Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism

Recent work in unsupervised language modeling demonstrates that training large neural language models advances the state of the art in Natural Language Processing applications. However, for very large models, memory constraints limit the size of models that can be practically trained. Model parallelism allows us to train larger models, because the parameters can be split across multiple processors. In this work, we implement a simple, efficient intra-layer model parallel approach that enables training state of the art transformer language models with billions of parameters. Our approach does not require a new compiler or library changes, is orthogonal and complimentary to pipeline model parallelism, and can be fully implemented with the insertion of a few communication operations in native PyTorch. We illustrate this approach by converging an 8.3 billion parameter transformer language model using 512 GPUs, making it the largest transformer model ever trained at 24x times the size of BERT and 5.6x times the size of GPT-2. We sustain up to 15.1 PetaFLOPs per second across the entire application with 76% scaling efficiency, compared to a strong single processor baseline that sustains 39 TeraFLOPs per second, which is 30% of peak FLOPs. The model is trained on 174GB of text, requiring 12 ZettaFLOPs over 9.2 days to converge. Transferring this language model achieves state of the art (SOTA) results on the WikiText103 (10.8 compared to SOTA perplexity of 16.4) and LAMBADA (66.5% compared to SOTA accuracy of 63.2%) datasets. We release training and evaluation code, as well as the weights of our smaller portable model, for reproducibility.

Practical Text Classification With Large Pre-Trained Language Models

Multi-emotion sentiment classification is a natural language processing (NLP) problem with valuable use cases on real-world data. We demonstrate that large-scale unsupervised language modeling combined with finetuning offers a practical solution to this task on difficult datasets, including those with label class imbalance and domain-specific context. By training an attention-based Transformer network (Vaswani et al. 2017) on 40GB of text (Amazon reviews) (McAuley et al. 2015) and fine-tuning on the training set, our model achieves a 0.69 F1 score on the SemEval Task 1:E-c multi-dimensional emotion classification problem (Mohammad et al. 2018), based on the Plutchik wheel of emotions (Plutchik 1979). These results are competitive with state of the art models, including strong F1 scores on difficult (emotion) categories such as Fear (0.73), Disgust (0.77) and Anger (0.78), as well as competitive results on rare categories such as Anticipation (0.42) and Surprise (0.37). Furthermore, we demonstrate our application on a real world text classification task. We create a narrowly collected text dataset of real tweets on several topics, and show that our finetuned model outperforms general purpose commercially available APIs for sentiment and multidimensional emotion classification on this dataset by a significant margin. We also perform a variety of additional studies, investigating properties of deep learning architectures, datasets and algorithms for achieving practical multidimensional sentiment classification. Overall, we find that unsupervised language modeling and finetuning is a simple framework for achieving high quality results on real-world sentiment classification.

Large Scale Language Modeling: Converging on 40GB of Text in Four Hours

Recent work has shown how to train Convolutional Neural Networks (CNNs) rapidly on large image datasets, then transfer the knowledge gained from these models to a variety of tasks. Following [Radford 2017], in this work, we demonstrate similar scalability and transfer for Recurrent Neural Networks (RNNs) for Natural Language tasks. By utilizing mixed precision arithmetic and a 32k batch size distributed across 128 NVIDIA Tesla V100 GPUs, we are able to train a character-level 4096-dimension multiplicative LSTM (mLSTM) for unsupervised text reconstruction over 3 epochs of the 40 GB Amazon Reviews dataset in four hours. This runtime compares favorably with previous work taking one month to train the same size and configuration for one epoch over the same dataset. Converging large batch RNN models can be challenging. Recent work has suggested scaling the learning rate as a function of batch size, but we find that simply scaling the learning rate as a function of batch size leads either to significantly worse convergence or immediate divergence for this problem. We provide a learning rate schedule that allows our model to converge with a 32k batch size. Since our model converges over the Amazon Reviews dataset in hours, and our compute requirement of 128 Tesla V100 GPUs, while substantial, is commercially available, this work opens up large scale unsupervised NLP training to most commercial applications and deep learning researchers. A model can be trained over most public or private text datasets overnight.