cuDNN 怎么读

cuDNN 在英语中的标准读法是 「coo dee en en」（/ˈkuː diː ɛn ɛn/）。

cuDNN is NVIDIA's CUDA Deep Neural Network library, a GPU-accelerated set of primitives (convolution, attention, matmul, pooling, normalization) that underpins PyTorch, TensorFlow, and JAX. Say it COO-dee-en-en: the lowercase 'cu' inherits CUDA's 'coo' sound (as in barra-CUDA), then D-N-N is spelled out letter by letter, with primary stress on COO. Non-native devs commonly misread the prefix as 'see-you' (C-U) or 'cud', which sounds wrong to practitioners. It's non-obvious because 'cu' is a CUDA abbreviation glued onto DNN, not a standalone English syllable.

来源：NVIDIA Developer — cuDNN (CUDA Deep Neural Network library)

读对项目名只是工程师文化的一部分 — 但开会时被人轻轻纠正、或者在 podcast 上被嘉宾礼貌打断的滋味，并不好受。Pronounce 是一个开源的开发者词汇发音词典，每条都标注来源、信心等级，并且有预渲染的音频可以直接听。