mirrors / deepspeedai--deepspeed

项目文件夹

deepspeedai--deepspeed
文件
main
T
wehub-resource-sync 4ececc111a chore: import upstream snapshot with attribution
2026-07-13 13:18:33 +08:00

296 B
原始文件 永久链接 Blame 文件历史

title: "ZeRO-Inference: 20X faster inference through weight quantization and KV cache offloading" excerpt: "" link: https://github.com/deepspeedai/DeepSpeedExamples/blob/master/inference/huggingface/zero_inference/README.md date: 2023-09-12 00:09:00 tags: inference ZeRO quantization English

在新工单中引用 查看 Git Blame 复制永久链接