Co-founder here, happy to answer questions. We link to more technical documentation from the blog.
Wow, nice. What do you think of Glow?
I’ve discussed Glow with our team, my recollection is that PlaidML is more directly comparable to Tensor Comprehensions and TVM than Glow. Basically PlaidML bridges the gap between high level tools like Keras or ONNX and most platforms (like Mac + Metal) without any other software. We’re open to collaborations though, maybe Nadav can comment in more detail.
My understanding is that Tensor Comprehensions is used to implement specific kernels, and Glow is used to implement things like Keras or ONNX.