shipwright

قبل سنتينdiscuss

Good thinking, and regarding Euro institutions there's lot of regional differences- for instance, brits have steps to take which are pretty set in stone with the Centres for Doctoral Training; basically it's the opposite of the process from the prior comment, it's the system that allocates someone for you. Denmark is free if you're native or from EU- Finland however is free for all (according to peers and the net) and would gauge they're both tad more flexible than UK in terms of requirements. Anyway, there's plenty options! Regarding the selection process, although they're a clear benefit perfect grades aren't a must and it's alright to be from outside of a top ranking school- but you'll have to bring something to the table to make up for it. Whether that is a solid paper, practical achievements along recommendations or whatever again depends on the place. I think there's just so many ways to show something that can be impressive without knowledge about the latest thing or dropping walls of equations for the sake of it haha. Best of luck!

shipwright

قبل سنتينdiscuss

That's not necessarily a deterrent :) it really depends where, also would you consider staying in the states for it? Because I know of some reputable places (west coast mostly) where pretty much if you are to the right person's liking and fit their requirements, your admission won't see objections- there's plenty of time to make a decision about it, because the applications (iirc) take place a bit before* the end of the year. Let me know if you need more details!

shipwright

قبل سنتينdiscuss

Sorry to hear, if both you and OP need help to keep those devices functional found a couple things which suggest they are quite easy to unbrick like https://support.ambiclimate.com/en/support/solutions/article...

shipwright

قبل سنتينdiscuss

Can wholeheartedly recommend https://www.feynmanlectures.caltech.edu/I_toc.html

Also, Physics Teaching for the 21st Century from UBC is a solid starter point regarding the applied aspects https://c21.phas.ubc.ca/article/heat-engines-in-the-real-wor...

shipwright

قبل سنتينdiscuss

I know someone who is on the second platform, they're labeled B2B professionals and can be filtered to your liking- the rate is between 70 and 160 $/hr which does fall within the initially mentioned budget it seems :)

shipwright

قبل سنتينdiscuss

Hah, so high tech- Turns out the said repos used to do the same, then one of them later added CodeFormer under the hood, conflated the strength with affinity and didn't bother renaming (calling codeformer --gfpgan_strength looks very unhinged to me but oh well haha). It's definitely doable though! What I had in mind as a GAN equivalent (used to apply in WSI context while back) is akin to Rayleigh EigenDirections in terms of latent space navigation https://www.ecva.net/papers/eccv_2022/papers_ECCV/html/7277_... There's some code out there, unsure if it's streamlined for this use case but can lend a hand if you decide to give it a shot :)

shipwright

قبل سنتينdiscuss

Sorry to hear, you did well though given no one has their guard up 24/7 and that's what they rely on in that case! There's a few solid guides on how to minimize the possibility of lateral movement, privilege escalation, forgery and so on without having to think through the inner workings of every attack vector under the sun (would say the most thorough is https://attack.mitre.org/versions/v14/mitigations). Could also provide some help beyond that this weekend (been a slow week) at no cost, e-mail's in the profile :) used to have all the bells and whistles a white hat could have like disclosed vulns, talks at conferences and whatnot years ago before going into ML if that matters haha.

shipwright

قبل سنتينdiscuss

Fortunately, that is indeed a thing :) several post-processing tools have some sort of gfpgan_strength (searching as-is should yield results on a few repos) param being passed, also it's been a while since diffusion-based restoration methods are equipped to diminish stochasticity enough to produce great results, worth the time as well! Also noteworthy: https://openmodeldb.info/?t=faces+restoration+texture-genera...

shipwright

قبل سنتينdiscuss

Great to hear! And if memory serves correct, long time no see (Belgium like 11 years ago? haha) :) the Greek Mozilla squad was tons of fun to hang out with, super interesting discussions that day- project I mentioned briefly back then got awarded by NASA one year later, wish what I did was half as impactful than what you guys got going on at Libre Space though! Hope all is well

shipwright

قبل سنتينdiscuss

You're likely eligible for Nvidia Inception which does offer a handsome amount of free credits for several providers, AWS included.

shipwright

قبل سنتينdiscuss

That's rad, lack of lock-in and sharing in a niche is a big green flag- Hope adding ultrasound to the list of modalities goes as planned, would gladly lend a hand at no cost if it's high on the priority list!

shipwright

قبل سنتينdiscuss

Such a hidden gem, gotta love the sketches and the layout overall is pacing the entire article so damn good!

shipwright

قبل سنتينdiscuss

We have a machine running Win7 that's just about that age at home, mostly for the sake of keeping old passion projects that rely on heap behavior being deterministic like some forensics plugins and a half baked try at application checkpointing for Windows- despite that meaning a wider attack surface for Win7 (like LFH buffer overflows or use-after-frees), there aren't that many major threats to be wary of especially if an eye is kept at ProcMon then and now, plus maybe a network packet analysis tool of choice with some rules on (anecdotally, said workstation still sees reckless unprotected use from relatives and it's just fine haha). XP on the other hand needs more precaution and it might be better off kept offline or replaced with a lightweight Linux distro like Lubuntu :)

shipwright

قبل سنتينdiscuss

It may be doable, a budget setup would be two half-duplex SDRs (synced with the same clock source between them), paired with diametrically opposed end-fire radiation pattern antennas in which one transceiver acts as Tx sends step-frequency continuous waveforms into the field and Rx collects the reflected signals, baring the snow the surroundings should be homogenous enough to pick up the wallet hopefully! Initially thought of a different approach than bouncing sfcw as well, but suggesting this because it has a properly documented success after a quick search: http://pe.org.pl/articles/2019/9/7.pdf

Best of luck :)

shipwright

قبل سنتينdiscuss

Getting to know some cuda is a solid step forward, I'd recommend familiarizing with Triton as a next move since you seem into that, besides fundamentals like trying out a scalable deployment workflow with a framework>training env>inference env>orchestration tooling of choice. Plenty of room to mix and match there (and can be done at no cost in many cases due to generous free trials/tiers) :). Simply finding something you love goes a long way, there's always going to be opportunities to prove your worth especially when it comes to infra: here's a little example, there's a lot of popular streamlined training tools which save checkpoints where tensors are nested oddly in subkeys- because of it deserialization gets iffy when converting to safetensors. That wouldn't be a hassle if it were a one off thing, but said models are really big and popular without even counting the sheer number of finetunes that also get deployed and when aware of it can just act accordingly but it's what stopped enough projects in their tracks for a while until pointed out haha. I'm sure that whatever you pick to do will be a good choice, best of luck :)

shipwright

قبل سنتينdiscuss

Cool idea! There are a few platforms which do tackle that kind of concept (I reckon SoundStorming is the most recent one that got some traction), though it's a really good moment to come in with a brand new take on it! The Web Audio API has matured a lot in the last few years, and the technical limitations regarding collaborative music are more a matter of execution right now- interfaces (and their dreaded ASIO drivers :)) also got better and it's easy to capitalize on that in so many ways while keeping latency low, but those without em or a quality mic get cast aside when it's now possible to post-process less aggressively if the conditions to do (most importantly placement) so are communicated. While ConvolverNode shenanigans in Web Audio won't come close to the granularity of popular reverb VSTs, there's plenty of musicians who would see much more value in creating with effects on the fly, especially those who didn't figure DAWs out yet to a good extent and I think paired with clever spatialization techniques for the tracks bearing in mind parameters related to effects and mixes (highlighting those timeline-style on top of spectrograms could make browsing tracks very interesting, along/alternatively filtering by vectorized features) are examples of what can easily popularize such a project given that people who use DAWs already share stems and mixes to collab asynchronously on Discord threads, it's a win-win to have the two demographics in one place readily available to create together.

Sorry about the long answer, most of it is just context in case someone else less accustomed with music finds this thread interesting (oh, also Audio Worklets are really cool but that would warrant another 3 paragraphs hahaha) :). Despite the amount of similar neat options available, hope you decide to make it anyway! It has potential and would recommend it to people for sure.

comments