Training has a lot of extra functionality like calculating how to update the weights of the model during training to make it more performant on the dataset(backpropagation and gradient descent) and much more
Meanwhile inference is mostly running the weights of the model as they are. The model isn't being adjusted in any way. And Nvidia holds a strong grip on training libraries through Cuda
That's all software, I don't think anyone is doing back prop in hardware. Your are right about cuda, but z.ai just demonstrated that moat is evaporating quick
7 Comments
geneva_convenience@lemmy.ml · 2 pts · 5d
Only for inference like the article says or also training?
yogthos@lemmy.ml · 5 pts · 5d
they might not need more training infrastructure at this point
m532@lemmy.ml · 1 pts · 5d
The article says "to run AI models" so it probably means both inference and training, not just inference.
geneva_convenience@lemmy.ml · 5 pts · 5d
Running refers to inference usually
embed_me@programming.dev · 0 pts · 5d
If they can be used for inference, I would assume they can be used for training
geneva_convenience@lemmy.ml · 3 pts · 5d
Training has a lot of extra functionality like calculating how to update the weights of the model during training to make it more performant on the dataset(backpropagation and gradient descent) and much more
Meanwhile inference is mostly running the weights of the model as they are. The model isn't being adjusted in any way. And Nvidia holds a strong grip on training libraries through Cuda
moonshine69@lemmy.nz · 1 pts · 5d
They can be used for training..
That's all software, I don't think anyone is doing back prop in hardware. Your are right about cuda, but z.ai just demonstrated that moat is evaporating quick