Subscribe to the gold package and get unlimited access to Shamra Academy
Register a new userUnderstanding the Difficulty of Training Transformers
90
0
0.0
(
0
)
Added by
Liyuan Liu
Publication date
2020
fields
Informatics Engineering
and research's language is
English
Ask ChatGPT about the research
No Arabic abstract
Transformers have proved effective in many NLP tasks. However, their training requires non-trivial efforts regarding designing cutting-edge optimizers and learning rate schedulers carefully (e.g., conventional SGD fails to train Transformers effectively). Our objective here is to understand $textit{what complicates Transformer training}$ from both empirical and theoretical perspectives. Our analysis reveals that unbalanced gradients are not the root cause of the instability of training. Instead, we identify an amplification effect that influences training substantially -- for each layer in a multi-layer Transformer model, heavy dependency on its residual branch makes training unstable, since it amplifies small parameter perturbations (e.g., parameter updates) and results in significant disturbances in the model output. Yet we observe that a light dependency limits the model potential and leads to inferior trained models. Inspired by our analysis, we propose Admin ($textbf{Ad}$aptive $textbf{m}$odel $textbf{in}$itialization) to stabilize stabilize the early stages training and unleash its full potential in the late stage. Extensive experiments show that Admin is more stable, converges faster, and leads to better performance. Implementations are released at: https://github.com/LiyuanLucasLiu/Transforemr-Clinic.
rate research
Read More
Abbreviation disambiguation is important for automated clinical note processing due to the frequent use of abbreviations in clinical settings. Current models for automated abbreviation disambiguation are restricted by the scarcity and imbalance of labeled training data, decreasing their generalizability to orthogonal sources. In this work we propose a novel data augmentation technique that utilizes information from related medical concepts, which improves our models ability to generalize. Furthermore, we show that incorporating the global context information within the whole medical note (in addition to the traditional local context window), can significantly improve the models representation for abbreviations. We train our model on a public dataset (MIMIC III) and test its performance on datasets from different sources (CASI, i2b2). Together, these two techniques boost the accuracy of abbreviation disambiguation by almost 14% on the CASI dataset and 4% on i2b2.
Transformers have been successfully applied to sequential, auto-regressive tasks despite being feedforward networks. Unlike recurrent neural networks, Transformers use attention to capture temporal relations while processing input tokens in parallel. While this parallelization makes them computationally efficient, it restricts the model from fully exploiting the sequential nature of the input. The representation at a given layer can only access representations from lower layers, rather than the higher level representations already available. In this work, we propose the Feedback Transformer architecture that exposes all previous representations to all future representations, meaning the lowest representation of the current timestep is formed from the highest-level abstract representation of the past. We demonstrate on a variety of benchmarks in language modeling, machine translation, and reinforcement learning that the increased representation capacity can create small, shallow models with much stronger performance than comparable Transformers.
Transformers-based models, such as BERT, have been one of the most successful deep learning models for NLP. Unfortunately, one of their core limitations is the quadratic dependency (mainly in terms of memory) on the sequence length due to their full attention mechanism. To remedy this, we propose, BigBird, a sparse attention mechanism that reduces this quadratic dependency to linear. We show that BigBird is a universal approximator of sequence functions and is Turing complete, thereby preserving these properties of the quadratic, full attention model. Along the way, our theoretical analysis reveals some of the benefits of having $O(1)$ global tokens (such as CLS), that attend to the entire sequence as part of the sparse attention mechanism. The proposed sparse attention can handle sequences of length up to 8x of what was previously possible using similar hardware. As a consequence of the capability to handle longer context, BigBird drastically improves performance on various NLP tasks such as question answering and summarization. We also propose novel applications to genomics data.
Transformers are being used extensively across several sequence modeling tasks. Significant research effort has been devoted to experimentally probe the inner workings of Transformers. However, our conceptual and theoretical understanding of their power and inherent limitations is still nascent. In particular, the roles of various components in Transformers such as positional encodings, attention heads, residual connections, and feedforward networks, are not clear. In this paper, we take a step towards answering these questions. We analyze the computational power as captured by Turing-completeness. We first provide an alternate and simpler proof to show that vanilla Transformers are Turing-complete and then we prove that Transformers with only positional masking and without any positional encoding are also Turing-complete. We further analyze the necessity of each component for the Turing-completeness of the network; interestingly, we find that a particular type of residual connection is necessary. We demonstrate the practical implications of our results via experiments on machine translation and synthetic tasks.
Catastrophic forgetting affects the training of neural networks, limiting their ability to learn multiple tasks sequentially. From the perspective of the well established plasticity-stability dilemma, neural networks tend to be overly plastic, lacking the stability necessary to prevent the forgetting of previous knowledge, which means that as learning progresses, networks tend to forget previously seen tasks. This phenomenon coined in the continual learning literature, has attracted much attention lately, and several families of approaches have been proposed with different degrees of success. However, there has been limited prior work extensively analyzing the impact that different training regimes -- learning rate, batch size, regularization method-- can have on forgetting. In this work, we depart from the typical approach of altering the learning algorithm to improve stability. Instead, we hypothesize that the geometrical properties of the local minima found for each task play an important role in the overall degree of forgetting. In particular, we study the effect of dropout, learning rate decay, and batch size, on forming training regimes that widen the tasks local minima and consequently, on helping it not to forget catastrophically. Our study provides practical insights to improve stability via simple yet effective techniques that outperform alternative baselines.
suggested questions
Log in to be able to interact and post comments
comments
Fetching comments
Sign in to be able to follow your search criteria
