16:10 - 16:30
AMAI: Algebraic Modelling of Distributed AI Computations for Symbolic Performance Analysis
Rubén Coll Sánchez
Huawei Technologies, France
Sorbonne University Paris VI, France
Thibaut Tachon, Pierre Leca, Chong Li
Huawei Technologies, France
Teng Su
Huawei Technologies, China
Developing large-scale AI systems requires extensive on-machine tuning: engineers evaluate hundreds of parallel configurations by trial and error, at considerable cost in accelerator rental and compute time. Existing performance models are either tied to specific frameworks or too coarse for modern distributed workloads. We present AMAI, a performance analysis framework featuring (i) a declarative domain-specific language based on the arrow pattern for describing distributed deep-learning computations, and (ii) compositional cost semantics that map programs into symbolic models of time and per-device memory. Models are parameterized by hardware constants and structural variables that define the parallelism configuration. As a result, descriptions can be analyzed at any scale and reused across machines. Predictions stay within 3 percent of measured step time and peak memory across batch-size and data-parallel scaling when analyzing the training of Qwen3.5 0.8B on a multi-node NPU cluster.
