To enable holistic optimization of ML training pipelines, we present Lara, a declarative domain- specific language for collections and matrices. Lara's inter-.
To enable holistic optimization of ML training pipelines, we present Lara, a declarative domain-specific language for collections and matrices. Lara's inter- ...
To enable holistic optimization of ML training pipelines, we present Lara, a declarative domain- specific language for collections and matrices. Lara's inter-.
Aug 19, 2019 · To enable holistic optimization of ML training pipelines, we present Lara, a declarative domain-specific language for collections and matrices.
To enable holistic optimization of ML training pipelines, we present Lara, a declarative domain-specific language for collections and matrices. Lara's inter- ...
To enable holistic optimization of ML training pipelines, we present Lara, a declarative domain-specific language for collections and matrices. Lara's common IR ...
Lara is presented, a declarative domainspecific language for collections and matrices with intermediate representation (IR) that reflects on the complete ...
[PDF] DistIR: An Intermediate Representation for Optimizing ...
timharris.uk › 2021-euromlsys-dist-ir
Apr 26, 2021 · Despite this apparent restriction, we show that DistIR can express the range of currently employed distribution strategies in deep learning.
People also ask
What are pipelines in machine learning?
Oct 30, 2018 · IR is useful as a platform-neutral asm-like language that's useful for JIT compiling. Your portable application can generate LLVM-IR, and feed that to LLVM.
In this paper, we study three broad classes of optimizations: Timing (e.g., Pipeline re-timing), Spatial (e.g., Compute tiling), and Higher-order Ops (e.g., ...