RaspberryPi 4B source install TensorFlow version issue #193
Comments
|
Thanks for the thorough report! Hmm, yeah this definitely looks like some problem related to the keras/tensorflow version. You might be able to get it working by using code off of PR #141 since that uses Tensorflow 2 which should resolve a lot of version related issues. Let me know if this helps. |
|
Thanks for the quick reply! Ok, got the #141 and tried running the ./setup.sh, this was the output: Should I uninstall/remove something before the new install? |
|
Try unpinning |
|
Ok, so tried unpinning from setup.py but didn't work. I found this guide on installing tensorflow-2.2.0 on RPi 4 and the installation went ok. Had to install it directly in the venv for it to work with precise though. Tried running I'm running After the training finishes, I'll test it out and post here how it went. |
|
Well, still no success... I'm following these steps to train my own word. Everything went well until I followed the "method 2" in Reducing False Activations. Downloaded all the files and ran the ffmpeg script on all of them. This is how my folder looks like after running all of these scripts: |
|
Hmm, I think I actually fixed that bug on my personal tf2 branch. Could you try checking it out with this?: git remote add matthew https://github.com/MatthewScholefield/mycroft-precise
git fetch matthew
git checkout feature/tf2 |
|
After changing to your tf2 branch I left the |
|
I saved the output of the Not sure if a specific message should end that script... |
|
Wanted to confirm I'm seeing the same issue/output on the same hardware/setup as @SandroMiccoli I have not yet tried upgrading the venv to TF2.2, were you able to get the incremental training to complete? |
|
Nope, I tried 3 times but on all of them I got the same |
|
Had success here. Trained the set with Of course, the app I'm using (Rhasspy) requires .pb/.pb.param files for the integration. Anybody have suggestions on that? Error from the
|
|
I should note: I'm using 141, not @MatthewScholefield repo/branch |
|
I haven't looked into @kbromer Looks like you can set |
|
Oh wow, I'll try that, thanks! |
|
@kbromer so you have the same setup? RPi 4B buster headless clean environment? Did you have to install TensorFlow 2.2.0 "by hand" or you had success with the standard mycroft-precise installation script? |
|
Yes, only changes to the environment were booting from SSD over USB. Used the link in your comment above to install TF 2.2. Only thing that regularly dies is the
Retraining it though with |
|
Thanks @kbromer, still no success here... Started with a new clean environment, installed TF2.2 via whl file, recorded some audio samples with I'm using the main branch for this repo though. Since I got a different error this time not sure which branch would be the most appropriate. And TF2.2 was installed successfully (in the .venv): |
|
@MatthewScholefield do you plan to upstrem the |
This is super late, but I think I know the issue here. I don't have a ton of experience with tensorflow, but by the error I get as well I'm assuming it's related to the save_format change: What it looks like is that Perhaps @MatthewScholefield could confirm this? If it is the case and has a reasonable fix then I'm happy to figure it out and push it as a merge to his branch. Edit: Looks like the fix to this would be to change the extension to .h5, and perhaps tell people to use the .h5 path (which appears to be a self-contained file?) to load the model instead of just ".net" |
|
Anyone reading this that wants to build from source, try these steps:
|
Describe the bug
I'm trying to train my own wake word on a RaspberryPi 4B. After installing via source, I used
precise-collectto record several samples, everything worked ok, but when I try toprecise-listenor toprecise-trainI get these WARNING messages:With precise-listen:
And precise-train:
Those two warnings appear at the beggninng but the rest of outputs seem ok. But when I try to
precise-listenmy new .net file, an Error appears:I've tried download the dev branch but the same error happens.
To Reproduce
Steps to reproduce the behavior:
./setup.shscriptsource .venv/bin/activateprecise-listenany model (I used some community data and also a test of my own after collecting several recordings and generating a .net withprecise-train)Expected behavior
I expected to run my created model but got an error there.
Environment
Additional context
I'm using a clean buster installation. I'm using voice2json and everything works fine there.
The text was updated successfully, but these errors were encountered: