ago

🌙 March 2025 - Open releases from the Chinese community

31 items • Updated Nov 20 • 13

upvoted a collection 10 months ago

RWKV v7

9 items • Updated Mar 17 • 6

upvoted 2 papers 10 months ago

How far can we go with ImageNet for Text-to-Image generation?

Paper • 2502.21318 • Published Feb 28 • 26

Chain of Draft: Thinking Faster by Writing Less

Paper • 2502.18600 • Published Feb 25 • 50

upvoted 3 papers 11 months ago

Large Language Diffusion Models

Paper • 2502.09992 • Published Feb 14 • 123

The Danger of Overthinking: Examining the Reasoning-Action Dilemma in Agentic Tasks

Paper • 2502.08235 • Published Feb 12 • 58

FoNE: Precise Single-Token Number Embeddings via Fourier Features

Paper • 2502.09741 • Published Feb 13 • 15

upvoted 2 collections 12 months ago

Aira

Aira is a series of chatbots developed as an experimentation playground for value alignment. • 27 items • Updated Jun 20, 2024 • 1

Loxa

a Loxa family models are best models to running on CPU and GPU with high quality(=>92% accuracy) • 5 items • Updated Feb 3 • 2