Discover and read the best of Twitter Threads about #MLIR
Most recents (2)
Today, we @zama_fhe are releasing Concrete 1.0, our open source homomorphic encryption compiler. It can convert python code into FHE, while guaranteeing the exactness of the result. 🧵 github.com/zama-ai/concre… #fhe #privacy #ai #cryptography #Python #mlir
Concrete converts Python code into TFHE circuits. TFHE is a homomorphic encryption scheme that support encrypted arithmetics and table lookups, using a new cryptographic operator called “programmable bootstrapping (pbs)”.
Homomorphic table lookups are a big deal, since any multivariate function can be represented as a linear combination of univariate functions (which is just a LUT!). This means we can now compute *any* function in FHE, without approximations: add, mul, div, comparisons, ReLu, …
Hey, internet: just in case you're not one of those engineers who hears "Hey, compilers!" and comes running delightedly, here's a thread on why I think #MLIR is going to transform the machine learning industry:
(1/n)
for more technical details, ref: ai.google/research/pubs/…
(1/n)
for more technical details, ref: ai.google/research/pubs/…
We're seeing an explosion of domain-specific accelerated hardware (yay!).
@GoogleAI's TPUs are an example of this, sure; but you also have @Apple's specialized chips for iPhones, @BaiduResearch's Kunlun, and literally *thousands* of others.
The way that you program for
(2/n)
@GoogleAI's TPUs are an example of this, sure; but you also have @Apple's specialized chips for iPhones, @BaiduResearch's Kunlun, and literally *thousands* of others.
The way that you program for
(2/n)
@GoogleAI @Apple @BaiduResearch these chips right now isn't easy.
Developer-facing tooling is essentially non-existent (or quite bad, even for TPUs); many models are built, and then found to be impossible to deploy to edge hardware; etc.
You also have very rad, but specialized types (e.g., bfloat16).
(3/n)
Developer-facing tooling is essentially non-existent (or quite bad, even for TPUs); many models are built, and then found to be impossible to deploy to edge hardware; etc.
You also have very rad, but specialized types (e.g., bfloat16).
(3/n)
