{ "cells": [ { "cell_type": "code", "execution_count": 1, "id": "7c02ae76-b540-4b16-9492-e9210b3b9249", "metadata": {}, "outputs": [], "source": [ "import os\n", "os.environ['CUDA_VISIBLE_DEVICE'] = ''\n", "import random\n", "\n", "%matplotlib inline\n", "import matplotlib.pyplot as plt\n", "\n", "import numpy as np\n", "from omegaconf import OmegaConf\n", "import torch\n", "%load_ext autoreload\n", "%autoreload 2\n", "\n", "from importlib.util import find_spec\n", "if find_spec(\"text_recognizer\") is None:\n", " import sys\n", " sys.path.append('..')" ] }, { "cell_type": "code", "execution_count": 2, "id": "ccdb6dde-47e5-429a-88f2-0764fb7e259a", "metadata": {}, "outputs": [], "source": [ "from hydra import compose, initialize\n", "from omegaconf import OmegaConf\n", "from hydra.utils import instantiate" ] }, { "cell_type": "code", "execution_count": 15, "id": "3cf50475-39f2-4642-a7d1-5bcbc0a036f7", "metadata": {}, "outputs": [], "source": [ "path = \"../training/conf/network/mammut_lines.yaml\"" ] }, { "cell_type": "code", "execution_count": 45, "id": "e52ecb01-c975-4e55-925d-1182c7aea473", "metadata": {}, "outputs": [], "source": [ "with open(path, \"rb\") as f:\n", " cfg = OmegaConf.load(f)" ] }, { "cell_type": "code", "execution_count": 46, "id": "f939aa37-7b1d-45cc-885c-323c4540bda1", "metadata": {}, "outputs": [ { "data": { "text/plain": [ "{'_target_': 'text_recognizer.network.mammut.MaMMUT', 'encoder': {'_target_': 'text_recognizer.network.vit.Vit', 'image_height': 56, 'image_width': 1024, 'patch_height': 56, 'patch_width': 8, 'dim': 512, 'encoder': {'_target_': 'text_recognizer.network.transformer.encoder.Encoder', 'dim': 512, 'heads': 12, 'dim_head': 64, 'ff_mult': 4, 'depth': 4, 'dropout_rate': 0.1}, 'channels': 1}, 'image_attn_pool': {'_target_': 'text_recognizer.network.transformer.attention.Attention', 'dim': 512, 'heads': 8, 'causal': False, 'dim_head': 64, 'ff_mult': 4, 'dropout_rate': 0.0, 'use_flash': True, 'norm_context': True, 'rotary_emb': None}, 'decoder': {'_target_': 'text_recognizer.network.transformer.decoder.Decoder', 'dim': 512, 'ff_mult': 4, 'heads': 12, 'dim_head': 64, 'depth': 6, 'dropout_rate': 0.1}, 'dim': 512, 'dim_latent': 512, 'num_tokens': 58, 'pad_index': 3, 'num_image_queries': 256}" ] }, "execution_count": 46, "metadata": {}, "output_type": "execute_result" } ], "source": [ "cfg" ] }, { "cell_type": "code", "execution_count": 47, "id": "aaeab329-aeb0-4a1b-aa35-5a2aab81b1d0", "metadata": {}, "outputs": [], "source": [ "net = instantiate(cfg)" ] }, { "cell_type": "code", "execution_count": 39, "id": "618b997c-e6a6-4487-b70c-9d260cb556d3", "metadata": {}, "outputs": [], "source": [ "from torchinfo import summary" ] }, { "cell_type": "code", "execution_count": 50, "id": "7daf1f49", "metadata": {}, "outputs": [ { "data": { "text/plain": [ "====================================================================================================\n", "Layer (type:depth-idx) Output Shape Param #\n", "====================================================================================================\n", "MaMMUT [1, 89, 58] 627,712\n", "├─Vit: 1-1 [1, 128, 512] --\n", "│ └─Sequential: 2-1 [1, 128, 512] --\n", "│ │ └─Rearrange: 3-1 [1, 128, 448] --\n", "│ │ └─LayerNorm: 3-2 [1, 128, 448] 896\n", "│ │ └─Linear: 3-3 [1, 128, 512] 229,888\n", "│ │ └─LayerNorm: 3-4 [1, 128, 512] 1,024\n", "│ └─Encoder: 2-2 [1, 128, 512] --\n", "│ │ └─ModuleList: 3-5 -- --\n", "│ │ │ └─Attention: 4-1 [1, 128, 512] 4,724,224\n", "│ │ │ └─Attention: 4-2 [1, 128, 512] 4,724,224\n", "│ │ │ └─Attention: 4-3 [1, 128, 512] 4,724,224\n", "│ │ │ └─Attention: 4-4 [1, 128, 512] 4,724,224\n", "│ │ └─LayerNorm: 3-6 [1, 128, 512] 1,024\n", "├─Attention: 1-2 [1, 257, 512] --\n", "│ └─LayerNorm: 2-3 [1, 257, 512] 1,024\n", "│ └─Linear: 2-4 [1, 257, 512] 262,144\n", "│ └─LayerNorm: 2-5 [1, 128, 512] 1,024\n", "│ └─Linear: 2-6 [1, 128, 1024] 524,288\n", "│ └─Attend: 2-7 [1, 8, 257, 64] --\n", "│ └─Linear: 2-8 [1, 257, 512] 262,144\n", "│ └─Sequential: 2-9 [1, 257, 512] --\n", "│ │ └─Linear: 3-7 [1, 257, 4096] 2,101,248\n", "│ │ └─SwiGLU: 3-8 [1, 257, 2048] --\n", "│ │ └─Linear: 3-9 [1, 257, 512] 1,049,088\n", "├─LayerNorm: 1-3 [1, 257, 512] 1,024\n", "├─Embedding: 1-4 [1, 89, 512] 29,696\n", "├─Decoder: 1-5 [1, 89, 512] --\n", "│ └─ModuleList: 2-10 -- --\n", "│ │ └─ModuleList: 3-10 -- --\n", "│ │ │ └─Attention: 4-5 [1, 89, 512] 4,724,224\n", "│ │ │ └─Attention: 4-6 [1, 89, 512] 4,724,224\n", "│ │ └─ModuleList: 3-11 -- --\n", "│ │ │ └─Attention: 4-7 [1, 89, 512] 4,724,224\n", "│ │ │ └─Attention: 4-8 [1, 89, 512] 4,724,224\n", "│ │ └─ModuleList: 3-12 -- --\n", "│ │ │ └─Attention: 4-9 [1, 89, 512] 4,724,224\n", "│ │ │ └─Attention: 4-10 [1, 89, 512] 4,724,224\n", "│ │ └─ModuleList: 3-13 -- --\n", "│ │ │ └─Attention: 4-11 [1, 89, 512] 4,724,224\n", "│ │ │ └─Attention: 4-12 [1, 89, 512] 4,724,224\n", "│ │ └─ModuleList: 3-14 -- --\n", "│ │ │ └─Attention: 4-13 [1, 89, 512] 4,724,224\n", "│ │ │ └─Attention: 4-14 [1, 89, 512] 4,724,224\n", "│ │ └─ModuleList: 3-15 -- --\n", "│ │ │ └─Attention: 4-15 [1, 89, 512] 4,724,224\n", "│ │ │ └─Attention: 4-16 [1, 89, 512] 4,724,224\n", "│ └─LayerNorm: 2-11 [1, 89, 512] 1,024\n", "├─Sequential: 1-6 [1, 89, 58] --\n", "│ └─LayerNorm: 2-12 [1, 89, 512] 1,024\n", "│ └─Linear: 2-13 [1, 89, 58] 29,696\n", "====================================================================================================\n", "Total params: 80,711,552\n", "Trainable params: 80,711,552\n", "Non-trainable params: 0\n", "Total mult-adds (M): 80.08\n", "====================================================================================================\n", "Input size (MB): 0.23\n", "Forward/backward pass size (MB): 131.05\n", "Params size (MB): 320.34\n", "Estimated Total Size (MB): 451.61\n", "====================================================================================================" ] }, "execution_count": 50, "metadata": {}, "output_type": "execute_result" } ], "source": [ "summary(net, ((1, 1, 56, 1024), (1, 89)), device=\"cpu\", depth=4)" ] }, { "cell_type": "code", "execution_count": 48, "id": "166bf656-aba6-4654-a530-dfce12666297", "metadata": {}, "outputs": [], "source": [ "t = net(torch.randn(1, 1, 56, 1024), torch.randint(1, 4, (1, 4)))" ] }, { "cell_type": "code", "execution_count": 49, "id": "43d9af25-9872-497d-8796-4835a65262ed", "metadata": { "tags": [] }, "outputs": [ { "data": { "text/plain": [ "torch.Size([1, 4, 58])" ] }, "execution_count": 49, "metadata": {}, "output_type": "execute_result" } ], "source": [ "t.shape" ] }, { "cell_type": "code", "execution_count": null, "id": "63ac7f1b-0eb1-4625-96b8-467846eb7ae6", "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": "3.9.4" } }, "nbformat": 4, "nbformat_minor": 5 }