模型

GPT-2

Q: GPT-2 最近有什么新动态？

traeai 已收录 2 篇与 GPT-2 相关的内容。最新一篇是「AI Paper Review: Language Models are Unsupervised Multitask Learners (GPT-2)」，由 freeCodeCamp.org 发布。

别名：gpt2

由 OpenAI 开发的开源语言模型，常用于教学与实验性推理任务。

AI Paper Review: Language Models are Unsupervised Multitask Learners (GPT-2)

freeCodeCamp.org5月11日3193 字 (约 13 分钟)

GPT-2 demonstrated that training a large language model solely on unsupervised next-word prediction enables emergent multitask capabilities, performing translation, QA, and summarization without task-specific fine-tuning.

入选理由：GPT-2在800万网页文本上训练，参数量达15亿，首次展示零样本迁移能力。

FeaturedArticle#GPT-2#Large Language Models#Zero-Shot Learning#Transformer英文

Serving Multiple Users at Once: How Continuous Batching Keeps LLM Inference Efficient

Machine Learning MasteryToday6661 字 (约 27 分钟)

Continuous batching resolves static batching’s padding-induced GPU idleness by enabling dynamic scheduling and ragged batching, significantly improving throughput and latency in multi-user LLM inference—real-world tests show 2–3x throughput gains and up to 50% lower average latency.

入选理由：静态批处理因固定长度填充导致短请求空等，最长请求决定整批完成时间，GPU 利用率常低于 60%

FeaturedArticle#LLM#Inference#Batching#GPU Optimization英文

跨材料问答 · GPT-2

回答基于：GPT-2 相关 2 条材料

GPT-2

相关材料

AI Paper Review: Language Models are Unsupervised Multitask Learners (GPT-2)

Serving Multiple Users at Once: How Continuous Batching Keeps LLM Inference Efficient

跨材料问答 · GPT-2