veganism.social is one of the many independent Mastodon servers you can use to participate in the fediverse.
Veganism Social is a welcoming space on the internet for vegans to connect and engage with the broader decentralized social media community.

Administered by:

Server stats:

230
active users

#machine

5 posts3 participants0 posts today

70% Size, 100% Accuracy: Lossless LLM Compression via Dynamic-Length Float

arxiv.org/abs/2504.11651

#HackerNews #Lossless #LLM #Compression #Dynamic-Length #Float #AI #Research #Machine #Learning

arXiv logo
arXiv.org70% Size, 100% Accuracy: Lossless LLM Compression for Efficient GPU Inference via Dynamic-Length FloatLarge Language Models (LLMs) have grown rapidly in size, creating significant challenges for efficient deployment on resource-constrained hardware. In this paper, we introduce Dynamic-Length Float (DFloat11), a lossless compression framework that reduces LLM size by 30% while preserving outputs that are bit-for-bit identical to the original model. DFloat11 is motivated by the low entropy in the BFloat16 weight representation of LLMs, which reveals significant inefficiency in existing storage format. By applying entropy coding, DFloat11 assigns dynamic-length encodings to weights based on frequency, achieving near information-optimal compression without any loss of precision. To facilitate efficient inference with dynamic-length encodings, we develop a custom GPU kernel for fast online decompression. Our design incorporates the following: (i) decomposition of memory-intensive lookup tables (LUTs) into compact LUTs that fit in GPU SRAM, (ii) a two-phase kernel for coordinating thread read/write positions using lightweight auxiliary variables, and (iii) transformer-block-level decompression to minimize latency. Experiments on recent models, including Llama-3.1, Qwen-2.5, and Gemma-3, validates our hypothesis that DFloat11 achieves around 30% model size reduction while preserving bit-for-bit exact outputs. Compared to a potential alternative of offloading parts of an uncompressed model to the CPU to meet memory constraints, DFloat11 achieves 1.9-38.8x higher throughput in token generation. With a fixed GPU memory budget, DFloat11 enables 5.3-13.17x longer context lengths than uncompressed models. Notably, our method enables lossless inference of Llama-3.1-405B, an 810GB model, on a single node equipped with 8x80GB GPUs. Our code and models are available at https://github.com/LeanModels/DFloat11.

Is the #brain a #machine? Much of the enthusiasm in the so-called "AI research field" is founded on a #Nietzsche-like #atheism. But a number of researchers over the past 100 years have argued that the non-mechanical features of #quantum #coherent structures may place a role in human consciousness (which are not in the proposed 'quantum computers'). It may be a question of degree: #EEG alpha has more coherence than EEG beta, and perhaps more sensitive to tiny fluctuations eg at the quantum level.

Apple's latest 🍏 "revolutionary" announcement is like a black box of buzzwords: Differential #Privacy, Aggregate Trends, and #Machine #Learning 🤖. Apparently, they're protecting your privacy by knowing just enough about you to sell you more stuff. It's like their #innovation department found a #thesaurus and thought: "Let's make privacy sound complicated and groundbreaking!" 🚀🙄
machinelearning.apple.com/rese #Apple #Buzzwords #HackerNews #ngated

Apple Machine Learning ResearchUnderstanding Aggregate Trends for Apple Intelligence Using Differential PrivacyAt Apple, we believe privacy is a fundamental human right. And we believe in giving our users a great experience while protecting their…