cuDNN, every mention
4 scenes · ← back to cuDNN
tap a year for its mentions
every year anyone Andrej Karpathy 4Quentin Anthony 1Chris Lattner 1
Verbatim, from the transcripts: the passages where cuDNN comes up
How Zyphra went all-in on AMD + Why Devs feel faster with AI but are slower — with Quentin Anthony
- ▶ 11:31 Quentin Anthony So I would use something like, um, uh, like QDNN or something like that with some, uh, gem backend.
The Shape of Compute (Chris Lattner of Modular)
- ▶ 31:23 Chris Lattner And again, our goal is meet and beat QDNN and TRT-LM and these things.
llm.c's Origin and the Future of LLM Compilers - Andrej Karpathy at CUDA MODE
- ▶ 14:24 Andrej Karpathy It turns out CodeEman has a very good flash retention implementation, so we switch to that.
- ▶ 18:17 Andrej Karpathy Uh, so, uh, you do need QDNN, which is the most heavy dependency, but QDNN is optional, so if you'd like to roll your own manual attention, that is possible in LLM.C, but QDNN is kind of like the hairiest dependency, but after that it's… 3 times in the scene