{ "cells": [ { "cell_type": "code", "execution_count": 1, "id": "a0d83737", "metadata": {}, "outputs": [], "source": [ "# !nvidia-smi" ] }, { "cell_type": "code", "execution_count": 2, "id": "88c3355d", "metadata": {}, "outputs": [], "source": [ "import os\n", "\n", "\n", "os.environ[\"CUDA_VISIBLE_DEVICES\"] = \"7\"" ] }, { "cell_type": "code", "execution_count": 3, "id": "b59c8785", "metadata": {}, "outputs": [], "source": [ "from suno_utils.audio import Audio\n", "from generation import preload_models, generate, _retrieve_models\n", "\n", "CKPT_DIR = \"/home/christian/code/neon/stable-audio-tools/harmonai_train/\"\n", "_ = preload_models(\n", " tokenizer_filepath=\"/home/georg/notebooks/gpu_nb/tmp/tokenizer_60k.json\",\n", " semantic_model_filepath=\"/home/georg/notebooks/gpu_nb/tmp/mert_25.pt\",\n", " semantic_clusters_filepath=\"/home/georg/notebooks/gpu_nb/tmp/mert_25_2x4k.npy\",\n", " ### 25hz ###\n", " codec_filepath=\"/home/georg/notebooks/gpu_nb/tmp/25hz_vae_peaq_kl_0.005.pth\",\n", " #\n", " #\n", " # 8 nodes 5s\n", " # dit_model_filepath=\"/app/suno/checkpoints/2025-01-13_22-30-17_s9391/last_ckpt_infer.pt\",\n", " # 16 nodes 5s single chunks\n", " # dit_model_filepath=\"/app/suno/checkpoints/2025-01-17_22-35-39_s2528/last_ckpt_infer.pt\",\n", " # 16 nodes 5s\n", " # dit_model_filepath=\"/app/suno/checkpoints/2025-01-15_04-45-00_s7938/last_ckpt_infer.pt\",\n", " # 16 nodes 1s\n", " # dit_model_filepath=\"/app/suno/checkpoints/2025-01-15_04-53-13_s8269/last_ckpt_infer.pt\",\n", " # 16 nodes 5s no text pad\n", " dit_model_filepath=\"/app/suno/checkpoints/2025-01-18_16-22-22_s4434/last_ckpt_infer.pt\",\n", " # single chunks only\n", " # dit_model_filepath=\"/app/suno/checkpoints/2025-01-21_00-14-22_s2829/last_ckpt_infer.pt\",\n", ")\n", "models = _retrieve_models()\n", "if models[\"dit_model\"].ctx_len is not None:\n", " model_duration_s = 6 * 60\n", "else:\n", " model_duration_s = models[\"dit_model\"].block_size // models[\"dit_model\"].io_hz\n", "duration_s = 2 * 60 if model_duration_s >= 2 * 60 else 30" ] }, { "cell_type": "code", "execution_count": 4, "id": "61510416", "metadata": {}, "outputs": [], "source": [ "offset_s = 0\n", "\n", "samples = {\n", " \"walking_down\": {\n", " \"audio_fp\": \"/home/georg/notebooks/samples/walking_down.mp3\",\n", " \"tags\": \"Country\",\n", " \"lyrics\": \"Walking down the street\\nFeeling so alive\\nGot my head in the clouds\\nGot a gleam in my eye\\nEvery step that i take\\nIs like a brand new start\\nNo matter where I'm going\\nI'll always find my part\\n\\nOooo\\n\\nLife is like a high wire act\\nwe're dancing in the sky\\nIt's a worry no need to ask why\\nWith a little bit or courage\\nWe can chase our dreams\\nNo matter what comes away\\nWe'll always be a team\\n\\nDon't stop oooo, yeah\",\n", " },\n", " # \"norah\": {\n", " # \"audio_fp\": \"/home/georg/notebooks/samples/norah.wav\",\n", " # \"tags\": \"jazz, pop, female vocal\",\n", " # \"lyrics\": \"[Verse 1]\\nI waited 'til I saw the sun\\nI don't know why I didn't come\\nI left you by the house of fun\\nI don't know why I didn't come\\nI don't know why I didn't come\\n\\n[Verse 2]\\nWhen I saw the break of day\\nI wished that I could fly away\\nInstead of kneeling in the sand\\nCatching teardrops in my hand\\n\\n[Refrain]\\nMy heart is drenched in wine\\nBut you'll be on my mind forever\\n\\n[Verse 3]\\nOut across the endless sea\\nI will die in ecstasy\\nBut I'll be a bag of bones\\nDriving down the road alone\\n\\n[Refrain]\\nMy heart is drenched in wine\\nBut you'll be on my mind\",\n", " # },\n", " \"turkey_time\": {\n", " \"audio_fp\": \"/home/georg/notebooks/samples/turkey_time.mp3\",\n", " \"tags\": \"Country\",\n", " \"lyrics\": \"[Verse]\\nGather 'round now, it's turkey time\\nCatchin' up with all your kin and dimes\\nFeastin' on that golden bird, so fine (so fine)\\nPass the gravy, pour some wine\\n\\n[Chorus]\\nTurkey time, it's a country rhyme\\nGrab your plate, fill it up, take your time\\nAnd when the sun starts settin', we'll gather 'round\\nSingin' songs, sippin' cider, dancin' all night sound (oooh)\\nTurkey time, it's a country rhyme\\n\\n[Verse]\\nWell, it's that time of year, with family near\\nGatherin' 'round the table, filled with cheer\\nTurkey's roasting, gravy's flowin'\\nMashed potatoes piled high, it's Thanksiving time\\n\\n[Chorus]\\nTurkey time, it's a country rhyme\\nGrab your plate, fill it up, take your time\\nAnd when the sun starts settin', we'll gather 'round\\nSingin' songs, sippin' cider, dancin' all night sound (oooh)\\nTurkey time, it's a country rhyme\\n\\n[Bridge] \\nTurkey time is here again\\nTurkey time is here again\\nTurkey time is here again\\nTurkey tiiiiime!\",\n", " },\n", " # \"mistletoe\": {\n", " # \"audio_fp\": \"/home/georg/notebooks/samples/mistletoe.mp3\",\n", " # \"tags\": \"Country\",\n", " # \"lyrics\": \"[intro]\\nooooh\\n(what is life, what is love?)\\n[Verse]\\nAnother year has come and gone, and I'm not where I long to be\\nSnowfall on the city streets, but it ain't Christmas to me\",\n", " # },\n", " \"halo\": {\n", " \"audio_fp\": \"/home/georg/notebooks/samples/halo.wav\",\n", " \"tags\": \"R&B, Ballad, Pop\",\n", " \"lyrics\": \"[Verse 1]\\nRemember those walls I built?\\nWell, baby, they're tumblin' down\\nAnd they didn't even put up a fight\\nThey didn't even make a sound\\nI found a way to let you in\\nBut I never really had a doubt\\nStandin' in the light of your halo\\nI got my angel now\\n\\n[Pre-Chorus]\\nIt's like I've been awakened\\nEvery rule, I had you breakin'\\nIt's the risk that I'm takin'\\nI ain't ever gonna shut you out\\n\\n[Chorus]\\nEverywhere I'm lookin' now\\nI'm surrounded by your embrace\\nBaby, I can see your halo\\nYou know you're my savin' grace\\nYou're everything I need and more\\nIt's written all over your face\\nBaby, I can feel your halo\\nPray it won't fade away\\n\\n[Post-Chorus]\\nI can feel your halo, halo, halo\\nI can see your halo, halo, halo\\nI can feel your halo, halo, halo\\nI can see your halo, halo, halo, ooh\\n\\n[Verse 2]\\nHit me like a ray of sun\\nBurnin' through my darkest night\\nYou're the only one that I want\\nThink I'm addicted to your light\\nI swore I'd never fall again\\nBut this don't even feel like fallin'\\nGravity can't begin\\nTo pull me back to\",\n", " },\n", "}\n", "\n", "# # Select current sample\n", "# current_sample = \"turkey_time\"\n", "# audio_fp = samples[current_sample][\"audio_fp\"]\n", "# tags = samples[current_sample][\"tags\"]\n", "# lyrics = samples[current_sample][\"lyrics\"]\n", "\n", "# if current_sample == \"halo\":\n", "# offset_s = 20 if duration_s <= 30.1 else 0\n", "\n", "# audio = (\n", "# Audio.from_file(audio_fp, n_channels=2)\n", "# .get_segment(from_s=offset_s, to_s=offset_s + duration_s + 0.01)\n", "# .normalize_volume()\n", "# )\n", "\n", "# # TODO: pad with silence for integer multiple of sliding\n", "# if models[\"dit_model\"].ctx_len is not None:\n", "# window_s = int(\n", "# round((models[\"dit_model\"].block_size - models[\"dit_model\"].ctx_len) / models[\"dit_model\"].io_hz)\n", "# )\n", "# if audio.duration_s % window_s > 0.1:\n", "# audio = Audio.concatenate(\n", "# [\n", "# audio,\n", "# Audio.from_silence(\n", "# window_s - audio.duration_s % window_s + 0.01,\n", "# sample_rate=audio.sample_rate,\n", "# n_channels=audio.n_channels,\n", "# ),\n", "# ]\n", "# )\n", "# audio.normalize_volume().play()\n", "# print(lyrics)" ] }, { "cell_type": "code", "execution_count": 5, "id": "9c6c8c80", "metadata": {}, "outputs": [], "source": [ "# steps = 16\n", "# cfg_coef = 1\n", "\n", "# audios = []\n", "# for n in range(1):\n", "# pred_audio = generate(\n", "# audio,\n", "# lyrics=lyrics,\n", "# tags=tags,\n", "# text_cfg_coef=2.0,\n", "# steps=steps,\n", "# seed=n,\n", "# ).normalize_volume()\n", "# audios.append(pred_audio)\n", "# pred_audio.play()" ] }, { "cell_type": "code", "execution_count": null, "id": "acda5fe8", "metadata": {}, "outputs": [], "source": [ "for name, sample in samples.items():\n", " audio = Audio.from_file(sample[\"audio_fp\"], n_channels=2)\n", " lyrics = sample[\"lyrics\"]\n", " # lyrics = \"cat cat cat cat cat cat cat cat cat cat cat cat cat cat cat cat cat cat cat cat cat cat cat cat cat cat cat \"\n", " tags = sample[\"tags\"]\n", " # Generate once for each sample\n", " pred_audio = generate(\n", " audio,\n", " lyrics=lyrics,\n", " tags=tags,\n", " text_cfg_coef=4.0,\n", " steps=32,\n", " seed=0,\n", " ).normalize_volume()\n", "\n", " print(name)\n", " pred_audio.play()\n", " print(lyrics)\n" ] }, { "cell_type": "code", "execution_count": null, "id": "5ea08dfd", "metadata": {}, "outputs": [], "source": [] } ], "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": "3.10.14" } }, "nbformat": 4, "nbformat_minor": 5 }