We’re leaving Kubernetes

Data scientists, MLOps engineers, or AI developers, can mount large language model weights or machine learning model weights in a pod alongside a model-server, so that they can efficiently serve them without including them in the model-server container image. They can package these in an OCI object to take advantage of OCI distribution and ensure efficient model deployment. This allows them to separate the model specifications/content from the executables that process them.

Art. 6719

Deploying Machine Learning Models with Flask and AWS Lambda: A Complete Guide

"production": {
    "app_function": "app.app",
    "exclude": [
        "boto3",
        "dateutil",
        "botocore",
        "s3transfer",
        "concurrent"
    ],
    "profile_name": null,
    "project_name": "flask-test-app",
    "runtime": "python3.10",
    "s3_bucket": "zappa-31096o41b"
}

# Convert the data into a numpy array
input_data = np.array(data['input']).reshape(1, -1)

# Make a prediction using the model
prediction = model.predict(input_data)

# Return the prediction as a JSON response
return jsonify({'prediction': int(prediction[0])})

I’m writing this on October 15th, 2024. Last week I would’ve said you probably shouldn’t be using uv’s Python in production, because you wouldn’t be getting security updates to OpenSSL. This week, I would tentatively say that it’s fine. This makes me a little uncomfortable, because there may well be other issues I haven’t thought of, and uv is still very new.

The uv-provided Python executable is slower than the one shipped by Ubuntu 24.04 LTS, but it’s faster than the “official” Docker image.

The ability to install Python with uv adds interesting possibilities for production packaging. For example, you can use an Ubuntu 24.04 base Docker image, download uv, and rely on uv to trivially install any Python version. Which is to say, you won’t be limited to the versions Ubuntu packages for you.

Unlike most Python packaging tools, uv doesn’t require Python to be installed to use it.

Optimizing Kubernetes Costs with Multi-Tenancy and Virtual Clusters

Cena złota przekroczyła 2700 USD za uncję Od początku tego roku złoto wyceniane w USD podrożało o blisko 32%, a w PLN o niecałe 33%. Metal zmierza do osiągnięcia najlepszego roku notowań od 45 lat.

Jak wynika z założeń planowanych zmian już niedługo korzystanie z 5% podatku dochodowego przez programistów może zostać ograniczone. Programista, który będzie chciał korzystać z IP BOX będzie musiał zatrudniać przynajmniej trzech pracowników na umowę o pracę przez 300 dni w skali roku. Możliwe będzie zatrudnienie 3 osób na umowę zlecenia, ale pod warunkiem, że miesięczne wynagrodzenie będzie wynosiło 3-krotność minimalnego wynagrodzenia (ok. 24 500 zł miesięcznie). Zmiany mogą zacząć obowiązywać już w 2025 roku, oznacza to, że wielu programistów prowadzących jednoosobowe działalności gospodarcze nie będzie mogło już korzystać z obniżonej 5% stawki podatku dochodowego.

Jeśli informatyk świadczy usługi związane z oprogramowaniem, nie ma prawa do 8,5 proc. zryczałtowanego podatku. Musi płacić 12 proc.

Niewielkie wzrosty odnotował też marketing i sprzedaż - o 1 proc. czy sektor IT - o 5 proc.

Remember to bring DVI/VGA-to-DisplayPort adapter if you own a Mac

Design, Setting, and Participants  In a population-based registry study, data on all Finnish citizens born between January 1, 1985, and December 31, 1997, whose demographic, health, and school information were linked from nationwide registers were included. Cohort members were followed up from August 1 in the year they completed ninth grade (approximately aged 16 years) until a diagnosis of mental disorder, emigration, death, or December 31, 2019, whichever occurred first. Data analysis was performed from May 15, 2023, to February 8, 2024.

Regularne spożywanie umiarkowanych ilości kawy i herbaty może chronić przed rozwojem wielu chorób kardiometabolicznych, w tym cukrzycy typu 2, choroby wieńcowej i udaru, tak przynajmniej wynika z nowych badań przeprowadzonych przez szwedzkich oraz chińskich naukowców.

Slashing Data Transfer Costs in AWS by 99%

The seasoned engineer learns that sometimes the best code is the code you never wrote. They become adept at delegating tasks, capitalizing on the strengths of their colleagues, and asking the dreaded question, "But why?" — a question that often leads to the heart of what needs to be solved, avoiding unnecessary work and focusing on what truly adds value.

Polacy ciągle się bogacą. Wzrasta liczba dobrze i bardzo dobrze zarabiających

Neither of the methods shown above are ideal in environments where you require several clusters or need them to be provisioned in a consistent way by multiple people.

Running a cluster directly on EC2 also gives you the choice of using any available Kubernetes distribution, such as Minikube, K3s, or standard Kubernetes as deployed by Kubeadm.

EKS is popular because it’s so simple to configure and maintain. You don’t need to understand the details of how Kubernetes works or how Nodes are joined to your cluster and secured. The EKS service automates cluster management procedures, leaving you free to focus on your workloads. This simplicity can come at a cost, though: you could find EKS becomes in-flexible as you grow, and it might be challenging to migrate from if you switch to a different cloud provider.

The EKS managed Kubernetes engine isn’t included in the free tier. You’ll always be billed $0.10 per hour for each cluster you create, in addition to the EC2 or Fargate costs associated with your Nodes. The basic EKS charge only covers the cost of running your managed control plane. Even if you don’t use EKS, you’ll still need to pay to run Kubernetes on AWS. The free tier gives you access to EC2 for 750 hours per month on a 12-month trial, but this is restricted to the t2.micro and t3.micro instance types. These only offer 1 GiB of RAM so they’re too small to run most Kubernetes distributions.

Some of the other benefits of Kubernetes on AWS include

Note that the Python documentation refers to these as special methods and notes the synonym "magic method" but very rarely uses the term "dunder method". However, "dunder method" is a fairly common Python colloquialism, as noted in my unofficial Python glossary.

python -m webbrowser https://pym.dev/p

["bash", "-i", "-c", "read -p 'Type a message: ' -t 10 && echo Attach $REPLY"],

Consequently, it’s one of the only commands that consistently allows interactions with users.

There are six available lifecycle script hooks

In one study led by researchers at The University of Oxford, participants with insomnia were divided into two groups and given fake or "sham" feedback on their sleep.One group was told they had a "positive" night's sleep, the other a "negative" night's sleep, and were then asked to rate their mood and sleepiness.Those who were given a fake "negative" score, rated themselves as much sleepier, and their mood significantly worse than those who were given a fake "positive" score, and vice versa.

Alerty BIK to więcej niż ochrona przed wyłudzeniem

At Google, an AI team member said the burnout is the result of competitive pressure, shorter timelines and a lack of resources, particularly budget and headcount. Although many top tech companies have said they are redirecting resources to AI, the required headcount, especially on a rushed timeline, doesn’t always materialize. That is certainly the case at Google, the AI staffer said.

A common feeling they described is burnout from immense pressure, long hours and mandates that are constantly changing. Many said their employers are looking past surveillance concerns, AI’s effect on the climate and other potential harms, all in the name of speed. Some said they or their colleagues were looking for other jobs or switching out of AI departments, due to an untenable pace.

Artificial intelligence engineers at top tech companies told CNBC that the pressure to roll out AI tools at breakneck speed has come to define their jobs.

Lesson 3: When executing a lot of requests to S3, make sure to explicitly specify the AWS region.

Lesson 2: Adding a random suffix to your bucket names can enhance security.

Lesson 1: Anyone who knows the name of any of your S3 buckets can ramp up your AWS bill as they like.

Google said Axion provides “up to 30% better performance than the fastest general-purpose Arm-based instances available in the cloud today” and “up to 50% better performance and up to 60% better energy-efficiency” than other general purpose Arm chips.

Google’s new AI chip is a rival to Nvidia, and its Arm-based CPU will compete with Microsoft and Amazon

Socially, we’re told, “Go work out. Go look good.” That’s a multi-player competitive game. Other people can see if I’m doing a good job or not. We’re told, “Go make money. Go buy a big house.” Again, external multiplayer competitive game. Training yourself to be happy is completely internal. There is no external progress, no external validation. You’re competing against yourself—it is a single-player game.

Replacing the lock icon with a neutral indicator prevents the misunderstanding that the lock icon is associated with the trustworthiness of a page, and emphasizes that security should be the default state in Chrome. Our research has also shown that many users never understood that clicking the lock icon showed important information and controls. We think the new icon helps make permission controls and additional security information more accessible, while avoiding the misunderstandings that plague the lock icon.

To address the issues of CAS, Karpenter uses a different approach. Karpenter directly interacts with the EC2 Fleet API to manage EC2 instances, bypassing the need for autoscaling groups.

The problem occurs when you want to move the pod to another node, in cases such as cluster rebalancing, spot interruptions, and other events. This is because the EBS volumes are zonal bound and can only be attached to EC2 instances within the zone they were originally provisioned in.This is a key limitation that CAS is not able to take into an account when provisioning a new node.

Since Karpenter can schedule nodes quicker, it will most often win this race and provide a new node for the pending workload. CAS will still attempt to create a new node, however will be slower and will most likely have to remove the node after some time, due to emptiness. This brings unnecessary costs to your cloud bill

It’s worth mentioning that Cluster Autoscaler and Karpenter can co-exist within the same cluster.

I recently chatted with a data science leader who described their company reaching this state. They couldn’t show any business impact from the past two years of their product releases, so the finance team identified a surefire way for R&D to make a business impact: laying off much of the R&D team.

Who am I speaking to?What do I want?What do they care about?How can I explain it to them in terms they care about?

To recap, I think these are my personal rebase rules I follow:

Besides communication, there are other soft skills:teamworklearning mindsetorganization/time managementemotional intelligence/empathyapproachabilitypersistence/patienceconfidence

You can think of it as the following cycle:software engineer writes codeusers get new featuresmore users use your productscompany profits from productsSo code is just a tool to get profit.

2) You will rarely get greenfield projects

However, as we want to do perform the bisection automatically using as criterion ./calc.py 14 0, we run git bisect run ./calc.py 14 0

Git Bisect! It allows us to find the commit that broke something. Given a “good” commit (a commit that is not broken, created before the introduction of the bug), and a “bad” commit (a commit that certainly is broken), Git will perform a binary search until the broken commit is found.

What are the tools that comes on your mind when someone say “debug”? Let me guess: a memory leak detector (e.g. Valgrind); a profiler (e.g. GNU gprof); a function that stops your program and gives you a REPL (e.g. Python’s breakpoint and Ruby’s byebug); something that we call a “debugger” (like GDB, or something similar embedded on the IDEs); or even our old friend, the print function. So, in this text I’ll try to convince you to add Git to your debug toolbelt.

The same LM can be a much more or less capable agent depending on the enhancements added. The researchers created and tested four different agents built on top of GPT-4 and Anthropic’s Claude:

The latest GPT-4 model from OpenAI, which is trained on human preferences using a technique called RLHFEstimated final training run compute cost: ~$50mModel version: gpt-4-0613

only 15% of European biopharma respondents reported no interest at all in AI in their organization

Additionally, students in the Codex group were more eager and excited to continue learning about programming, and felt much less stressed and discouraged during the training.

On code-authoring tasks, students in the Codex group had a significantly higher correctness score (80%) than the Baseline (44%), and overall finished the tasks significantly faster. However, on the code-modifying tasks, both groups performed similarly in terms of correctness, with the Codex group performing slightly better (66%) than the Baseline (58%).

OpenAI is offering limited access to a text-to-voice generation platform it developed called Voice Engine, which can create a synthetic voice based on a 15-second clip of someone’s voice.

Changing the login URL is a feature we do not include in Wordfence. Though it is something that many people swear by and can help a little in certain situations it’s ultimately not very beneficial. These are the reasons why:

By default, curl uses HTTP/1.1 for the http scheme and HTTP/2 for https. You can change this with flags

Sekurak – 4373 Niebezpiecznik – 4171 Z3S – 3383

docker init will scan your project and ask you to confirm and choose the template that best suits your application. Once you select the template, docker init asks you for some project-specific information, automatically generating the necessary Docker resources for your project.

The result? Our runtime image just got 6x smaller! Six times! From > 1.1 GB to 170 MB.

This final trick is not known to many as it’s rather newer compared to the other features I presented. It leverages Buildkit cache mounts, which basically instruct Buildkit to mount and manage a folder for caching reasons. The interesting thing is that such cache will persist across builds!By plugging this feature with Poetry cache (now you understand why I did want to keep caching?) we basically get a dependency cache that is re-used every time we build our project. The result we obtain is a fast dependency build phase when building the same image multiple times on the same environment.

At a minimum, each ADR should define the context of the decision, the decision itself, and the consequences of the decision for the project and its deliverables

We’ve (painstakingly) manually reviewed 310 live MLOps positions, advertised across various platforms in Q4 this year

W cenie 165 euro + podatek (prawie tysiąc złotych) otrzymujemy – teraz nie boję się tego napisać – zabawkę. Fakt, najeżoną techniką, ale zabawkę. W dodatku do jakiegokolwiek działania wymaga ona sporej wiedzy i umiejętności programowania. Niestety w głównej mierze „działająca z pudełka” funkcja tego urządzenia sprowadza się do uniwersalnego pilota telewizyjnego.

console.table($$('a'),['innerHTML','href'])

LocalStack is a cloud service emulator that runs AWS services solely on your laptop without connecting to a remote cloud provider .

The most common reasons I hear for hating on JIRA are:  - It's too complicated  - I spend more time tracking tickets than doing workTo that I say...You hate your micro-Manager

setuptools is the most popular (at 50k packages), Poetry is second at 41k, Hatchling is third at 8.1k. Other tools to cross 500 users include Flit (4.4k), PDM (1.3k), Maturin (1.3k, build backend for Rust-based packages).

Rick was a very talented developer. Rick could solve complex business logic problems and create sophisticated architectures to support his lofty designs. Rick could not solve the problem of how to work effectively on a team.

I dove into the source code. Rick was right: no-one could possibly understand what Rick had created. Except for Rick. It was a reflection of the workings of his own mind. Some of it was very clever, a lot of it was copy-pasta, it was all very idiosyncratic, and it was not at all documented.

“MLX” is more than just a technical solution; it is an innovative and user-friendly framework inspired by popular frameworks like PyTorch, Jax, and ArrayFire. It facilitates the training and deployment of AI models on Apple devices without sacrificing performance or compatibility.

Kubeflow 1.8 adds ARM processor support which simplifies adoption for Apple Silicon users and IoT servers.

RUN poetry install --without dev && rm -rf $POETRY_CACHE_DIR

You can now run Amazon EKS clusters on a Kubernetes version for up to 26 months from the time the version is generally available on Amazon EKS.

Rosetta is now Generally Available for all users on macOS 13 or later. It provides faster emulation of Intel-based images on Apple Silicon. To use Rosetta, see Settings. Rosetta is enabled by default on macOS 14.1 and later.

PHP would serve WordPress when it's run as a standalone Wasm application.

WebAssembly brings true portability to the picture. You can build a binary once and run it everywhere.

However, on top of the big image size, traditional containers are also bound to the architecture of the platform on which they run.

Wasm container images are much smaller than the traditional ones. Even the alpine version of the php container is bigger than the Wasm one.

With WASI SDK we can build a Wasm module out of PHP's codebase, written in C. After that, it takes a very simple Dockerfile based on scratch for us to make an OCI image that can be run with Docker+Wasm.

Docker Desktop now includes support for WebAssembly. It is implemented with a containerd shim that can run Wasm applications using a Wasm runtime called WasmEdge. This means that instead of the typical Windows or Linux containers which would run a separate process from a binary in the container image, you can now run a Wasm application in the WasmEdge runtime, mimicking a container. As a result, the container image does not need to contain OS or runtime context for the running application - a single Wasm binary suffices.

We now have WebAssembly. Its technical features and portability make it possible to distribute the application, without requiring shipping OS-level dependencies and can run with strict security constraints.

If WASM+WASI existed in 2008, we wouldn't have needed to create Docker. That's how important it is. WebAssembly on the server is the future of computing.

There are Wasm runtimes that can run outside of the browser, including traditional operating systems such as Linux, Windows and macOS. Because they cannot rely on a JavaScript engine being available they communicate with the outside world using different interfaces, such as WASI, the WebAssembly System Interface. These runtimes allow Wasm applications to interact with their host system in a similar (but not quite the same) way as POSIX. Projects like WASI SDK and wasi-libc help people compile existing POSIX-compliant applications to WebAssembly.

Browser engines integrate a Wasm virtual machine, usually called a Wasm runtime, which can run the Wasm binary instructions. There are compiler toolchains (like Emscripten) that can compile source code to the Wasm target. This allows for legacy applications to be ported to a browser and directly communicate with the JS code that runs in client-side Web applications.

the new Docker+Wasm integration allows you to run a Wasm application alongside your Linux containers at much faster speed.

Docker Desktop and CLI can now manage both Linux containers and Wasm containers side by side.

The results showed that the group asked to reduce their social media use had an average 15% improvement in immune function, including fewer colds, flu, warts, and verrucae, a 50% improvement in sleep quality, and 30% fewer depressive symptoms.

Adolescents who spend more than 3 hours per day on social media may be at heightened risk for mental health problems, particularly internalizing problems.

We wrześniu 2023 roku w większości badanych zawodów zanotowano spadki liczby ofert pracy rok do roku Największy widoczny jest w branży IT – pracodawcy opublikowali o 52 proc. mniej ofert rok do roku

How to assess durability

Mandel’s system was simple — but incredibly complex from a logistical standpoint

merge queue prevents semantic merge conflicts by automating the rebase process during merge, and ensuring that the trunk branch stays “green.”

When I create I learn. When I consume I just relax

We all know the old saying practice makes perfect. The more we use a certain region of our brain, the more our brain "prioritizes" and "hones" it. That is what leads to myelin: activity induces myelination, which leads to increased strength of connectivity and efficiency along those very neurons. It’s a self-reinforcing process.

The fact of the matter is that digital products make it uniquely easy to trick yourself into thinking that you’re learning when you are actually being entertained.

learning must be effortful in order for it to happen

engineering blogs focus on problems where the solution is a necessary but not sufficient part of what they do. And, ideally, they focus on problems that are complementary to scale that only the publisher of that post has.

cat requirements.txt | grep -E '^[^# ]' | cut -d= -f1 | xargs -n 1 poetry add

What happened here is that the file 'somefile.txt' is encoded in UTF-16, but your terminal is (probably) by default set to use UTF-8.  Printing the characters from the UTF-16 encoded text to the UTF-8 encoded terminal doesn't show an apparent problem since the UTF-16 null characters don't get represented on the terminal, but every other odd byte is just a regular ASCII character that looks identical to its UTF-8 encoding.

Writing to the database may fail (e.g. it will not respond). When that happens, the process handling outbox pattern will try to resend the event after some time and try to do it until the message is correctly marked as sent in the database.

staff are more open to returning to the office if it is out of choice, rather than forced

Unispace finds that nearly half (42%) of companies that mandated office returns witnessed a higher level of employee attrition than they had anticipated. And almost a third (29%) of companies enforcing office returns are struggling with recruitment. Imagine that — nearly half!

python -m calendar

python -m site, which outputs useful information about your installation

sudo softwareupdate -ia installs all available updates.

The results from both Midjourney and Stable Diffusion seem to be the most convincing and realistic if I was to judge from a human point of view and if I didn't know they were AI generated, I would believe their results.

For a new project, I’d just immediately start with Ruff; for existing projects, I would strongly recommend trying it as soon as you start getting annoyed about how long linting is taking in CI (or even worse, on your computer).

A documentation-first culture does not mean everyone is busy writing documents all day. It means that everyone appreciates the value of documenting and sharing their experiences.

I’m a big fan of documentation. I think it’s my favorite boring thing to do after coding. It brings the business so much long-term value that every hour invested into documentation by anyone saves literally x100 productivity hours across the company.

All of these values, including the precious contents of the private key file, can be seen via ps when these commands are running. ps finds them via /proc/<pid>/cmdline, which is globally readable for any process ID.

Digital nomads must earn at least €2,800 per month to qualify for its new visa, around four times Portugal’s minimum wage. According to Nomad List nearly 16,000 people were remote working in Lisbon last December, where they now find themselves blamed for rocketing rents and house prices.

According to a March survey, 36 per cent of digital nomads have an annual income of between $100,000 and $250,000. Another eight per cent earn between $250,000 and one million. Attracted by these bank balances, dozens of countries have now introduced so-called “digital nomad visas” (permitting extended stays to work remotely).

This is the script, which I’ve named docker and put before the real Docker CLI in my PATH

There are better parameters to evaluate quality, not quantity, of the time spent staring at your screens

The key to hacking yourself is to increase your awareness of your emotional state. When you become aware that you are angry, the anger is losing the grip it has over you. When you are angry, you are sometimes doing things you would not have done if you were not angry. (Sometimes anger is healthy, it may also be a signal to us that our boundaries have been violated.)

Examples of frontends include: pip, build, poetry, hatch, pdm, flit Examples of backends include: setuptools (>=61), poetry-core, hatchling, pdm-backend, flit-core

pyproject.toml-based builds are the future, and they promote better practices for reliable package builds and installs. You should prefer to use them!

Did you say setuptools? Yes! You may be familiar with setuptools as the thing that used your setup.py files to build packages. Setuptools now also fully supports pyproject.toml-based builds since version 61.0.0. You can do everything in pyproject.toml and no longer need setup.py or setup.cfg.

// save to tar filedocker save nodeversion > nodeversion.tar// load from tar filedocker load < nodeversion.tar

Developers often speak of "getting into the flow" or "being in the zone." Such statements colloquially describe the concept of flow state, a mental state in which a person performing an activity is fully immersed in a feeling of energized focus, full involvement, and enjoyment.

With this dataclass, I have an explicit description of what the function returns.

Today is 9th Feb. The oldest segment – segment 100 – still can’t be deleted by the 7-day topic retention policy, because the most recent message in that segment is only 5 days old. The result is that they can see messages dating back to 28th January on this topic, even though the 28th Jan is now 12 days ago. In a couple of days, all the messages in segment 100 will be old enough to exceed the retention threshold so that segment file will be deleted.

Host machine: docker run -it -p 8888:8888 image:version Inside the Container : jupyter notebook --ip 0.0.0.0 --no-browser --allow-root Host machine access this url : localhost:8888/tree‌

Armed with all this knowledge, we realise that we can construct an almost unlimited number of different path strings that all refer to the same directory

UNC stands for Universal Naming Convention and describes paths that start with \\, commonly used to refer to network drives. The first segment after the \\ is the host, which can be either a named server or an IP address

On any Unix-derived system, a path is an admirably simple thing: if it starts with a /, it’s a path. Not so on Windows

study that found that friends living within a mile of each other are 25 percent more likely to feel happy

Em dash Works better than commas to set apart a unique idea from the main clause of a sentence

The hyphen does not indicate a range of numbers, like a date range, which is the job of an en dash

As I see it, there are two hard things about getting into flow: loading the state of the system / problem / abstractions into your head (i.e. filling your L1 and L2 cache with everything you need to know to work on the problem) and building momentum and confidence for yourself.

To return to information overload: this means treating your "to read" pile like a river (a stream that flows past you, and from which you pluck a few choice items, here and there) instead of a bucket (which demands that you empty it). After all, you presumably don't feel overwhelmed by all the unread books in the British Library – and not because there aren't an overwhelming number of them, but because it never occurred to you that it might be your job to get through them all.

If you suddenly transform the social lives of girls, putting them onto platforms that prioritize social comparison and performance, platforms where we know that heavy users are three times more likely to be depressed than light users, might that have some impact on the mental health of girls around the world? We think so

Kilka przykładów z wykorzystujących Conventional Commits:

Conventional Commits – type

Commity o treści refactor, added XXX czy cr fixes, to smutna i nudna rzeczywistość.

If you install a package with pip’s --user option, all its files will be installed in the .local directory of the current user’s home directory.

rg . | fzf: Fuzzy search every line in every file

Spend your career hanging out with people you like working with, doing work you enjoy, trying new experiences, and having fewer regrets.That’s how you retire one day & make it to your deathbed a happy human.