00:58 polygon2: The conflict resolution which happened in 1996 was there were abortion typical ones picking a fight from a guy, who was called a gangster but in my life it is not confirmed status, cause in estonia we have a tradition that every abortion leftover who can not deal with sports and life issues thrashes a guy who can, so what happened was that guy was playing billiards with his wife and a set of porn comments annoyed him, which is
00:58 polygon2: what such incapable people do, all night it continued two assaults, and the so called gangster beat them all up in a row, but later he was gunned down for this, so that is what estonian abortion leftovers perform, they gun down powers, that is what their miserable life work is. For me it it did not succeed, so they used doctors to torn me, cause all thosde leftovers they were brought into this world by parents who are fucked
00:58 polygon2: up personalities too.
00:59 polygon2: but the guy was so strong before he was gunned down, in other episodes he just took the kick boxer from the neck and rolled back and forth on the table, he was picked a lot of fights before the killoff
01:00 polygon2: was not an athlete, typical country dude who claimingly did just some work in the garden, like wrestler palusalu, moving stones, but was not associated with sports.
01:04 polygon2: normally 99.9 percent of the time, it is another way around, especially in technical sports, real trained athletes can not be beaten this way
01:05 polygon2: no matter what is the event, but i have seen such exception yes
01:12 polygon2: it went so viral they claim something entire nonsense based to the finnish and europe and it has halted my life, along with the notorious treatments to use my cells around the world and putting extra touch to crack up my health while harvesting those by intention.
01:12 polygon2: brutal leftovers
01:14 polygon2: i am not just paranoid without a reason, i did not cause them being scum, very dangerous people you do not even understand why they are so dangerous, cause they are internally so weak mostly and unstable.
01:20 polygon2: there just is no other solution than trying to force or inject some interest into top end science to have passions, to have hobbies to have interests
01:20 polygon2: i do not know how far i have come with this progress, if that substantially brought results
01:20 polygon2: but other solutions if the results did not work i do not have
01:29 polygon2: i am afraid to touch those subjects, some nuclear scientists came to talk with me, or guitarrists i spotted that maybe there is something they do not realize, talking about triton, so i generalized that yes nucleus can be enriched, but i understood that this guy was either upset or picking a fight from me claiming that he knows more, and i would insult him talking about that other path
01:29 polygon2: so i added for safety this discussion to the IRC
01:30 polygon2: and it does not matter, if it is somewhat crooked or needs to improve on theory
01:30 polygon2: it was impossible to do it before without conflicts
01:30 polygon2: my intention was not to insult at all, he himself took everything personal
01:41 polygon2: but i am afraid this worked out bad too, from the same day another crew near by ended up the next day intoxicated likely with stabbing show again, such unstable releases you know
01:44 polygon2: yeah the radeon i have gone over along with all gpus and everything is fine to me, but ideally i wanted to port code to xilinx now amd or altera intel too, similar flow to divine
01:44 polygon2: so my career in reading is indeed rich, i know how those things work nowdays
01:46 polygon2: i'd just like to confirm if my vision was great, i would rank xilinx and altera fpga's more performant in this paradigm than gpus
01:46 polygon2: and overall circuitwise they should be also best you can ever get , implementable on any underlying hw
01:52 polygon2: i am just saying that my full dedication will not be given there, it can't be given cause i am not relatedly employed, but little wise test i could in theory perform to just know if i saw it correctly
01:54 polygon2: rshell actually has the login password option automatic too, i just did not use it the first time, once something flickered , display went off on bootup two machines interconnected and battery sign changed on icon, i am struggling with having momentum and time a bit, i went over the code, and it looks fine to me
01:56 polygon2: bunch of pretty much best computers actually i have at home, they are very good cause the FET curve puts them to leading processes
01:56 polygon2: on steady states
01:58 polygon2: you can get that curve from ericsson science labs and such other companies had it too
02:02 polygon2: i mean memfault the pure uart with micropython and swedish embedded labs qemu solitions like SLIP should both work in little bit different fashion, but the last is even more reliable, but slower a bit
02:03 polygon2: i just did not feel good to compile more, i am not so big time knowing the renode stack, i thing @gpio X in conf should do that with or without some csharp hacks
02:03 polygon2: but francois baldeissar had another stm thread where it was written, i did not compile this, he already mapped the gpios
02:06 polygon2: physical uart on renode should be from there possible too, just like ble but renode has to have some conf for this which i do not know how it goes
02:06 polygon2: but on qemu i know, more documentation
02:09 polygon2: there is one sample where twi which likely stands for something like wireless interfaces is passed through that way, only one sample
02:09 polygon2: so it has to be worked out from there its nano board was it n33 or something arduino based
02:11 polygon2: but my errors likely might had come cause i did not link the softdevice in to the final executable too
02:11 polygon2: so i must rework it, but i delay that to later
02:14 polygon2: I got predominantly frustrated cause i got into a conflict on osdev and they additionally started to intrude into my systems and surroundings where i exposed some of the vunlerabilities, many of those methods might harm human health, i am not sure how much i can handle this, there are vulnerabilities that backends need to built for
02:15 polygon2: i got several severe bursts beamed at me, and i was nearly gone
02:15 polygon2: they were too strong to handle for most, but i kept running or walking around like a lunatic and survived
02:18 polygon2: the vulnerability is very simple, it is a hw one, you get into point to point communication and to the network interfaces
02:18 polygon2: and you have such a packet that writes with memory jam into interrupt handler
02:20 polygon2: the real defense there is sort of like military backend which is a complex thing on wireless cause that would require a military convoy
02:22 polygon2: convoy taken like from war movies, so you have this scramblers for bus and wireless , with airforce guidance and they take energy which way they agreed to defend their lead , you know stuff like that
02:23 polygon2: it's over my shadow, and i can not full fill this goal well
02:25 polygon2: I have no danger to get jailed, cause these people who make those wireless frauds, they do not understand tech well
02:25 polygon2: they have eyes on them too
02:25 polygon2: not possible to frame me in it
02:26 polygon2: every GSM mast as said has this logging capability
02:26 polygon2: as/has
02:26 polygon2: they will understand even if my computer was bounced to make the duplicate transfer from, who actually performed it
02:27 polygon2: those are measuring instruments that do that
02:30 polygon2: this is heisenberg paradox, but the real definition also includes it does not account measuring instruments
02:41 polygon2: that is why those syndicate members along with this horrible slut laura keskinen , were not athletes neither real scientists
02:41 polygon2: they wanted to frame me, but they lack understanding how tech works, those on the air do not interact, but they interact with eachother on the time samppling based sensor
02:42 polygon2: which filters out other signals
02:42 polygon2: and it can maintain the log
02:53 polygon2: come on , you are not so stupid the invitations alike where you pay and get into troubles are always built to find a tank driver whom to kill, they filthy viking and porn criminals , absolutely deluded people, my kill off just did not work out well
02:53 polygon2: and by such gangster wannabes who are entirely nuts from the head, it is very expected thing, they do not know how to perform anything and how life works
02:56 polygon2: so it was some effort and try but i got some injuries but it just never fully worked out, they have to possibly enter the jail their own and be responsible from the shit they do in life their own
02:59 polygon2: frustrating of course to be a devil but i am not so stupid for letting leftovers to succeed
03:03 polygon2: the biggest question remains , when you are desperate of killing someone with estonian scammers and finnish scammer gangsta crews
03:03 polygon2: then hat you want to achieve your own
03:03 polygon2: ?
03:04 polygon2: hat/what
03:04 polygon2: it's just philosophical
03:04 polygon2: cause i lit up some more weed , cause i can not handle those vibes without anger bursts
03:05 polygon2: is that an utter lack of philosophy , understandings , and even aiming something
03:06 polygon2: aiming at perhaps but, the person may not realize that one is totally off
03:06 polygon2: cause killing is not an achievement
03:06 polygon2: or goal or dream
03:09 polygon2: if you would not fight with cancerous thoughts and be prone to have deluded self-satisfying thoughts and self-favoring without any reason
03:09 polygon2: you know how smart and successful you would be?
03:13 polygon2: i look at possible amazing talent and you wanna poison yourself cause of me and my internals understandings that got shaped into place?
03:14 polygon2: that is very stupid in my opinion, cause my understandings and past is different, i allow people to be different in parentheses don't be too criminal cunt
03:16 polygon2: cause my understandings were shaped by restraining orders and blocking strategy and even Kill attempts at me.
03:16 polygon2: I am very glad that you do not have to be restrained by such options and go to heights
03:19 polygon2: btw boys has anyone noticed what polyurethane stem scaffolds do?
03:19 polygon2: or nanotransfection , or 3d bioprintings at upnano?
03:20 polygon2: I can get well at anytime if i want, it's remote 1s procedure for nano transfection
03:20 polygon2: this will heal my nerve , muscles as tendon
03:21 polygon2: me i was so much powerful and talented and better than you, when i get better it all happens again
03:21 polygon2: me, i do not have any troubles with that, and let you go to heights too instead
03:21 polygon2: as additional benefit
03:22 polygon2: i like to understand people a lot
03:22 polygon2: and the count with them
03:23 polygon2: cause you grow so much stronger if you do not have any restrictions placed to your mindset
03:24 polygon2: cause i had only restrictions, you grow much much stronger than me, cause you would not have any restrictions
03:24 polygon2: i would allow you to do that, cause it's fun to me
03:27 polygon2: even if i had only blocks and was better than you, you grow lot stronger without keeping my state
03:28 polygon2: you need me for nothing, but some who truely love me, elect me back appareantly
03:29 polygon2: otherwise i am already done my show in this planets history is done, by genes will no long appear in the future
03:29 polygon2: cause you do not need so restraining and poisonous state
03:30 polygon2: i graduate my history and my line
03:30 polygon2: cause none wants my state to the future
03:30 polygon2: my lineup ends with me
03:30 polygon2: first time in history i am very calm and peaceful
03:31 polygon2: cause the transaction was already done, i have two choices:
03:31 polygon2: 1. to end my career in the universe
03:31 polygon2: 2. go to another planet to extend it
03:32 polygon2: or graduate my existence as genes in this universe, genes and cells are not going forward as a new guy like me
03:32 polygon2: in other words, my genes will die out, i am done on the planet
03:34 polygon2: there is only two ways for me to continue in universe
03:35 polygon2: when my genes go ahead or when i switch the planet for staying
03:35 polygon2: first two are decided against me, then the last one is to die out in universe
03:35 polygon2: but this basis into what and what really would happen there, i just do not at all know
03:38 polygon2: and why that block is so severe that i die out as result, is that no one needs restraining state
03:39 polygon2: that immediately adds to the fact, that my departure is the most severe
03:40 polygon2: me i am going to execute my last 40 years as when i appear the last, i no longer worry about anything
03:41 polygon2: cause i am in the deepest water in a bag already
03:41 polygon2: i only live my last 40years through, and then i am done
17:35 dcunit3d: hey, can i get some advice on building a Dockerfile to support both pytorch and tensorflow for ROCm?
17:37 dcunit3d: i'm on archlinux, which I've been regretting for some time. The Garuda distro works well for gaming and I do use AUR a lot. I also have a Guix install, but it's nigh impossible to run OpenCL there or on Nix.
17:38 dcunit3d: I've gotten the tensorflow image working, confirming that OpenCL works. Then I started looking into the python signatory library for path signatures, but it requires pytorch and benefits quite a bit from OpenCL.
17:39 dcunit3d: so then I need that. I could probably just run my torch notebooks in that image and the TF notebooks elsewhere, but I would rather build one image.
17:39 dcunit3d: and particularly, I would like to learn about building the python dependencies anyways.
17:40 dcunit3d: however, I'm having trouble finding the Dockerfiles that ROCmSoftwarePlatform projects actually build to produce their images on DockerHub.
17:40 dcunit3d: and then I'm likely going to have some issues synthesizing whatever Dockerfiles for TF/Python i reference anyways, since my Docker Fu isn't so great.
17:41 dcunit3d: i've got this repository, which lets me clone the 100 most recently updated ROCm repo's in one go. i basically just ripgrep on that.
17:42 dcunit3d: i also don't know much about python. so i needed to use python 3.10 instead of 3.11 to download a pytorch wheel, since the latter was briefly offline. I'm worried that issues like these are just going to be common anyways and they may interfere with Docker images I build that don't build the python wheels from source.
17:44 dcunit3d: and I really hate nvidia. i had a macbook pro from 2013 and they kept making the 400 series drivers unavailable for my hardware which borked my linux installs when i was more noobish. I may buy some of their hardware in the future, but it will never be because i need to. it will be because i want to.
17:46 dcunit3d: the gaming experience on Linux with AMD has been great, but i bought an RDNA chip, since I didn't know about that. I've gotten past that I think, but the ecosystem on linux is borked. it's not even the AMD ROCm software that causes the confusion. Instead, the issues are related to higher-level projects requiring GPU, usually python. it's every corporate vender or open source project that just treats AMD compute like a second class
17:46 dcunit3d: citizen that causes the problems.
17:47 dcunit3d: it's extremely frustrating. i started learning ML in 2012 and I've never been able to actually use it. and on one hand, i'm learning about path signatures and gaussian processes ... but I go to use stuff or pick up an old project and then ... it's broken or can't be updated.
17:48 dcunit3d: so the solution seems to be to just build the python deps or use C/C++ and Rust ... but I don't have experience with that.
17:50 dcunit3d: anyways, I have a lot of questions, but the ROCm ecosystem is a bit daunting. if i could connect with someone who could provide general advice or describe workflows, that would help alot.
18:04 Venemo: dcunit3d: what exactly are you trying to do? and what are you using ROCm for?
18:04 Venemo: for running OpenCL apps, I think rusticl will be in pretty great shape soon enough
18:22 dcunit3d: thanks Venemo i need the dev libraries in the AUR opencl-amd-dev package for python libraries like tfga (geometric algebra) and signatory (path signatures)
18:22 Venemo: so, if you wanna use opencl I recommend giving a try to rusticl
18:23 Venemo: if you insist on using ROCm, I don't know anything about that unfortunately, but maybe someone else on the channel can help you
18:23 dcunit3d: at a high-level, i was working on the Kaggle competition for ASL recognition. that involved mediapipe, parquet, tflite and a few other things. i'm seriously considering just blowing away archlinux and trying out silverblue/ublue.
18:25 dcunit3d: yeh but there's just not many people in there. i'm looking at a few discord servers to see if i can some people. i get the feeling that most people in ML and data science just have enough money to buy a nvidia card anyways.
18:26 dcunit3d: Venemo: have you used rusticl before or have you gotten feedback on it? i like lower-level, compiled code anyways. i just don't know anything about rust. i've actually been looking at zig because i'm also interested in embedded programming.
18:28 kchibisov: dcunit3d: from what I remember torch provides builds for rocm on their site via pip.
18:28 kchibisov: At least I've used it without much issues on arch like half a year ago.
18:28 dcunit3d: my impressions on rust make it seem like it has some of the flexibility of scripting languages while being compiled. it looks like it's a language that's bindable to C/C++, but i'm not sure whether how much FFI overhead is necessary. i've heard a lot of good things about compilers, but i just haven't faced a problem where rust beats out what i have experience with.
18:29 dcunit3d: kchibisov: the maintainer on AUR just released a build of opencl-amd-dev that pulls in ROCm 5.5 and the pytorch wheels for that aren't available on pip yet
18:29 Venemo: dcunit3d: I only used rusticl a few months ago when Karol helped me get it running on zink + radv, for fun. I do not use OpenCL regularly myself
18:29 kchibisov: But you can downgrade?
18:29 dcunit3d: also, i want to ensure that i've got cupy pulled in
18:30 kchibisov: Just use previous version of rocm stuff if it doesn't work with torch yet.
18:30 Venemo: rusticl is an opencl implementation just like any other, you don't need to write your app in rust to use it.
18:30 dcunit3d: not ... really. there's probably some way to reference a git commit for `yay -Syu`, but i'm not sure. i could pull it down and `makepkg` myself, but the problem will just happen again.
18:31 dcunit3d: at some point, something dependency won't have the wheels on pip for ROCm 5.x.x.
18:32 kchibisov: There's even a package in aur for python-pytorch-rocm (updated like hour ago) ¯\_(ツ)_/¯
18:32 dcunit3d: Venemo: i see. i'll look into it.
18:33 kchibisov: Ah, it's not even aur, it's mainline.
18:33 dcunit3d: yeh. i saw that, but i haven't looked at the pkgsource yet. it might work, but it requires the system python
18:34 dcunit3d: yeh, so there are options, but i'd rather just use silverblue or ublue and have more package stability if i'm going to keep doing this. having a second GPU and just using a VM would also work i guess, but then i'm CPU and IO bound
18:35 kchibisov: I just know that with DL you basically have HIP/CUDA, never seen anyone using OpenCL.
18:42 dcunit3d: kchibisov: what's DL?
18:42 kchibisov: Deep Learning.
18:43 dcunit3d: I think some of the libraries built on TF expect to use XLA and either require custom builds to avoid that or they require the ROCm dev libraries for reflection and LLVM
18:46 dcunit3d: so there are the rocm-opencl-sdk and rocm-hip-sdk which are mainline, but they don't have the devleopment libraries. i could be wrong about whether the way that XLA/triton use LLVM requires the dev libraries. I just don't really know, so i'd rather have it all, then it doesn't matter (i think?)