@apoth3osis_io@SpeakezTech@JustDeezGuy Sounds the same. From the Lean perspective, MLIR is a monolith accessible only through the formal dialect IR. On the MLIR side, you lower the formal IR dialect to the target architecture.
The other end is targeting the architecture without an intermediate dialect.
@apoth3osis_io@SpeakezTech@JustDeezGuy I would keep a separate verification dialect, follow existing work. In this case the Inet dialect *is* the formal semantics, and lowering to GPU, CPU ot anything else is user choice.
https://t.co/yeV78VNjQD
@apoth3osis_io@SpeakezTech@JustDeezGuy That's great to know 😁
The graph rewriting should be implemented both at compile time (Inet dialect) and at runtime (not done yet) in order to implement runtime behavior. Are you using HVM as runtime?
@apoth3osis_io@SpeakezTech@JustDeezGuy There's lots of us :)
The dialect is in the same state as I presented in the design meeting, and I would be happy to go back to it for a real world use case. Further development is awaiting community interest.
https://t.co/dXpaV7bQ9j
https://t.co/lu1Jq6WMq5
I've been implementing Mytilus support in mini-swe-agent and replicating SWE-bench. Models understand the production rules and write shell command pipelines using YAML as effectively as Bash scripts.
https://t.co/OpAuH0fyxf
https://t.co/LlXntM9V3v
I'm working on a C compiler where the intermediate representation is based on the single-instruction Run language by Dusko Pavlovic.
Most features are grammar transformations defined in Lex and Yacc files, and C for the remaining implementation.
https://t.co/gClVLcf511
I'm actively looking for cofounders to build general-purpose, massively parallel computer architectures on the basis of Interaction Nets. If you're talented and have some experience in hardware design, compiler engineering or models of computation feel free to DM me
"Tensor Species" by @andrewdudzik 🧮🐒
A great talk at the @ToposInstitute Colloquium, overviewing our team's category-theoretic vision for understanding AI systems through CS primitives...
...casually algebrafying einsums and softmax in the process 🐈
https://t.co/gbSag2Bqb8
📢If you’re interested in Temporal Networks, consider joining the TENET satellite @NetSciConf in Maastricht!🔎We accept one-page abstracts concerning all aspects of temporal network theory, as well as real-world applications.
📅Deadline: February 7
info 👉🏻 https://t.co/9e4il86pZt
Another year, another edition!💥
Join #TopoNets25 at @NetSciConf in Maastricht and send us your work on higher-order networks: structure, dynamics, methods, inference, software, topology, #TDA, etc.
📅 Submit by Feb 7th to give a contributed talk:
https://t.co/It3wzvQaKT