Search for: All records

Over the past decade, machine learning model complexity has grown at an extraordinary rate, as has the scale of the systems training such large models. However there is an alarmingly low hardware utilization (5-20%) in large scale AI systems. The low system utilization is a cumulative effect of minor losses across different layers of the stack, exacerbated by the disconnect between engineers designing different layers spanning across different industries. To address this challenge, in this work we designed a cross-stack performance modelling and design space exploration framework. First, we introduce CrossFlow, a novel framework that enables cross-layer analysis all the way from the technology layer to the algorithmic layer. Next, we introduce DeepFlow (built on top of CrossFlow using machine learning techniques) to automate the design space exploration and co-optimization across different layers of the stack. We have validated CrossFlow’s accuracy with distributed training on real commercial hardware and showcase several DeepFlow case studies demonstrating pitfalls of not optimizing across the technology-hardware-software stack for what is likely, the most important workload driving large development investments in all aspects of computing stack. 

Numerical simulations have revolutionized material design. However, although simulations excel at mapping an input material to its output property, their direct application to inverse design has traditionally been limited by their high computing cost and lack of differentiability. Here, taking the example of the inverse design of a porous matrix featuring targeted sorption isotherm, we introduce a computational inverse design framework that addresses these challenges, by programming differentiable simulation on TensorFlow platform that leverages automated end-to-end differentiation. Thanks to its differentiability, the simulation is used to directly train a deep generative model, which outputs an optimal porous matrix based on an arbitrary input sorption isotherm curve. Importantly, this inverse design pipeline leverages the power of tensor processing units (TPU)—an emerging family of dedicated chips, which, although they are specialized in deep learning, are flexible enough for intensive scientific simulations. This approach holds promise to accelerate inverse materials design. 

Abstract Nitrous acid (HONO) plays pivotal roles in various metal‐free as well as metal‐mediated routes relevant to biogeochemistry, atmospheric chemistry, and mammalian physiology. While the metastable nature of HONO hinders the detailed investigations into its reactivity at a transition metal site, this report herein utilizes a heteroditopic copper(II) cryptate [oC]Cu^IIfeaturing a proton‐responsive second‐coordination‐sphere located at a suitable distance from a [Cu^II](ONO) core, thereby enabling isolation of a [Cu^II](κ¹‐ONO⋅⋅⋅H⁺) complex (2H‐NO₂). A set of complementary analytical studies (UV‐vis,¹⁴N/¹⁵N FTIR,¹⁵N NMR, HRMS, EPR, and CHN) on2H‐NO₂and its¹⁵N‐isotopomer (2H‐¹⁵NO₂) reveals the formulation of2H‐NO₂as {[oCH]Cu^II(κ¹‐ONO)}(ClO₄)₂. Non‐covalent interaction index (NCI) based on reduced density gradient (RDG) analysis on {[oCH]Cu^II(κ¹‐ONO)}²⁺discloses a H‐bonding interaction between the apical 3° ammonium site and the nitrite anion bound to the copper(II) site. The FTIR spectra of [Cu^II](κ¹‐ONO⋅⋅⋅H⁺) species (2H‐NO₂) shows a shift of ammonium NH vibrational feature to a lower wavenumber due to the H‐bonding interaction with nitrite. The reactivity profile of [Cu^II](κ¹‐ONO⋅⋅⋅H⁺) species (2H‐NO₂) towards anaerobic nitration of substituted phenol (2,4‐DTBP) is distinctly different relative to that of the closely related tripodal [Cu^II]‐nitrite complexes (1‐NO₂/3‐NO₂/4‐NO₂).