{ "cells": [ { "cell_type": "code", "execution_count": 1, "id": "ffdbda4b", "metadata": { "ExecuteTime": { "end_time": "2023-09-06T17:16:50.543968Z", "start_time": "2023-09-06T17:16:50.542484Z" } }, "outputs": [], "source": [ "import os\n", "\n", "os.environ[\"CUDA_VISIBLE_DEVICES\"] = \"\"" ] }, { "cell_type": "code", "execution_count": 2, "id": "dd41c10c", "metadata": { "ExecuteTime": { "end_time": "2023-09-06T17:16:53.667092Z", "start_time": "2023-09-06T17:16:50.544832Z" } }, "outputs": [], "source": [ "import random\n", "import json\n", "import numpy as np\n", "import tqdm\n", "import torch\n", "import funcy\n", "import time\n", "import gc\n", "from scipy.io import wavfile\n", "import tempfile\n", "import collections\n", "from collections import defaultdict\n", "from joblib import Parallel, delayed\n", "\n", "from suno_utils.utils.s3 import _apply_mp\n", "from suno_utils.audio import Audio\n", "from suno_utils.tasks.data_loader import load_audio_mp\n", "from suno_utils.utils.text import write_jsonl, read_jsonl, write_json, read_json\n", "from suno_utils.utils.s3 import read_from_s3, check_s3_file_exists, open_from_s3\n", "from suno_utils.audio.conversion import convert_audio_files\n", "\n", "SAMPLE_RATE = 24_000\n", "EMBEDDING_RATE = 25\n", "N_CODEBOOKS = 8\n", "\n", "IN_DATA_DIR = \"/app/suno/data/mert_25hz\"\n", "IN_AUDIO_DIR = os.path.join(IN_DATA_DIR, \"audio\")\n", "IN_TSV_DIR = os.path.join(IN_DATA_DIR, \"audio_tsv\")\n", "IN_LABEL_DIR = os.path.join(IN_DATA_DIR, \"label\")\n", "\n", "OUT_DATA_DIR = \"/app/suno/data/mert_25hz_short\"\n", "OUT_AUDIO_DIR = os.path.join(OUT_DATA_DIR, \"audio\")\n", "OUT_TSV_DIR = os.path.join(OUT_DATA_DIR, \"audio_tsv\")\n", "OUT_LABEL_DIR = os.path.join(OUT_DATA_DIR, \"label\")\n", "OUT_TEMP_DIR = os.path.join(OUT_DATA_DIR, \"temp\")\n", "\n", "os.makedirs(OUT_DATA_DIR, exist_ok=True)\n", "os.makedirs(OUT_AUDIO_DIR, exist_ok=True)\n", "os.makedirs(OUT_TSV_DIR, exist_ok=True)\n", "os.makedirs(OUT_LABEL_DIR, exist_ok=True)\n", "os.makedirs(OUT_TEMP_DIR, exist_ok=True)" ] }, { "cell_type": "code", "execution_count": 3, "id": "195c03f2", "metadata": { "ExecuteTime": { "end_time": "2023-09-06T17:16:53.670018Z", "start_time": "2023-09-06T17:16:53.668304Z" } }, "outputs": [], "source": [ "MIN_SIZE = SAMPLE_RATE * 5\n", "MAX_SIZE = SAMPLE_RATE * 12" ] }, { "cell_type": "markdown", "id": "c729d86d", "metadata": {}, "source": [ "# File loading" ] }, { "cell_type": "code", "execution_count": 4, "id": "66631638", "metadata": { "ExecuteTime": { "end_time": "2023-09-06T17:16:54.300102Z", "start_time": "2023-09-06T17:16:53.671659Z" } }, "outputs": [], "source": [ "# load current data\n", "tsv_info = []\n", "with open(os.path.join(IN_TSV_DIR, \"train.tsv\"), \"r\") as f:\n", " for line in f.read().strip().split(\"\\n\"):\n", " if len(line.strip()) == 0:\n", " continue\n", " tsv_info.append(line.strip().split(\"\\t\"))" ] }, { "cell_type": "code", "execution_count": 5, "id": "cad52faa", "metadata": { "ExecuteTime": { "end_time": "2023-09-06T17:16:54.303887Z", "start_time": "2023-09-06T17:16:54.301923Z" } }, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "436804\n" ] } ], "source": [ "print(len(tsv_info))" ] }, { "cell_type": "code", "execution_count": 6, "id": "741d5c23", "metadata": { "ExecuteTime": { "end_time": "2023-09-06T17:16:54.354280Z", "start_time": "2023-09-06T17:16:54.304983Z" } }, "outputs": [], "source": [ "tsv_data_path = tsv_info[0]\n", "tsv_data = tsv_info[1:]" ] }, { "cell_type": "markdown", "id": "d17fd56f", "metadata": {}, "source": [ "## Slice the TSV files" ] }, { "cell_type": "code", "execution_count": 7, "id": "c4906e01", "metadata": { "ExecuteTime": { "end_time": "2023-09-06T17:17:06.346433Z", "start_time": "2023-09-06T17:16:54.355730Z" } }, "outputs": [ { "name": "stderr", "output_type": "stream", "text": [ "100%|████████████████████████████████████████████████████████████████████████████████████████████████████| 436803/436803 [00:11<00:00, 36572.24it/s]\n" ] } ], "source": [ "new_tsv_data = []\n", "# list of list of tuples,\n", "# each index is the audio's codec spllit list (start, end)\n", "split_times = []\n", "for n_row, (audio_fn, audio_size) in tqdm.tqdm(\n", " enumerate(tsv_data), total=len(tsv_data)\n", "):\n", " new_splits = []\n", " for n in range(int(np.ceil(int(audio_size) / MAX_SIZE))):\n", " start_idx = n * MAX_SIZE\n", " end_idx = min(int(audio_size), (n + 1) * MAX_SIZE)\n", " if end_idx - start_idx < MIN_SIZE:\n", " continue\n", " out_audio_fn = f\"{n_row}_{n}.wav\"\n", " new_tsv_data.append((audio_fn, out_audio_fn, start_idx, end_idx))\n", " start_idx_embed = int(round(start_idx / SAMPLE_RATE * EMBEDDING_RATE))\n", " end_idx_embed = int(round(end_idx / SAMPLE_RATE * EMBEDDING_RATE))\n", " new_splits.append((start_idx_embed, end_idx_embed))\n", " split_times.append(new_splits)" ] }, { "cell_type": "markdown", "id": "4defb74e", "metadata": {}, "source": [ "## Slice the codebook labels" ] }, { "cell_type": "code", "execution_count": 8, "id": "a0b384f7", "metadata": { "ExecuteTime": { "end_time": "2023-09-06T17:17:06.351314Z", "start_time": "2023-09-06T17:17:06.347717Z" } }, "outputs": [ { "data": { "text/plain": [ "436803" ] }, "execution_count": 8, "metadata": {}, "output_type": "execute_result" } ], "source": [ "len(split_times)" ] }, { "cell_type": "code", "execution_count": 9, "id": "c7a427db", "metadata": { "ExecuteTime": { "end_time": "2023-09-06T17:17:06.401042Z", "start_time": "2023-09-06T17:17:06.352357Z" } }, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "[(0, 300), (300, 600), (600, 900), (900, 1200), (1200, 1500), (1500, 1800), (1800, 2100), (2100, 2400), (2400, 2700), (2700, 3000), (3000, 3300), (3300, 3437)]\n" ] } ], "source": [ "print(split_times[0])" ] }, { "cell_type": "code", "execution_count": null, "id": "9065a223", "metadata": { "ExecuteTime": { "start_time": "2023-09-07T04:20:15.949Z" } }, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "codebook: 0\n" ] } ], "source": [ "for codebook in range(N_CODEBOOKS):\n", " print(\"codebook:\", codebook)\n", " bunchsize = 100\n", " bunch = []\n", " with open(os.path.join(OUT_LABEL_DIR, f\"train.codec_{codebook}\"), \"w\") as f_out:\n", " with open(os.path.join(IN_LABEL_DIR, f\"train.codec_{codebook}\"), \"r\") as f:\n", " n_curr = 0\n", " for n_row, line in tqdm.tqdm(enumerate(f.readlines())):\n", " line = line.split()\n", " split_time_row = split_times[n_row]\n", " for (start_idx_embed, end_idx_embed) in split_time_row:\n", " bunch.append(\" \".join(line[start_idx_embed:end_idx_embed]) + \"\\n\")\n", " n_curr += 1\n", " if n_curr == bunchsize:\n", " n_curr = 0\n", " f_out.writelines(bunch)\n", " bunch = []\n", " # J also do the last one\n", " f_out.writelines(bunch)" ] }, { "cell_type": "code", "execution_count": 133, "id": "5e34ae67", "metadata": { "ExecuteTime": { "end_time": "2023-09-07T04:03:17.895825Z", "start_time": "2023-09-07T04:03:17.828079Z" } }, "outputs": [], "source": [ "for codebook in range(N_CODEBOOKS):\n", " with open(os.path.join(OUT_LABEL_DIR, f\"dict.codec_{codebook}.txt\"), \"w\") as f:\n", " for n in range(N_CODEBOOKS):\n", " f.write(f\"{n} 1\\n\")" ] }, { "cell_type": "code", "execution_count": 134, "id": "32fa51c5", "metadata": { "ExecuteTime": { "end_time": "2023-09-07T04:04:45.218614Z", "start_time": "2023-09-07T04:04:45.215554Z" } }, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "['0 1\\n', '1 1\\n', '2 1\\n', '3 1\\n', '4 1\\n', '5 1\\n', '6 1\\n', '7 1\\n']\n" ] } ], "source": [ "with open(os.path.join(OUT_LABEL_DIR, f\"dict.codec_{codebook}.txt\"), \"r\") as f:\n", " print(f.readlines())" ] }, { "cell_type": "markdown", "id": "58b4c431", "metadata": {}, "source": [ "## Write the train TSV" ] }, { "cell_type": "code", "execution_count": 13, "id": "d7231209", "metadata": { "ExecuteTime": { "end_time": "2023-09-06T17:42:46.340970Z", "start_time": "2023-09-06T17:42:43.897349Z" } }, "outputs": [], "source": [ "with open(os.path.join(OUT_TSV_DIR, f\"train.tsv\"), \"w\") as f:\n", " f.write(OUT_AUDIO_DIR + \"\\n\")\n", " f.write(\n", " \"\\n\".join(\n", " [\n", " fn + \"\\t\" + str(int(idx_end - idx_start))\n", " for _, fn, idx_start, idx_end in new_tsv_data\n", " ]\n", " )\n", " )" ] }, { "cell_type": "markdown", "id": "7a85ab3f", "metadata": {}, "source": [ "## Generate the split audios" ] }, { "cell_type": "code", "execution_count": 14, "id": "8aed8c10", "metadata": { "ExecuteTime": { "end_time": "2023-09-06T17:42:58.446167Z", "start_time": "2023-09-06T17:42:52.603704Z" } }, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "436803 work_items\n" ] } ], "source": [ "# write new audios to disk\n", "def _write_new_wav_files(work_item):\n", " from_fn, to_work_items = work_item\n", " _, audio_arr = wavfile.read(os.path.join(IN_AUDIO_DIR, from_fn))\n", " for to_fn, (start_idx, end_idx) in to_work_items:\n", " wavfile.write(\n", " os.path.join(OUT_AUDIO_DIR, to_fn),\n", " SAMPLE_RATE,\n", " audio_arr[start_idx:end_idx],\n", " )\n", "\n", "\n", "work_items = defaultdict(list)\n", "for from_fn, to_fn, start_idx, end_idx in new_tsv_data:\n", " work_items[from_fn].append((to_fn, (start_idx, end_idx)))\n", "work_items = [(k, v) for k, v in work_items.items()]\n", "print(len(work_items), \"work_items\")" ] }, { "cell_type": "code", "execution_count": 19, "id": "7a3b3f60", "metadata": { "ExecuteTime": { "end_time": "2023-09-06T18:09:37.251591Z", "start_time": "2023-09-06T18:09:37.248430Z" } }, "outputs": [ { "data": { "text/plain": [ "7085918" ] }, "execution_count": 19, "metadata": {}, "output_type": "execute_result" } ], "source": [ "len(new_tsv_data)" ] }, { "cell_type": "code", "execution_count": 21, "id": "5e941403", "metadata": { "ExecuteTime": { "end_time": "2023-09-06T18:10:31.486681Z", "start_time": "2023-09-06T18:10:15.743945Z" } }, "outputs": [], "source": [ "# import json\n", "# with open(\"train_origin.tsv\", \"w\") as fp:\n", "# json.dump(new_tsv_data, fp)" ] }, { "cell_type": "code", "execution_count": 18, "id": "9cab4abe", "metadata": { "ExecuteTime": { "end_time": "2023-09-06T18:09:32.086642Z", "start_time": "2023-09-06T18:09:32.084300Z" } }, "outputs": [], "source": [ "# _ = _apply_mp(\n", "# _write_new_wav_files,\n", "# work_items,\n", "# chunksize=640,\n", "# n_cores=32,\n", "# joblib_backend=\"threads\",\n", "# )" ] }, { "cell_type": "code", "execution_count": 12, "id": "738138b8", "metadata": { "ExecuteTime": { "end_time": "2023-09-06T17:42:30.677232Z", "start_time": "2023-09-06T17:42:30.625842Z" } }, "outputs": [], "source": [ "# codec_data = defaultdict(list)\n", "# for codebook in range(N_CODEBOOKS):\n", "# # print(i)\n", "# with open(os.path.join(IN_LABEL_DIR, f\"train.codec_{codebook}\"), \"r\") as f:\n", "# for line in f.read().strip().split(\"\\n\"):\n", "# if len(line.strip()) == 0:\n", "# continue\n", "# codec_data[codebook].append(line.strip().split())\n", "# # break\n", "# for codebook in range(N_CODEBOOKS):\n", "# with open(os.path.join(OUT_LABEL_DIR, f\"train.codec_{codebook}\"), \"w\") as f:\n", "# f.write(\"\\n\".join(new_codec_data[codebook]))" ] }, { "cell_type": "code", "execution_count": 88, "id": "7428b9fc", "metadata": { "ExecuteTime": { "end_time": "2023-09-07T02:31:34.064279Z", "start_time": "2023-09-07T02:24:03.889046Z" } }, "outputs": [], "source": [ "files = os.listdir(\"/app/suno/data/mert_25hz_short/audio\")" ] }, { "cell_type": "code", "execution_count": 89, "id": "4baa2a98", "metadata": { "ExecuteTime": { "end_time": "2023-09-07T02:31:34.069021Z", "start_time": "2023-09-07T02:31:34.066514Z" } }, "outputs": [ { "data": { "text/plain": [ "5603230" ] }, "execution_count": 89, "metadata": {}, "output_type": "execute_result" } ], "source": [ "len(files)" ] }, { "cell_type": "code", "execution_count": 90, "id": "e2b8431c", "metadata": { "ExecuteTime": { "end_time": "2023-09-07T02:31:34.324238Z", "start_time": "2023-09-07T02:31:34.070111Z" } }, "outputs": [], "source": [ "selected = [f for f in files if \"_\" in f]" ] }, { "cell_type": "code", "execution_count": 113, "id": "5e5eb810", "metadata": { "ExecuteTime": { "end_time": "2023-09-07T02:53:07.574317Z", "start_time": "2023-09-07T02:53:07.571549Z" } }, "outputs": [ { "data": { "text/plain": [ "(5229011, 7085918)" ] }, "execution_count": 113, "metadata": {}, "output_type": "execute_result" } ], "source": [ "len(selected), len(new_tsv_data)" ] }, { "cell_type": "code", "execution_count": 117, "id": "3f38d6d5", "metadata": { "ExecuteTime": { "end_time": "2023-09-07T02:54:28.014439Z", "start_time": "2023-09-07T02:54:28.011765Z" } }, "outputs": [ { "data": { "text/plain": [ "('1.wav', '0_4.wav', 1152000, 1440000)" ] }, "execution_count": 117, "metadata": {}, "output_type": "execute_result" } ], "source": [ "new_tsv_data[4]" ] }, { "cell_type": "code", "execution_count": 118, "id": "ad7d2b4e", "metadata": { "ExecuteTime": { "end_time": "2023-09-07T02:55:04.584294Z", "start_time": "2023-09-07T02:55:03.087804Z" } }, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "436803\n" ] } ], "source": [ "must_unique_f = set(int(i[0].split(\".wav\")[0]) for i in new_tsv_data)\n", "print(len(must_unique_f))" ] }, { "cell_type": "code", "execution_count": 131, "id": "ea9c7030", "metadata": { "ExecuteTime": { "end_time": "2023-09-07T03:49:41.953599Z", "start_time": "2023-09-07T03:49:41.939992Z" } }, "outputs": [ { "data": { "text/plain": [ "(1, 463758)" ] }, "execution_count": 131, "metadata": {}, "output_type": "execute_result" } ], "source": [ "min(must_unique_f), max(must_unique_f)" ] }, { "cell_type": "code", "execution_count": 120, "id": "9d88ad2f", "metadata": { "ExecuteTime": { "end_time": "2023-09-07T02:55:24.369334Z", "start_time": "2023-09-07T02:55:22.836407Z" } }, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "436803\n" ] } ], "source": [ "must_unique_f_1 = set(int(i[1].split(\"_\")[0]) for i in new_tsv_data)\n", "print(len(must_unique_f_1))" ] }, { "cell_type": "code", "execution_count": 121, "id": "0859be0f", "metadata": { "ExecuteTime": { "end_time": "2023-09-07T02:57:25.211055Z", "start_time": "2023-09-07T02:57:25.208342Z" } }, "outputs": [ { "data": { "text/plain": [ "(5229011, 7085918)" ] }, "execution_count": 121, "metadata": {}, "output_type": "execute_result" } ], "source": [ "len(selected), len(new_tsv_data)" ] }, { "cell_type": "code", "execution_count": 122, "id": "ceb1c434", "metadata": { "ExecuteTime": { "end_time": "2023-09-07T02:57:39.399387Z", "start_time": "2023-09-07T02:57:38.296763Z" } }, "outputs": [], "source": [ "unique_f = set(int(f.split(\"_\")[0]) for f in selected)" ] }, { "cell_type": "code", "execution_count": 123, "id": "8b093aed", "metadata": { "ExecuteTime": { "end_time": "2023-09-07T02:57:39.403453Z", "start_time": "2023-09-07T02:57:39.401030Z" } }, "outputs": [ { "data": { "text/plain": [ "337756" ] }, "execution_count": 123, "metadata": {}, "output_type": "execute_result" } ], "source": [ "len(unique_f)" ] }, { "cell_type": "code", "execution_count": 107, "id": "67dab7b1", "metadata": { "ExecuteTime": { "end_time": "2023-09-07T02:46:11.437039Z", "start_time": "2023-09-07T02:46:11.391474Z" } }, "outputs": [], "source": [] }, { "cell_type": "code", "execution_count": 132, "id": "8ee0c8f2", "metadata": { "ExecuteTime": { "end_time": "2023-09-07T03:50:41.738230Z", "start_time": "2023-09-07T03:50:41.680965Z" } }, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "112756 184900\n", "247918 274821\n", "162244 500000\n" ] } ], "source": [ "c = 0\n", "s = None\n", "for i in range(500000):\n", " if i not in unique_f:\n", " c += 1\n", " if s is None:\n", " s = i\n", " else:\n", " continue\n", " if i in unique_f:\n", " if s is not None:\n", " print(s, i)\n", " s = None\n", " # print(i)\n", "print(c, c + len(unique_f))" ] }, { "cell_type": "code", "execution_count": 49, "id": "4541b7ba", "metadata": { "ExecuteTime": { "end_time": "2023-09-06T21:59:37.014277Z", "start_time": "2023-09-06T21:59:37.012201Z" } }, "outputs": [ { "data": { "text/plain": [ "112756" ] }, "execution_count": 49, "metadata": {}, "output_type": "execute_result" } ], "source": [ "len(unique_f)" ] }, { "cell_type": "code", "execution_count": 50, "id": "53769df5", "metadata": { "ExecuteTime": { "end_time": "2023-09-06T21:59:37.236355Z", "start_time": "2023-09-06T21:59:37.232184Z" } }, "outputs": [ { "data": { "text/plain": [ "[112736,\n", " 112737,\n", " 112738,\n", " 112739,\n", " 112740,\n", " 112741,\n", " 112742,\n", " 112743,\n", " 112744,\n", " 112745,\n", " 112746,\n", " 112747,\n", " 112748,\n", " 112749,\n", " 112750,\n", " 112751,\n", " 112752,\n", " 112753,\n", " 112754,\n", " 112755]" ] }, "execution_count": 50, "metadata": {}, "output_type": "execute_result" } ], "source": [ "list(sorted(unique_f))[-20:]" ] }, { "cell_type": "code", "execution_count": 51, "id": "2f7dd898", "metadata": { "ExecuteTime": { "end_time": "2023-09-07T00:51:23.568034Z", "start_time": "2023-09-07T00:51:23.564721Z" } }, "outputs": [ { "data": { "text/plain": [ "('1.wav', '0_0.wav', 0, 288000)" ] }, "execution_count": 51, "metadata": {}, "output_type": "execute_result" } ], "source": [ "new_tsv_data[0]" ] }, { "cell_type": "code", "execution_count": 52, "id": "7b26ee0b", "metadata": { "ExecuteTime": { "end_time": "2023-09-07T00:51:54.245019Z", "start_time": "2023-09-07T00:51:53.512686Z" } }, "outputs": [], "source": [ "new_wav = wavfile.read(os.path.join(OUT_AUDIO_DIR, '0_0.wav'))" ] }, { "cell_type": "code", "execution_count": 64, "id": "b9e62de4", "metadata": { "ExecuteTime": { "end_time": "2023-09-07T00:54:22.565356Z", "start_time": "2023-09-07T00:54:22.562256Z" } }, "outputs": [ { "data": { "text/plain": [ "(288000, 1)" ] }, "execution_count": 64, "metadata": {}, "output_type": "execute_result" } ], "source": [ "new_wav[1].reshape(-1, 1).shape" ] }, { "cell_type": "code", "execution_count": 55, "id": "3d73fab0", "metadata": { "ExecuteTime": { "end_time": "2023-09-07T00:52:29.427411Z", "start_time": "2023-09-07T00:52:27.396720Z" } }, "outputs": [], "source": [ "from nnAudio import features as nnAudioFeatures" ] }, { "cell_type": "code", "execution_count": 60, "id": "1722728f", "metadata": { "ExecuteTime": { "end_time": "2023-09-07T00:53:28.148373Z", "start_time": "2023-09-07T00:53:27.640337Z" } }, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "CQT kernels created, time used = 0.5050 seconds\n" ] } ], "source": [ "spec_layer = nnAudioFeatures.cqt.CQT(sr=new_wav[0], hop_length=new_wav[0]//EMBEDDING_RATE, fmin=32.7, \n", " fmax=None, n_bins=336, bins_per_octave=336//7, \n", " filter_scale=1, norm=1, window='hann', center=True, \n", " pad_mode='constant', trainable=False, \n", " output_format='Magnitude', verbose=True)" ] }, { "cell_type": "code", "execution_count": 75, "id": "8fcbc412", "metadata": { "ExecuteTime": { "end_time": "2023-09-07T01:00:35.908643Z", "start_time": "2023-09-07T01:00:35.904899Z" } }, "outputs": [ { "data": { "text/plain": [ "torch.Size([288000])" ] }, "execution_count": 75, "metadata": {}, "output_type": "execute_result" } ], "source": [ "input_tensor = torch.from_numpy(new_wav[1]).float()\n", "input_tensor.shape" ] }, { "cell_type": "code", "execution_count": 76, "id": "85f56f93", "metadata": { "ExecuteTime": { "end_time": "2023-09-07T01:00:37.935224Z", "start_time": "2023-09-07T01:00:37.932492Z" } }, "outputs": [ { "data": { "text/plain": [ "300" ] }, "execution_count": 76, "metadata": {}, "output_type": "execute_result" } ], "source": [ "288000 // new_wav[0] * EMBEDDING_RATE" ] }, { "cell_type": "code", "execution_count": 77, "id": "6ad79910", "metadata": { "ExecuteTime": { "end_time": "2023-09-07T01:00:38.572993Z", "start_time": "2023-09-07T01:00:38.459900Z" } }, "outputs": [], "source": [ "output_tensor = spec_layer(input_tensor)" ] }, { "cell_type": "code", "execution_count": 78, "id": "8e0f8508", "metadata": { "ExecuteTime": { "end_time": "2023-09-07T01:00:39.115600Z", "start_time": "2023-09-07T01:00:39.112815Z" } }, "outputs": [ { "data": { "text/plain": [ "torch.Size([1, 336, 301])" ] }, "execution_count": 78, "metadata": {}, "output_type": "execute_result" } ], "source": [ "output_tensor.shape" ] }, { "cell_type": "code", "execution_count": 79, "id": "2b405932", "metadata": { "ExecuteTime": { "end_time": "2023-09-07T01:00:43.598526Z", "start_time": "2023-09-07T01:00:43.540744Z" } }, "outputs": [ { "data": { "text/plain": [ "tensor([[[ 49.1279, 52.6179, 55.2570, ..., 2552.8076, 1880.0348,\n", " 1353.1643],\n", " [ 52.0929, 58.9065, 66.4197, ..., 1992.2830, 1287.6638,\n", " 714.6923],\n", " [ 51.8104, 59.5420, 68.0058, ..., 2252.4263, 1565.3777,\n", " 1002.5173],\n", " ...,\n", " [ 0.0000, 0.0000, 0.0000, ..., 330.8419, 308.4335,\n", " 925.4127],\n", " [ 0.0000, 0.0000, 0.0000, ..., 837.1494, 576.7127,\n", " 765.4628],\n", " [ 0.0000, 0.0000, 0.0000, ..., 868.1017, 926.3876,\n", " 564.1667]]])" ] }, "execution_count": 79, "metadata": {}, "output_type": "execute_result" } ], "source": [ "output_tensor" ] }, { "cell_type": "code", "execution_count": 80, "id": "75d4bfab", "metadata": { "ExecuteTime": { "end_time": "2023-09-07T01:01:44.413557Z", "start_time": "2023-09-07T01:01:44.349371Z" } }, "outputs": [], "source": [ "output_tensor_half = spec_layer(input_tensor[:144000])" ] }, { "cell_type": "code", "execution_count": 81, "id": "36a04b60", "metadata": { "ExecuteTime": { "end_time": "2023-09-07T01:01:46.820207Z", "start_time": "2023-09-07T01:01:46.816609Z" } }, "outputs": [ { "data": { "text/plain": [ "tensor([[[ 49.1279, 52.6179, 55.2570, ..., 149.2399, 147.4749, 144.6790],\n", " [ 52.0929, 58.9065, 66.4197, ..., 125.9033, 125.5422, 124.2411],\n", " [ 51.8104, 59.5420, 68.0058, ..., 95.9982, 95.1028, 93.6845],\n", " ...,\n", " [ 0.0000, 0.0000, 0.0000, ..., 45.0109, 18.7750, 76.7034],\n", " [ 0.0000, 0.0000, 0.0000, ..., 26.5843, 46.0399, 61.8331],\n", " [ 0.0000, 0.0000, 0.0000, ..., 6.8963, 67.4006, 36.9114]]])" ] }, "execution_count": 81, "metadata": {}, "output_type": "execute_result" } ], "source": [ "output_tensor_half" ] }, { "cell_type": "code", "execution_count": 86, "id": "7a576cfb", "metadata": { "ExecuteTime": { "end_time": "2023-09-07T01:05:44.769568Z", "start_time": "2023-09-07T01:05:44.764225Z" } }, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "0 True\n", "1 True\n", "2 True\n", "3 True\n", "4 True\n", "5 True\n", "6 True\n", "7 True\n", "8 True\n", "9 True\n", "10 True\n", "11 True\n", "12 True\n", "13 True\n", "14 True\n", "15 True\n", "16 True\n", "17 True\n", "18 True\n", "19 True\n", "20 True\n", "21 True\n", "22 True\n", "23 True\n", "24 True\n", "25 True\n", "26 True\n", "27 True\n", "28 True\n", "29 True\n", "30 True\n", "31 True\n", "32 True\n", "33 True\n", "34 True\n", "35 True\n", "36 True\n", "37 True\n", "38 True\n", "39 True\n", "40 True\n", "41 True\n", "42 True\n", "43 True\n", "44 True\n", "45 True\n", "46 True\n", "47 True\n", "48 True\n", "49 True\n", "50 True\n", "51 True\n", "52 True\n", "53 True\n", "54 True\n", "55 True\n", "56 True\n", "57 True\n", "58 True\n", "59 True\n", "60 True\n", "61 True\n", "62 True\n", "63 True\n", "64 True\n", "65 True\n", "66 True\n", "67 True\n", "68 True\n", "69 True\n", "70 True\n", "71 True\n", "72 True\n", "73 True\n", "74 True\n", "75 True\n", "76 True\n", "77 True\n", "78 True\n", "79 True\n", "80 True\n", "81 True\n", "82 True\n", "83 True\n", "84 True\n", "85 True\n", "86 True\n", "87 True\n", "88 True\n", "89 True\n", "90 True\n", "91 True\n", "92 True\n", "93 True\n", "94 True\n", "95 True\n", "96 True\n", "97 True\n", "98 True\n", "99 True\n", "100 True\n", "101 True\n", "102 True\n", "103 True\n", "104 True\n", "105 True\n", "106 True\n", "107 True\n", "108 True\n", "109 True\n", "110 True\n", "111 True\n", "112 True\n", "113 True\n", "114 True\n", "115 True\n", "116 True\n", "117 True\n", "118 True\n", "119 True\n", "120 True\n", "121 True\n", "122 True\n", "123 True\n", "124 False\n", "125 False\n", "126 False\n", "127 False\n", "128 False\n", "129 False\n", "130 False\n", "131 False\n", "132 False\n", "133 False\n", "134 False\n", "135 False\n", "136 False\n", "137 False\n", "138 False\n", "139 False\n", "140 False\n", "141 False\n", "142 False\n", "143 False\n", "144 False\n", "145 False\n", "146 False\n", "147 False\n", "148 False\n", "149 False\n", "150 False\n" ] } ], "source": [ "for i in range(151):\n", " print(i, torch.equal(output_tensor_half[:, :, i], output_tensor[:, :, i]))" ] }, { "cell_type": "code", "execution_count": 87, "id": "91ca160e", "metadata": { "ExecuteTime": { "end_time": "2023-09-07T02:04:23.054565Z", "start_time": "2023-09-07T02:04:23.051847Z" } }, "outputs": [ { "data": { "text/plain": [ "7085918" ] }, "execution_count": 87, "metadata": {}, "output_type": "execute_result" } ], "source": [ "len(new_tsv_data)" ] }, { "cell_type": "code", "execution_count": null, "id": "e319cb90", "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": "3.10.12" }, "toc": { "base_numbering": 1, "nav_menu": {}, "number_sections": true, "sideBar": true, "skip_h1_title": false, "title_cell": "Table of Contents", "title_sidebar": "Contents", "toc_cell": false, "toc_position": {}, "toc_section_display": true, "toc_window_display": false } }, "nbformat": 4, "nbformat_minor": 5 }