{ "cells": [ { "cell_type": "code", "execution_count": 25, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "625\n", "tensor([ 15, 91, 3, 91, 64, 68, 65, 13, 87, 90, 15, 41,\n", " 65, 67, 53, 68, 83, 30, 71, 4, 18, 55, 95, 42,\n", " 60, 91, 14, 16, 73, 69, 46, 44, 49, 77, 65, 20,\n", " 76, 44, 34, 67, 8, 80, 32, 88, 1, 36, 71, 91,\n", " 70, 47, 45, 63, 10, 6, 81, 13, 64, 31, 52, 41,\n", " 9, 50, 77, 62, 55, 20, 67, 29, 46, 74, 28, 47,\n", " 9, 8, 98, 92, 35, 55, 43, 76, 19, 25, 86, 15,\n", " 59, 47, 57, 13, 27, 57, 50, 13, 2, 57, 52, 25,\n", " 53, 14, 20, 42, 24, 3, 44, 18, 67, 79, 8, 44,\n", " 88, 71, 71, 16, 17, 10, 80, 48, 72, 50, 95, 89,\n", " 76, 95, 73, 54, 69, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000])\n" ] } ], "source": [ "import torch\n", "import random\n", "\n", "semantic_n_tokens = 750\n", "semantic_pad_token = 4000\n", "semantic_rate_hz = 25\n", "\n", "semantic_codes = torch.randint(0, 100, (semantic_n_tokens,))\n", "\n", "dice_roll_2 = 0.67\n", "if dice_roll_2 <= 0.33:# 33% chance pad with random amount at the end\n", " n = random.randint(1, semantic_n_tokens - 1)\n", "elif dice_roll_2 <= 0.66: # 33% chance to pad exactly 15s at the end (15s chunk)\n", " n = semantic_rate_hz * 15\n", "else: # 33% chance to pad exactly 25s at the end (5s chunk)\n", " n = semantic_rate_hz * 25\n", "\n", "print(n)\n", "semantic_codes[-n:] = semantic_pad_token\n", "\n", "print(semantic_codes)" ] }, { "cell_type": "code", "execution_count": 49, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "0.6652893338666702\n", "tensor([4000, 4000, 4000, 4000, 4000, 14, 12, 36, 4000, 4000, 74, 4000,\n", " 60, 4000, 4000, 4000, 4000, 94, 20, 4000, 4000, 4000, 88, 4000,\n", " 3, 4000, 4000, 51, 4000, 15, 4000, 4000, 5, 4000, 4000, 82,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 83, 4000, 4000, 4000, 4000, 53,\n", " 43, 10, 4000, 43, 4000, 4000, 4000, 4000, 55, 4000, 4000, 6,\n", " 4000, 9, 4000, 30, 4000, 4000, 4000, 7, 31, 79, 4000, 60,\n", " 84, 4000, 4000, 4000, 0, 4000, 4000, 31, 4000, 4000, 4000, 4000,\n", " 4000, 44, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 8, 4000, 4000,\n", " 4000, 4000, 4000, 86, 4000, 59, 4000, 3, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 44, 4000, 4000, 69, 4000, 4000, 81,\n", " 4000, 4000, 54, 4000, 4000, 4000, 4000, 74, 4000, 37, 4000, 58,\n", " 4000, 21, 71, 4000, 4000, 23, 4000, 4000, 4000, 4000, 6, 31,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 63, 52, 5, 4000, 22,\n", " 14, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 89, 4000, 61, 4000,\n", " 4000, 4000, 4000, 4000, 54, 4000, 80, 78, 4000, 4000, 15, 76,\n", " 4000, 14, 4000, 4000, 4000, 17, 9, 65, 4000, 4000, 4000, 72,\n", " 4000, 88, 4000, 69, 4000, 54, 17, 4000, 4000, 4000, 15, 4000,\n", " 4000, 28, 4000, 4000, 63, 12, 4000, 4000, 4000, 4000, 4000, 73,\n", " 67, 4000, 4000, 4000, 82, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4, 4000, 4000, 20, 64, 4000, 4000, 4000, 4000,\n", " 56, 4000, 4000, 4000, 86, 4000, 25, 4000, 4000, 4000, 97, 4000,\n", " 4000, 4000, 4000, 93, 4000, 4000, 4000, 4000, 22, 4000, 30, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 4000, 4000, 72, 4000, 49, 53, 32,\n", " 4000, 4000, 4000, 38, 4000, 4000, 89, 9, 4000, 68, 4000, 4000,\n", " 4000, 76, 4000, 4000, 4000, 4000, 4000, 4000, 78, 4000, 4000, 4000,\n", " 4000, 4000, 25, 90, 4000, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 17, 4000, 4000, 71, 4000, 57, 4000, 35, 4000, 4000, 4000, 4000,\n", " 95, 54, 4000, 79, 4000, 63, 4000, 1, 74, 4000, 73, 66,\n", " 70, 36, 4000, 4000, 4000, 4000, 62, 4000, 53, 4000, 4000, 4000,\n", " 4000, 40, 4000, 4000, 51, 4000, 4000, 4000, 4000, 17, 5, 4000,\n", " 67, 4000, 4000, 53, 4000, 4000, 4000, 60, 12, 4000, 4000, 4000,\n", " 53, 82, 4000, 97, 22, 4000, 89, 4000, 4000, 4000, 19, 4000,\n", " 85, 54, 4000, 4000, 4000, 79, 4000, 4000, 18, 4000, 4000, 4000,\n", " 4000, 19, 4000, 4000, 4000, 74, 4000, 96, 4000, 4000, 4000, 50,\n", " 4000, 4000, 29, 16, 4000, 68, 4000, 4000, 43, 62, 99, 4000,\n", " 4000, 1, 69, 4000, 4000, 4000, 67, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 4000, 97, 4000, 69, 4000, 4000, 4000, 62, 4000, 11, 4000,\n", " 24, 99, 4000, 4000, 4000, 4000, 4000, 4000, 65, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 91, 46, 4000, 4000, 70, 4000, 11, 4000, 4000,\n", " 59, 4000, 4000, 4000, 4000, 45, 4000, 4000, 4000, 4000, 53, 4000,\n", " 4000, 4000, 4000, 4000, 80, 69, 41, 4000, 7, 98, 4000, 4000,\n", " 4000, 4000, 14, 4000, 4000, 12, 4000, 41, 4000, 96, 4000, 4000,\n", " 58, 4000, 4000, 42, 4000, 79, 4000, 4000, 4000, 4000, 65, 4000,\n", " 34, 4000, 72, 4000, 55, 4000, 4000, 4000, 86, 9, 4000, 4000,\n", " 4000, 4000, 4000, 99, 58, 47, 4000, 55, 4000, 58, 4000, 4000,\n", " 18, 4000, 4000, 12, 46, 53, 4000, 4000, 7, 4000, 4000, 4000,\n", " 27, 4000, 4000, 68, 84, 4000, 4000, 4000, 87, 4000, 4000, 4000,\n", " 4000, 39, 4000, 85, 4000, 56, 29, 98, 65, 4000, 4000, 4000,\n", " 4000, 4000, 4000, 4000, 4000, 32, 4000, 4000, 13, 4000, 4000, 4000,\n", " 4000, 82, 4000, 87, 33, 44, 98, 4000, 4000, 22, 18, 4000,\n", " 4000, 4000, 9, 4000, 4000, 74, 4000, 14, 4000, 4000, 4000, 17,\n", " 4000, 3, 4000, 4000, 4000, 4000, 4000, 4000, 13, 4000, 37, 20,\n", " 4000, 22, 4000, 4000, 96, 4000, 4000, 24, 89, 4000, 58, 4000,\n", " 4000, 4000, 4000, 30, 32, 4000, 4000, 4000, 24, 4000, 4000, 22,\n", " 4000, 4000, 4000, 4000, 1, 4000, 4000, 3, 4000, 58, 4000, 4000,\n", " 4000, 4000, 4000, 4, 4000, 4000, 4000, 4000, 4000, 4000, 11, 66,\n", " 34, 4000, 4000, 4000, 83, 52, 4000, 4000, 39, 4000, 15, 19,\n", " 4000, 4000, 4000, 19, 91, 4000, 77, 4000, 55, 5, 8, 4000,\n", " 4000, 4000, 30, 4000, 4000, 4000, 4000, 4000, 4000, 9, 4000, 14,\n", " 4000, 60, 4000, 4000, 4000, 22, 11, 90, 4000, 4000, 49, 60,\n", " 85, 4000, 57, 82, 57, 4000, 4000, 4000, 4000, 4000, 4000, 4000,\n", " 4000, 48, 4000, 77, 82, 4000, 4000, 4000, 4000, 44, 4000, 4000,\n", " 4000, 4000, 4000, 86, 4000, 15])\n" ] } ], "source": [ "semantic_codes = torch.randint(0, 100, (semantic_n_tokens,))\n", "\n", "# Sample random masking probability between 0 and 1\n", "p_mask = random.random()\n", "mask = torch.bernoulli(torch.full((semantic_n_tokens,), 1-p_mask))\n", "semantic_codes[mask == 0] = semantic_pad_token\n", "print(semantic_codes)" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [] } ], "metadata": { "kernelspec": { "display_name": "suno_env", "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": "3.10.9" } }, "nbformat": 4, "nbformat_minor": 2 }