Hi there, I am trying to install Open AI on my PiDog. I am up to Step 3 on (AI Interaction Using GPT-4O — SunFounder PiDog Kit 1.0 documentation) which includes running the command ‘pip3 install openai-whisper’
I can get every other package to install except for Whisper. It appears to be a dependency issue but I’ve not been able to solve it.
Any advice is appreciated!
Maybe by removing triton and torch packages. Then install whisper again and see if it automatically gets the satisfied torch and triton packages?
I’m not an expert by all means. Just started out this summer with all these things and as such a noob but this is where I would start at least.
Is this a zero 2W? If it is I think you’ll need to build from source. Or change the recognizer
Thanks Lucas! I am not sure how to remove packages but I will try to find out 
Hi there! Thanks for the reply - I’m using a Pi 4
Ah sorry. Hoped that would help. Over to Sunfounder, I guess
We went to test and verify, and did not encounter the problem you said Oh.
Hello!
I’m using a Raspberry Pi 4, and using a 64bit OS
I have installed the required codebase as per the instructions. The only step I have not done is under step 1:
Let me know if you think I should do this.
I have also followed the AI Interaction using GPT-40 instructions. Under Step 1 (install required packages and dependencies), every other step and command works, with the exception of the ‘pip3 install openai-whisper’ one.
I too am having the same Whisper installation problem. Apparently the Whisper package is not needed to run gpt_dog in keyboard mode. Keyboard mode has been working great. Amazing that this is even possible! So cool!! But I cannot get the listen mode to work, apparently without succesfull install of OpenAI Whisper. Any help you can provide will be appreciated!
It should be an update to upgrade the system kernel version that causes the installation to report an error.
We reproduced the problem.
If you install the Raspberry Pi system without executing the sudo apt upgrade command, then the system kernel is version 6.6.51, and this version can be installed successfully when you install openai-whisper.
If you execute the sudo apt upgrade command, update the kernel to 6.6.62, and then install openai-whisper, you will get an installation error.
We will fix this issue later.
For now, we recommend that you use the 6.6.51 kernel, do not upgrade the kernel, and go ahead and install all the library code, it will work fine.
Kernel version 6.6.51 is the initial version of the system.
Thanks so much for the suggestion. I rebuilt a new SD card with the 6.6.51 system kernel and was successful in installing all the packages including Whisper. I was able to run an audio listening test and receive audio response a couple of times from ChatGPT. Unfortunately the audio output has stopped. Speach recognition still works but no sound comes out of the pidog nor are there any error messages. I tried rerunning the i2samp.sh configuration and test multiple times but without success. I saw in another post that there might be a problem with the speaker on the PiHat. I took the screws from the board and lifted it enough to take a picture of the speaker and it looked OK.
I think it must be a problem with the hardware on the Pidog Hat since I reinstalled the previous SD card that had good sound output and it does not work anymore either. Is there a warranty on the sound card hardware?
Please contact our after sales team at service@sunfounder.com
They will send you a replacement speaker.
Thanks so much for the quick response. Do you have any instructions for replacing the speaker on the Pihat board?
Thanks,
Jon
Connecting the speakers is simple. Just connect the speaker connectors to the 2IPN holders on the back of the robot hat and tape the speakers in place.
Remove the old speakers. Unplug the wiring from the seat. Separate the speaker from the robot hat with the help of some hardware’s tools such as a one screwdriver, as they are taped with double sided tape.
