It would be even easier to run it on a normal system. A good chance to read a trainable LM implemented in 700 lines of C.
I would like to see that one day. I want something that runs in Windows or Linux and does something I can edit and use.
You can run it. The model isn't very good. But you can edit it or give it new training text. And if you give it more learned parameters to work with it should get better.
But it is limited by the SD card. What if I used a gigabyte of training data?
More training data is usually more better, ish. I think it trains on your computer and then exports the result to the micro controller.
So is it worth the $8 for the ESP card? Do I telnet a shell to it over USB like a serial cable?
It would be even easier to run it on a normal system. A good chance to read a trainable LM implemented in 700 lines of C.
I would like to see that one day. I want something that runs in Windows or Linux and does something I can edit and use.
You can run it. The model isn't very good. But you can edit it or give it new training text. And if you give it more learned parameters to work with it should get better.
But it is limited by the SD card. What if I used a gigabyte of training data?
More training data is usually more better, ish. I think it trains on your computer and then exports the result to the micro controller.
So is it worth the $8 for the ESP card? Do I telnet a shell to it over USB like a serial cable?