RecursionError: maximum recursion depth exceeded in comparison encountered while finetuning

❓ Questions and Help

Before asking:

1. search the issues.
2. search the docs.

What is your question?

Question 1 : As a part of Lightning-Flash , do you have provision to prepare datamodules from dataframe

Code

Like we have the from_folders

datamodule = ImageClassificationData.from_folders(
    train_folder="data/hymenoptera_data/train/",
    val_folder="data/hymenoptera_data/val/",
    test_folder="data/hymenoptera_data/test/",
)

Question 1 : Do we have ImageClassificationData.from_dataframe() ?

Question 2 : Encounter RecursionError: while performing finetuning

# 2. Load the data
datamodule = ImageClassificationData.from_folders(
    train_folder="/content/drive/MyDrive/dphi/Kvasir/The Kvasir-Capsule/TRAIN/",
    val_folder="/content/drive/MyDrive/dphi/Kvasir/The Kvasir-Capsule/VAL/",
    predict_folder="/content/drive/MyDrive/dphi/Kvasir/The Kvasir-Capsule/test/",
)

# 3 Build the model
model = ImageClassifier(backbone="densenet121", num_classes=datamodule.num_classes)

# 4. Train the model
trainer.finetune(model, datamodule=datamodule, strategy=FreezeUnfreeze(unfreeze_epoch=1))

Error Encountered:-

RecursionError: maximum recursion depth exceeded in comparison

What have you tried?

What’s your environment?

Lightning-flash version :0.4.1dev Lightning version : 1.3.7post0 Coding in Google Colab

• OS: [e.g. iOS, Linux, Win]
• Packaging [e.g. pip, conda]
• Version [e.g. 0.5.2.1]

[ethanwharris]

Hi @plaban1981 This turned out to be an issue in lightning. The PR PyTorchLightning/lightning-flash#8170 has been merged now so your issue should be fixed when lightning 1.3.8 is released (should be tomorrow). Hope that helps 😃

[ethanwharris]

PyTorch Lightning 1.3.8 is out so this should now be resolved 😃 Feel free to re-open if not