AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
Is the Diffusion Model a Good Approach to Machine Translation?
The diffusion model is kind of weer, if i'm understanding i correctly. Its kind of going through that that noising and then de noising. The decoter takes the result of that and kind of up scales it back to the uncompressed model. And this model was was trained on approximately a hundred and 20 million image text pairs. It took 256 a one hundreds, about a hundred n 50 k hours, which would kind of equal, at least that at market price, around 600 k.