Ya, es ist möglich fur lokale KI betrieben auf einen PC. Aber ich habe kein experience KI-Modelle trainien. Ich vernemm dasz nimt viel mehr Rechenleistung.

Hallo miteinander,

haben wir hier jemanden, der sich mit KI auskennt oder vielleicht sogar schon eigene KI-Modelle trainiert hat?

Ich habe mir die Frage gestellt, ob man nicht mithilfe der LinuxCNC-Dokumentation, vorhandener Konfigurationen und Beispiele eine eigene, lokal laufende KI aufbauen könnte.

Die Idee wäre, eine KI zu haben, die speziell auf LinuxCNC ausgerichtet ist und beispielsweise bei Fragen, Problemen, der Konfiguration sowie bei der Erstellung und Fehlersuche in HAL- und INI-Dateien unterstützt.

Interessant wäre auch, wenn solche Modelle durch Erfahrungen und Lösungen der Nutzer immer weiter verbessert werden könnten. Die verbesserten Modelle könnte man anschließend untereinander austauschen, sodass die gesamte Gemeinschaft davon profitiert.

Das Ganze sollte möglichst vollständig lokal und ohne Internetverbindung funktionieren.

Leider kenne ich mich mit dem Trainieren eigener KI-Modelle bisher nicht wirklich aus und habe auch noch kein eigenes Modell eingerichtet.

Deshalb einfach mal die Frage in die Runde:

Wäre so etwas technisch und mit vertretbarem Aufwand umsetzbar? Hat vielleicht schon jemand Erfahrungen damit gesammelt?

Gruß
Mücke

I am installing a M.2 SSD. Can I copy my current boot image from the SD card to my SSD?
I am running this on Raspberry PI5

thanks for your help 

Yes, Stable does not necessarily mean bug free. But it also does not mean that there are not bug fixes when bugs are found. This isn't Debian. In the case of Linuxcnc, it only means that it is no longer getting new features and major changes, and it will be much more bug free than the development branch. The Master development branch in Linuxcnc is more analogous to Sid in Debian. Linuxcnc does not have a Testing stabilizing branch like Debian.

It didn't on the last machine I set up about 6 months ago. (But that does not mean that it didn't change since then.) It would be nice if it did do that.

Tuxcnc sums sit up "working model", "zero practical value". If the goal is making things then a drill press hand tools and a jig saw, might be the better options. Do not make the mistake thinking a CNC machine is "the tool" for making stuff.

Having seen a TTC-40h in person I would stay away from this "brand"

Considering what your goals are is the first priority. "I do not need it to be fast" "I only want to do hobby work" that is what I hear all the time. Most probably the "more hobby" it is, the more babysitting a machine needs.
Hobby machines lack reliability and most of the time for one off parts I have to supervise the process full time. G-code from any cam system needs to be veryfied and or observed. (never trust the CAM).

Thinking about just spending more than 100$ on a few small endmills and one M6 tap last weekend. It becomes almost ridiculous how many discussion I see about making a controller electronics costs 50$ less than another option.
If you use a machine a lot, other costs add up quickly. 
p.s. the machine price is the least concern.

I thought the spindle selected was remembered until another spindle was chosen so you don't need to set the spindle with every M3 command

I understood the "Show Your Stuff" section as a place where members could share their own projects and technical work, so I thought it would be appropriate to share mine there.

I'm still exploring LinuxCNC and learning from the community. I may not have much to contribute to LinuxCNC itself at this stage, but I wanted to share what I've been working on and exchange ideas with others.

There was no intention to misuse the forum or disregard its purpose. If my post is not appropriate for this section, I understand.

"Stable" does not mean that the program is really stable and bugs-free.
It just means that it's old and there's something missing.
For example, Debian is released every three years and Ubuntu every six months. People prefer Ubuntu because they probably don't know it's unstable and should use Debian...
Coming back to LinuxCNC, I use version 2.10 because I don't want to wait a few years for functionality that I can have today.
And you must always be careful, even if something is called stable.

Muchisimas Gracias, de todas maneras, comparto lo que tengo Placa Intel D865DKD, procesador Intel Celeron 2.53Hz )1 cpu lógico), 998MB RAM, kernel 3.4.9-rtai-686.pae. Test de latencia: Max Jitter servo 1ms us a 405us, con wifi apagado y BIOS con APM desactivado. ¿Con período de base de 100000 ( 100 us), en Stepconf, se queja de latencia?. Que hardware maneja usted?, Aún me gustaría saber lo de JT, quien me confirmó, que usaba ese Ubuntu hacías años, sin problemas. Muchas Gracias por haberme respondido y su paciencia. Me encantaría que ésta viejita de cpu, me anduviera bien....

I think I can configure it myself, but I decided to start this topic because, firstly, maybe someone has a ready-made solution, and secondly, it may be useful to someone else.

Well, I bought a Chinese rotary axis. It is of good quality and works properly, but the designer did not provide homing limit switches.
I decided to add them in such a way as to interfere with the original structure as little as possible.
I came up with an idea that if you cleverly place miniature inductive sensors, they can work with small pieces of metal attached to moving elements, while the sensors themselves will be attached to a stationary base.
I fiddled with it a bit and it worked.
The video shows a complete makeshift, the elements are attached with double-sided tape, and the Chinese TL-W3MC1 sensors work with plates from a canned food can. Of course, ultimately it needs to be done better.

As you can see, the B axis (tilt) is referenced first, then it is set at the appropriate angle relative to the sensors (in this case +90 deg from machine zero), and only then the A axis (rotary table) is referenced. No special combinations are needed here, just set the HOME, HOME_OFFSET and HOME_SEQUENCE parameters correctly in the ini file.

As I wrote above and as you can see in the video, it works properly, but there is a problem that needs to be solved.
Well, everything is fine when we base all axes at once.
If we try to reference only the A axis, it will probably fail, because the sensor and the mating metal will be in such positions that they will never come close to each other. Therefore, when homing only the A axis, should be first check whether the B axis is referenced and is in the appropriate position, and then decide whether to perform homing or display an appropriate message instead.
It seems to me that LinuxCNC has the appropriate pins and HAL components to do this, but I haven't thought about the details yet, because I only started this project today and, objectively speaking, I've already done a lot.

If someone would like to help me, that would be nice.

Yes indeed it was a wiring issue. As I was debugging today I discovered the power connector to the micro stepper was loose. When will I ever learn?!