Identifying and Solving Issues

Encountering problems with Modlee? This page provides solutions to the most common issues users face. Follow these steps to resolve problems or get guidance on where to find further assistance.

Installation Issues

Problem: Installation Errors

Issue: Error messages during installation or failure to install Modlee.

Solutions:

Check Python Version: Ensure you are using a compatible Python version. Modlee supports Python 3.10 and above.
Upgrade Pip: Sometimes, updating pip can resolve installation issues. Run this command:
```
pip install --upgrade pip
```
Correct Installation Command: Use the correct installation command to download the Modlee package:
```
pip install modlee
```
Dependencies: Verify that all dependencies are installed. You can find the list of required dependencies on Modlee’s Github.

Problem: Incompatible Package Versions

Issue: Errors related to conflicting versions of packages.

Solutions:

Check Dependency Versions: Ensure that all required packages are installed in compatible versions. Refer to Modlee’s requirements.txt for the correct versions.
Use Virtual Environment: Consider using a virtual environment to manage package versions and avoid conflicts. Create one and activate it before installing Modlee. Run the following command to create a virtual environment:
```
python -m venv myenv
```
Activate the virtual environment using this command:
- On Windows:
```
myenv\Scripts\activate
```
- On macOS/Linux:
```
source myenv/bin/activate
```

Setup Problems

Problem: Configuration Errors

Issue: Issues with the initial setup or configuration of Modlee.

Solutions:

Follow Setup Instructions: Review the Quickstart page for detailed instructions on configuring the Modlee package.

Environment Variables: Verify that necessary environment variables are correctly set. For example, ensure that any API keys are properly configured.

import os
import modlee

# Set your API key
os.environ['MODLEE_API_KEY'] = "replace-with-your-api-key"

# Initialize the Modlee package
modlee.init(api_key=os.environ.get('MODLEE_API_KEY'))

Problem: Unable to Access Modlee API

Issue: Errors related to API key.

Solutions:

Verify API Keys: Ensure that your API keys are valid and correctly configured in your environment.
Generate New Key: Refresh the Modlee dashboard and generate a new API key.

Model Issues

Problem: Training Process Fails or Crashes

Issue: Training process fails, crashes, or does not complete as expected.

Solutions:

Inspect Training Logs: Review training logs to identify any error messages or warnings.
Check Data Integrity: Ensure that your data is correctly formatted and preprocessed before training.
Verify Model Configuration: Double-check your model configuration, including hyperparameters and architecture, to ensure they are correctly set.

Problem: None Model Recommended

Issue: The model recommender does not return a valid model.

Solutions:

Check Input Parameters: Ensure that the parameters passed to the recommender function are correct. For example, verify that the num_classes parameter is properly specified and supported by Modlee.
```
recommender = modlee.recommender.ImageClassificationRecommender(
    num_classes=10
)
```
Verify Dataset Compatibility: Ensure that the dataset you are using is compatible with the recommender’s modality and task. The recommender might not be able to suggest a model if the dataset or task does not align with available models.
Restart Session: Sometimes, restarting your development environment or Jupyter Notebook session can resolve issues. Restart your session and try running the recommender again.

Environment Issues

Problem: Kaggle Notebook Environment

Issue: Encountering problems while using Modlee in a Kaggle Notebook environment.

Solutions:

Check Accelerator Type: Ensure that the accelerator type in the Kaggle Notebook is set correctly. Go to the notebook’s Session Options and select the appropriate accelerator (e.g., GPU, TPU, or CPU) depending on your project requirements.
Install Dependencies: Verify that all necessary packages and dependencies are installed. You might need to install or update the libraries using:
```
pip install modlee
```
Check File Paths: Make sure the file paths for data and models are correctly specified, especially when accessing datasets stored in Kaggle’s input directory. If needed, load datasets directly from the Kaggle datasets:
```
dataset = pd.read_csv('/kaggle/input/your-dataset/your-dataset.csv')
```

Problem: Kaggle Notebook, Run Out of GPU Compute

Issue: Errors while connecting to GPU on Kaggle Notebooks.

Solutions:

Reduce Batch Size: Decrease the batch size to reduce memory usage. For example
```
train_loader = DataLoader(dataset, batch_size=4, shuffle=True)
```
Free Up GPU Memory: Clear unused variables and force garbage collection to free up GPU memory
Upgrade Runtime: If possible, upgrade your Kaggle account to a Pro tier for more GPU resources.

Model Limitations

Problem: Errors Using LSTMs

Issue: Attempting to use LSTM layers or recurrent operations results in errors.

Explanation : Currently, Modlee’s modlee.model.TimeseriesClassificationModleeModel does not support recurrent layers like LSTMs.

Solutions:

Use Non-Recurrent Models: Instead, try using models designed to handle time series data without recurrence, such as convolutional neural networks (CNNs) and transformers. These architectures can capture sequential dependencies effectively without relying on LSTM layers.
Consider Alternative Libraries: If your project requires LSTM functionality, consider using an alternative library that supports recurrent layers specifically for time series tasks.

Still Need Help?

If you continue to experience issues or need further assistance, please email us at support@modlee.ai or join our community forum. You can also raise an issue on our GitHub repository for additional support.