Рубрика: AI

Image generated by Author with DALL·E 3

For many professional data scientists, Jupyter Notebook has become their staple working environment. Even for me, it’s always the first place I go to for any data science experiment and workflow.

As a data scientist workplace, Jupyter Notebook is a unique IDE as the code can be executed independently in each cell. At the same time, the author could explain each cell. This distinction allows the notebook to be reused by others and become a project template.

In this article, we will discuss five free templates for building a data science project on Jupyter Notebook. So, what are these Jupyter Notebook templates? Let’s get into them.

1. Cookiecutter Template for Python Data Science Projects

The first template that we discuss is not necessarily a complete code project that we can fill in already. The template we would discuss is not only the Jupyter Notebook but also complete projects that support the Jupyter Notebook. What we would have is the Python data science projects by AWS.

This template creates a complete data science project structure that is ready to be used for your actual project. Using Cookiecutter CLI, you can generate the directory structure of the data science projects, which is similar to the structure below.

|-- bin/  |-- notebooks                    # A directory to place all notebooks files.  |   |-- *.ipynb  |   `-- my_nb_path.py            # Imported by *.ipynb to treat src/ as PYTHONPATH  |-- requirements/  |-- src  |   |-- my_custom_module         # Your custom module  |   |-- my_nb_color.py           # Imported by *.ipynb to colorize their outputs  |   `-- source_dir               # Additional codes such as SageMaker source dir  |-- tests/                       # Unit tests  |-- MANIFEST.in                  # Required by setup.py (if module name specified)  |-- setup.py                     # To pip install your Python module (if module name specified)    # These sample configuration files are auto-generated too:  |-- .editorconfig                # Sample editor config (for IDE / editor that supports this)  |-- .gitattributes               # Sample .gitattributes  |-- .gitleaks.toml               # Sample Gitleaks config (if pre_commit is advanced)  |-- .gitignore                   # Sample .gitignore  |-- .pre-commit-config.yaml      # Sample precommit hooks  |-- LICENSE                      # Boilperplate (auto-generated)  |-- README.md                    # Template for you to customize  |-- pyproject.toml               # Sample configurations for Python toolchains  `-- tox.ini                      # Sample configurations for Python toolchains  

If you are interested in how the template is applied to a real project, check its actual usage above this Reinforcement Learning Energy Storage use case.

2. Data Science Notebooks Templates by Coen Meintjes

The next Jupyter Notebook template we will discuss is the one by Coen Meintjes. It is a basic Jupyter Notebook collection from Data Exploration to Model Evaluation. It’s not a project-specific kind of template; in fact, it mostly consists of the essential code, nothing more. But I would say it is good. Why is that?

It’s a basic template that everyone can use for different kinds of projects with little tweaks. You can use this Jupyter Notebook template to develop any project idea. Moreover, the templates here explain in depth many of the processes in the notebook, so any beginner or professional could benefit from them.

3. Data Science Projects by Yusuf Cinarci

Let’s move on to a more project-specific template, the Data Science Projects Jupyter Notebook templates by Yusuf Cinarci. In these templates, you can use them to develop simple projects for your portfolio or any business needs.

There are many project templates you can choose from. You can choose from simple Salary Data Exploration with Python to the development of a Fake News Detection and Movie Recommendation System. The notebook is perfect for you who want to kickstart your projects easily.

What I like about the Yusuf Cinarci template collections is that they are not overly complicated, so beginners can start their projects when they learn about data science. However, many of the projects are for beginners, so they might lack data science projects if you are looking for one.

4. Data Science Projects by Sukman Singh

If you need a more complex project Jupyter Notebook template, then the Data Science Projects Jupyter Notebook template by Sukman Singh could be the one for you. It’s perfect for those who want to develop prediction models easily but need inspiration for their ideas.

The template collection contains many data science project templates, including customer churn prediction, loan approval prediction, and claim fraud projects. These are standard business projects that you can use to enrich your data project portfolio.

The project might seem one-dimensional, but you can extend the project as well. It’s a template that you can use to develop your project and use another dataset that you feel is appropriate for your business problem.

5. Awesome Notebooks by Jupyter Naas

Lastly, we will discuss Awesome Notebooks by Jupyter Naas. Awesome Notebook is a project by Jupyter Naas to create the largest catalogue of production-ready Jupyter Notebook templates. There is an abundance of free Jupyter Notebook templates from which you can choose.

The project consists of many professional Jupyter Notebook templates with specific use cases. From AI development to analysing business funnels to YouTube video download, there are so many templates you can choose from.

Many of the templates require you to understand how to work as a data scientist, so learning how to use Python as a data scientist would help you use these templates. Once you understand what you need, this template collection will help your work.

Conclusion

Jupyter Notebook is an environment used by many professional data scientists as many data science breakthroughs happen in this platform. One of its nice features is that it is easily shareable and can become a template that many can use.

In this article, we have discussed five free Jupyter Notebook templates that can be used to boost your data science activity. The templates are:

1. Cookiecutter Template for Python data science projects
2. Data Science Notebooks Templates by Coen Meintjes
3. Data Science Projects by Yusuf Cinarci
4. Data Science Projects by Sukman Singh
5. Awesome Notebooks by Jupyter Naas

Cornellius Yudha Wijaya is a data science assistant manager and data writer. While working full-time at Allianz Indonesia, he loves to share Python and data tips via social media and writing media. Cornellius writes on a variety of AI and machine learning topics.

More On This Topic

• 10 Jupyter Notebook Tips and Tricks for Data Scientists
• Python in Finance: Real Time Data Streaming within Jupyter Notebook
• How to Setup Julia on Jupyter Notebook
• Jupyter Notebook Magic Methods Cheat Sheet
• An Overview of Mercury: Creating Data Science Portfolio and…
• Analyze Python Code in Jupyter Notebooks

Stone Ridge Technology (SRT) pioneered the use of GPUs for high performance reservoir simulation (HPC) nearly a decade ago with ECHELON, its flagship software product. ECHELON, the first of its kind, engineered from the outset to harness the full potential of massively parallel GPUs, stands apart in the industry for its power, efficiency, and accuracy. Now, ECHELON has added support for AMDInstinct accelerators into its simulation engine, offering new flexibility and optionality to its clients.

Reservoir simulators model the flow of hydrocarbons and water in the subsurface of the earth in the presence of wells. Energy companies use them to create and assess field development strategies. Emily Fox, SRT Director of Communications remarks, “SRT pioneered the integration of GPU accelerators in high-performance reservoir simulations and continues to lead by now offering AMD Instinct as a computational platform.” SRT’s CTO and ECHELON developer Ken Esler adds, “Many in the field were skeptical about the efficacy of GPUs, however, ECHELON’s GPU-native design overcame these doubts and marked a significant leap in simulation speed and performance, firmly establishing SRT’s position in the industry.”

Esler explains, “Over a decade ago we set out to supercharge the capability of a CPU-based in-house simulator at a major oil company. We could see the enormous potential of GPUs with their high memory-bandwidth and floating-point performance. We also realized that to unlock the true potential of GPUs, we needed to build a new simulator from scratch, specifically designed for GPU acceleration. This was the genesis of ECHELON, around 2013. Since then, we’ve continuously enhanced its features, robustness, and performance, first alone, and since 2018 together with our consortium partner Eni S.p.A”

Expanding GPU Options to Meet Customer Demands

SRT’s latest development effort was to port ECHELON from CUDA to the AMD HIP platform, enabling ECHELON to use AMD Instinct GPUs like the MI210, MI250X, and the upcoming MI300 Series. This strategic decision broadens ECHELON’s hardware compatibility and offers clients increased flexibility and choice for their high-performance computing needs. Vincent Natoli, SRT’s CEO says, “Companies need flexibility in selecting their hardware technology and should not be locked into a single vendor. We want our clients to have convenient access to the hardware of their choice when it comes to implementing business critical workflows.”

The timing of SRT’s decision in Ken Esler’s words: “The impressive specifications of AMD’s Instinct processors, particularly the MI210 and MI300, presented an increasingly compelling, competitive solution that caught our attention. The MI210’s memory bandwidth at 1.6 terabytes per second was quite competitive, and the MI300’s subsequent leap to over five terabytes per second is even more exciting. AMD’s innovative approach in processor packaging and its use of chiplets have allowed for larger GPUs without compromising yields, resulting in highly competitive products. The engineering behind these developments is quite impressive.”

The Right Software for a Smooth Port

SRT had been contemplating the idea of adapting ECHELON for use on AMD platforms for a while. “The maturation of the ROCm and HIP ecosystem significantly lowered the barriers to adopting AMD GPUs,” says Esler. The integration with ROCm, AMD’s open software platform, was essential in ensuring that ECHELON could fully leverage AMD’s GPU capabilities. “I appreciated AMD’s strategy with ROCm,” says Esler. “Instead of creating a new, proprietary and incompatible language for accelerated computing, AMD embraced existing frameworks. That significantly reduced the effort needed to adapt our existing code, allowing us to avoid more complex alternatives that are even further removed from ECHELON.”

“We develop most of our code internally, but we do rely on Thrust. AMD’s rocThrust turned out to be an effective drop-in replacement. We were also impressed with the ROCm LLVM compiler and Clang in combination with HIP extensions, which enhanced productivity. The support for debuggers and profilers in ROCm has been beneficial. Overall, we are seeing impressive progress in AMD tool development.”

Collaboration Yields Fast Results

The project began in earnest in spring 2023. Erik Greenwald of SRT described porting from CUDA to HIP as initially straightforward. “We created a wrapper to retarget each build. The initial version was created relatively quickly,” he says. Although there were a few challenges, like adjusting from CUDA’s 32 warp to AMD’s 64 warp, Greenwald found these issues manageable. “It was quite painless and quick to achieve initial results,” he reflects. Esler adds, “With AMD’s support, we steadily enhanced ECHELON’s performance and progress in a very acceptable timeframe. We were pleased and satisfied with the performance.”

ECHELON Gets to Work on AMD Instinct accelerators

ECHELON is developed in a Consortium framework, with charter members SRT and Eni, S.p.A, the Italian integrated energy company. The ECHELON Consortium, is a collaboration of industry partners committed to advancing high-performance subsurface flow simulation and is currently open to participation by new member organizations that would like a role in shaping the development of ECHELON.

Eni S.p.A recently announced its new HPC6 high-performance computing (HPC) system at its Green Data Center. Each of the 3472 computing nodes in HPC6 comprises a 64-core AMD EPYC CPU (AMDEPYC) and four high-performance AMD Instinct MI250X GPUs, offering unmatched computational efficiency and versatility for a wide array of applications.

“With a peak computing power of over 600 PetaFlop/s, HPC6 reaffirms Eni’s leadership position in the field of supercomputing among industrial entities,” says Sergio Zazzera, Head of Technical Computing for Geosciences & Subsurface Operations, Eni. “It enables highly complex simulations with enormous volumes of data, such as those needed for studying new geological basins, forecasting sub-surface flows in complex geologies, researching new materials for CO2 capture, and ensuring plasma stability in the field of magnetically confined fusion. Additionally, HPC6 will take advantage of specialized Generative AI solutions in the energy sector.”

So, what’s ahead for the world’s fastest GPU-powered reservoir simulation software? According to Esler, “We still have the lead in performance relative to our competitors by a good margin, but we are always looking to enhance ECHELON’s performance and extend its features.” There are also exciting prospects in fields beyond traditional hydrocarbon recovery, such as CO₂ and hydrogen storage, emerging areas that represent new frontiers for ECHELON as the energy sector moves towards greener technologies.

Segment Anything with AMD GPUs

Do you ever wonder what happened to Alexa? Mihail Eric, a former research scientist from Alexa AI, wrote a tell-all post about why Alexa is no longer at the forefront of voice assistants, which is particularly true in this era of rapid advancements.

While Siri got ChatGPT integration, Eric undertook to find out why its prime competitor, Alexa, failed. A key reason that came up was it getting embroiled in bureaucracy due to its commitment to pushing an ethical product.

“We had all the resources, talent, and momentum to become the unequivocal market leader in conversational AI. But most of that tech never saw the light of day and never received any noteworthy press,” Eric said on X.

How Alexa dropped the ball on being the top conversational system on the planet
—
A few weeks ago OpenAI released GPT-4o ushering in a new standard for multimodal, conversational experiences with sophisticated reasoning capabilities.
Several days later, my good friends at PolyAI…

— Mihail Eric (@mihail_eric) June 11, 2024

Eric was part of the Conversational Modelling team at AlexaAI in 2019. He worked on improving Alexa’s capabilities through the power of AI. However, since this was much before ChatGPT, the necessary requirements for compute and other infrastructure were not met.

However, this is not what effectively killed Alexa as an AI competitor. While only a few years ago, parents were panicking about their kids learning the name “Alexa” before the word “mama”, things nosedived rapidly.

Eric says this could be attributed to several issues, most notably to several bureaucratic hurdles, including a huge focus on ethically sourcing user data.

Alexa Serves as a Cautionary Tale

While the commonly held belief at the moment is that AI should be built on ethical frameworks, Alexa’s story shows that it is easier said than done.

Data privacy is one of the most significant issues raised regarding ethical AI. Eric stated that in a bid to ensure data privacy was preserved, the company inadvertently set up several roadblocks for itself, effectively halting any advancements in training the voice assistant.

“Definitely a crucial practice, but one consequence was that the internal infrastructure for developers was agonisingly painful to work with. It would take weeks to get access to any internal data for analysis of experiments,” he said.

Interestingly, his next point goes into another major problem that companies face today – the issue of poorly annotated data. Eric stated that despite repeated attempts to ensure that data was properly annotated, this was again bogged down by layers of bureaucracy, leading to further delays in development.

Data Plays a Major Factor Too

Currently, major AI companies have begun a mad scramble for functional datasets. Data as a product (DaaP) has slowly become a point of consideration, especially with customer-facing companies that accrue data.

Meanwhile, companies like OpenAI and Google have struck several partnerships with media companies to have reliable datasets that they can train their LLMs on. However, Alexa largely relied on crowd-sourced data, as well as data taken from Alexa users and employees to train it. This data needed to be properly annotated.

“I remember, on one occasion, our team did an analysis demonstrating that the annotation scheme for some subset of utterance data was completely wrong, leading to incorrect data labels,” Eric said.

However, correcting this proved to be even worse, as it required approval from several teams and a proper justification for why it needed to be done, apart from just “it’s scientifically the right thing to do and could lead to better models for some other team”.

This proved to be true, as accurate datasets are hard to come by and worth their weight in gold. This is precisely why OpenAI has partnered with several media organisations over the past year. They’ve always needed quality datasets that can be reliably used to train ChatGPT.

Meanwhile, Google was an example of what poor datasets could do, as their integration of the Reddit API led to baffling answers to some really innocuous queries from users.

Google AI overview suggests adding glue to get cheese to stick to pizza, and it turns out the source is an 11 year old Reddit comment from user F*cksmith pic.twitter.com/uDPAbsAKeO

— Peter Yang (@petergyang) May 23, 2024

Obviously, there were several other issues with the company that ultimately led to the fall of Alexa. This included a lack of communication between teams as well as a mindset that leaned towards the consumer side rather than the scientific side.

As Eric put it, “The success metric imposed by senior leadership had no scientific grounding and was borderline impossible to achieve.”

With Google, OpenAI and Apple all announcing major upgrades to their multi-modal capabilities, it seems that Alexa is nowhere in the race. However, not all is lost.

How Do You Avoid This?

Eric’s post, while critical of Alexa, also proves to be a valuable lesson for AI companies and startups. As mentioned before, one of the biggest talking points surrounding the industry is keeping in mind ethical AI, but again, this is easier said than done.

However, this doesn’t mean that ethical AI as a whole is impossible and should be abandoned. Eric believes that better data infrastructures need to be put into place to ensure that better models are built.

Further, he said that rapid advancements mean that both companies and startups feel the pressure to ship products quickly. “Of course, you should conduct research aggressively, but don’t have delivery cycles measured in quarters, as this will produce inferior systems to meet the deadlines,” he concluded.

With everyone rushing to stay ahead in the AI race, this can also be remedied if startups work to ensure their products stay relevant and sustainable in the long run.

These days, everyone seems to have strong opinions on LLMs. While some are grounded in research by experts such as Yann LeCun, others just follow the hype to criticise it. Some say they’re our ticket to the AGI, while others think they’re just glorified text-producing algorithms with a fancy name.

One of the biggest arguments against LLMs achieving AGI is that they’re just not like us. As one user puts it in a Reddit discussion, “Human intelligence develops from small amounts of data, in real-time, on 20W of power, using metacognition. By contrast, LLMs work with massive amounts of data, are pre-trained, use massive power, and operate without any cognitive awareness. Therefore, AGI requires a different paradigm.”

Meanwhile, LeCun advises people getting into the AI space to work on something apart from LLMs. “If you are a student interested in building the next generation of AI systems, don’t work on LLMs. This is in the hands of large companies, there’s nothing you can bring to the table,” said LeCun.

Similarly, Francois Chollet, the creator of Keras, also recently shared similar thoughts about this. “OpenAI has set back the progress towards AGI by 5-10 years because frontier research is no longer being published and LLMs are an offramp on the path to AGI,” he said in an interview.

Interestingly, the discussion on alternatives of LLM-based models has been ongoing for a while now. Recently, Mufeed VH, the young creator of Devin’s alternative Devika, spoke about how people should move away from Transformer models and start building new architectures and how we should do more innovation.

LLMs are the projections of the world

But here’s the kicker according to people on Reddit: human intelligence also developed with tons of data and energy. Our cognitive architecture is like the universe’s ultimate information superhighway, built over hundreds of millions of years and encoded in our DNA.

Ilya Sutskever, the former chief scientist at OpenAI, has noted several times that text is the projection of the world. But how much of that is linked with LLMs is still questionable. LLMs, in a way, are building cognitive architecture from scratch, echoing the evolutionary and real-time learning processes, albeit with a bit more electricity.

One insightful comment noted, “An essential similarity between the human brain and LLMs is that they are essentially compression algorithms… compressing massive amounts of world data into worldviews that provide predictive models to guide action.”

The brain’s architecture is that of a finely tuned machine, learning efficiently from small amounts of data in real time. LLMs, on the other hand, need vast amounts of data and computational power to get anywhere close to that level of performance.

On the other hand, it’s a misconception that LLMs are all about scaling up datasets. In fact, progress is happening with ever-smaller datasets and clever techniques like synthetic data generation with positive-feedback cycles, and even very small models such as Phi-3 and Llama.

One user wisely noted, “Our brain isn’t one architecture for all processes. There are different parts of our brain that process things differently to do different things in life.” Indeed, while LLMs emulate parts of the brain related to language, there’s still much work to be done.

Diffusion models, tackling visual processing, and retrieval-augmented generation (RAG), are trying to mimic the hippocampus. But the road to AGI is long and winding, with many more brain regions to cover.

Then what to focus on? A little bit more of LLMs

Just as LeCun suggests, there are many people working in different fields of AI to figure out things other than LLMs.

A Reddit thread sparked this conversation, asking, “What is it like to work on niche topics that aren’t LLM or Vision?” One user shared, “I shifted my focus from computer vision to ML theory. Now I’m working on kernel methods which are nascent but hold promise to explain and interpret large and over-parameterised networks.”

While LLMs grab the headlines, the world of AI research is full of unsung heroes working on everything from speech synthesis to climate modelling. As one comment wisely put it, “Most of the cooler AI applications aren’t chatbots or generating bounding boxes on camera feeds.”

The bottom line is – LLMs are algorithms that utilise extensive datasets, rely on unsupervised learning, generalise skills not explicitly trained for, and have broad applicability to various downstream tasks. This approach resembles human intelligence, but current LLMs are not upgrading themselves every second like humans.

This is one discussion that can go on for the longest time. Humans can continuously learn, while models trained on static datasets with current training methods cannot. But it might be too soon to write off LLMs.

The world is built on spreadsheets. Even before the introduction of modern machine learning tools, most of the data of the world was stored on Excel spreadsheets. Cut to today, and that’s still the case.

Just like any other tool, though, with the advent of applications such as ChatGPT, it was thought that there would be no need for Excel sheets anymore, since data science could easily be done by just allowing LLMs to make sense of the data.

Probably true … pic.twitter.com/s2aSya4vxl

— Wall Street Silver (@WallStreetSilv) June 12, 2024

The AI Revolution in Excel

One of the strongest arguments against Excel was that it was becoming obsolete. Excel is not user-friendly, and the application rounds off very large figures with accurate computations, which reduces its accuracy.

Excel is also a stand-alone application that is not fully integrated with other corporate systems. It does not provide sufficient control because users do not have a clear and consistent view of the quotations sent by their representatives, as well as the history of those quotes.

So, what does the future of Excel look like? We can anticipate a significant role for AI. Microsoft has been steadily integrating AI aspects into Excel, and with its ever-expanding capabilities, it is set to continue revolutionising the software, sparking excitement about the future possibilities.

Microsoft unveiled Copilot, its innovative AI tool, in mid-March 2023. This cutting-edge technology is set to revolutionise the functionality of Microsoft Office programs, including Word, Excel, PowerPoint, Outlook, and Teams.

Consider the practical implications of Copilot in Excel. By harnessing natural language processing (NLP) AI techniques, it empowers users to ask questions in plain language and receive accurate, context-aware answers. This transformative technology enhances Excel’s usability, providing valuable recommendations and precise results.

By integrating Copilot into the Microsoft 365 suite, Microsoft is placing generative AI tools in front of over a billion of its users, possibly changing how large segments of the global workforce communicate with one another.

One example is the Data Analyse feature found in the most recent versions of Excel. NLP can also recommend functions, formulas, and features the user may not know, making it easier to identify the best answer.

Python in Excel

The introduction of Python in Excel has given the application a major boost. Python in Excel uses Anaconda, a prominent repository notable for allowing developers to run multiple Python environments.

Now users can do advanced data analysis within the familiar Excel interface leveraging Python, which is available on the Excel ribbon.

Access to Python allows users to use Python objects within cell functions and calculations. Consider a Python object being referenced or its data used in a PivotTable.

Even popular libraries such as scikit-learn, Seaborn, and Matplotlib can be utilised with Excel. This allows Python-created visualisations, data models, and statistical calculations to be combined with Excel functions and plugins.

The integration of Python cements Excel’s position in data analytics, suggesting that Excel’s utility in the workplace is far from diminishing. Last year, Microsoft announced that they would be experimenting with GPT in their office applications.

Interestingly, Microsoft has been following this path the entire time in acquiring Github, acquiring OpenAI, which helped them make Copilot for writing Python code and use ChatGPT’s Code Interpreter, AI assistance on PowerBI, and, now, Python in Excel.

Looking at the progress of Excel, a future where AI writing Python on our Excel sheets cannot be ruled out. Moreover, the talk of AGI being built around Python makes the future of Excel more secure.

Why AI in Excel Is a Win-Win Situation

Microsoft Excel is the world’s most popular spreadsheet software. Approximately 54% of organisations use Excel. According to a recent report by corporate planning company Board International, 55% of organisations undertake their enterprise planning, including budgeting and sales forecasting, on spreadsheets.

Four out of five Fortune 500 companies use Excel, and over two billion people worldwide use spreadsheets.

Excel and spreadsheets became popular in the 1980s, and despite attempts by competitors, Excel and its ilk continue to dominate. Office workers today utilise tools like Excel to visualise, analyse, organise, distribute, and present chunks of corporate data—whether exported from databases or created on the fly.

Businesses continue to rely on Excel, so much so that financial specialists claim you’ll “have to pry Excel out of their cold, dead hands” before they ever stop using it.