An official website of the United States government
Hyperparameter tuning is essential to achieving state-of-the-art accuracy in machine learning (ML), but requires substantial compute resources to perform. Existing systems primarily focus on effectively allocating resources for a hyperparameter tuning job under fixed resource constraints. We show that the available parallelism in such jobs changes dynamically over the course of execution and, therefore, presents an opportunity to leverage the elasticity of the cloud. In particular, we address the problem of minimizing the financial cost of executing a hyperparameter tuning job, subject to a time constraint. We present RubberBand---the first framework for cost-efficient, elastic execution of hyperparameter tuning jobs in the cloud. RubberBand utilizes performance instrumentation and cloud pricing to model job completion time and cost prior to runtime, and generate a cost-efficient, elastic resource allocation plan. RubberBand is able to efficiently execute this plan and realize a cost reduction of up to 2x in comparison to static allocation baselines.
more »
« less
Elastic Hyperparameter Tuning on the Cloud
Hyperparameter tuning is a necessary step in training and deploying machine learning models. Most prior work on hyperparameter tuning has studied methods for maximizing model accuracy under a time constraint, assuming a fixed cluster size. While this is appropriate in data center environments, the increased deployment of machine learning workloads in cloud settings necessitates studying hyperparameter tuning with an elastic cluster size and time and monetary budgets. While recent work has leveraged the elasticity of the cloud to minimize the execution cost of a pre-determined hyperparameter tuning job originally designed for fixed-cluster sizes, they do not aim to maximize accuracy. In this work, we aim to maximize accuracy given time and cost constraints. We introduce SEER---Sequential Elimination with Elastic Resources, an algorithm that tests different hyperparameter values in the beginning and maintains varying degrees of parallelism among the promising configurations to ensure that they are trained sufficiently before the deadline. Unlike fixed cluster size methods, it is able to exploit the flexibility in resource allocation the elastic setting has to offer in order to avoid undesirable effects of sublinear scaling. Furthermore, SEER can be easily integrated into existing systems and makes minimal assumptions about the workload. On a suite of benchmarks, we demonstrate that SEER outperforms both existing methods for hyperparameter tuning on a fixed cluster as well as naive extensions of these algorithms to the cloud setting.
more »
« less
- Award ID(s):
- 1730628
- PAR ID:
- 10310452
- Date Published:
- Journal Name:
- SoCC '21: Proceedings of the ACM Symposium on Cloud Computing
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Tuning hyperparameters is a crucial but arduous part of the machine learning pipeline. Hyperparameter optimization is even more challenging in federated learning, where models are learned over a distributed network of heterogeneous devices; here, the need to keep data on device and perform local training makes it difficult to efficiently train and evaluate configurations. In this work, we investigate the problem of federated hyperparameter tuning. We first identify key challenges and show how standard approaches may be adapted to form baselines for the federated setting. Then, by making a novel connection to the neural architecture search technique of weight-sharing, we introduce a new method, FedEx, to accelerate federated hyperparameter tuning that is applicable to widely-used federated optimization methods such as FedAvg and recent variants. Theoretically, we show that a FedEx variant correctly tunes the on-device learning rate in the setting of online convex optimization across devices. Empirically, we show that FedEx can outperform natural baselines for federated hyperparameter tuning by several percentage points on the Shakespeare, FEMNIST, and CIFAR-10 benchmarks, obtaining higher accuracy using the same training budget.more » « less
-
Solving a bilevel optimization problem is at the core of several machine learning problems such as hyperparameter tuning, data denoising, meta- and few-shot learning, and training-data poisoning. Different from simultaneous or multi-objective optimization, the steepest descent direction for minimizing the upper-level cost in a bilevel problem requires the inverse of the Hessian of the lower-level cost. In this work, we propose a novel algorithm for solving bilevel optimization problems based on the classical penalty function approach. Our method avoids computing the Hessian inverse and can handle constrained bilevel problems easily. We prove the convergence of the method under mild conditions and show that the exact hypergradient is obtained asymptotically. Our method's simplicity and small space and time complexities enable us to effectively solve large-scale bilevel problems involving deep neural networks. We present results on data denoising, few-shot learning, and training-data poisoning problems in a large-scale setting. Our results show that our approach outperforms or is comparable to previously proposed methods based on automatic differentiation and approximate inversion in terms of accuracy, run-time, and convergence speed.more » « less
-
null (Ed.)While cloud platforms enable users to rent computing resources on demand to execute their jobs, buying fixed resources is still much cheaper than renting if their utilization is high. Thus, optimizing cloud costs requires users to determine how many fixed resources to buy versus rent based on their workload. In this paper, we introduce the concept of a waiting policy for cloud-enabled schedulers, which is the dual of a scheduling policy, and show that the optimal cost depends on it. We define multiple waiting policies and develop simple analytical models to reveal their tradeoff between fixed resource provisioning, cost, and job waiting time. We evaluate the impact of these waiting policies on a year-long production batch workload consisting of 14M jobs run on a 14.3k-core cluster, and show that a compound waiting policy decreases the cost (by 5%) and mean job waiting time (by 7x) compared to a fixed cluster of the current size.more » « less
-
null (Ed.)While cloud platforms enable users to rent computing resources on demand to execute their jobs, buying fixed resources is still much cheaper than renting if their utilization is high. Thus, optimizing cloud costs requires users to determine how many fixed resources to buy versus rent based on their workload. In this paper, we introduce the concept of a waiting policy for cloud-enabled schedulers, which is the dual of a scheduling policy, and show that the optimal cost depends on it. We define multiple waiting policies and develop simple analytical models to reveal their tradeoff between fixed resource provisioning, cost, and job waiting time. We evaluate the impact of these waiting policies on a year-long production batch workload consisting of 14Mjobs run on a 14.3k-core cluster, and show that a compound waiting policy decreases the cost (by 5%) and mean job waiting time (by 7×) compared to a fixed cluster of the current size.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- https://doi.org/10.1145/3472883.3486989
