Ask HN: Why hasn't x86 caught up with Apple M series?
397 points by stephenheron 1d ago 570 comments
Ask HN: Is there a temp phone number like temp email?
8 points by piratesAndSons 14h ago 11 comments
Ask HN: Best codebases to study to learn software design?
100 points by pixelworm 3d ago 89 comments
Ask HN: Are AI filters becoming stricter than society itself?
29 points by tsevis 3d ago 16 comments
Nvidia announces 4-bit training with NVFP4
2 opcode84 1 8/25/2025, 4:54:35 PM developer.nvidia.com ↗
Comments (1)
opcode84 · 1d ago
A version of the 12B Hybrid Mamba-Transformer model was initially trained with 8-bit precision—FP8, which has been shown in previous studies to closely match 16-bit precision, and hence served as our baseline for comparison. We then successfully trained this same 12B model from scratch using NVFP4, demonstrating that this new low-precision format can support full pretraining at trillion-token scale. The NVFP4 run exhibited stable convergence without the training instabilities or divergence issues that typically plague ultra-low precision training.