Forge-SID-Model

This repository contains a pre-trained RQVAE (Residual Quantized Variational Autoencoder) model designed for SID (Speaker Identity/Structure) generation tasks. It is part of the FORGE ecosystem.

The model weights are stored in final_sid_rq_model.pth.

Usage

1. Download the Model

You can download the model files locally using the huggingface_hub library:

import os
# Optional: Use mirror for faster download in some regions (e.g., China)
os.environ["HF_ENDPOINT"] = "https://hf-mirror.com" 
os.environ["KMP_DUPLICATE_LIB_OK"] = "True"

from huggingface_hub import snapshot_download

snapshot_download(
    repo_id='AL-GR/Forge-SID-Model',
    local_dir='./Forge-SID-Model',  # Replace with your desired local path
    local_dir_use_symlinks=False,
)

2. Run Inference

To use this model for inference, you need to update the checkpoint path in the official inference script provided by the al_sid repository.

Step 1: Clone or download the inference code: https://github.com/selous123/al_sid/blob/main/SID_generation/infer_SID.py

Step 2: Open infer_SID.py and locate Line 23.

Step 3: Modify the CKPT_PATH variable to point to your downloaded .pth file:

# Original line:
# CKPT_PATH = 'output_model/checkpoint-7.pth'

# Update to (example):
CKPT_PATH = './Forge-SID-Model/final_sid_rq_model.pth' 

Note: Ensure the path matches the actual location where you saved the final_sid_rq_model.pth file.

---

For more details about the training setup or the FORGE framework, please refer to the main repository: AL-GR/FORGE.