The motivation to make inference cheaper doesn’t seem to be mentioned in the Switch Transformer paper nor in the original Shazeer paper. They do mention improving training cost, training time (from being much easier to parallelize), and peak accuracy.
I’m not sure what you mean. They refer all over the place to greater computational efficiency and the benefits of constant compute cost even as one scales up experts. And this was front and center in the original MoE paper emphasizing the cheapness of the forward pass and positioning it as an improvement on the GNMT NMT RNN Google Translate had just rolled out the year before or so (including benchmarking the actual internal Google Translate datasets), and which was probably a major TPUv1 user (judging from the % of RNN workload reported in the TPU paper). Training costs are important, of course, but a user like Google Translate, the customer of the MoE work, cares more about the deployment costs because they want to serve literally billions of users, while the training doesn’t happen so often.
I’m not sure what you mean. They refer all over the place to greater computational efficiency and the benefits of constant compute cost even as one scales up experts. And this was front and center in the original MoE paper emphasizing the cheapness of the forward pass and positioning it as an improvement on the GNMT NMT RNN Google Translate had just rolled out the year before or so (including benchmarking the actual internal Google Translate datasets), and which was probably a major TPUv1 user (judging from the % of RNN workload reported in the TPU paper). Training costs are important, of course, but a user like Google Translate, the customer of the MoE work, cares more about the deployment costs because they want to serve literally billions of users, while the training doesn’t happen so often.