PyTorch
PulseAugur coverage of PyTorch — every cluster mentioning PyTorch across labs, papers, and developer communities, ranked by signal.
13 天有情绪数据
-
Hugging Face 推出用于高效 LLM 的先进量化技术
研究人员正在开发先进的量化技术，以提高大型语言模型 (LLM) 的效率。AutoRound、LATMiX 和 GSQ 等新方法旨在减小模型大小和计算需求，从而能够在功能较弱的硬件上进行部署。这些方法侧重于优化模型权重和激活在较低比特宽度下的表示方式，其中一些方法已达到与更高精度模型相当的准确性。创新包括用于训练后量化的新颖校准策略和用于提高鲁棒性的可学习仿射变换。
-
Safetensors library audited as secure, set to become default for ML models
The safetensors library, developed by Hugging Face in collaboration with EleutherAI and Stability AI, has undergone a security audit by Trail of Bits, confirming its safety. This audit allows the organizations to move t…
-
Eugene Yan reviews 2022, detailing career growth, writing goals, and investment thesis
Eugene Yan's 2022 review highlights personal and professional achievements, including writing 18 posts on technical topics like text-to-image and machine learning techniques. He was promoted from L5 to L6, focusing on M…
-
Replit 通过新的缓存推出启动速度提高 100 倍的 Python repl
Replit 通过实施新的缓存机制，显著提高了新 Python repl 的启动速度。此次更新解决了之前导致某些 Python 环境无法使用的大型软件包尺寸和漫长安装时间的问题。新系统利用软件包内单个文件的内容寻址缓存，允许使用符号链接而不是完整副本，这大大减少了磁盘空间使用量并加快了 repl 的初始化速度。
-
Eugene Yan shares strategies for continuous machine learning education
Eugene Yan's essay offers practical advice for staying current in the rapidly evolving field of machine learning. He suggests actively experimenting with new tools and techniques in projects, sharing learnings with coll…
-
Eugene Yan: MOOCs offer diminishing returns; real learning comes from doing
Eugene Yan argues that while Massive Open Online Courses (MOOCs) can be useful for initial learning, they often lead to diminishing returns and can even become a form of procrastination. He suggests that true learning, …
-
Data scientists can avoid role mismatches by carefully vetting job descriptions and interview questions.
Eugene Yan's article advises data science professionals on how to navigate potential mismatches between their job title and actual responsibilities. He suggests carefully reviewing job descriptions, asking targeted ques…
-
Spark+AI Summit 2020：笔记涵盖特征工程、数据质量和模型效率
Eugene Yan 撰写的 Spark+AI Summit 2020 笔记涵盖了深度学习和数据工程中的实际应用和通用性会谈。特定应用会话重点介绍了 Airbnb 的 Zipline 等特征工程框架和 Sputnik 数据工程框架，以及 Gojek 的 Feast 和 Netflix 的数据质量方法。通用性会谈则侧重于通过模型剪枝、量化和蒸馏等技术提高深度学习效率，并引用了 IBM 和 Instagram 的示例。
-
OpenAI standardizes on PyTorch to boost research productivity and GPU performance
OpenAI has announced its standardization on the PyTorch deep learning framework to enhance research productivity and streamline the development of optimized model implementations. This strategic shift aims to reduce ite…
-
Eugene Yan enhances recommender systems using graph and NLP techniques
Eugene Yan's blog posts detail methods for building recommender systems that outperform baseline matrix factorization models. The approach involves using Natural Language Processing (NLP) techniques, specifically word2v…
-
探讨机器学习研究进展、系统设计模式及战略性问题选择
Eugene Yan 的系列文章探讨了在实际系统中应用机器学习的实用方面。他强调在实施机器学习之前，应先从启发式方法开始项目，设计模式对于高效的数据处理和系统维护的重要性，以及基于成本效益分析仔细选择问题的必要性。Yan 还详细介绍了部署机器学习模型后遇到的常见挑战，如数据污染和反馈循环，并提出了有效的项目管理和系统维护策略。