huggingface.co Training Cluster as a service: Train your LLM at scale on our infrastructure
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
Instance Notes
Please review our community rules and introduce yourself!
Useful links
Members
2
Posts
35
Active Today
1
Created
2 yr. ago
-
Machine Learning - Training | Fine Tuning @lemmy.intai.tech manitcor @lemmy.intai.tech 
codeconfessions.substack.com LZ77 Is All You Need? Why Gzip + KNN Works for Text ClassificationDecoding the Success of Gzip + KNN: The Central Role of LZ77
cross-posted from: https://lemmy.world/post/2706141
Decoding the Success of Gzip + KNN: The Central Role of LZ77
-
Machine Learning - Training | Fine Tuning @lemmy.intai.tech manitcor @lemmy.intai.tech 
github.com GitHub - NVIDIA/cutlass: CUDA Templates for Linear Algebra SubroutinesCUDA Templates for Linear Algebra Subroutines. Contribute to NVIDIA/cutlass development by creating an account on GitHub.
-
Machine Learning - Training | Fine Tuning @lemmy.intai.tech manitcor @lemmy.intai.tech New trick scales LLMs even longer! - GitHub - jquesnelle/scaled-rope
NTK-Aware Scaled RoPE allows LLaMA models to have extended (8k+) context size without any fine-tuning and minimal perplexity degradation.
News I've seen the posts about SuperHOT and just recently, the paper from Meta which uses RoPE interpolation, and I've noticed an immediate improvement that can be brought to this method. Basically if you apply Neural Tangent Kernel (NTK) theory to this problem, it becomes clear that simply interpolating the RoPE's fourier space "linearly" is very sub-optimal, as it prevents the network to distinguish the order and positions of tokens that are v
-
Machine Learning - Training | Fine Tuning @lemmy.intai.tech manitcor @lemmy.intai.tech 
github.com GitHub - teknium1/GPTeacher: A collection of modular datasets generated by GPT-4, General-Instruct - Roleplay-Instruct - Code-Instruct - and ToolformerA collection of modular datasets generated by GPT-4, General-Instruct - Roleplay-Instruct - Code-Instruct - and Toolformer - GitHub - teknium1/GPTeacher: A collection of modular datasets generated ...
-
Machine Learning - Training | Fine Tuning @lemmy.intai.tech manitcor @lemmy.intai.tech 
www.preemo.io Preemo | Fine Tune Foundational ModelsFine tune and deploy generative models via a single API
-
Machine Learning - Training | Fine Tuning @lemmy.intai.tech manitcor @lemmy.intai.tech 
github.com GitHub - google-research/google-research: Google ResearchGoogle Research. Contribute to google-research/google-research development by creating an account on GitHub.
All datasets in this repository are released under the CC BY 4.0 International license, which can be found here: https://creativecommons.org/licenses/by/4.0/legalcode. All source files in this repository are released under the Apache 2.0 license, the text of which can be found in the LICENSE file.
-
Machine Learning - Training | Fine Tuning @lemmy.intai.tech manitcor @lemmy.intai.tech Relationship between LLM model size and emergent power
zhuanlan-zhihu-com.translate.goog 大语言模型的涌现能力：现象与解释注1：本文整理自我在今年3 月 11 日 “中国人工智能学会”主办的「ChatGPT 及大模型专题研讨会」上《大型语言模型的涌现能力：现象与解释》的现场分享，介绍了大语言模型中的涌现现象，以及关于涌现能力背后原因的…
-
Machine Learning - Training | Fine Tuning @lemmy.intai.tech manitcor @lemmy.intai.tech 
github.com GitHub - ggerganov/ggml: Tensor library for machine learningTensor library for machine learning. Contribute to ggerganov/ggml development by creating an account on GitHub.
-
Machine Learning - Training | Fine Tuning @lemmy.intai.tech manitcor @lemmy.intai.tech 
github.com GitHub - microsoft/LMOps: General technology for enabling AI capabilities w/ LLMs and MLLMsGeneral technology for enabling AI capabilities w/ LLMs and MLLMs - GitHub - microsoft/LMOps: General technology for enabling AI capabilities w/ LLMs and MLLMs
-
Machine Learning - Training | Fine Tuning @lemmy.intai.tech manitcor @lemmy.intai.tech GitHub - dadukhankevin/Finch: A Keras style GA genetic algorithm library
github.com GitHub - dadukhankevin/Finch: Finch is a genetic algorithm framework for python. It is also GPU compatible, making it probably the fastest genetic algorithm framework ever. Inspired by Keras.Finch is a genetic algorithm framework for python. It is also GPU compatible, making it probably the fastest genetic algorithm framework ever. Inspired by Keras. - GitHub - dadukhankevin/Finch: Fin...
-
Machine Learning - Training | Fine Tuning @lemmy.intai.tech manitcor @lemmy.intai.tech 
github.com GitHub - philipturner/metal-flash-attention: Faster alternative to Metal Performance ShadersFaster alternative to Metal Performance Shaders. Contribute to philipturner/metal-flash-attention development by creating an account on GitHub.
