Author name cluster

Liam Fedus

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

7 papers

2 author rows

ICLR Conference 2024 Conference Paper

Mixture-of-Experts Meets Instruction Tuning: A Winning Combination for Large Language Models

Sheng Shen 0001
Le Hou
Yanqi Zhou
Nan Du 0002
Shayne Longpre
Jason Wei
Hyung Won Chung
Barret Zoph

Sparse Mixture-of-Experts (MoE) is a neural architecture design that adds learnable parameters to Large Language Models (LLMs) without increasing computational complexity (FLOPs). Instruction tuning is a technique for training LLMs to follow instructions. We advocate combining these two approaches, as we find that MoE models benefit more from instruction tuning than dense models. In particular, we conduct empirical studies across three experimental setups: (i) Direct finetuning on individual downstream tasks devoid of instruction tuning; (ii) Instruction tuning followed by in-context few-shot or zero-shot generalization on downstream tasks; and (iii) Instruction tuning supplemented by further finetuning on individual downstream tasks. In the first scenario, MoE models overall underperform dense models of identical computational capacity. This narrative, however, dramatically changes with the introduction of instruction tuning (in the second and third scenarios), used independently or in conjunction with task-specific finetuning. Our most powerful model, FLAN-MoE-32B, surpasses the performance of Flan-PaLM-62B on four benchmark tasks, while using only a third of the FLOPs. The advancements embodied by FLAN-MoE inspire a reevaluation of the design principles of large-scale, high-performance language models in the framework of task-agnostic learning.

Details

JMLR Journal 2023 Journal Article

PaLM: Scaling Language Modeling with Pathways

Aakanksha Chowdhery
Sharan Narang
Jacob Devlin
Maarten Bosma
Gaurav Mishra
Adam Roberts
Paul Barham
Hyung Won Chung

Large language models have been shown to achieve remarkable performance across a variety of natural language tasks using few-shot learning, which drastically reduces the number of task-specific training examples needed to adapt the model to a particular application. To further our understanding of the impact of scale on few-shot learning, we trained a 540-billion parameter, densely activated, Transformer language model, which we call Pathways Language Model (PaLM). We trained PaLM on 6144 TPU v4 chips using Pathways, a new ML system which enables highly efficient training across multiple TPU Pods. We demonstrate continued benefits of scaling by achieving state-of-the-art few-shot learning results on hundreds of language understanding and generation benchmarks. On a number of these tasks, PaLM 540B achieves breakthrough performance, outperforming the finetuned state-of-the-art on a suite of multi-step reasoning tasks, and outperforming average human performance on the recently released BIG-bench benchmark. A significant number of BIG-bench tasks showed discontinuous improvements from model scale, meaning that performance steeply increased as we scaled to our largest model. PaLM also has strong capabilities in multilingual tasks and source code generation, which we demonstrate on a wide array of benchmarks. We additionally provide a comprehensive analysis on bias and toxicity, and study the extent of training data memorization with respect to model scale. Finally, we discuss the ethical considerations related to large language models and discuss potential mitigation strategies. [abs] [ pdf ][ bib ] &copy JMLR 2023. ( edit, beta )

PDF Details

ICML Conference 2022 Conference Paper

GLaM: Efficient Scaling of Language Models with Mixture-of-Experts

Nan Du 0002
Yanping Huang
Andrew M. Dai
Simon Tong
Dmitry Lepikhin
Yuanzhong Xu
Maxim Krikun
Yanqi Zhou

Scaling language models with more data, compute and parameters has driven significant progress in natural language processing. For example, thanks to scaling, GPT-3 was able to achieve strong results on in-context learning tasks. However, training these large dense models requires significant amounts of computing resources. In this paper, we propose and develop a family of language models named \glam (\textbf{G}eneralist \textbf{La}nguage \textbf{M}odel), which uses a sparsely activated mixture-of-experts architecture to scale the model capacity while also incurring substantially less training cost compared to dense variants. The largest \glam has 1. 2 trillion parameters, which is approximately 7x larger than GPT-3. It consumes only 1/3 of the energy used to train GPT-3 and requires half of the computation flops for inference, while still achieving better overall fewshot performance across 29 NLP tasks.

Details

ICLR Conference 2022 Conference Paper

Scale Efficiently: Insights from Pretraining and Finetuning Transformers

Yi Tay
Mostafa Dehghani 0001
Jinfeng Rao
Liam Fedus
Samira Abnar
Hyung Won Chung
Sharan Narang
Dani Yogatama

There remain many open questions pertaining to the scaling behaviour of Transformer architectures. These scaling decisions and findings can be critical, as training runs often come with an associated computational cost which have both financial and/or environmental impact. The goal of this paper is to present scaling insights from pretraining and finetuning Transformers. While Kaplan et al. presents a comprehensive study of the scaling behaviour of Transformer language models, the scope is only on the upstream (pretraining) loss. Therefore, it is still unclear if these set of findings transfer to downstream task within the context of the pretrain-finetune paradigm. The key findings of this paper are as follows: (1) we show that aside from only the model size, model shape matters for downstream fine-tuning, (2) scaling protocols operate differently at different compute regions, (3) widely adopted T5-base and T5-large sizes are Pareto-inefficient. To this end, we present improved scaling protocols whereby our redesigned models achieve similar downstream fine-tuning quality while having 50\% fewer parameters and training 40\% faster compared to the widely adopted T5-base model. We publicly release over 100 pretrained checkpoints of different T5 configurations to facilitate future research and analysis.

Details

ICLR Conference 2020 Conference Paper

Language GANs Falling Short

Massimo Caccia
Lucas Caccia
Liam Fedus
Hugo Larochelle
Joelle Pineau
Laurent Charlin

Traditional natural language generation (NLG) models are trained using maximum likelihood estimation (MLE) which differs from the sample generation inference procedure. During training the ground truth tokens are passed to the model, however, during inference, the model instead reads its previously generated samples - a phenomenon coined exposure bias. Exposure bias was hypothesized to be a root cause of poor sample quality and thus many generative adversarial networks (GANs) were proposed as a remedy since they have identical training and inference. However, many of the ensuing GAN variants validated sample quality improvements but ignored loss of sample diversity. This work reiterates the fallacy of quality-only metrics and clearly demonstrate that the well-established technique of reducing softmax temperature can outperform GANs on a quality-only metric. Further, we establish a definitive quality-diversity evaluation procedure using temperature tuning over local and global sample metrics. Under this, we find that MLE models consistently outperform the proposed GAN variants over the whole quality-diversity space. Specifically, we find that 1) exposure bias appears to be less of an issue than the complications arising from non-differentiable, sequential GAN training; 2) MLE trained models provide a better quality/diversity trade-off compared to their GAN counterparts, all while being easier to train, easier to cross-validate, and less computationally expensive.

Details

ICLR Conference 2020 Conference Paper

On Bonus Based Exploration Methods In The Arcade Learning Environment

Adrien Ali Taïga
Liam Fedus
Marlos C. Machado
Aaron C. Courville
Marc G. Bellemare

Research on exploration in reinforcement learning, as applied to Atari 2600 game-playing, has emphasized tackling difficult exploration problems such as Montezuma's Revenge (Bellemare et al., 2016). Recently, bonus-based exploration methods, which explore by augmenting the environment reward, have reached above-human average performance on such domains. In this paper we reassess popular bonus-based exploration methods within a common evaluation framework. We combine Rainbow (Hessel et al., 2018) with different exploration bonuses and evaluate its performance on Montezuma's Revenge, Bellemare et al.'s set of hard of exploration games with sparse rewards, and the whole Atari 2600 suite. We find that while exploration bonuses lead to higher score on Montezuma's Revenge they do not provide meaningful gains over the simpler epsilon-greedy scheme. In fact, we find that methods that perform best on that game often underperform epsilon-greedy on easy exploration Atari 2600 games. We find that our conclusions remain valid even when hyperparameters are tuned for these easy-exploration games. Finally, we find that none of the methods surveyed benefit from additional training samples (1 billion frames, versus Rainbow's 200 million) on Bellemare et al.'s hard exploration games. Our results suggest that recent gains in Montezuma's Revenge may be better attributed to architecture change, rather than better exploration schemes; and that the real pace of progress in exploration research for Atari 2600 games may have been obfuscated by good results on a single domain.

Details

ICML Conference 2020 Conference Paper

Revisiting Fundamentals of Experience Replay

Liam Fedus
Prajit Ramachandran
Rishabh Agarwal
Yoshua Bengio
Hugo Larochelle
Mark Rowland 0001
Will Dabney

Experience replay is central to off-policy algorithms in deep reinforcement learning (RL), but there remain significant gaps in our understanding. We therefore present a systematic and extensive analysis of experience replay in Q-learning methods, focusing on two fundamental properties: the replay capacity and the ratio of learning updates to experience collected (replay ratio). Our additive and ablative studies upend conventional wisdom around experience replay {—} greater capacity is found to substantially increase the performance of certain algorithms, while leaving others unaffected. Counterintuitively we show that theoretically ungrounded, uncorrected n-step returns are uniquely beneficial while other techniques confer limited benefit for sifting through larger memory. Separately, by directly controlling the replay ratio we contextualize previous observations in the literature and empirically measure its importance across a variety of deep RL algorithms. Finally, we conclude by testing a set of hypotheses on the nature of these performance benefits.

Details