RAVE-Latent-Diffusion_Training_Template--Kaggle.ipynb

{"metadata":{"kernelspec":{"display_name":"Python 3 (ipykernel)","language":"python","name":"python3"},"language_info":{"codemirror_mode":{"name":"ipython","version":3},"file_extension":".py","mimetype":"text/x-python","name":"python","nbconvert_exporter":"python","pygments_lexer":"ipython3","version":"3.11.5"},"kaggle":{"accelerator":"none","dataSources":[],"isInternetEnabled":true,"language":"python","sourceType":"notebook","isGpuEnabled":false}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"markdown","source":"# RAVE-Latent Diffusion","metadata":{"_cell_guid":"b2c77c4e-dd54-4f77-bef7-09dce78f4683","_uuid":"e07ef263-3824-40b2-a4de-a11b51a32a7e","collapsed":false,"jupyter":{"outputs_hidden":false}}},{"cell_type":"markdown","source":"**RAVE-Latent Diffusion** is a denoising diffusion model designed to generate new RAVE latent codes with a large context window, faster than realtime, while maintaining music structural coherency. Code by [Moisés Horta Valenzuela](https://github.com/moiseshorta) \\ 𝔥𝔢𝔵𝔬𝔯𝔠𝔦𝔰𝔪𝔬𝔰 \n\nhttps://github.com/moiseshorta/RAVE-Latent-Diffusion\n\n**RAVE** is a variational autoencoder for fast and high-quality neural audio synthesis by Antoine Caillon and Philippe Esling. [Article on arxiv](https://arxiv.org/abs/2111.05011) & [Source code on Github](https://github.com/acids-ircam/RAVE)\n\n*Note that this notebook has been tested with RAVE models until version 2.2.2 - using models created with version >= 2.3. might not lead to a successful training.*\n\n----\n\nNotebook author: [Martin Heinze](https://github.com/devstermarts)\n\nLast updated: 23.11.2024","metadata":{"_cell_guid":"5da54055-3717-4ae4-902b-935b416effc7","_uuid":"d8fa2601-d25e-45bb-be07-b3ddc3f45d98","collapsed":false,"jupyter":{"outputs_hidden":false}}},{"cell_type":"markdown","source":"## Install Miniconda, RAVE-Latent Diffusion, dependencies","metadata":{}},{"cell_type":"code","source":"## Install Miniconda\n!mkdir /kaggle/temp/\n%cd /kaggle/temp\n!curl -L https://repo.anaconda.com/miniconda/Miniconda3-py39_4.12.0-Linux-x86_64.sh -o miniconda.sh\n!chmod +x miniconda.sh\n!sh miniconda.sh -b -p /kaggle/temp/miniconda\n\n## Install RAVE-Latent Diffusion and dependencies\n!git clone https://github.com/moiseshorta/RAVE-Latent-Diffusion.git\n%cd /kaggle/temp/RAVE-Latent-Diffusion\n!/kaggle/temp/miniconda/bin/pip install -r requirements.txt","metadata":{"_cell_guid":"62b06402-07bb-471b-9d99-28108af1fd60","_uuid":"1fde1f10-e403-4701-959e-28edc82510c7","collapsed":false,"jupyter":{"outputs_hidden":false}},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"## Preprocess RAVE model and audio dataset\n\n***This section is only needed once before initial training. Disable for training resume or output creation.***","metadata":{}},{"cell_type":"code","source":"!mkdir /kaggle/working/latents/\n!/kaggle/temp/miniconda/bin/python preprocess.py \\\n--rave_model \"/kaggle/input/your_rave_model.ts\" \\\n--audio_folder \"/kaggle/input/your_dataset\" \\\n--sample_rate 44100 \\\n--latent_length 4096 \\\n--latent_folder \"/kaggle/working/latents\"","metadata":{"_cell_guid":"8f3edb2b-9d8e-4274-8264-1d14ebd2e00a","_uuid":"f86d8e17-818c-4dab-ad78-5b656d597a9a","collapsed":false,"jupyter":{"outputs_hidden":false}},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"## Start training\n\n***This section is for initial training. Disable for training resume or output creation.***\n\nFor all available parameters, check train.py ","metadata":{}},{"cell_type":"code","source":"!mkdir /kaggle/working/checkpoints/\n!/kaggle/temp/miniconda/bin/python train.py \\\n--name your_training_name \\\n--latent_folder \"/kaggle/working/latents\" \\\n--save_out_path \"/kaggle/working/checkpoints\" \\\n--save_interval 100 #Default=50. LD checkpoints are big, to avoid memory depletion, this is set to higher value","metadata":{"_cell_guid":"19caaeae-4e81-4711-8f1d-48497a0fb1ab","_uuid":"530e10b4-7834-4f64-a666-68bc5a771a72","collapsed":false,"jupyter":{"outputs_hidden":false}},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"## Resume training\n\n***This section is for training resume. Disable for initial training or output creation. Make sure to use the same configuration as in your initial training.***\n\nFor all available parameters, check train.py ","metadata":{}},{"cell_type":"code","source":"#Copy contents of earlier training to /kaggle/working folder.\n!cp -r /kaggle/input/root_folder_of_your_earlier_training/* /kaggle/working\n\n!/kaggle/temp/miniconda/bin/python train.py \\\n--name your_training_name \\\n--latent_folder \"/kaggle/working/latents\" \\\n--save_out_path \"/kaggle/working/checkpoints\" \\\n--checkpoint_path \"/kaggle/working/checkpoints\" \\\n--save_interval 100 #Default=50. LD checkpoints are big, to avoid memory depletion, this is set to higher value","metadata":{"_cell_guid":"a9c91dbc-f367-496a-ac92-694221a36f1a","_uuid":"cc707430-e770-4a98-9785-ef7a11ce6492","collapsed":false,"jupyter":{"outputs_hidden":false}},"outputs":[],"execution_count":null},{"cell_type":"markdown","source":"## Generate output wave files\n\n***This section is for output creation. Disable for initial training or resume training.***\n\nFor all available parameters, check generate.py ","metadata":{}},{"cell_type":"code","source":"!mkdir /kaggle/working/output/\n\n#The below command is exemplary for one output file, change to your liking and c/p with different parameters as you seem fit.\n!/kaggle/temp/miniconda/bin/python generate.py \\\n--model_path \"/kaggle/input/root_folder_of_your_earlier_training/checkpoints/model_with_best_suffix.pt\" \\\n--rave_model \"/kaggle/input/your_rave_model.ts\" \\\n--sample_rate 44100 \\\n--diffusion_steps 100 \\\n--seed 91827536 \\\n--temperature=1 \\\n--latent_length 8192 \\\n--length_mult 1 \\\n--lerp=True \\\n--seed_a=91827536 \\\n--seed_b=19283574 \\\n--output_path \"/kaggle/working/output\" \\\n--name your_filename","metadata":{"_cell_guid":"bdb5089c-bc52-4ca2-90d3-5fdbcb4f81d9","_uuid":"99a450f7-c6b4-48c7-8eb7-ef55fc0a849d","collapsed":false,"jupyter":{"outputs_hidden":false}},"outputs":[],"execution_count":null}]}