Given incomplete input data, is it possible to use tvm to do as much computation for a model as is possible so that we can evaluate the model faster when the rest of the input data arrives? For instance, given a contiguous block of pixels in a larger unknown image, could we calculate the values of only those neurons whose receptive fields are contained in the known block of pixels, then use the result of that computation to speed up computation on subsequent blocks of pixels as they arrive in a streaming fashion? Could I use tvm as part of a solution to this problem?
--- [Visit Topic](https://discuss.tvm.apache.org/t/partial-neural-network-evaluation-for-streaming-data/8322/1) to respond. You are receiving this because you enabled mailing list mode. To unsubscribe from these emails, [click here](https://discuss.tvm.apache.org/email/unsubscribe/491cb2f06201a07deeafbf6b9a4873e3ad937dbbc40bb608870fe09836a05791).
