Bingbi AI 已开源 BitCPM-CANN，这是一个专为国内 AI 加速器设计的训练框架。该框架支持 1.58 位模型训练，据称与传统的全精度训练相比，可将推理内存需求降低多达六倍。该技术与华为 Ascend 等芯片兼容，使得在本地硬件上进行高级 AI 训练更加便捷。 AI
影响 使得在国内硬件上进行更高效的 AI 模型训练成为可能，从而可能降低成本并提高可及性。
排序理由 该集群描述了一个新的 AI 模型训练框架的发布，属于 AI 领域的研发范畴。
AI 生成摘要 · Google Gemini · 来自 3 个来源。 我们如何撰写摘要 →
Model Best has open-sourced BitCPM-CANN, a complete training framework enabling 1.58-bit model training on domestic AI accelerators, reportedly reducing inference memory requirements by up to six times compared to full-precision training.
Bingbi AI has open-sourced BitCPM-CANN, a complete training framework enabling 1.58-bit model training on domestic AI accelerators, reportedly reducing inference memory requirements by up to six times compared to full-precision training.
Bingbi AI has released BitCPM-CANN, a training framework that enables 1.58-bit model training on domestic AI chips including Huawei Ascend. The approach cuts inference memory requirements by up to six times versus full-precision training. https:// pandaily.com/bingbi-ai-bitcpm- c…