HNHacker News
TopNewBestAskShowJobs

ilovefood

839 karma · joined December 5, 2013

karim dot jedda at gmail dot com https://karimjedda.com
submissionscomments
ilovefood··on Data Engineering with Rust – An online book
Hi all, after a lot of feedback I've pushed the new version of my Data Engineering with Rust online book.

It's free for now and I've ported the first chapters from the previous version, would love to hear what you think !

Karim

ilovefood··on The C++ Killers (Not You, Rust)
Beautiful and well written article. Unfortunately I made a small mistake in the second quizz :P
ilovefood··on Pandoc
I love Pandoc!

I recently learned you can use LUA to write custom plugins and change some of the converting behavior. I'm using it for example to create slides similar to the "sent" program.

It helps me bootstrap new presentations and talks very quickly: https://github.com/KarimJedda/justslides

ilovefood··on It's time to put REST to rest
The article is good, and I get the point. However, from experience, this:

> POST on /queries/enlisted-students-on-joining-date/version/1 { "date": "2023-09-22" } to retrieve all students that joined on a given date.

always ends up in a complete and absolute mess, where every possible query gets it's own random name, different parameters, ending up with duplicates all over the place. Also, while it can be possible to cache these POST requests (responses), it's additional work and more friction compared to REST. I'm not convinced the tradeoff is worth it in this particular case.

All in all, I don't know if the current proposal can be an alternative either but the idea of a "REST-lite" goes in the right direction and it's a great start.

However this is a complete no for me:

> When a requested student is nonexistent, your API can return 200 OK HTTP status code with a { "user": null, "message": "No user exists with the specified ID" } response body.

If nothing is found, I want a 404! :D

ilovefood··on Pushing ChatGPT's Structured Data Support to Its Limits
> There is promise in constraining output to be valid JSON. One new trick that the open-source llama.cpp project has popularized is generative grammars

This has been working for months now and is the best method for this type of stuff, a thing for moat-lovers. Too bad it wasn't explored here, the text-based methods turned out to be mainly an unreliable waste of time.

ilovefood··on Ask HN: What's the coolest non standard application of LLMs you've seen?
Thank you very much for the supporting words, I've been getting lots of positive feedback from this. The end of year workload has made it such though that I need to be mindful of time. I think one of the two first options will be the way to go. I'll post an update here as I always do with my small projects.
ilovefood··on Ask HN: What's the coolest non standard application of LLMs you've seen?
I would say, bring it on! Nothing will make it past my phi-2 or mistral-7B-v0.1 ^^, at least for now.

I think what this could lead to is homogenization of the content serving layer, since all you'd really need is to get content to the user that can move their filters from one site to the other, the display layer being less relevant (and differentiating). But let's see, exciting times.

ilovefood··on Ask HN: What's the coolest non standard application of LLMs you've seen?
I'm using it to filter out the content that's displayed in my browser screen as I browse: https://karimjedda.com/llms-in-the-middle-content-aware-clie...

Essentially, I wrote a small browser extension, that takes the content of LinkedIn, Twitter, YouTube posts/titles, and filters them out based on if they are clickbait, low effort, etc.

It's liberating :D

ilovefood··on Show HN: DuckDB-WASM, execute queries in a browser, and share them as links
Amazing, I was eagerly waiting for this one. Loading extensions in previous DuckDB-WASM releases didn't work seamlessly. Looks like now it's the case :D

ref: https://github.com/duckdb/duckdb-wasm/issues/1542#issuecomme...

Thanks!!

ilovefood··on LLMs in the middle: Content aware client-side filtering
Full agreement, but since these sort of approaches are new, I'm treading very carefully here. It was very important to me to have this working locally first and prove the concept is sound. One thing is sure: my browser activity changed significantly since I started using this ~2 weeks ago. I'll share an analysis of this down the line.
ilovefood··on Ask HN: Speech to text models, are they usable yet?
I have been using this with a lot of success for a while now: https://github.com/KoljaB/RealtimeSTT/tree/master , it works in real time, without any delays on an old Nvidia card.

I tried it with German & English without issues. It should also work for French but might need a bit of tweaking. The code is very straightforward, but depending on the context I'd recommend experimenting with the parameters that would suit you.

It's using a model called "Whisper" under the hood.

Have fun :)

ilovefood··on Efficient streaming language models with attention sinks
It's okay I have to say. I just ran out of memory on my 4090, so I had to retry on an A100. Here's an extract: https://pastebin.com/pzLfCFWt

I think something might be off with the example. Can't wait for this stuff to work on llama.cpp. Going to try it with mistral & stable lm now, thankfully tomorrow is a holiday in Germany :)

ilovefood··on Efficient streaming language models with attention sinks
This is working relatively well, the code is really worth a read. If you run it locally, consider the open PR and install sentencepiece as well. It's been generating text for the past 10 minutes now :D

Some of the instructions are ignored though so I'd be careful there, one instruction is to rewrite the previous response by "starting every sentence with the letter A" which is a bit of a hit or miss right now.

ilovefood··on Guidance: A guidance language for controlling large language models
I've been using this library a lot, it's amazing. However, I noticed a very considerable degradation (time taken + generation quality) with versions > 0.0.58 when used with local LLMs.

I haven't taken time to compare between the different releases but if anyone is having the same type of issues, I recommend downgrading even if it might mean less features.

ilovefood··on Nginx Unit – Universal web app server
I've been tinkering quite a bit with WasmCloud and WasmEdge lately, for some data workloads (scraping, ingesting, apis etc) mainly as an excuse to learn WebAssembly.

What I want to see for myself is how Unit compares to those for this reduced scope (api's/services) and how fast I can be productive with it. This could be a good starting point for a lot of new applications, kind of an "airflow (the data orchestration thing) on Webassembly".

So main points are performance (is it good enough) and developer experience (is it easy to maintain/change).

Nothing really ground breaking, just trying out stuff.

ilovefood··on Nginx Unit – Universal web app server
This is really great! What I'm really look forward to trying is the server side WebAssembly: https://unit.nginx.org/configuration/#configuration-wasm.

Huge potential!

ilovefood··on Automatic Generation of Visualizations and Infographics with LLMs
Thanks for the positivity, I'll give it a shot.
ilovefood··on Automatic Generation of Visualizations and Infographics with LLMs
Thank you for the kind words. I wouldn't even know where to start. There's a lot I can code and do, but making something into a product is something I have no experience with.

What's your recommended approach? Even if open-source, curious to learn.

ilovefood··on Automatic Generation of Visualizations and Infographics with LLMs
I took this a step further, turning charts into infographics using stable diffusion (SDXL)

https://karimjedda.com/beautiful-data-visualizations-powered...

ilovefood··on Plain old gzip+kNN outperforms BERT and other DNNs
This is completely unbelievable, I love it! The paper is also written in a very accessible way.
ilovefood··on NodePad: LLM-aided graph based note-taking and brainstorming tool
This is really cool and quite fun to use! I very much enjoy having all the information on one screen and pair with an LLM to brainstorm and test ideas.

I'm wondering if there is any way of running this locally, against my own models?

ilovefood··on Rust, RR, Neovim: A perfect debug combination
Sanctioned by the foundation. According to the February 2023 Rust Foundation Board meeting notes (https://foundation.rust-lang.org/static/minutes/2023-02-14-m..., Paragraph 7.), the board reviewed and believes this is fine, this has been in discussions for months.

It is based on the results of a previous survey: https://old.reddit.com/r/rust/comments/wkv8gv/rust_foundatio.... Looking at where we are now it's safe to assume that not all community feedback was given the same level of care, the (vocal) community is against this, as am I.

ilovefood··on UK Treasury trying to find a Head of Cyber Security for £55k
Could it simply be, that these types of jobs (high rank, low pay) are reserved to wealthy people? It might be so that the only ones that can apply are people for which the salary is a mere formality - it's just listed because it might be mandatory for the paperwork. Kind of similar to how artificially high housing cost is some sort of gatekeeping/community control and not actually reflecting demand in some places...

What the job actually gets you (mandate, network, access, decision power, ...) is the real value money couldn't buy otherwise.

But then again this argument might be wrong, because if it were the case why list the job in the first place?

ilovefood··on Llama.rs – Rust port of llama.cpp for fast LLaMA inference on CPU
Great job porting the C++ code! Seems like the reasoning was to provide the code as a library to embed in a HTTP Server, cannot wait to see that happen and try it out.

Looking at how the inference runs, this shouldn't be a big problem, right? https://github.com/setzer22/llama-rs/blob/main/llama-rs/src/...

ilovefood··on Large language models are having their Stable Diffusion moment
This is really great, very good write-up.

Seems it now also supports AVX2 for x86 architectures too. https://twitter.com/ggerganov/status/1634588951821393922

ilovefood··on Ask HN: What Are You Working On? (March 2023)
It's still very modest and in it's starting phases (not ready for Show HN), but I'm working on small website/guide on how to get started using Rust for Data Engineering. I've been using Rust in my new job with quite some success over the past months and decided to share what I learned. As someone who's used to Python for almost a decade now, it is a very nice complement and quite surprisingly fits the bill for things like Data Engineering.

https://datawithrust.com

ilovefood··on Ask HN: Share your personal site
Focusing currently on mobile computing, Rust and other things as they pop up in my head https://karimjedda.com
ilovefood··on Carefully exploring Rust as a Python developer
There are plenty of resources out there to get started, the one you mention is a very good one and can get you pretty far. My personal recommendation would be to decide on a small hobby project, simple in scope and complexity, to get started and figure out how to solve that using Rust. Just like with any other tool, the only way to learn how to use it is to use it. This takes some time but for me it's the best way to learn. So far there is no challenge Rust presented I couldn't solve using the tools at my disposal: Stackoverflow, Google and YouTube. For example, "a tool that takes some data from an API, does some stuff and stores the result in a database" or "an application that gets the latest 10 closing prices for a stock and sends a telegram message with a prediction". The most important part though (at least to me) is to write about it and share your experience, it helps cement the knowledge and might help others who want to start out too.
ilovefood··on Carefully exploring Rust as a Python developer
Hey everyone, author here, happy you're liking my article. I have gotten a lot of feedback on this since I posted it. If there's anything I can add or update or even improve, let me know. I've been using Rust more and more every day and if I can motivate even just one person to give it a try then I'm satisfied. :) Have a great Sunday!
ilovefood··on From Pythonista to Rustacean
This has been a great help to get started but was missing the "guide me through it" part for me. I wrote an article with some code samples for writing Rust code coming from a Python background (data engineering). It's great to see many "Pythonistas" getting interested / curious about Rust. I'm certain there is a lot of good things in both languages even though the onboarding experience is a bit more hands on with Python.

Might be helpful for someone, my small contribution to how to get started with Rust by "keeping it close to Python" https://karimjedda.com/carefully-exploring-rust/

← PreviousPage 2 of 8Next →