moose
moose copied to clipboard
tf-encrypted
Secure distributed dataflow framework for encrypted machine learning and data processing
The way that `Hash` and `Eq` are currently implemented is not ideal, the former going through the (slower) textual representation and the latter not supposed to be implemented manually as...
Maybe it's just a docstring, or maybe there's a good default, but helping the user choose a value for `chunks` would be nice.
We currently have two computation formats: `moose::Computation` with its typed operators, and the textual as defined in `moose::textual`. The former could be seen as primarily our in-memory representation although we...
We should satisfy this invariant but are currently not checking it. It could potentially be a `debug_assert`.
This is a discussion issue. We currently have three kernel flavours: `hybrid`, `transparent`, and `concrete`. We are looking at getting rid of `hybrid` since the use of `Into`/`TryInto` is causing...
We're currently always compiling kernels to closures, with some performance overhead as a result. Letting the kernel traits return enums instead would probably not solve this, but we might be...
The current approach creates a single bit tensor on host placements by stacking together several bit tensors. But we often call `index` on the stacked tensor afterwards to separate them...
It should be possible, which would also simplify trait bounds. Maybe the only blocking thing is that there must be alignment on the underlying host ring, meaning maybe we need...
The current implementations of storage modules is using eg `HostFloat64Tensor::from` to lift a (raw) `ndarray` tensor into a Moose tensor (and in turn a `Value`), yet these `From` impl do...
The use of `TryFrom` in the kernel macros can lead to confusion since it's a very general pattern that doesn't convey what's happening very clearly. We are currently using `TryFrom`...
Metadata
Owner
Metadata
Secure distributed dataflow framework for encrypted machine learning and data processing