mpeg

16 ms·

271.
by mpeg 2y ago
You must be joking, the title of this post literally says "build fast LLMs from scratch" but the code is neither an LLM, nor particularly fast. Python dicts are actually different from JS objects, Python uses a hashmap behind the
272.
by mpeg 2y ago
Bit of an oversimplification though, no one would look at Postgres and go “it’s just a couple lookup tables put together”
273.
by mpeg 2y ago
Exactly my thoughts, this isn’t really an LLM. I understand it’s just a personal project but it’s probably good to know LLMs are not just lookup tables. Also, I’m not sure I understand the explanation of why js objects are so great for this
274.
by mpeg 3y ago
I’ve had good results in the past by fine tuning YOLO for image classification and object detection within images, you can find a rough guide here on how to create a good training dataset and such [0] [1] YMMV though, ultimately accuracy is
275.
by mpeg 3y ago
… thanks GPT
276.
by mpeg 3y ago
it's https://huggingface.co/InfiniFlow/deepdoc and the code for usage is in https://github.com/infiniflow/ragflow/blob/main/deepdoc/READ... – it took me a bit of trial and
277.
by mpeg 3y ago
Took me some time to figure out how to run it, but the layout recogniser model hosted on huggingface is pretty good! It correctly identifies tables that even paid models like the AWS Textract Document Analysis API fails to – for instance ta
278.
by mpeg 3y ago
nothing really, probably just me, but it always felt a bit ugly to nest function definitions, plus it can separate the (often single-use) function from the place it is being used in
279.
by mpeg 3y ago
I tend to use inner functions (nested def) when I need something more than a simple lambda, but don't want to come up with a good name or pollute the class / module namespace It works okay, but I do wish there was a better way
280.
by mpeg 3y ago
It's other visitors online at the same time, your mouse pointer is sent via websockets every few seconds and received by all other online visitors. I agree it's very distracting
281.
by mpeg 3y ago
Nice, I didn't know about it – I took a quick look at the templating options and it seems like it might do the job. Thanks!
282.
by mpeg 3y ago
Does it export to powerpoint? I have the same issue where my client wants weekly data driven decks and I’ve automated the data part but the presentation layer I offload to other people who often make mistakes copy/pasting or with the c
283.
by mpeg 3y ago
The scale on that bar chart for "Programming (Human Eval)" is wild. Manager: "looks ok, but can you make our numbers pop? just make the LLaMa bar smaller"
284.
by mpeg 3y ago
Yeah I too try to avoid positional args as much as possible, huge source of bugs and time wasting especially when refactoring code Re inspect.getsource, I'm not sure if it'd be a huge performance impact, but if it's in the wr
285.
by mpeg 3y ago
Weird comment... yes, they are, but what is faster for me to type `rm .cache_dir/function_name*.pickle` or just to delete the one sqlite file in my file manager / vscode file tree. Regardless, there are other reasons why sqlite is
286.
by mpeg 3y ago
This is also why in my custom cache I back it with sqlite – much easier to delete one db file than thousands of pickle files.
287.
by mpeg 3y ago
I recently wrote a version of this that I use in my projects, some things I do differently that you may or may not care about: - from your code it seems you're not sorting kwargs, I would strongly recommend sorting them so that whether
288.
by mpeg 3y ago
What you're suggesting is basically the same thing as what the law is achieving. Yes, users could block all cookies but this will break functionality on a lot of sites, so it's not reasonable. And yes, sites could communicate whic
289.
by mpeg 3y ago
You should probably know your HR team is insta-rejecting applications for reasons such as "having moved into management too quickly" with 6 years IC 8 years management. No hard feelings, I wasn't even that invested in it, I j
290.
by mpeg 3y ago
Treat him as an advisor, don’t give him founder-level equity (the idea by itself is not worth a bigger chunk of equity) and instead give him a bigger sales bonus that depends on how much business he can close for you.
291.
by mpeg 3y ago
I guess I wasn’t expecting this process for a management role, I know that at the level I was interviewing I’d be one of hundreds/thousands in Meta but I would have expected a bit more of a personal touch for a role that required 10-15
292.
by mpeg 3y ago
Exactly, I was pretty highly paid for the UK market so I was worried that I’d go through the whole process and then they’d present a total comp that was not a significant upgrade. As the OP says in the article, I was also worried there was
293.
by mpeg 3y ago
Last year in a moment of career uncertainty I applied for a job at Meta, I think it was a M1 role in data engineering – I ended up giving up a few interviews in when I realised I still had 3 more interviews to go, I think for a total of 6,
294.
by mpeg 3y ago
If I had to guess, I would say the reason is that they're getting a high amount of chargebacks because the website marketing copy promises "14 days money back guaranteed" and even highlights that their competitors don't
295.
by mpeg 3y ago
Yep! In my case I switched to a processor that had higher fees but had less of an issue with chargebacks. Having said that, I think these days there are ways to limit your exposure, for example requiring 3D Secure for every card payment, as
296.
by mpeg 3y ago
Hence why I said the remainder. In my case they held about $10k and I got back most of it, they take back any chargeback amounts + a flat fee per chargeback
297.
by mpeg 3y ago
Ok so I know a bit about this as I had the same exact thing happen to me in the past. The reason this probably happened is that their business was getting a high % of chargebacks from paying customers, paypal is not a specialist payment pro
298.
by mpeg 3y ago
In the case of an NFT, I can prove that I own it, and I can prove the provenance of it all in one go. Sure, the artist could do something dishonest like sell the same piece twice, but I'm assuming good faith from the artist, or at leas
299.
by mpeg 3y ago
The way I look at it is that I can buy a print of a van Gogh for $10 but the real painting costs millions. The key difference there is that the real painting is real and the art house will some sort of provenance record that lists all previ
300.
by mpeg 3y ago
They don't at all! In fact, there have been successful NFT collections that handled payments via fiat for stable pricing and used a non-custodial OAuth wallet (eg a wallet that gets uniquely linked to your google/twitter/meta

More ›