File size: 5,506 Bytes
c6919c4 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 |
{
"cells": [
{
"cell_type": "markdown",
"id": "90641144",
"metadata": {},
"source": [
"# Bark Memory Profiling\n",
"Bark has two ways to reduce GPU memory: \n",
" - Small models: a smaller version of the model. This can be set by using the environment variable `SUNO_USE_SMALL_MODELS`\n",
" - offloading models to CPU: Holding only one model at a time on the GPU, and shuttling the models to the CPU in between generations. \n",
"\n",
"## NOTE: this requires a GPU to run\n",
"\n",
"# $ \\\\ $\n",
"## First, we'll use the most memory efficient configuration"
]
},
{
"cell_type": "code",
"execution_count": 1,
"id": "39ea4bed",
"metadata": {},
"outputs": [],
"source": [
"import os\n",
"\n",
"os.environ[\"CUDA_VISIBLE_DEVICES\"] = \"0\"\n",
"os.environ[\"SUNO_USE_SMALL_MODELS\"] = \"1\"\n",
"os.environ[\"SUNO_OFFLOAD_CPU\"] = \"1\"\n",
"\n",
"from bark.generation import preload_models\n",
"from bark import generate_audio, SAMPLE_RATE\n",
"\n",
"import torch"
]
},
{
"cell_type": "code",
"execution_count": 2,
"id": "66b0c006",
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"100%|██████████████████████████████████████████████████████████████████████| 100/100 [00:01<00:00, 62.17it/s]\n",
"100%|████████████████████████████████████████████████████████████████████████| 10/10 [00:03<00:00, 2.74it/s]\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"max memory usage = 2396MB\n"
]
}
],
"source": [
"torch.cuda.reset_peak_memory_stats()\n",
"preload_models()\n",
"audio_array = generate_audio(\"madam I'm adam\", history_prompt=\"v2/en_speaker_5\")\n",
"max_utilization = torch.cuda.max_memory_allocated()\n",
"print(f\"max memory usage = {max_utilization / 1024 / 1024:.0f}MB\")"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "9922dd2d",
"metadata": {},
"outputs": [],
"source": []
},
{
"cell_type": "code",
"execution_count": null,
"id": "bdbe578e",
"metadata": {},
"outputs": [],
"source": []
},
{
"cell_type": "markdown",
"id": "213d1b5b",
"metadata": {},
"source": [
"# Memory Profiling:\n",
"We can profile the memory consumption of 4 scenarios\n",
" - Small models, offloading to CPU\n",
" - Large models, offloading to CPU\n",
" - Small models, not offloading to CPU\n",
" - Large models, not offloading to CPU"
]
},
{
"cell_type": "code",
"execution_count": 1,
"id": "417d5e9c",
"metadata": {},
"outputs": [],
"source": [
"import os\n",
"from bark.generation import preload_models\n",
"from bark import generate_audio, SAMPLE_RATE\n",
"import torch\n",
"import time"
]
},
{
"cell_type": "markdown",
"id": "f4d19d60",
"metadata": {},
"source": []
},
{
"cell_type": "code",
"execution_count": 2,
"id": "cd83b45d",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Small models True, offloading to CPU: True\n",
"\tmax memory usage = 2949MB, time 3s\n",
"\n",
"Small models False, offloading to CPU: True\n",
"\tmax memory usage = 7826MB, time 4s\n",
"\n",
"Small models True, offloading to CPU: False\n",
"\tmax memory usage = 5504MB, time 2s\n",
"\n",
"Small models False, offloading to CPU: False\n",
"\tmax memory usage = 7825MB, time 5s\n",
"\n"
]
}
],
"source": [
"offload_models = True\n",
"use_small_models = True\n",
"\n",
"for offload_models in (True, False):\n",
" for use_small_models in (True, False):\n",
" torch.cuda.reset_peak_memory_stats()\n",
" preload_models(\n",
" text_use_small=use_small_models,\n",
" coarse_use_small=use_small_models,\n",
" fine_use_small=use_small_models,\n",
" force_reload=True,\n",
" )\n",
" t0 = time.time()\n",
" audio_array = generate_audio(\"madam I'm adam\", history_prompt=\"v2/en_speaker_5\", silent=True)\n",
" dur = time.time() - t0\n",
" max_utilization = torch.cuda.max_memory_allocated()\n",
" print(f\"Small models {use_small_models}, offloading to CPU: {offload_models}\")\n",
" print(f\"\\tmax memory usage = {max_utilization / 1024 / 1024:.0f}MB, time {dur:.0f}s\\n\")"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "bfe5fa06",
"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": "3.9.16"
}
},
"nbformat": 4,
"nbformat_minor": 5
}
|