Long text summarization, Staggered Attention and fine-tuning PEGASUS-X on E2E Cloud

February 14, 2023

Large pretrained Transformer models have proven to be extremely capable at tackling natural language tasks. However, handling long textual sequences continues to be a significant challenge for Transformer models.

What is the problem?

Quadratic growth in memory consumption of the attention computation in Transformers.

What is this article all about?

The empirical investigation conducted in paper “Investigating Efficiently Extending Transformers for Long Input Summarization” by Jason Phang, Yao Zhao and Peter J. Liu  into the architectural changes, model configurations and pretraining schemes to identify the better approaches for training Transformer models in tackling long input summarization. The experiments conducted and their findings are discussed in this post.

What are the key findings?

They evaluate a set of efficient Transformer variants and propose a simpler blockwise local Transformer architecture with staggered blocks and global tokens that strikes a good balance of performance and memory efficiency.


Based on the findings from our empirical investigation, we adapt the pretrained PEGASUS Large model in tackling long input summarization on up to 16K input tokens. The resulting model, which we call PEGASUS-X attains top scores on long summarization tasks, outperforming much larger models like LongT5 (Guo et al., 2021) in some cases, and sets the state of the art of two tasks: GovReport and PubMed.

For Code and Weights:-https://github.com/google-research/pegasus

Let's dive deep into each of these segments:

What are the key challenges faced in Long Input Summarization:-

  • Computational Challenge:- Let’s understand the quadratic scaling with an example - Consider a summarization model with 12 encoder and 12 decoder layers, pre trained on an input length of 512 and fine tuned on a task with input sequence length 16384, using output length of 512 in both cases. Pre training is typically done with shorter sequences, while fine-tuning uses long inputs for summaries. Fine-tuning can now be more resource intensive and slower than pretraining, which is contrary to the conventional paradigm. Since the encoder inputs have increased 32 times, the memory requirements will increase by 1024 times. Even efficient transformers that achieve linear scaling of memory can still consume 32 times memory. This leads to several design choices which are evaluated through experimentation.
  • Dataset Challenge:-The main issues in current datasets are: relative simplicity of summarization, lack of diverse inputs, potential leakage of data due to the data collection procedures, and low quantity of examples for training.

Experimentations and their findings:-

1.Encoder Architecture:-

The investigation begins with the efficacy of swapping the encoder for an efficient Transformer encoder. Big Bird takes the approach of using sparse attention computation, combining sliding window attention, random attention and a set of global-attention tokens.

Conversely, Performer takes the approach of factoring attention matrices through orthogonal random features.

In addition, we also introduce two simple variants of local attention Transformer encoders. First, we use a simple block-local Transformer (Local), where encoder input tokens are divided into non overlapping blocks, tokens can only attend to other tokens within the block.

Second, we extend this local Transformer by adding a set of global tokens with learnable embeddings, that can attend to as well as be attended from every encoder’s token (GlobalLocal).

However, we opt for the simpler block-local attention rather than sliding window attention, and compensate for the lack of overlapping blocks by staggering the local attention blocks.

Among the short tasks, the full-attention Transformer performs best, followed by BigBird. On the long tasks, BigBird and Global-Local models perform the best, but BigBird consumes significantly more memory and trains much slower than the other architectures.

On the other hand, we find that the Local and Global-Local encoders strike a good balance of both performance and efficiency. The simple local attention encoder, which uses a block-local attention mechanism, attains performance surprisingly close to that of BigBird.

Local and Global-local configurations:-

Staggering:-First staggering of local attention blocks was introduced. Unlike in sliding window attention, in block-local attention, tokens can only attend other tokens within the same block. If the input tokens are divided up into the same blocks in every layer, it means that no information is exchanged across blocks through the entire encoder. To address this pitfall, we introduce a small architectural change wherein we stagger the block allocation across alternating layers. We have shown an example of this in Figure 2. Concretely, we stagger attention blocks by shifting the block boundaries by half a block every other layer:

We show the results of both of these changes in Table 2. We have found that staggering local blocks improves performance in both Local and Global-Local models by a noticeable amount. We have highlighted this improvement in the performance even with the Global-Local models.

Global-Local: Block Size and Number of Global Tokens:-

Larger block sizes and/or numbers of global tokens leads to improved performance, although the effect saturates.

For the remainder of the ablation experiments, they stick to a block size of 64 and 32 global tokens for consistency.

Positional Encoding Scheme:- Sinusoidal position encodings still remain to be a good choice for  long input Transformers.

Scaling Encoder and Decoder Layer:-They fix the total number of layers to 24, and consider both encoder-heavy and decoder-heavy distributions, for both Local and Global-Local models. They observe that the impact of distribution of encoder and decoder layers on performance is relatively smaller. For Local models, they see a slight boost from decoder-heavy models. For Global-Local models, we observe that a balanced encoder-decoder outperforms encoder-heavy and decoder-heavy models, both of which perform comparably.

Pretraining vs Fine-tuning Architectures:-For Local models, they found that pre-training with local attention using small block sizes tends to discourage the performance, but at moderate block sizes (e.g. 64) there is little difference between the two approaches. In contrast, we found that for Global-Local, pretraining with the efficient architecture tends to perform better. We hypothesize that this difference arises because of the presence of the learned global embedding tokens, which are randomly initialized when adapting from a pre-trained Transformer and hence may benefit from pre-training and being jointly trained with the local attention.

Pre-training Scheme:-We consider two setups for pretraining: short-input pretraining, with 512 input tokens and 256 output tokens, and long-input pretraining, with 4096 input tokens and 256 output tokens.

Given a fixed compute budget, allocating some portion of training to long-input training can improve performance, although the precise optimal allocation is difficult to determine. Exclusively long pre-training results in worse performance.

Partial Cross Attention:-

Another major memory bottleneck is the encoder-decoder cross-attention. It is found that reducing the number of cross-attention layers leads to a drop in performance, but the impact on performance is smaller than expected. For instance, with only cross-attention on the first and sixth layer, the Global-Local model still outperforms a Local model. The reduction of cross-attention layers also leads to a corresponding improvement in training step and reduction in memory consumption.

PEGASUS-X:-The experimentation with two model sizes PEGASUS-X (PEGASUS eXtended), based on PEGASUSLarge; and PEGASUS-XBase, based on a newly trained PEGASUSBase model which we call PEGASUSBase+. 

Only two new sets of parameters were introduced: the global token embeddings, and a separate LayerNorm for the global input representations in each Transformer layer. This is approximately 1M more parameters for PEGASUS-XBase and 2M more for PEGASUS-X. 

In Table 11 we can see the performance of PEGASUS models to those of PEGASUS-X on three long-input summarization tasks: arXiv, Big Patent and PubMed. In all three tasks there is significant amount of improvement in performance of PEGASUSXBase over PEGASUSBase+, and PEGASUS-X over PEGASUSLarge. 

Fine-tuning  PEGASUS-X-base JAX/Flax on E2E Cloud:-

1.Launch A100 GPU on E2E Cloud

If need any help to launch the node please follow https://docs.e2enetworks.com/computes/nodes/launchnode.html#how-to-launch-nodes-from-myaccount-portal

2.Install dependency:- To install the dependency run the following commands one by one.

git clone https://github.com/google-research/pegasus

cd pegasus
pip3 install -r requirements.txt

3.Tokenizer Download:-

wget “https://storage.googleapis.com/pegasus_ckpt/c4.unigram.newline.10pct.96000.model

4.Download Checkpoints fine tuned on  :-

wget “https://storage.googleapis.com/pegasus_ckpt/px/tuned/large/scrolls_summscreen.ckpt


First, we need to prepare the data for fine-tuning using TFDS.

python -m tensorflow_datasets.scripts.download_and_prepare --datasets=scrolls/summ_screen_fd

To fine-tune the model, we need to modify a config file

wget “https://github.com/google-research/pegasus/blob/main/pegasus/flax/configs/examples/summscreen_eval.py

You can modify this file if want any changes and then run this command:-

python pegasus/flax/main.py \    --config
    pegasus/flax/configs/examples/summscreen_eval.py \    --workdir

E2E Networks is the leading accelerated Cloud Computing player which provides the latest Cloud GPUs at a great value. Connect with us at sales@e2enetworks.com





Latest Blogs
This is a decorative image for: A Complete Guide To Customer Acquisition For Startups
October 18, 2022

A Complete Guide To Customer Acquisition For Startups

Any business is enlivened by its customers. Therefore, a strategy to constantly bring in new clients is an ongoing requirement. In this regard, having a proper customer acquisition strategy can be of great importance.

So, if you are just starting your business, or planning to expand it, read on to learn more about this concept.

The problem with customer acquisition

As an organization, when working in a diverse and competitive market like India, you need to have a well-defined customer acquisition strategy to attain success. However, this is where most startups struggle. Now, you may have a great product or service, but if you are not in the right place targeting the right demographic, you are not likely to get the results you want.

To resolve this, typically, companies invest, but if that is not channelized properly, it will be futile.

So, the best way out of this dilemma is to have a clear customer acquisition strategy in place.

How can you create the ideal customer acquisition strategy for your business?

  • Define what your goals are

You need to define your goals so that you can meet the revenue expectations you have for the current fiscal year. You need to find a value for the metrics –

  • MRR – Monthly recurring revenue, which tells you all the income that can be generated from all your income channels.
  • CLV – Customer lifetime value tells you how much a customer is willing to spend on your business during your mutual relationship duration.  
  • CAC – Customer acquisition costs, which tells how much your organization needs to spend to acquire customers constantly.
  • Churn rate – It tells you the rate at which customers stop doing business.

All these metrics tell you how well you will be able to grow your business and revenue.

  • Identify your ideal customers

You need to understand who your current customers are and who your target customers are. Once you are aware of your customer base, you can focus your energies in that direction and get the maximum sale of your products or services. You can also understand what your customers require through various analytics and markers and address them to leverage your products/services towards them.

  • Choose your channels for customer acquisition

How will you acquire customers who will eventually tell at what scale and at what rate you need to expand your business? You could market and sell your products on social media channels like Instagram, Facebook and YouTube, or invest in paid marketing like Google Ads. You need to develop a unique strategy for each of these channels. 

  • Communicate with your customers

If you know exactly what your customers have in mind, then you will be able to develop your customer strategy with a clear perspective in mind. You can do it through surveys or customer opinion forms, email contact forms, blog posts and social media posts. After that, you just need to measure the analytics, clearly understand the insights, and improve your strategy accordingly.

Combining these strategies with your long-term business plan will bring results. However, there will be challenges on the way, where you need to adapt as per the requirements to make the most of it. At the same time, introducing new technologies like AI and ML can also solve such issues easily. To learn more about the use of AI and ML and how they are transforming businesses, keep referring to the blog section of E2E Networks.

Reference Links




This is a decorative image for: Constructing 3D objects through Deep Learning
October 18, 2022

Image-based 3D Object Reconstruction State-of-the-Art and trends in the Deep Learning Era

3D reconstruction is one of the most complex issues of deep learning systems. There have been multiple types of research in this field, and almost everything has been tried on it — computer vision, computer graphics and machine learning, but to no avail. However, that has resulted in CNN or convolutional neural networks foraying into this field, which has yielded some success.

The Main Objective of the 3D Object Reconstruction

Developing this deep learning technology aims to infer the shape of 3D objects from 2D images. So, to conduct the experiment, you need the following:

  • Highly calibrated cameras that take a photograph of the image from various angles.
  • Large training datasets can predict the geometry of the object whose 3D image reconstruction needs to be done. These datasets can be collected from a database of images, or they can be collected and sampled from a video.

By using the apparatus and datasets, you will be able to proceed with the 3D reconstruction from 2D datasets.

State-of-the-art Technology Used by the Datasets for the Reconstruction of 3D Objects

The technology used for this purpose needs to stick to the following parameters:

  • Input

Training with the help of one or multiple RGB images, where the segmentation of the 3D ground truth needs to be done. It could be one image, multiple images or even a video stream.

The testing will also be done on the same parameters, which will also help to create a uniform, cluttered background, or both.

  • Output

The volumetric output will be done in both high and low resolution, and the surface output will be generated through parameterisation, template deformation and point cloud. Moreover, the direct and intermediate outputs will be calculated this way.

  • Network architecture used

The architecture used in training is 3D-VAE-GAN, which has an encoder and a decoder, with TL-Net and conditional GAN. At the same time, the testing architecture is 3D-VAE, which has an encoder and a decoder.

  • Training used

The degree of supervision used in 2D vs 3D supervision, weak supervision along with loss functions have to be included in this system. The training procedure is adversarial training with joint 2D and 3D embeddings. Also, the network architecture is extremely important for the speed and processing quality of the output images.

  • Practical applications and use cases

Volumetric representations and surface representations can do the reconstruction. Powerful computer systems need to be used for reconstruction.

Given below are some of the places where 3D Object Reconstruction Deep Learning Systems are used:

  • 3D reconstruction technology can be used in the Police Department for drawing the faces of criminals whose images have been procured from a crime site where their faces are not completely revealed.
  • It can be used for re-modelling ruins at ancient architectural sites. The rubble or the debris stubs of structures can be used to recreate the entire building structure and get an idea of how it looked in the past.
  • They can be used in plastic surgery where the organs, face, limbs or any other portion of the body has been damaged and needs to be rebuilt.
  • It can be used in airport security, where concealed shapes can be used for guessing whether a person is armed or is carrying explosives or not.
  • It can also help in completing DNA sequences.

So, if you are planning to implement this technology, then you can rent the required infrastructure from E2E Networks and avoid investing in it. And if you plan to learn more about such topics, then keep a tab on the blog section of the website

Reference Links



This is a decorative image for: Comprehensive Guide to Deep Q-Learning for Data Science Enthusiasts
October 18, 2022

A Comprehensive Guide To Deep Q-Learning For Data Science Enthusiasts

For all data science enthusiasts who would love to dig deep, we have composed a write-up about Q-Learning specifically for you all. Deep Q-Learning and Reinforcement learning (RL) are extremely popular these days. These two data science methodologies use Python libraries like TensorFlow 2 and openAI’s Gym environment.

So, read on to know more.

What is Deep Q-Learning?

Deep Q-Learning utilizes the principles of Q-learning, but instead of using the Q-table, it uses the neural network. The algorithm of deep Q-Learning uses the states as input and the optimal Q-value of every action possible as the output. The agent gathers and stores all the previous experiences in the memory of the trained tuple in the following order:

State> Next state> Action> Reward

The neural network training stability increases using a random batch of previous data by using the experience replay. Experience replay also means the previous experiences stocking, and the target network uses it for training and calculation of the Q-network and the predicted Q-Value. This neural network uses openAI Gym, which is provided by taxi-v3 environments.

Now, any understanding of Deep Q-Learning   is incomplete without talking about Reinforcement Learning.

What is Reinforcement Learning?

Reinforcement is a subsection of ML. This part of ML is related to the action in which an environmental agent participates in a reward-based system and uses Reinforcement Learning to maximize the rewards. Reinforcement Learning is a different technique from unsupervised learning or supervised learning because it does not require a supervised input/output pair. The number of corrections is also less, so it is a highly efficient technique.

Now, the understanding of reinforcement learning is incomplete without knowing about Markov Decision Process (MDP). MDP is involved with each state that has been presented in the results of the environment, derived from the state previously there. The information which composes both states is gathered and transferred to the decision process. The task of the chosen agent is to maximize the awards. The MDP optimizes the actions and helps construct the optimal policy.

For developing the MDP, you need to follow the Q-Learning Algorithm, which is an extremely important part of data science and machine learning.

What is Q-Learning Algorithm?

The process of Q-Learning is important for understanding the data from scratch. It involves defining the parameters, choosing the actions from the current state and also choosing the actions from the previous state and then developing a Q-table for maximizing the results or output rewards.

The 4 steps that are involved in Q-Learning:

  1. Initializing parameters – The RL (reinforcement learning) model learns the set of actions that the agent requires in the state, environment and time.
  2. Identifying current state – The model stores the prior records for optimal action definition for maximizing the results. For acting in the present state, the state needs to be identified and perform an action combination for it.
  3. Choosing the optimal action set and gaining the relevant experience – A Q-table is generated from the data with a set of specific states and actions, and the weight of this data is calculated for updating the Q-Table to the following step.
  4. Updating Q-table rewards and next state determination – After the relevant experience is gained and agents start getting environmental records. The reward amplitude helps to present the subsequent step.  

In case the Q-table size is huge, then the generation of the model is a time-consuming process. This situation requires Deep Q-learning.

Hopefully, this write-up has provided an outline of Deep Q-Learning and its related concepts. If you wish to learn more about such topics, then keep a tab on the blog section of the E2E Networks website.

Reference Links



This is a decorative image for: GAUDI: A Neural Architect for Immersive 3D Scene Generation
October 13, 2022

GAUDI: A Neural Architect for Immersive 3D Scene Generation

The evolution of artificial intelligence in the past decade has been staggering, and now the focus is shifting towards AI and ML systems to understand and generate 3D spaces. As a result, there has been extensive research on manipulating 3D generative models. In this regard, Apple’s AI and ML scientists have developed GAUDI, a method specifically for this job.

An introduction to GAUDI

The GAUDI 3D immersive technique founders named it after the famous architect Antoni Gaudi. This AI model takes the help of a camera pose decoder, which enables it to guess the possible camera angles of a scene. Hence, the decoder then makes it possible to predict the 3D canvas from almost every angle.

What does GAUDI do?

GAUDI can perform multiple functions –

  • The extensions of these generative models have a tremendous effect on ML and computer vision. Pragmatically, such models are highly useful. They are applied in model-based reinforcement learning and planning world models, SLAM is s, or 3D content creation.
  • Generative modelling for 3D objects has been used for generating scenes using graf, pigan, and gsn, which incorporate a GAN (Generative Adversarial Network). The generator codes radiance fields exclusively. Using the 3D space in the scene along with the camera pose generates the 3D image from that point. This point has a density scalar and RGB value for that specific point in 3D space. This can be done from a 2D camera view. It does this by imposing 3D datasets on those 2D shots. It isolates various objects and scenes and combines them to render a new scene altogether.
  • GAUDI also removes GANs pathologies like mode collapse and improved GAN.
  • GAUDI also uses this to train data on a canonical coordinate system. You can compare it by looking at the trajectory of the scenes.

How is GAUDI applied to the content?

The steps of application for GAUDI have been given below:

  • Each trajectory is created, which consists of a sequence of posed images (These images are from a 3D scene) encoded into a latent representation. This representation which has a radiance field or what we refer to as the 3D scene and the camera path is created in a disentangled way. The results are interpreted as free parameters. The problem is optimized by and formulation of a reconstruction objective.
  • This simple training process is then scaled to trajectories, thousands of them creating a large number of views. The model samples the radiance fields totally from the previous distribution that the model has learned.
  • The scenes are thus synthesized by interpolation within the hidden space.
  • The scaling of 3D scenes generates many scenes that contain thousands of images. During training, there is no issue related to canonical orientation or mode collapse.
  • A novel de-noising optimization technique is used to find hidden representations that collaborate in modelling the camera poses and the radiance field to create multiple datasets with state-of-the-art performance in generating 3D scenes by building a setup that uses images and text.

To conclude, GAUDI has more capabilities and can also be used for sampling various images and video datasets. Furthermore, this will make a foray into AR (augmented reality) and VR (virtual reality). With GAUDI in hand, the sky is only the limit in the field of media creation. So, if you enjoy reading about the latest development in the field of AI and ML, then keep a tab on the blog section of the E2E Networks website.

Reference Links




Build on the most powerful infrastructure cloud

A vector illustration of a tech city using latest cloud technologies & infrastructure