{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "fed7c348",
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "\n",
    "os.environ[\"CUDA_VISIBLE_DEVICES\"] = \"3\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "c046f7b4",
   "metadata": {},
   "outputs": [],
   "source": [
    "from suno_utils.diffusion import generation as diffusion_gen\n",
    "from suno_utils.tasks.upsample_engine import UpsampleEngine, Request, Job\n",
    "import torch\n",
    "from suno_utils.audio import Audio"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "5fbd9b54",
   "metadata": {},
   "outputs": [],
   "source": [
    "from suno_utils.tasks.dac_vae_fixed_25hz import preload_models as preload_codec_models, decode, encode\n",
    "\n",
    "preload_codec_models(\"s3://suno-data/minz/models/dac_vae_tuned_25hz.pth\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "a718fb2b",
   "metadata": {},
   "outputs": [],
   "source": [
    "dit_model_filepath = \"/app/suno/checkpoints/2025-02-17_16-54-01_s7787/last_ckpt.pt\"  # base\n",
    "dit_model_filepath = \"/app/suno/checkpoints/2025-03-26_13-24-59_s1608/last_ckpt_infer.pt\"\n",
    "dit_model_filepath = \"/app/suno/checkpoints/2025-03-29_03-49-41_s6886/step_40000_infer.pt\"\n",
    "diffusion_gen.preload_models(\n",
    "    dit_model_filepath=dit_model_filepath,\n",
    ")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "cf696991",
   "metadata": {},
   "outputs": [],
   "source": [
    "import json\n",
    "import numpy as np\n",
    "from suno_utils.utils.s3 import read_from_s3\n",
    "\n",
    "# gen_id = \"99bffa17-7e59-47b4-a048-5528cbda05d5\" # sister\n",
    "# gen_id = \"081d73c4-7805-4212-9c80-8db1137ca3c4\" # friends\n",
    "# gen_id = \"562f762d-6ced-4080-9af1-910ee3d0a5dc\" # something real\n",
    "# gen_id = \"23c15c62-494d-422d-8a60-8b0454044322\" # rubber duck\n",
    "# gen_id = \"4b140a9e-964b-422c-85b5-5861ad1a9d38\" # once\n",
    "# gen_id = \"7b214347-fa38-4e9b-96f4-f7ec65adea45\" # rock n roll\n",
    "gen_id = \"a5e2198a-f352-4abb-9a24-7f81b143ded3\"  # stone\n",
    "# gen_id = \"05ffc918-0bde-4418-8828-5105cd5717e1\"  # sara bad click\n",
    "# gen_id = \"7f5e7819-b7a1-449d-a76c-9c38950c4274\" # canada\n",
    "# gen_id = \"27135940-957f-4152-9e40-72e5c10a46bd\" # drunk\n",
    "# gen_id = \"caf1d7ae-ce81-4c75-b737-fd41e47a38ba\" # linger\n",
    "# gen_id = \"ac5aa29a-f18f-406f-a904-c47493f1beec\" # linger vox\n",
    "# gen_id = \"b04719c3-012a-4a4a-9b02-3cb2875d1744\" # courtney\n",
    "# gen_id = \"7407c7b3-8aa2-4f0a-82e8-fa19933f6e94\" # opera\n",
    "# gen_id = \"5b9624ef-b416-4095-a5f3-4945f09bba86\" # butterflies\n",
    "# gen_id = \"e757aa19-5c33-4945-aaa1-1519b7890f0b\" # bubblegum\n",
    "# gen_id = \"7f774078-1672-4858-a37f-acad373c5a84\" # golden\n",
    "# gen_id = \"1cee79db-a0ee-44e4-9646-d312cd620997\"  # chinese\n",
    "# gen_id = \"e404a4f0-64e0-423a-93d9-486b64c606ce\"  # mikey click\n",
    "\n",
    "# gen_id = \"7f35ab8a-5ae2-4586-868f-db8f85643385\"  # auk\n",
    "# gen_id = \"7f35ab8a-5ae2-4586-868f-db8f85643385\"  # auk\n",
    "# gen_id = \"512bc95e-2253-4adc-a2df-9df4427fafa0\"  # eating plastic\n",
    "# gen_id = \"f36bd16b-c81a-4c0d-a978-1eda0f024e4f\"  # too many people\n",
    "gen_id = \"053e61f7-2d4a-4f7a-ac92-e25c6d84c089\"  # guitar\n",
    "\n",
    "s3_filepath = f\"s3://suno-data-uploads/studio/uploads/{gen_id}.npz\"\n",
    "mp3_filepath = f\"s3://suno-data-uploads/studio/uploads/{gen_id}.mp3\"\n",
    "vae_filepath = f\"s3://suno-data-uploads/studio/uploads/{gen_id}_vae.npz\"\n",
    "print(s3_filepath)\n",
    "data = read_from_s3(s3_filepath, read_f=np.load)\n",
    "vae_data = read_from_s3(vae_filepath, read_f=np.load)\n",
    "audio = Audio.from_s3(mp3_filepath, n_channels=2)\n",
    "\n",
    "# if \"v3.0_raw\" in data:\n",
    "#     codes = data[\"v3.0_raw\"]\n",
    "# elif \"v3.5_raw\" in data:\n",
    "#     codes = data[\"v3.5_raw\"]\n",
    "# elif \"v4.0_raw\" in data:\n",
    "#     codes = data[\"v4.0_raw\"]\n",
    "# else:\n",
    "#     raise ValueError(\"No codes found\")\n",
    "audio.play()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "5ce61455",
   "metadata": {},
   "outputs": [],
   "source": [
    "# print(list(vae_data.keys()))\n",
    "# vae_latents = vae_data[\"vae_latents_v_vae_25_tuned_2\"]\n",
    "# vae_latents.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "c0c34cd3",
   "metadata": {},
   "outputs": [],
   "source": [
    "engine = UpsampleEngine(compile=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "b9002570",
   "metadata": {},
   "outputs": [],
   "source": [
    "vae_latents = encode(audio)[0 * 25 :]\n",
    "vae_latents.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "37b3fcef",
   "metadata": {},
   "outputs": [],
   "source": [
    "gen_cfg = diffusion_gen.DiffusionGenerationConfig(\n",
    "    # tags=\"add French Horn\",\n",
    "    tags=\"extract Guitar\",\n",
    "    # tags=\"extract Lead Vocal, Backing Vocals\",\n",
    "    # tags=\"remove Lead Vocal, Backing Vocals\",\n",
    "    text_cfg_coef=1.0,\n",
    "    ctx_cfg_coef=1.0,\n",
    "    steps=12,\n",
    "    codec_scale_factor=0.4,\n",
    "    seed=1,\n",
    "    noise_ctx_level=0.3,\n",
    "    scale_ctx_vector=0.4,\n",
    "    # noise_ctx_pad_len=1*25,\n",
    "    noise_stem_ctx_level=0.0,\n",
    ")\n",
    "request = Request(\n",
    "    id=\"dummy\",\n",
    "    generation_config=gen_cfg,\n",
    "    tokens=np.full((vae_latents.shape[0]), engine.dit_config[\"cond_semantic_n_vocab\"] - 1),\n",
    "    stem_ctx_latents=vae_latents,\n",
    "    input_tokens_finished=True,\n",
    ")\n",
    "\n",
    "result = engine.run_request(request)\n",
    "out_vae_latents = torch.concat(result.vae_latents)\n",
    "stem_audio = decode(out_vae_latents)\n",
    "stem_audio.play()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "d0aeca0a",
   "metadata": {},
   "outputs": [],
   "source": [
    "mixed_audio = audio.mix(stem_audio)\n",
    "mixed_audio.play()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "8628f293",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "language_info": {
   "name": "python"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}
