"LLM"的搜索結果
2026-03-17
13:08

Tether 推出 BitNet LoRA 框架,支持手机端大模型训练

Gate News report states that on March 17, Tether's QVAC Fabric launched the world's first cross-platform LoRA fine-tuning framework for Microsoft BitNet (1-bit LLM), significantly lowering the threshold for VRAM and computing power required for large model training. The framework supports LoRA fine-tuning and inference acceleration on Intel, AMD, Apple Silicon M series, and mobile GPUs (including Adreno, Mali, and Apple Bionic).
展開
09:32

AMI Labs 聯合創始人謝賽寧:LLM 某種程度上是「反 Bitter Lesson」,世界模型才是未來

Xiexai Ning criticized the current AI development route centered on large language models in an interview, believing it limits AI's learning of the real world and emphasizing the advantages of world models. He believes that building intelligent agents that can survive in reality is more challenging and mentioned that AMII Labs focuses on real data collection, with plans to launch AI smart glasses and robots in the future.
展開
03:07

Andrej Karpathy 開源 autoresearch 項目,AI 自動運行百次 LLM 訓練實驗

Eureka Labs創始人Andrej Karpathy於3月8日開源項目autoresearch,採用「人寫Markdown,AI寫代碼」模式,允許開發者通過program.md定義研究方向,AI自主優化代碼。每小時可進行約12次實驗,展示中83次實驗中有15次有效改進,該項目基於NVIDIA GPU和PyTorch,已在H100測試並遵循MIT協議。
展開