đ´2. Model Training | Applio
The following tutorial is currently for Applio, please follow the Applio tutorial before.
Before you begin, you should have a dataset ready. You can follow the tutorial for creating a dataset 1. Dataset Creation
Step 1: Dataset Preparation
Create a folder with the name of your choice inside Applio-RVC-Fork/datasets/
.
Download the audio files you intend to use for training into this folder.
Step 2: Dataset Processing
Select the folder you created in step 1.
Step 3: Feature Extraction
Wait for the "All-Feature-Done
" message to confirm the completion of feature extraction.
Step 4: Model Training
Configure the training parameters according to your needs..
Save Frequency: Set this value between 10 and 50. It determines how often the model's state is saved during training. This helps in reverting to a previous point in case of overfitting.
Training Epochs: The number of epochs needed varies based on your dataset. Choose a value that seems appropriate and monitor the progress using TensorBoard. Typically, models start performing well around 100-200 epochs.
Batch Size: Adjust the batch size based on your GPU's VRAM. For example, if you have 8 GB of VRAM, use a batch size between 6 and 8. Consider the CUDA cores of your GPU when experimenting with higher batch sizes.
Monitor the progress in the Applio console with each epoch.
Final Step: Model Saving and Index File Generation
If you have followed the steps correctly, it should resemble the example below:
To save your model, select a preferred saving method and click the corresponding button. The .zip
file containing your model will be located inside the Applio-RVC-Fork/logs/finished/
folder.
Last updated