Hi Henry, Thanks for asking the question. It depends on how and to what extent AWS is gonna release its low level API. To put on my Amazon hat (and I work closely with inferentia team), this is still not clear. We need more time to investigate.
It is definitely feasible If we only treat TVM as a model converter for neuron-sdk. Though it also requires neuron-sdk to clean and expose some high level APIs. On Wed, Jan 15, 2020 at 11:59 AM Henry Saputra <henry.sapu...@gmail.com> wrote: > > Hi All, > > I remember someone asked about AWS Neuron SDK [1] during TVM summit. > AFAIK it is used to support AWS new Inferentia chip that focus for low > power and low precision inferences jobs. > > Does anyone have followup whether would make sense to add support in TVM > for the new AWS Inferentia chip [2]? > > Thanks, > > Henry > > > [1] https://github.com/aws/aws-neuron-sdk > [2] https://aws.amazon.com/machine-learning/inferentia/ -- Yizhi Liu --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@tvm.apache.org For additional commands, e-mail: dev-h...@tvm.apache.org
