Error when using custom classification block

Help
#1

Question/Issue:
I created a custom classification block on VScode and it has the same structure as when you select edit block locally. I’ve been reading the documentation and followed the steps. Once the training was done and the rained model was stored in the out folder, I created a block and pushed it with edge-impulse-blocks push with the cmd directory on the same path as the project folder. It created a custom block but when using it in edge impulse and error appears and it says that there is no metadata, and block has not been trained. How can i solve this?
Here it is the screen cap:

no_metadata

Here is my block in Machine Learning blocks page:

image
image1601×123 8.45 KB

The training terminal output is this:

Creating job... OK (ID: 4737273)

Scheduling job in cluster...
Job started
Copying features from processing blocks...
Copying features from DSP block...
Copying features from DSP block OK
Copying features from processing blocks OK

Scheduling job in cluster...
Container image pulled!
Job started
Splitting data into training and validation sets...
Splitting data into training and validation sets OK
Scheduling job in cluster...
Container image pulled!
Job started
Application exited with code 126 (Unknown)

Job failed (see above)

Project ID:
156428
Context/Use case:

#2

Hi @JosuGaztelu,

Did you use the CLI tool to create your block template (Blocks - Edge Impulse Documentation)?

You also might want to check the metadata given in extract_dataset.py as compared to a known good learning block, such as this one: GitHub - edgeimpulse/example-custom-ml-block-ti-yolox: YOLOX block from TI

#3

Yes, I followed the steps and didn’t work.
Anyway, I used the template from edgeimpulse/example-custom-ml-block-keras: Custom Keras ML block example for Edge Impulse (github.com) and it did work so I just changed the structure of the network for the actual project and it worked.

2 Likes