Today I am exploring the Korg Volca FM driven by a M5Stack Core2 computer and a M5Stack Midi Unit with DIN Connector. I also set the MIDI notes using a M5Stack 8-encoder and the master volume with a potentiometer. The Korg Volca FM is a re-creation of the most famous vintage synth: the Yamaha DX-7.
This is a project for STEM students and they can also learn a little about music, maths and synthesizers.
See the video on my Youtube channel: dsrc12 ( https://lnkd.in/gXhTMeM5 )
Exploring AI for STEM Education.
My M5Stack off-line Large Language Model (LLM) module arrived yesterday.
It is another example of how China is excelling in AI. Everyone has heard about DeepSeeker. It got the cost down and my LLM only cost $US49.95 while the Core3 computer is another $US59.
I have coupled the LLM module with my M5Stack Core3 computer and did the programming with UIFlow2 - a simple blocky language. I added an applications layer to the LLM.
The model embedded in the modules's firmware is based on qwen 0.5B (Billion), so you can hardly call this a large language.
My whole model is contained within the eMMC memory of the module's computer which is limited to 32G while the RAM is 4G. The computer that runs the LLM and the host are both based on Linux.
Compare it to the trillions for Chat-GPT4 which has an estimated 170T (Trillion) tokens.
Despite that I had my tiny LLM (with the prompt "Hi Jimmy!") answer several questions and requests including:
"I want to stay heathly, please help me"
"How do you make noodles?"
"I want to apply for a job as a teacher, please help"
In addition but not shown on the video I also asked Jimmy to provide the Arduino Code to flash an LED every second. Jimmy gave me a python answer and not C++ for the Arduino IDE.
Despite this, my tiny LLM is interesting and my goal is to develop a hybrid AI (artificial intelligence) model to control a robot with both a tiny LLM and an AI Camera. The camera will pass on objects recognised in images and these will be used to formulate a robot navigation plan. Just what NASA needs for its Mars rovers!
I am also interested in other hybrid models on tiny applications that can route requests that are beyond their context. They would route the requests onto OpenAI API server. Perhaps for a simple system the tinny LLM could answer 80% of either my or the robot's requests.
Other applications include developing models for people with language disabilities. Instead of Jimmy answering he or his assistant will ask the questions to encourage a response.
I have already loaded a qwen 1.5B model and can not wait to test this. That is a x3 improvement on what am have already tried.
Ref
A STEM project for Christmas.
In this project I map MIDI musical notes and their Boomwhacker colours to a M5Stack CoreS3 (ESP32 based) computer's screen.
Boomwackers are great to teach students music and I used them in STEM clubs and also to make music fun for people with disabilities. You can also use this hardware & software configuration to learn the piano keyboard as you get feedback as to what keys you are playing.
There is a wide range of material to support Boomwhackers including musical instruments like xylophones, bells and tubes and there are also plenty of books with the Boomwhacker colour codes for songs.
I go a step further and show both the notes from the C Major scale and its associated Boomwhacker colour on the computer and external RG LED strip.
So for Christmas connect your tree lights to Christmas music and let STEM students learn coding and music.
This makes a great STEM project particularly if the students share their results with other students with a disability.
This project is on Youtube: https://youtu.be/gPvJRJ3iCL8
see more on my Youtube channel: dsrc12
References:
Youtube:
M5Stack MIDI ( https://shop.m5stack.com/products/midi-unit-with-din-connector-sam2695 )
M5Stack CoreS3 ( https://shop.m5stack.com/products/m5stack-cores3-esp32s3-lotdevelopment-kit )
Boomwhackers ( https://boomwhackers.com/ )
Exploring AI for STEM Education.
My M5Stack off-line Large Language Model (LLM) module arrived yesterday.
It is another example of how China is excelling in AI. Everyone has heard about DeepSeeker. It got the cost down and my LLM only cost $US49.95 while the Core3 computer is another $US59.
I have coupled the LLM module with my M5Stack Core3 computer and did the programming with UIFlow2 - a simple blocky language. I added an applications layer to the LLM.
The model embedded in the modules's firmware is based on qwen 0.5B (Billion), so you can hardly call this a large language.
My whole model is contained within the eMMC memory of the module's computer which is limited to 32G while the RAM is 4G. The computer that runs the LLM and the host are both based on Linux.
Compare it to the trillions for Chat-GPT4 which has an estimated 170T (Trillion) tokens.
Despite that I had my tiny LLM (with the prompt "Hi Jimmy!") answer several questions and requests including:
"I want to stay heathly, please help me"
"How do you make noodles?"
"I want to apply for a job as a teacher, please help"
In addition but not shown on the video I also asked Jimmy to provide the Arduino Code to flash an LED every second. Jimmy gave me a python answer and not C++ for the Arduino IDE.
Despite this, my tiny LLM is interesting and my goal is to develop a hybrid AI (artificial intelligence) model to control a robot with both a tiny LLM and an AI Camera. The camera will pass on objects recognised in images and these will be used to formulate a robot navigation plan. Just what NASA needs for its Mars rovers!
I am also interested in other hybrid models on tiny applications that can route requests that are beyond their context. They would route the requests onto OpenAI API server. Perhaps for a simple system the tinny LLM could answer 80% of either my or the robot's requests.
Other applications include developing models for people with language disabilities. Instead of Jimmy answering he or his assistant will ask the questions to encourage a response.
I have already loaded a qwen 1.5B model and can not wait to test this. That is a x3 improvement on what am have already tried.
Ref
A STEM project for Christmas.
In this project I map MIDI musical notes and their Boomwhacker colours to a M5Stack CoreS3 (ESP32 based) computer's screen.
Boomwackers are great to teach students music and I used them in STEM clubs and also to make music fun for people with disabilities. You can also use this hardware & software configuration to learn the piano keyboard as you get feedback as to what keys you are playing.
There is a wide range of material to support Boomwhackers including musical instruments like xylophones, bells and tubes and there are also plenty of books with the Boomwhacker colour codes for songs.
I go a step further and show both the notes from the C Major scale and its associated Boomwhacker colour on the computer and external RG LED strip.
So for Christmas connect your tree lights to Christmas music and let STEM students learn coding and music.
This makes a great STEM project particularly if the students share their results with other students with a disability.
This project is on Youtube: https://youtu.be/gPvJRJ3iCL8
see more on my Youtube channel: dsrc12
References:
Youtube:
M5Stack MIDI ( https://shop.m5stack.com/products/midi-unit-with-din-connector-sam2695 )
M5Stack CoreS3 ( https://shop.m5stack.com/products/m5stack-cores3-esp32s3-lotdevelopment-kit )
Boomwhackers ( https://boomwhackers.com/ )
Today I am exploring the Korg Volca FM driven by a M5Stack Core2 computer and a M5Stack Midi Unit with DIN Connector. I also set the MIDI notes using a M5Stack 8-encoder and the master volume with a potentiometer. The Korg Volca FM is a re-creation of the most famous vintage synth: the Yamaha DX-7.
This is a project for STEM students and they can also learn a little about music, maths and synthesizers.
See the video on my Youtube channel: dsrc12 ( https://lnkd.in/gXhTMeM5 )
@j_doe I need help with this one. Cannot get any UART program to work on CoreS3 with UIFlow2. I am using CoreS3 as a co-processor to a Raspberry Pi using serial coms but my initial unit testing is being done on a Windows PC working with Putty. Here is the program:
What I want is an equivalent program to this which uses an original Core computer. It is based on UIFlow 1 (V12.1.9)
I am using the latest firmware & on-line UIFlow IDEs (Alpha-28)
On the Core program I had to decode before I could access text from the UART - might need a similar solution for UIFlow2?
On the UIFlow1 program I accessed the Core UART using pins TX:1 & RX:3. I am guessing that the equivalent pins for CoreS3 are TX:43 & RX:44?
I tried to verify that these unit worked using M5Stack demo software at:https://docs.m5stack.com/en/unit/poesp32.
I tried examples with Core, M5Stack Plus and Atom for both TCP Client, HTTP Client and MQTT Client. None of these worked.
All of the examples resulted in the M5tack core device hanging with a serial debug message
"wait device connect
device connected"
Taking one example using the ESP32 Ethernet Unit with Core + Base to get access to the ports. I found a potential error in the published code. The published code uses the configuration for line 23:
"eth.Init(&Serial2, 9600, 32, 26);"
If the 32 & 36 parameters refer to the Core ports then this is in error. For Port B this should be "eth.Init(&Serial2, 9600, 36, 26);". Alternatively if Port C is used then the replacement code is: Line 23 "eth.Init(&Serial2, 9600, 16, 17);" Do the parameters in "eth.Init(&Serial2, 9600, 36, 26);" refer to the host Core or the embedded ESP32 in the ESP32 Ethernet Unit?
With the new code this results in progressing execution beyond the "device connect" and it now hangs waiting for an ethernet connecitons - debug message:
"wait device connect
device connected
wait ethernet connect"
I could not get it to progress beyond this stage.
My hardware configuration was that I connected the ESP32 Ethernet Unit to an ethernet cable which was connected at the other end to a Tp-link PoE Injector TL-POE 15OS with a 12V DC supply. The input connection to the Tp-link PoE Injector came from a Ethernet Switch connected to my home network. There was a Grove cable connecting the ESP32 Ethernet Unit to the M5Stack Core on Port C.
Is there any revised and update to date code for these examples?
Do anyone else get the example code to work?
I also could not get the UIFlow example to work.
I am having trouble getting recent modules like the Red 7-Segment Digit Clock Unit and the CO2 Unit with Temperature and humidity Sensor (SCD40) to work. None of the demo work - either in Arduino IDE or the on-line UIFlow (1.10.6)
I am using the latest burner (2.5.1) with firm version 1.9.8 on Core, Fire and M5StickC Plus. Tucked away in the on-line UIFlow for the Co2 Unit is a notice that these demos only work in firmware 1.10.4
The problem is that it is impossible to update to this firmware version as the M5Burner can only support up to 1.9.8
I see that there is a beta version of the M5Burner (V3) but the M5Stack server fails to download this.
I am assuming that M5Burner(V3 -Beta) supports access to getting firmware 10.10.4 and I am assuming that this is required to get the latest units to work. There is no announcement on this anywhere and it is total guess work!
M5Stack needs to improve its quality of support
Please sort this out otherwise none of the latest units work!
I have both the GPS Unit and the GPS module. Both give incorrect readings using UIFlow. Why is this the case?
@zontex I am in Australia. I also had problems with downloads yesterday - extremely slow download speeds and often downloading just hung. Occasionally core firmware downloads would get accepted by M5Burner after 10 attempts but drivers for the latest M5Stack cores could not be updated from the Arduino IDE using the Board Manager with a reference to 'M5Stack'. At the end of attempting downloads if it got that far it produced a CRC error and the Arduino IDE abandoned it.
Today everything changed - downloads were fast, only one CRC error but a second download was successful. On a second Windows PC the first download was successful. I now have the latest drivers on my Arduino IDE for all M5Stack Cores and I have the latest firmware versions. I could develop test software for Fire, Core 2 and Core Ink using UIFlow IDE on Windows and downloaded onto all of these platforms but I was only able to execute code - Fire. While Core Ink accepted my firmware it would not execute it. Both Core Ink and Core 2 are hard to control compared to Fire and I could only get my firmware to download and execute on Fire.
I suggest that M5Stack do more alpha testing with users from a few countries before officially announcing a new version. This is a regular issue with M5Stack and happens after each new version is released. While these failures were occurring I could not do any software development on M5Stack from Arduino IDE to UIFlow.
@m5stack I had problems downloading firmware updates for cores and drivers for the Arduino IDE over the last 5 days but today it all went well. All downloads worked and my test code on Fire also worked. I was able to download the latest version for Core Ink: 1.6.6.1 and it was able to connect and accept firmware from my Windows PC UIFlow IDE on-line (V1.6.6) but it would not execute any simple programs on Core Ink. It had problems with Core 2 and ended up abandoned it and completed my latest projects with Fire - it is a good stable platform and works with the latest firmware (1.6.6).
@dsrc12: I have now got this to work and it works best if I discount the USB link to Fire. The code I had difficult understanding is based on a C++ Ternary Operator in C. Returns either ECG_Data_LENGTH if data equals 0xAA or ECG_IDLE if false. At least this is a start. Still need to decode the meaning of the parameters.