Autotrain image classification 500 error

🤗AutoTrain
1

I am trying to use autotrain’s image classification but everytime I upload the data and it processes it, it throws a 500 error and this in the logs - I am using the file format mentioned in the docs here Image Classification

INFO Task: llm:sft
INFO: 10.16.34.18:39034 - “GET /params/llm%3Asft HTTP/1.1” 200 OK
INFO: 10.16.18.44:16135 - “GET /model_choices/llm%3Asft HTTP/1.1” 200 OK
INFO Task: image-classification
INFO: 10.16.18.44:36314 - “GET /params/image-classification HTTP/1.1” 200 OK
INFO: 10.16.18.44:23915 - “GET /model_choices/image-classification HTTP/1.1” 200 OK
INFO hardware: A100 Large
INFO Dataset: autotrain-images-category-10 (image_multi_class_classification)
Train data: <tempfile.SpooledTemporaryFile object at 0x7fd1d0fa9090>
Valid data: <tempfile.SpooledTemporaryFile object at 0x7fd1d0faa470>

I also see this in the console -
Error: 500
Uncaught (in promise) SyntaxError: JSON.parse: unexpected end of data at line 1 column 1 of the JSON data

I’m thinking it’s because of the column mapping? It’s expecting {“image”: “image”, “label”: “label”} which is different to what the docs say which is just to use a zipped folder containing the classes that then contain the images?

2

cc @abhishek

3

I just ran a training and it seems to be working as expected:

image
image2762Ă—1516 442 KB

When the zip is unzipped, it must produce all subfolders with different classes, not one folder. please also add logs so we know whats happening :slight_smile: