{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "import os\n",
    "import torch\n",
    "import torchaudio\n",
    "from tqdm import tqdm"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "audio_dir = \"/app/suno/christian/data/outputs/v45_2b_step_2_600_000/discogs_subset_sampled_metas\"\n",
    "chunk_audio_dir = \"/app/suno/christian/data/outputs/v45_2b_step_2_600_000/discogs_subset_sampled_metas_chunks_5s\"\n",
    "manifest_path = \"/home/christian/code/christian/metadata/ear/upsample_manifest_tr.txt\"\n",
    "\n",
    "CHUNK_SIZE_S = 5.0\n",
    "\n",
    "if not os.path.exists(chunk_audio_dir):\n",
    "    os.makedirs(chunk_audio_dir)\n",
    "\n",
    "example_ids = []\n",
    "with open(manifest_path, \"r\") as f:\n",
    "    for line in f:\n",
    "        example_ids.append(line.strip())\n",
    "\n",
    "print(len(example_ids))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "def chunk_audio(filepath, chunk_size_s=5.0):\n",
    "    audio, sr = torchaudio.load(filepath)\n",
    "    chunk_size = int(chunk_size_s * sr)\n",
    "    chunks = audio.unfold(-1, chunk_size, chunk_size)\n",
    "    num_chunks = chunks.shape[1]\n",
    "    for i in range(num_chunks):\n",
    "        chunk = chunks[:, i, :]\n",
    "        output_filepath = filepath.replace(\".mp3\", f\"_{i}.mp3\")\n",
    "        torchaudio.save(output_filepath, chunk, sr)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def process_example(example_id, chunk_size_s=5.0):\n",
    "    for audio_type in [\"original\", \"diff_ctx\", \"diff_no_ctx\", \"codec\"]:\n",
    "        audio_path = os.path.join(audio_dir, f\"{example_id}_{audio_type}.mp3\")\n",
    "        audio, sr = torchaudio.load(audio_path)\n",
    "        chunk_size = int(chunk_size_s * sr)\n",
    "        chunks = audio.unfold(-1, chunk_size, chunk_size)\n",
    "        num_chunks = chunks.shape[1]\n",
    "        for i in range(num_chunks):\n",
    "            chunk = chunks[:, i, :]\n",
    "            chunk_path = os.path.join(chunk_audio_dir, f\"{os.path.basename(audio_path)}\".replace(\".mp3\", f\"_{i}.mp3\"))\n",
    "            torchaudio.save(chunk_path, chunk, sr)\n",
    "\n",
    "from joblib import Parallel, delayed\n",
    "\n",
    "# Set number of processes to use (adjust based on your system)\n",
    "num_processes = 64\n",
    "print(f\"Using {num_processes} CPU cores for parallel processing\")\n",
    "\n",
    "# Process all examples in parallel with progress bar\n",
    "Parallel(n_jobs=num_processes, verbose=1)(\n",
    "    delayed(process_example)(example_id, chunk_size_s=CHUNK_SIZE_S) \n",
    "    for example_id in example_ids\n",
    ")\n"
   ]
  },
  {
   "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
}