simple-shapes-dataset-inference.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "396f5519-ed15-4b67-87dd-0d9d57e6f7a6",
   "metadata": {},
   "source": [
    "# Vision to Text Translations using a Global Workspace\n",
    "Benjamin Devillers\n",
    "\n",
    "[![Open in Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/ruflab/shimmer-tutorials/blob/main/simple-shapes-dataset-inference.ipynb)\n",
    "\n",
    "\n",
    "In this notebook, we will see how to use `shimmer` and our pretrained models. We will use our pretrained model to translate visual images of shapes from the [simple-shapes-datset](https://github.com/ruflab/simple-shapes-dataset) to their text descriptions."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8450ae66-426c-4543-b723-de6dbc904591",
   "metadata": {},
   "source": [
    "For this tutorial, we will need to install the [shimmer-ssd](https://github.com/ruflab/shimmer-ssd) package."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "20c0875a-689c-43ab-9cfc-381bde9ed73a",
   "metadata": {},
   "outputs": [],
   "source": [
    "!pip install \"git+https://github.com/ruflab/shimmer-ssd.git\""
   ]
  },
  {
   "cell_type": "markdown",
   "id": "46ab231b-f1bc-4477-8f4a-c490a8c773f5",
   "metadata": {},
   "source": [
    "This package depends on [simple-shapes-dataset](https://github.com/ruflab/simple-shapes-dataset) and provides all of its commands. You can then use all of its commands.\n",
    "\n",
    "For instance, we can download the dataset directly with:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "3fdb5133-2fb0-4a2d-9790-753bbf8e144e",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Save labels...\n",
      "Saving training set...\n",
      "100%|███████████████████████████████████████████| 32/32 [00:00<00:00, 38.69it/s]\n",
      "Saving validation set...\n",
      "100%|███████████████████████████████████████████| 32/32 [00:00<00:00, 56.60it/s]\n",
      "Saving test set...\n",
      "100%|█████████████████████████████████████████████| 1/1 [00:00<00:00, 58.35it/s]\n",
      "Saving captions...\n",
      "100%|█████████████████████████████████████████| 32/32 [00:00<00:00, 3837.64it/s]\n",
      "/home/bdevillers/.cache/pypoetry/virtualenvs/shimmer-tutorials-RRvxJ_Ue-py3.11/lib/python3.11/site-packages/transformers/tokenization_utils_base.py:1601: FutureWarning: `clean_up_tokenization_spaces` was not set. It will be set to `True` by default. This behavior will be depracted in transformers v4.45, and will be then set to `False` by default. For more details check this issue: https://github.com/huggingface/transformers/issues/31884\n",
      "  warnings.warn(\n",
      "100%|█████████████████████████████████████████████| 1/1 [00:00<00:00,  1.59it/s]\n",
      "100%|█████████████████████████████████████████| 32/32 [00:00<00:00, 5156.67it/s]\n",
      "100%|█████████████████████████████████████████████| 1/1 [00:00<00:00, 12.90it/s]\n",
      "100%|███████████████████████████████████████████| 1/1 [00:00<00:00, 2321.14it/s]\n",
      "100%|█████████████████████████████████████████████| 1/1 [00:00<00:00, 31.77it/s]\n"
     ]
    }
   ],
   "source": [
    "!shapesd create --output_path \"simple_shapes_dataset\" --ntrain 32 --nval 32 --ntest 1"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "05d1a776-cfb7-4da2-b49d-74bec9d264fd",
   "metadata": {},
   "source": [
    "Note that `shapesd download` automatically migrates the dataset so that it is correctly formatted. If you downloaded the dataset manually, use `shapesd migrate -p PATH_TO_DATASET` to migrate manually."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9e48fcb6-43e4-427d-a63d-e8c4000a20a7",
   "metadata": {},
   "source": [
    "Our evaluation will also require the simple-shapes dataset tokenizer data, so let's download this from the `shimmer-ssd` repository:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "3a90f345-7b3d-4d85-8ea4-59682ad935d6",
   "metadata": {
    "editable": true,
    "slideshow": {
     "slide_type": ""
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "!ssd download tokenizer"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1c8ac2ba-d66a-4f62-90e7-971c71aadb2a",
   "metadata": {},
   "source": [
    "Let's now import what we will need in this notebook."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "75a2017e-7de7-4568-87dd-23bf4f93a0cc",
   "metadata": {
    "editable": true,
    "slideshow": {
     "slide_type": ""
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "from collections.abc import Mapping\n",
    "from pathlib import Path\n",
    "from typing import cast\n",
    "\n",
    "import matplotlib\n",
    "import matplotlib.pyplot as plt\n",
    "import torch\n",
    "from lightning.pytorch import seed_everything\n",
    "from shimmer.modules.global_workspace import GlobalWorkspaceFusion\n",
    "from shimmer.modules.selection import SelectionBase\n",
    "from shimmer_ssd import DEBUG_MODE, PROJECT_DIR\n",
    "from shimmer_ssd.config import DomainModuleVariant, LoadedDomainConfig, load_config\n",
    "from shimmer_ssd.dataset.pre_process import TokenizeCaptions\n",
    "from shimmer_ssd.logging import batch_to_device\n",
    "from shimmer_ssd.modules.domains import load_pretrained_domains\n",
    "from shimmer_ssd.modules.domains.visual import VisualLatentDomainModule\n",
    "from tokenizers.implementations.byte_level_bpe import ByteLevelBPETokenizer\n",
    "from torchvision.utils import make_grid\n",
    "\n",
    "from simple_shapes_dataset import SimpleShapesDataModule, get_default_domains\n",
    "\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "dff5b171-42c0-426f-96f0-2d258ef33dca",
   "metadata": {},
   "source": [
    "## Config\n",
    "\n",
    "Let's first generate the config folder for the rest of the scripts.\n",
    "This will create a `config` folder with different yaml files used by the different scripts and in the notebook."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "6de5cbe7-d857-4baa-886a-74484200eedc",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Config folder already exists. Skipping.\n"
     ]
    }
   ],
   "source": [
    "!ssd config create"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fa2daeff-21a4-4ed7-b02b-bf15bbd282d4",
   "metadata": {},
   "source": [
    "This will create a `config` folder. This contains many file, but in this tutorial, only `main.yaml` will interest us.\n",
    "\n",
    "You can start by taking a look at the default values which should be mostly set correctly for this tutorial. But you can try and make some changes to see the outcome.\n",
    "\n",
    "<div class=\"alert alert-info\">\n",
    "Anytime you make a change to the config, don't forget to reload it with the following cell!\n",
    "</div>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "6fb97e2b-3ed9-45c5-a10c-c475c80f9502",
   "metadata": {},
   "outputs": [],
   "source": [
    "# We don't use cli in the notebook, but consider using it in normal scripts.\n",
    "config = load_config(\"./config\", use_cli=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "79ab9ee9-84f8-4de9-b7d6-fdba50744cc8",
   "metadata": {},
   "source": [
    "## Make the dataloader\n",
    "\n",
    "Thanks to `shapesd download`, we should now have our dataset in the folder \"simple_shapes_dataset\". This should be the path set in `main.yaml`:\n",
    "```yaml\n",
    "dataset:\n",
    "    path: \"./simple_shapes_dataset\"\n",
    "```\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0d37d4d8-d2c5-481e-a947-5f1dfc253dfc",
   "metadata": {},
   "source": [
    "We now need to define which modality we will load, and in what proportion. Open `config/main.yaml` and update `domain_proportions` with selected values:\n",
    "\n",
    "```yaml\n",
    "domain_proportions: \n",
    "    -   domains: [\"v\"]  # unimodal visual passes use 100% of the available data \n",
    "        proportion: 1.0\n",
    "    -   domains: [\"t\"]  # unimodal text passes use 100% of the available data\n",
    "        proportion: 1.0\n",
    "    -   domains: [\"v\", \"t\"]  # paired passes uses 100% of the available data\n",
    "        proportion: 1.0\n",
    "```\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1a9d083d-847d-41bb-84c6-1623912c8239",
   "metadata": {},
   "source": [
    "To use this proportion, we need to create a split in the dataset. A dataset split depends on a seed and the proportion of each group of domain.\n",
    "We only need to generate this split once.\n",
    "\n",
    "This can be done with the `shapesd alignment add` command. It needs the following arguments:\n",
    "- `--dataset_path \"DATASET_PATH\"`: the location where the dataset is stored\n",
    "- `--seed SEED` the split seed\n",
    "- `--domain_alignment DOMAIN_1,DOMAIN_2,...DOMAIN_N PROP` the proportion for each domain group. This corresponds to what has been defined in `domain_proportion`\n",
    "\n",
    "When running this command, it will create a file containing the indices of the items available in the train set (update so that it matches what we set in the config file).\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "ea1d63eb-c931-484f-81da-58edfc090122",
   "metadata": {},
   "outputs": [],
   "source": [
    "!shapesd alignment add --dataset_path \"simple_shapes_dataset\" --seed 0 --domain_alignment t 1.0 --domain_alignment v 1.0 --domain_alignment t,v 1.0"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "75208a58-9ef8-4272-8e52-e15c0c7b5d84",
   "metadata": {},
   "source": [
    "We also need to get the class used to load the different domains. For that, we can use `get_default_domains` and give as a parameter the list of domain loaders that we need. \n",
    "\n",
    "Here are the main available loaders, listed in the `DomainType` enum:\n",
    "- `DomainType.v`: visual model using a pretrained VAE.\n",
    "- `DomainType.v_latents`: same as `v`, but starting from the VAE latent representation. This is used for faster inference by short-circuiting the VAE.\n",
    "- `DomainType.attr`: attribute (proto-language) module using a VAE.\n",
    "- `DomainType.attr_legacy`: legacy version (from Devillers et al., 2024) of the attibute module using the attribute as the latent representation directly (no VAE).\n",
    "- `DomainType.t`: natural language module.\n",
    "\n",
    "Here, we will focus on vision to text, and we will select `v_latents` and `t`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "82998628-f97e-4c17-9d9f-9bd7aa405ec4",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{DomainDesc(base='v', kind='v_latents'): <class 'simple_shapes_dataset.domain.SimpleShapesPretrainedVisual'>, DomainDesc(base='t', kind='t'): <class 'simple_shapes_dataset.domain.SimpleShapesText'>}\n"
     ]
    }
   ],
   "source": [
    "# Then we get the data domain that will load the data for our two different domain\n",
    "# formats:\n",
    "# - `t` will load the text (with bert representations of the sentences\n",
    "# - `v_latents` will load pre-saved latent representations of the visual VAE.\n",
    "domain_classes = get_default_domains([\"v_latents\", \"t\"])\n",
    "print(domain_classes)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b9bef0c1-2e95-4e56-9c0d-ca7b75b426d3",
   "metadata": {},
   "source": [
    "`v_latents` will allow us to use the pretrained model provided in the [`shimmer-ssd`](https://github.com/ruflab/shimmer-ssd) repo.\n",
    "However, it requires pre-saved latent representations of the visual VAE.\n",
    "\n",
    "We can first download the checkpoinst, then extract the visual latents with:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "bb0ca2ea-217d-4962-9ccd-639a969e551e",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Downloading in checkpoints.\n",
      "  0%|                                                | 0.00/288M [00:00<?, ?B/s]\n",
      "  0%|                                        | 246k/288M [00:00<02:20, 2.05MB/s]\n",
      "  2%|▌                                      | 4.42M/288M [00:00<00:11, 23.6MB/s]\n",
      "  5%|██                                     | 15.2M/288M [00:00<00:04, 60.5MB/s]\n",
      "  8%|███                                    | 22.5M/288M [00:00<00:04, 65.2MB/s]\n",
      "2745it [00:00, 7773.54it/s]\u001b[A\n",
      " 14%|█████▎                                 | 39.4M/288M [00:00<00:03, 74.0MB/s]\n",
      "4813it [00:00, 9075.22it/s]\u001b[A\n",
      " 19%|███████▌                               | 56.0M/288M [00:00<00:03, 76.6MB/s]\n",
      " 23%|█████████▏                             | 67.5M/288M [00:00<00:02, 87.8MB/s]\n",
      " 27%|██████████▌                            | 78.3M/288M [00:01<00:02, 93.9MB/s]\n",
      " 31%|████████████                           | 89.4M/288M [00:01<00:02, 99.0MB/s]\n",
      " 35%|██████████████▎                          | 101M/288M [00:01<00:01, 103MB/s]\n",
      " 39%|███████████████▊                         | 111M/288M [00:01<00:01, 101MB/s]\n",
      " 42%|█████████████████▎                       | 121M/288M [00:01<00:01, 100MB/s]\n",
      " 46%|██████████████████▉                      | 133M/288M [00:01<00:01, 104MB/s]\n",
      " 50%|████████████████████▌                    | 144M/288M [00:01<00:01, 107MB/s]\n",
      " 54%|██████████████████████                   | 155M/288M [00:01<00:01, 107MB/s]\n",
      " 58%|███████████████████████▌                 | 165M/288M [00:01<00:01, 105MB/s]\n",
      " 61%|████████████████████████▍               | 176M/288M [00:02<00:01, 90.6MB/s]\n",
      " 65%|█████████████████████████▉              | 186M/288M [00:02<00:01, 93.7MB/s]\n",
      " 68%|███████████████████████████▏            | 196M/288M [00:02<00:00, 93.1MB/s]\n",
      " 71%|████████████████████████████▌           | 205M/288M [00:02<00:00, 89.2MB/s]\n",
      " 75%|█████████████████████████████▉          | 216M/288M [00:02<00:00, 92.7MB/s]\n",
      " 79%|███████████████████████████████▌        | 227M/288M [00:02<00:00, 97.5MB/s]\n",
      " 82%|████████████████████████████████▉       | 237M/288M [00:02<00:00, 99.5MB/s]\n",
      "28934it [00:02, 12037.80it/s]\u001b[A\n",
      " 86%|██████████████████████████████████▎     | 247M/288M [00:02<00:00, 98.2MB/s]\n",
      " 89%|███████████████████████████████████▋    | 257M/288M [00:02<00:00, 95.0MB/s]\n",
      " 96%|██████████████████████████████████████▍ | 276M/288M [00:03<00:00, 62.0MB/s]\n",
      "100%|███████████████████████████████████████▊| 286M/288M [00:03<00:00, 70.8MB/s]\n",
      "35108it [00:03, 10426.02it/s][A\n",
      "100%|████████████████████████████████████████| 288M/288M [00:03<00:00, 85.4MB/s]\n",
      "Extracting archive...\n"
     ]
    }
   ],
   "source": [
    "!ssd download checkpoints\n",
    "!mv checkpoints/checkpoints/* checkpoints/\n",
    "!rm -rf checkpoints/checkpoints"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "e66f3b72-b506-46a7-8512-192dea756478",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Latent file already exists. Skipping.\n"
     ]
    }
   ],
   "source": [
    "!ssd extract v \"checkpoints/domain_v.ckpt\" -p \"simple_shapes_dataset\""
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a8ccd794-79fe-4708-9eb5-5d07f7b31482",
   "metadata": {},
   "source": [
    "We now need to set the location of the presaved latent vectors in the config file. You can see that `domain_data_args` is defined as follows:\n",
    "\n",
    "```yaml\n",
    "domain_data_args:\n",
    "    v_latents:\n",
    "        presaved_path: domain_v.npy  # same name as `domain_v.ckpt`\n",
    "```"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b115e4aa-a683-48a9-939b-703a359757b9",
   "metadata": {},
   "source": [
    "And now let's define text transforms to change the text input as a token sequence.\n",
    "\n",
    "You can already change the path to the tokenizer vocab and merges file in:\n",
    "```yaml\n",
    "domain_modules:\n",
    "    text:\n",
    "        vocab_path: ./tokenizer/vocab.json\n",
    "        merges_path: ./tokenizer/merges.txt\n",
    "```\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "ce9f3a2e-a462-4eb0-861e-bbbc047798d3",
   "metadata": {},
   "outputs": [],
   "source": [
    "seq_length = 64  # we will use a sequence length of 64\n",
    "\n",
    "# And define a transform for the text domain to tokenize the text string\n",
    "additional_transforms = {\n",
    "    \"t\": [TokenizeCaptions(config.domain_modules.text.vocab_path, config.domain_modules.text.merges_path, seq_length)]\n",
    "}\n",
    "tokenizer = ByteLevelBPETokenizer(config.domain_modules.text.vocab_path, config.domain_modules.text.merges_path)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "06905ab3-b465-41fa-b532-fe51f7c12570",
   "metadata": {},
   "source": [
    "Finally, let's define some arguments:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "9c4db0d3-b30c-40e2-9a88-06afb31a1658",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Seed set to 0\n"
     ]
    }
   ],
   "source": [
    "seed_everything(config.seed, workers=True)\n",
    "\n",
    "# initilize the lightning data module\n",
    "data_module = SimpleShapesDataModule(\n",
    "    config.dataset.path,\n",
    "    domain_classes,\n",
    "    config.domain_proportions,\n",
    "    config.training.batch_size,\n",
    "    seed=config.seed,\n",
    "    domain_args=config.domain_data_args,\n",
    "    additional_transforms=additional_transforms,\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f8e7ecaa-cb07-4b7b-b0c1-7c3df93df0ac",
   "metadata": {},
   "source": [
    "## Load pretrained Model"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "eade4a9d-830b-4b5b-84ae-306e93fe38ac",
   "metadata": {},
   "source": [
    "The checkpoints are stored in \"./checkpoints\" after `ssd download checkpoints`."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "57aa948b-081a-4c74-99b2-cebf525a0399",
   "metadata": {},
   "source": [
    "Let's now define which pretrained domain module we will use for the model. You can update the `domains` section of the config:\n",
    "```yaml\n",
    "domains:\n",
    "    - checkpoint_path: \"#{default_root_dir}/domain_v.ckpt\"\n",
    "      domain_type: v_latents\n",
    "    - checkpoint_path: \"#{default_root_dir}/domain_t.ckpt\"\n",
    "      domain_type: t\n",
    "```\n",
    "\n",
    "Note here that we define the path relative to `#{default_root_dir}` which is replaced by the value set in `default_root_dir` (\"./checkpoints\").\n",
    "\n",
    "We also need to update the dimensions of the encoders and decoders that will translate the unimodal latent representations through the global workspace.\n",
    "\n",
    "```yaml\n",
    "global_workspace:\n",
    "    latent_dim: 16  # This changed compared to the default config!\n",
    "\n",
    "    encoders:\n",
    "        hidden_dim: 32\n",
    "        n_layers: 3\n",
    "\n",
    "    decoders:\n",
    "        hidden_dim: 32\n",
    "        n_layers: 3\n",
    "```"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "727ebb42-7a6a-4ee1-a72a-37c1b5b6a153",
   "metadata": {},
   "outputs": [],
   "source": [
    "# After some config changes, don't forget to reload the changes!\n",
    "config = load_config(\"./config\", use_cli=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "740bf5f9-b90d-4ebb-bfa1-dc23a71ee0ef",
   "metadata": {},
   "source": [
    "To make sure this tutorial runs without manually changing the config, the following line will force the correct value to be set.\n",
    "You don't need to run it if you made the correct change in the config."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "13e83063-5613-4f5c-a232-25568d9d2fba",
   "metadata": {},
   "outputs": [],
   "source": [
    "config.global_workspace.latent_dim = 16  # latent dim of the pretrained gw"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "1677290a-a24f-4e83-8fc2-b8c297c709ba",
   "metadata": {},
   "outputs": [],
   "source": [
    "# we load the pretrained domain modules and define the associated GW encoders and\n",
    "# decoders\n",
    "domain_modules, gw_encoders, gw_decoders = load_pretrained_domains(\n",
    "    config.domains,\n",
    "    config.global_workspace.latent_dim,\n",
    "    config.global_workspace.encoders.hidden_dim,\n",
    "    config.global_workspace.encoders.n_layers,\n",
    "    config.global_workspace.decoders.hidden_dim,\n",
    "    config.global_workspace.decoders.n_layers,\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f71dff64-c032-4a1e-ad46-264998532659",
   "metadata": {},
   "source": [
    "Let's now load a pre-trained global workspace trained to align vision and texts."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "b62469f5-be42-4b6c-87d4-a27d6417c314",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "GlobalWorkspaceFusion(\n",
       "  (gw_mod): GWModule(\n",
       "    (domain_mods): ModuleDict(\n",
       "      (v_latents): VisualLatentDomainModule(\n",
       "        (visual_module): VisualDomainModule(\n",
       "          (vae): VAE(\n",
       "            (encoder): RAEEncoder(\n",
       "              (layers): Sequential(\n",
       "                (0): Conv2d(3, 64, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)\n",
       "                (1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "                (2): ReLU()\n",
       "                (3): Conv2d(64, 128, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)\n",
       "                (4): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "                (5): ReLU()\n",
       "                (6): Conv2d(128, 256, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)\n",
       "                (7): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "                (8): ReLU()\n",
       "                (9): Conv2d(256, 512, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)\n",
       "                (10): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "                (11): ReLU()\n",
       "              )\n",
       "              (q_mean): Linear(in_features=2048, out_features=12, bias=True)\n",
       "              (q_logvar): Linear(in_features=2048, out_features=12, bias=True)\n",
       "            )\n",
       "            (decoder): RAEDecoder(\n",
       "              (layers): Sequential(\n",
       "                (0): ConvTranspose2d(12, 512, kernel_size=(8, 8), stride=(1, 1), bias=False)\n",
       "                (1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "                (2): ReLU()\n",
       "                (3): ConvTranspose2d(512, 256, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)\n",
       "                (4): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "                (5): ReLU()\n",
       "                (6): ConvTranspose2d(256, 128, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)\n",
       "                (7): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "                (8): ReLU()\n",
       "              )\n",
       "              (out_layer): Sequential(\n",
       "                (0): ZeroPad2d((0, 1, 0, 1))\n",
       "                (1): Conv2d(128, 3, kernel_size=(4, 4), stride=(1, 1), padding=(1, 1))\n",
       "                (2): Sigmoid()\n",
       "              )\n",
       "            )\n",
       "          )\n",
       "        )\n",
       "      )\n",
       "      (t): GRUTextDomainModule(\n",
       "        (projector): Sequential(\n",
       "          (0): Linear(in_features=768, out_features=256, bias=True)\n",
       "          (1): ReLU()\n",
       "          (2): Linear(in_features=256, out_features=64, bias=True)\n",
       "          (3): Tanh()\n",
       "        )\n",
       "        (embeddings): Embedding(822, 64)\n",
       "        (decoder): GRU(64, 256, num_layers=2, batch_first=True)\n",
       "        (text_head): Linear(in_features=256, out_features=822, bias=True)\n",
       "      )\n",
       "    )\n",
       "    (gw_encoders): ModuleDict(\n",
       "      (v_latents): GWEncoder(\n",
       "        (0): Linear(in_features=12, out_features=32, bias=True)\n",
       "        (1): ReLU()\n",
       "        (2): Linear(in_features=32, out_features=32, bias=True)\n",
       "        (3): ReLU()\n",
       "        (4): Linear(in_features=32, out_features=32, bias=True)\n",
       "        (5): ReLU()\n",
       "        (6): Linear(in_features=32, out_features=32, bias=True)\n",
       "        (7): ReLU()\n",
       "        (8): Linear(in_features=32, out_features=16, bias=True)\n",
       "      )\n",
       "      (t): GWEncoder(\n",
       "        (0): Linear(in_features=64, out_features=32, bias=True)\n",
       "        (1): ReLU()\n",
       "        (2): Linear(in_features=32, out_features=32, bias=True)\n",
       "        (3): ReLU()\n",
       "        (4): Linear(in_features=32, out_features=32, bias=True)\n",
       "        (5): ReLU()\n",
       "        (6): Linear(in_features=32, out_features=32, bias=True)\n",
       "        (7): ReLU()\n",
       "        (8): Linear(in_features=32, out_features=16, bias=True)\n",
       "      )\n",
       "    )\n",
       "    (gw_decoders): ModuleDict(\n",
       "      (v_latents): GWDecoder(\n",
       "        (0): Linear(in_features=16, out_features=32, bias=True)\n",
       "        (1): ReLU()\n",
       "        (2): Linear(in_features=32, out_features=32, bias=True)\n",
       "        (3): ReLU()\n",
       "        (4): Linear(in_features=32, out_features=32, bias=True)\n",
       "        (5): ReLU()\n",
       "        (6): Linear(in_features=32, out_features=32, bias=True)\n",
       "        (7): ReLU()\n",
       "        (8): Linear(in_features=32, out_features=12, bias=True)\n",
       "      )\n",
       "      (t): GWDecoder(\n",
       "        (0): Linear(in_features=16, out_features=32, bias=True)\n",
       "        (1): ReLU()\n",
       "        (2): Linear(in_features=32, out_features=32, bias=True)\n",
       "        (3): ReLU()\n",
       "        (4): Linear(in_features=32, out_features=32, bias=True)\n",
       "        (5): ReLU()\n",
       "        (6): Linear(in_features=32, out_features=32, bias=True)\n",
       "        (7): ReLU()\n",
       "        (8): Linear(in_features=32, out_features=64, bias=True)\n",
       "      )\n",
       "    )\n",
       "  )\n",
       "  (selection_mod): RandomSelection()\n",
       "  (loss_mod): GWLosses(\n",
       "    (gw_mod): GWModule(\n",
       "      (domain_mods): ModuleDict(\n",
       "        (v_latents): VisualLatentDomainModule(\n",
       "          (visual_module): VisualDomainModule(\n",
       "            (vae): VAE(\n",
       "              (encoder): RAEEncoder(\n",
       "                (layers): Sequential(\n",
       "                  (0): Conv2d(3, 64, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)\n",
       "                  (1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "                  (2): ReLU()\n",
       "                  (3): Conv2d(64, 128, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)\n",
       "                  (4): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "                  (5): ReLU()\n",
       "                  (6): Conv2d(128, 256, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)\n",
       "                  (7): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "                  (8): ReLU()\n",
       "                  (9): Conv2d(256, 512, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)\n",
       "                  (10): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "                  (11): ReLU()\n",
       "                )\n",
       "                (q_mean): Linear(in_features=2048, out_features=12, bias=True)\n",
       "                (q_logvar): Linear(in_features=2048, out_features=12, bias=True)\n",
       "              )\n",
       "              (decoder): RAEDecoder(\n",
       "                (layers): Sequential(\n",
       "                  (0): ConvTranspose2d(12, 512, kernel_size=(8, 8), stride=(1, 1), bias=False)\n",
       "                  (1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "                  (2): ReLU()\n",
       "                  (3): ConvTranspose2d(512, 256, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)\n",
       "                  (4): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "                  (5): ReLU()\n",
       "                  (6): ConvTranspose2d(256, 128, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)\n",
       "                  (7): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "                  (8): ReLU()\n",
       "                )\n",
       "                (out_layer): Sequential(\n",
       "                  (0): ZeroPad2d((0, 1, 0, 1))\n",
       "                  (1): Conv2d(128, 3, kernel_size=(4, 4), stride=(1, 1), padding=(1, 1))\n",
       "                  (2): Sigmoid()\n",
       "                )\n",
       "              )\n",
       "            )\n",
       "          )\n",
       "        )\n",
       "        (t): GRUTextDomainModule(\n",
       "          (projector): Sequential(\n",
       "            (0): Linear(in_features=768, out_features=256, bias=True)\n",
       "            (1): ReLU()\n",
       "            (2): Linear(in_features=256, out_features=64, bias=True)\n",
       "            (3): Tanh()\n",
       "          )\n",
       "          (embeddings): Embedding(822, 64)\n",
       "          (decoder): GRU(64, 256, num_layers=2, batch_first=True)\n",
       "          (text_head): Linear(in_features=256, out_features=822, bias=True)\n",
       "        )\n",
       "      )\n",
       "      (gw_encoders): ModuleDict(\n",
       "        (v_latents): GWEncoder(\n",
       "          (0): Linear(in_features=12, out_features=32, bias=True)\n",
       "          (1): ReLU()\n",
       "          (2): Linear(in_features=32, out_features=32, bias=True)\n",
       "          (3): ReLU()\n",
       "          (4): Linear(in_features=32, out_features=32, bias=True)\n",
       "          (5): ReLU()\n",
       "          (6): Linear(in_features=32, out_features=32, bias=True)\n",
       "          (7): ReLU()\n",
       "          (8): Linear(in_features=32, out_features=16, bias=True)\n",
       "        )\n",
       "        (t): GWEncoder(\n",
       "          (0): Linear(in_features=64, out_features=32, bias=True)\n",
       "          (1): ReLU()\n",
       "          (2): Linear(in_features=32, out_features=32, bias=True)\n",
       "          (3): ReLU()\n",
       "          (4): Linear(in_features=32, out_features=32, bias=True)\n",
       "          (5): ReLU()\n",
       "          (6): Linear(in_features=32, out_features=32, bias=True)\n",
       "          (7): ReLU()\n",
       "          (8): Linear(in_features=32, out_features=16, bias=True)\n",
       "        )\n",
       "      )\n",
       "      (gw_decoders): ModuleDict(\n",
       "        (v_latents): GWDecoder(\n",
       "          (0): Linear(in_features=16, out_features=32, bias=True)\n",
       "          (1): ReLU()\n",
       "          (2): Linear(in_features=32, out_features=32, bias=True)\n",
       "          (3): ReLU()\n",
       "          (4): Linear(in_features=32, out_features=32, bias=True)\n",
       "          (5): ReLU()\n",
       "          (6): Linear(in_features=32, out_features=32, bias=True)\n",
       "          (7): ReLU()\n",
       "          (8): Linear(in_features=32, out_features=12, bias=True)\n",
       "        )\n",
       "        (t): GWDecoder(\n",
       "          (0): Linear(in_features=16, out_features=32, bias=True)\n",
       "          (1): ReLU()\n",
       "          (2): Linear(in_features=32, out_features=32, bias=True)\n",
       "          (3): ReLU()\n",
       "          (4): Linear(in_features=32, out_features=32, bias=True)\n",
       "          (5): ReLU()\n",
       "          (6): Linear(in_features=32, out_features=32, bias=True)\n",
       "          (7): ReLU()\n",
       "          (8): Linear(in_features=32, out_features=64, bias=True)\n",
       "        )\n",
       "      )\n",
       "    )\n",
       "    (selection_mod): RandomSelection()\n",
       "    (domain_mods): ModuleDict(\n",
       "      (v_latents): VisualLatentDomainModule(\n",
       "        (visual_module): VisualDomainModule(\n",
       "          (vae): VAE(\n",
       "            (encoder): RAEEncoder(\n",
       "              (layers): Sequential(\n",
       "                (0): Conv2d(3, 64, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)\n",
       "                (1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "                (2): ReLU()\n",
       "                (3): Conv2d(64, 128, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)\n",
       "                (4): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "                (5): ReLU()\n",
       "                (6): Conv2d(128, 256, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)\n",
       "                (7): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "                (8): ReLU()\n",
       "                (9): Conv2d(256, 512, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)\n",
       "                (10): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "                (11): ReLU()\n",
       "              )\n",
       "              (q_mean): Linear(in_features=2048, out_features=12, bias=True)\n",
       "              (q_logvar): Linear(in_features=2048, out_features=12, bias=True)\n",
       "            )\n",
       "            (decoder): RAEDecoder(\n",
       "              (layers): Sequential(\n",
       "                (0): ConvTranspose2d(12, 512, kernel_size=(8, 8), stride=(1, 1), bias=False)\n",
       "                (1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "                (2): ReLU()\n",
       "                (3): ConvTranspose2d(512, 256, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)\n",
       "                (4): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "                (5): ReLU()\n",
       "                (6): ConvTranspose2d(256, 128, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)\n",
       "                (7): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "                (8): ReLU()\n",
       "              )\n",
       "              (out_layer): Sequential(\n",
       "                (0): ZeroPad2d((0, 1, 0, 1))\n",
       "                (1): Conv2d(128, 3, kernel_size=(4, 4), stride=(1, 1), padding=(1, 1))\n",
       "                (2): Sigmoid()\n",
       "              )\n",
       "            )\n",
       "          )\n",
       "        )\n",
       "      )\n",
       "      (t): GRUTextDomainModule(\n",
       "        (projector): Sequential(\n",
       "          (0): Linear(in_features=768, out_features=256, bias=True)\n",
       "          (1): ReLU()\n",
       "          (2): Linear(in_features=256, out_features=64, bias=True)\n",
       "          (3): Tanh()\n",
       "        )\n",
       "        (embeddings): Embedding(822, 64)\n",
       "        (decoder): GRU(64, 256, num_layers=2, batch_first=True)\n",
       "        (text_head): Linear(in_features=256, out_features=822, bias=True)\n",
       "      )\n",
       "    )\n",
       "    (contrastive_fn): ContrastiveLoss()\n",
       "  )\n",
       ")"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\n",
    "\n",
    "# And now we load the GW checkpoint\n",
    "checkpoint_path = Path(\"./checkpoints\")\n",
    "checkpoint = checkpoint_path / \"gw-t-v_e6sm51xy.ckpt\"\n",
    "\n",
    "global_workspace = GlobalWorkspaceFusion.load_from_checkpoint(\n",
    "    checkpoint,\n",
    "    domain_mods=domain_modules,\n",
    "    gw_encoders=gw_encoders,\n",
    "    gw_decoders=gw_decoders,\n",
    ")\n",
    "global_workspace.to(device)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "83abaa0e-3179-4b31-b05b-05ef8bb4566d",
   "metadata": {},
   "source": [
    "## Play with the data"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b63de48b-36f3-4cc6-bf7f-ec8b0c4cf3da",
   "metadata": {},
   "source": [
    "Let's fetch some samples from the dataset."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "0ec5bcca-7c39-4d18-b385-d006cebd20f8",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "dict_keys([frozenset({'t'}), frozenset({'v_latents'}), frozenset({'v_latents', 't'})])\n"
     ]
    }
   ],
   "source": [
    "# Let's get 32 samples from our dataset\n",
    "train_samples = data_module.get_samples(\"train\", 32)\n",
    "train_samples = batch_to_device(train_samples, device)\n",
    "print(train_samples.keys())"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3f2f7e6c-15ad-4211-b491-f3257919faf6",
   "metadata": {},
   "source": [
    "This contains three groups: 2 unimodal data groups, and one paired group.\n",
    "It is packaged as a dict, with a `frozenset` containing the domains of the group for the keys.\n",
    "\n",
    "This corresponds to the groups defined in `domain_proportions`.\n",
    "<div class=\"alert alert-info\">\n",
    "NOTE:\n",
    "This is only the case for the train split. The train and test splits only have the paired data group because everything is paired.   \n",
    "</div>\n",
    "\n",
    "For each group, the value contains anothe dict with the samples for each domain of the group."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "880e52bc-627e-4292-958b-632715aa076b",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'v_latents': tensor([[-6.3928e-02,  1.7979e-02,  1.9808e+00,  3.8199e-01, -4.5171e-01,\n",
      "         -3.0649e-01, -1.1606e-01, -3.1035e-01, -3.7965e-01,  2.9842e-03,\n",
      "          3.8199e-01,  5.3416e-02],\n",
      "        [ 1.2547e+00,  3.9291e-01, -8.0259e-01,  1.1118e-01, -2.5586e-01,\n",
      "         -1.1708e+00,  5.5127e-01,  1.1952e+00, -1.8277e+00, -3.9273e-02,\n",
      "          1.2012e+00,  5.5203e-01],\n",
      "        [-9.1288e-01,  1.8649e+00,  2.5332e-01,  7.5205e-01, -4.7010e-01,\n",
      "         -6.5149e-01, -2.0799e-01,  1.3994e+00, -9.5400e-01,  5.6362e-02,\n",
      "         -9.6622e-01,  6.8637e-01],\n",
      "        [-6.8283e-01, -1.6199e+00,  7.7749e-03,  1.9025e-02,  5.0190e-02,\n",
      "         -1.5576e-01, -3.7625e-02, -3.8195e-01,  8.0247e-01, -3.3602e-01,\n",
      "          5.5143e-01,  1.6605e+00],\n",
      "        [-8.6961e-02,  1.4944e+00,  6.6780e-01, -2.5980e-01, -1.1766e+00,\n",
      "         -3.5819e-01,  1.3969e+00, -4.0627e-01, -8.1232e-01,  4.4536e-01,\n",
      "         -5.3927e-01, -8.7352e-01],\n",
      "        [ 1.7148e+00,  1.3238e+00,  5.0477e-02, -8.1989e-02,  2.8140e+00,\n",
      "          3.7121e-01,  1.9507e+00, -5.8689e-01, -2.6242e-01,  5.6079e-01,\n",
      "         -2.9084e-01, -5.9677e-01],\n",
      "        [ 2.3198e-01,  3.0275e-02,  1.3394e+00, -1.4894e+00, -3.4325e-01,\n",
      "         -1.8714e-01, -5.9854e-02, -9.1483e-01, -5.7876e-01, -2.5054e-03,\n",
      "          1.2024e+00,  7.0115e-01],\n",
      "        [ 3.3676e-01, -1.5078e-01,  1.9305e+00,  3.0152e-01,  4.9536e-01,\n",
      "         -3.3191e+00,  5.3110e-02,  1.7538e+00, -7.7098e-01,  1.6761e-01,\n",
      "          1.1439e+00,  8.1445e-01],\n",
      "        [-2.7266e-02,  5.4585e-02,  7.6377e-01,  3.6025e-01,  7.2443e-01,\n",
      "         -1.0457e+00,  1.3295e+00, -3.8332e-01, -4.4969e-02, -4.2042e-01,\n",
      "         -2.8337e-01,  4.5770e-01],\n",
      "        [ 9.9179e-01, -5.1080e-01, -3.4790e-01, -1.4628e+00,  2.6069e-01,\n",
      "          1.4694e+00,  2.5055e+00,  5.9781e-01, -2.2249e-01, -5.8937e-01,\n",
      "         -3.9404e-01, -1.1884e-01],\n",
      "        [ 9.1346e-01, -3.9392e-02,  5.5915e-01,  9.5799e-01,  4.6973e-01,\n",
      "         -3.7715e-01,  5.2489e-01,  1.5246e+00,  3.4803e-01,  2.3866e-02,\n",
      "          1.0882e-01, -7.3691e-01],\n",
      "        [-3.6449e-01,  1.0638e+00, -1.0372e+00, -5.3002e-01, -8.4987e-01,\n",
      "          1.0291e-01,  3.5096e-02,  7.1564e-01, -4.0058e-01,  2.5168e-01,\n",
      "         -4.8627e-01,  1.5757e+00],\n",
      "        [-5.1719e-01, -6.4303e-01, -4.2195e-01,  8.6021e-02, -2.2018e-02,\n",
      "         -6.6143e-01,  2.5480e+00, -2.4315e-01,  1.4628e+00,  9.7181e-01,\n",
      "         -1.1896e+00, -6.6998e-01],\n",
      "        [ 4.2327e-01, -1.5514e+00, -3.3742e-01, -1.0034e+00, -1.7558e-01,\n",
      "          6.9477e-02,  4.2813e-02, -2.3034e-02,  6.8400e-01,  1.2369e+00,\n",
      "         -3.7505e-01,  1.6835e+00],\n",
      "        [-7.6863e-01,  8.5076e-01,  1.5691e-01,  2.4191e-02, -3.5614e-01,\n",
      "         -7.1601e-03, -2.5627e-01, -7.9055e-01, -9.2622e-01,  3.7636e-01,\n",
      "         -1.1317e+00, -9.3924e-01],\n",
      "        [-8.3273e-01, -3.3544e-01, -6.4065e-02, -4.6170e-01, -9.6924e-01,\n",
      "         -6.6464e-01, -1.4594e+00,  5.3178e-01,  1.4020e+00, -9.3466e-01,\n",
      "          5.7892e-01, -6.6272e-02],\n",
      "        [-5.4079e-01,  9.4771e-01,  1.2024e+00,  4.5769e-03, -1.3732e+00,\n",
      "          6.1877e-01,  3.1188e-01,  6.8294e-01,  5.2274e-01,  1.9901e-01,\n",
      "         -1.4820e+00, -1.8745e-01],\n",
      "        [ 1.1094e+00, -2.1941e-01,  3.2950e-02,  8.6057e-01, -7.3020e-01,\n",
      "          1.0045e+00,  1.6674e+00, -1.6534e+00, -8.0033e-01,  5.2651e-01,\n",
      "         -6.0344e-01, -1.9040e+00],\n",
      "        [ 5.5105e-01,  2.3406e-01,  5.1031e-01,  7.0845e-02, -1.2879e-01,\n",
      "         -5.3809e-01, -8.6631e-01,  6.0552e-02, -6.8649e-01, -1.0303e-01,\n",
      "         -1.1780e+00,  3.0259e-01],\n",
      "        [ 2.8670e-01,  2.9182e-02, -2.2236e-01, -1.9941e-01, -6.0746e-01,\n",
      "         -1.7714e-01,  5.1434e-01, -8.9214e-01, -6.1120e-01,  6.9888e-01,\n",
      "         -1.2601e+00, -2.2056e+00],\n",
      "        [ 9.8493e-01, -2.9848e-01, -4.1500e-01,  1.1959e+00,  3.1640e-01,\n",
      "         -2.7854e-01,  9.4939e-01,  1.5752e+00, -2.3489e-01,  1.3423e+00,\n",
      "         -5.2182e-01, -3.9396e-01],\n",
      "        [-1.0746e+00, -2.2334e-01, -2.0777e-01,  4.5332e-02, -7.0257e-01,\n",
      "          5.7060e-01, -8.1416e-01,  7.3448e-02,  1.4308e+00,  6.4969e-01,\n",
      "         -4.6003e-01,  4.8005e-01],\n",
      "        [-9.0497e-01, -1.2269e+00, -6.1564e-01,  2.1630e-01,  9.5606e-01,\n",
      "         -8.8404e-01,  6.4231e-01,  2.1414e+00, -3.3274e-01,  1.9813e+00,\n",
      "          1.4187e+00,  1.4869e-01],\n",
      "        [ 7.2998e-01,  4.5307e-01,  9.4265e-01,  3.5672e-01, -5.1396e-01,\n",
      "         -3.1769e-01,  1.0273e+00,  7.7223e-01, -3.9297e-02, -7.7819e-01,\n",
      "         -2.2957e-01, -9.0947e-01],\n",
      "        [ 1.4934e+00, -8.4289e-01, -8.6126e-01, -2.3361e-01, -4.3484e-02,\n",
      "         -1.8705e-01, -1.7037e+00, -3.9826e-01,  3.2829e-01, -7.6484e-01,\n",
      "         -5.0435e-01, -1.0500e+00],\n",
      "        [ 3.0122e-01,  1.7605e+00,  5.9346e-01, -2.8462e-02, -2.1002e+00,\n",
      "          1.8279e+00, -1.2911e+00,  6.2273e-01, -6.8656e-01,  4.7369e-02,\n",
      "         -5.9019e-01,  9.9844e-01],\n",
      "        [ 2.7105e-01,  9.3937e-01, -1.3282e+00, -9.2970e-01, -8.0474e-01,\n",
      "          3.6067e-02,  1.4618e+00, -4.2825e-01, -5.5758e-01,  3.7801e-01,\n",
      "         -1.3903e-01,  2.8066e-01],\n",
      "        [ 1.6104e+00,  1.3230e+00, -4.3547e-01,  1.8035e-01, -3.1234e-01,\n",
      "          8.4827e-03,  1.5701e+00, -5.7374e-02, -1.5976e-01,  2.3567e+00,\n",
      "         -5.1504e-01, -2.8488e-01],\n",
      "        [-1.4887e+00, -9.9639e-02,  1.0083e+00,  1.3789e-01, -2.0938e-01,\n",
      "          1.7462e+00,  2.8090e-01,  1.3085e-02, -3.5452e-01, -1.0990e+00,\n",
      "          1.7643e+00, -4.2319e-01],\n",
      "        [ 1.6198e-01, -2.1687e-02,  1.1985e+00,  1.3665e-02, -1.3551e+00,\n",
      "         -6.9465e-02, -1.1977e-01,  6.7679e-01, -7.5276e-01,  1.3666e-01,\n",
      "         -2.5534e+00,  1.7544e-01],\n",
      "        [ 7.6733e-01, -5.7256e-01,  2.8714e-02, -3.4376e-01, -3.5904e-01,\n",
      "          1.8210e-01,  1.2124e-01, -1.0181e+00,  3.4599e-01, -4.2224e-01,\n",
      "          1.2930e+00,  6.4342e-01],\n",
      "        [ 6.2586e-02,  1.1047e+00,  9.6798e-01, -1.5116e-01, -8.1225e-01,\n",
      "         -3.3050e-02,  3.0416e-01, -4.5777e-01, -6.8666e-01, -4.6432e-01,\n",
      "          7.6802e-01,  5.3879e-01]], device='cuda:0'), 't': {'bert': tensor([[-0.8998,  0.6153, -0.4713,  ...,  0.5043, -1.1603,  0.9502],\n",
      "        [-0.0699,  1.7032, -0.5868,  ...,  0.3386,  0.7757, -0.8087],\n",
      "        [-2.1292, -0.6447,  1.2002,  ..., -2.0594,  0.7870, -0.1977],\n",
      "        ...,\n",
      "        [-0.2175, -0.1037, -1.3829,  ...,  0.6464,  1.3163,  1.3407],\n",
      "        [-0.3428, -0.0204, -0.5938,  ...,  0.0970,  0.6054, -0.3208],\n",
      "        [ 0.1212,  0.3881, -1.5330,  ..., -0.0510,  0.8997, -0.2159]],\n",
      "       device='cuda:0'), 'tokens': tensor([[314, 346, 441,  ...,   0,   0,   0],\n",
      "        [429, 265, 261,  ...,   0,   0,   0],\n",
      "        [ 35, 358, 447,  ...,   0,   0,   0],\n",
      "        ...,\n",
      "        [ 35, 426, 425,  ...,   0,   0,   0],\n",
      "        [ 35, 426, 425,  ...,   0,   0,   0],\n",
      "        [ 35, 426, 425,  ...,   0,   0,   0]], device='cuda:0')}}\n"
     ]
    }
   ],
   "source": [
    "train_paired_samples = train_samples[frozenset([\"v_latents\", \"t\"])]\n",
    "print(train_paired_samples)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "32bcdc20-3900-4291-a233-a7057e017565",
   "metadata": {},
   "source": [
    "### Vision"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "dd450237-cea5-4ff1-831a-e785126a5e7a",
   "metadata": {},
   "source": [
    "This group contains two items: one of each domain. Each item in the domain is paired\n",
    "Let's look at the images first."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "5d2a5fb5-7ec8-4bb1-a55e-3ac41596ee52",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "torch.Size([32, 12])"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_paired_samples[\"v_latents\"].size()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "c3e9c3fe-d665-4de3-8d04-48d5b0c536ea",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Extract the `v_latents` domain module\n",
    "visual_module = cast(VisualLatentDomainModule, global_workspace.domain_mods[\"v_latents\"])\n",
    "\n",
    "# `train_paired_samples` contains the latent representation, we can use `decode_images`\n",
    "# of the visual module to get back the actual images:\n",
    "train_images = visual_module.decode_images(train_paired_samples[\"v_latents\"])\n",
    "\n",
    "\n",
    "def plot_image_grid(samples: torch.Tensor, nrow=8):\n",
    "    grid = make_grid(samples, nrow=nrow, pad_value=1).permute(1, 2, 0).detach().cpu().numpy()\n",
    "    plt.imshow(grid)\n",
    "    plt.show()\n",
    "\n",
    "\n",
    "plot_image_grid(train_images)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e19d88dc-4c04-486c-ba7a-dd133e7a4ab8",
   "metadata": {},
   "source": [
    "### Text"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1688ccb4-d1d8-4e30-8a52-7d96bc771131",
   "metadata": {},
   "source": [
    "Let's now look at the text domain:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "f092ed7b-1f31-416c-b2af-2dd41aa92877",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'bert': tensor([[-0.8998,  0.6153, -0.4713,  ...,  0.5043, -1.1603,  0.9502],\n",
       "         [-0.0699,  1.7032, -0.5868,  ...,  0.3386,  0.7757, -0.8087],\n",
       "         [-2.1292, -0.6447,  1.2002,  ..., -2.0594,  0.7870, -0.1977],\n",
       "         ...,\n",
       "         [-0.2175, -0.1037, -1.3829,  ...,  0.6464,  1.3163,  1.3407],\n",
       "         [-0.3428, -0.0204, -0.5938,  ...,  0.0970,  0.6054, -0.3208],\n",
       "         [ 0.1212,  0.3881, -1.5330,  ..., -0.0510,  0.8997, -0.2159]],\n",
       "        device='cuda:0'),\n",
       " 'tokens': tensor([[314, 346, 441,  ...,   0,   0,   0],\n",
       "         [429, 265, 261,  ...,   0,   0,   0],\n",
       "         [ 35, 358, 447,  ...,   0,   0,   0],\n",
       "         ...,\n",
       "         [ 35, 426, 425,  ...,   0,   0,   0],\n",
       "         [ 35, 426, 425,  ...,   0,   0,   0],\n",
       "         [ 35, 426, 425,  ...,   0,   0,   0]], device='cuda:0')}"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_paired_samples[\"t\"]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c68bf7fb-6748-4086-a903-b677cd95355a",
   "metadata": {},
   "source": [
    "It contains the bert latent representations of the text, and the token sequence. Let's decode the tokens and look at the sentences:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "5beac4be-ee02-4849-937c-77646922b714",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1 The image represents a four-sided shape, and is big, it is in dark slate grey color, and is located at the lower side, on the left, and is pointing to the east-northeast.\n",
      "2 There is a light blue bullet, it is pointing towards the bottom-right, it is tiny, it is located at the center.\n",
      "3 A tiny blue violet trapezoidal shape and it is located slightly lower side, on the left side and it is pointing towards the right top-right corner.\n",
      "4 A kind of tiny medium seagreen colored object at the upper side, slightly right, it is pointing to the west, it looks like a bullet.\n",
      "5 A tiny shape pointing to the southwest, it is located at the top, on the very left, it is dark orchid colored and looks like an isosceles triangle that has round corners.\n",
      "6 There is a big orchid colored shape located slightly upper side, at the left and it is pointing towards the bottom and looks like an isosceles triangle.\n",
      "7 It is a tiny shape in sienna slightly top, at the middle, pointing to the northwest. It looks like a four-sided shape.\n",
      "8 A kind of turquoise colored triangle and is large and is pointing towards the bottom-left and is located slightly bottom, on the left side.\n",
      "9 The image contains a sienna four-sided shape. It is at the top left side. It is little. It is pointing towards the northwest.\n",
      "10 There is a medium gainsboro colored object at the center left side, pointing to the bottom-right and it looks like a trapezoidal shape.\n",
      "11 The image contains a big dark slate grey trapezoidal shape, and is in the center, on the left side, and is pointing towards the east.\n",
      "12 This is a small object in crimson color located at the upper side, at the right, pointing west, it looks like a triangle.\n",
      "13 It is a big gainsboro bullet and it is located at the top left side and it is pointing towards the bottom bottom-left corner.\n",
      "14 It is a little shape in lime green color pointing towards the right. It is located at the upper side right side. It looks like a triangle.\n",
      "15 There is a small shape in rebecca purple at the bottom right, pointing to the bottom-left, it looks like a triangle.\n",
      "16 There is a slate grey big kite. It is pointing towards the northeast. It is located at the lower side.\n",
      "17 A medium sized purple object pointing southeast. It is at the bottom left side, it looks like an oval shaped structure.\n",
      "18 The image represents a bullet, little, and it is in floral white color, at the very top, at the very left side, pointing northeast.\n",
      "19 A kind of tiny shape in dark slate grey pointing to the southeast, and it is located at the lower side, on the left side, and looks like an isosceles triangle that has round corners.\n",
      "20 It is a tiny gainsboro colored object at the bottom right and is pointing towards the south-southwest. It looks like an egg.\n",
      "21 A medium sized shape pointing to the top-right, located in the center left, in grey, it looks like a diamond.\n",
      "22 The image represents a medium sized shape at the very top, on the right side and is pointing towards the top-left and is in sienna, it looks like a triangle-like shape with rounded vertices.\n",
      "23 The image represents a pale turquoise colored quadrilateral, medium sized, located in the middle, pointing east.\n",
      "24 The image is a dark slate grey colored diamond. It is pointing towards the southwest. It is big. It is located in the center left side.\n",
      "25 There is a silver colored tiny bullet and is at the lower side and is pointing towards the southeast.\n",
      "26 A kind of deep pink colored triangle. It is located at the lower side left side. It is average sized. It is pointing towards the left side.\n",
      "27 A tiny shape pointing to the bottom-left corner and is at the top, on the left and is in steel blue and it looks like an arrow-shaped polygon.\n",
      "28 This is a big rebecca purple triangle. It is pointing towards the northeast. It is located at the upper side left.\n",
      "29 It is a medium tan colored diamond and it is located at the middle right and it is pointing towards the southeast.\n",
      "30 A kind of isosceles triangle that has round corners, and is tiny, it is silver, and is pointing towards the southeast, and is located at the bottom, on the very left side.\n",
      "31 A kind of dark olive green small water droplet, it is pointing to the bottom bottom-right corner, it is located at the center.\n",
      "32 A kind of maroon colored little isosceles triangle, and is slightly top, at the middle, and is pointing to the west.\n"
     ]
    }
   ],
   "source": [
    "def get_text_samples(samples: dict[str, torch.Tensor]) -> list[str]:\n",
    "    \"\"\"\n",
    "    Decode the token sequence into a full sentence.\n",
    "    \"\"\"\n",
    "    return tokenizer.decode_batch(\n",
    "        samples[\"tokens\"].detach().cpu().tolist(),\n",
    "        skip_special_tokens=True,\n",
    "    )\n",
    "\n",
    "\n",
    "train_text_samples = get_text_samples(train_paired_samples[\"t\"])\n",
    "for k, text in enumerate(train_text_samples, 1):\n",
    "    print(k, text.replace(\"<pad>\", \"\"))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3982f65c-524d-4933-b05e-9bd98bd1a723",
   "metadata": {},
   "source": [
    "## Text to Image Translations"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "cc1c3e8d-d384-4092-914d-aeeb688d29a5",
   "metadata": {},
   "source": [
    "Now that we are familiar with the data, let's use the global workspace to do some translations.\n",
    "\n",
    "We will first encode the raw data into the unimodal latent representations with `encode_domain`:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "0baf3400-f305-4d28-919a-a884abd597fd",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "torch.Size([32, 64])\n"
     ]
    }
   ],
   "source": [
    "latent_unimodal_text = global_workspace.encode_domain(train_paired_samples[\"t\"], \"t\")\n",
    "print(latent_unimodal_text.size())"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "87a0b0e2-234d-4d63-b496-0657d7f17f1b",
   "metadata": {},
   "source": [
    "Then, we will encode the unimodal text representation into the GW reprensentation. First we encode the reprensentations"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "52acf10d-3c2f-47d6-97d0-0411812b9659",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "torch.Size([32, 16])\n"
     ]
    }
   ],
   "source": [
    "t_gw_latent = global_workspace.gw_mod.encode({\"t\": latent_unimodal_text})\n",
    "print(t_gw_latent[\"t\"].size())"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e895c15b-28a7-4f1f-b575-9d8022742d20",
   "metadata": {},
   "source": [
    "Then we fuse the latents together. This requires the coefficients that we use for each domain. As we only have one domain, we put all weights to the text domain with:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "cbfdf418-7f4a-43cf-8629-e97cba2ee348",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "torch.Size([32, 16])\n"
     ]
    }
   ],
   "source": [
    "t_gw_latent_fused = global_workspace.gw_mod.fuse(t_gw_latent, {\"t\": torch.ones(t_gw_latent.size(0)).to(device)})\n",
    "print(t_gw_latent_fused.size())"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0da79075-4028-4eb4-ae5d-c932ee32a4e1",
   "metadata": {},
   "source": [
    "Finally, we will decode the GW representation:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "4dbce724-f6c6-4b85-9bdd-4b035b40c2bc",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "dict_keys(['v_latents', 't'])\n"
     ]
    }
   ],
   "source": [
    "decoded_latents = global_workspace.gw_mod.decode(t_gw_latent_fused)\n",
    "print(decoded_latents.keys())"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "cb7782ae-1b2c-41e3-8260-58a6b6fd3aae",
   "metadata": {},
   "source": [
    "As you can see, this decodes to both domains. We will look now at the predicted visual latents."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "a44a7a8c-3335-4440-bd75-2613d0bdc81f",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "torch.Size([32, 12])\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "print(decoded_latents[\"v_latents\"].size())\n",
    "decoded_images = visual_module.decode_images(decoded_latents[\"v_latents\"])\n",
    "plot_image_grid(decoded_images)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c8f43605-9078-4d0b-be6e-342170a56913",
   "metadata": {},
   "source": [
    "This is the predicted visual translation from text."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5d21d6b0-1d0d-458a-8a9f-dd72f3ebea7a",
   "metadata": {},
   "source": [
    "## Image to Text translations"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c9861386-4099-4e13-b869-4fd83d90d796",
   "metadata": {},
   "source": [
    "You could do the same as in the other side, but we will here look at another method. When the previous method is useful to translate starting from domains. This method can take domain groups and will do the translation for each group.\n",
    "\n",
    "We here use `encode_domains` instead of `encode_domain` with all groups:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "85bea5b7-a747-40f6-a9e2-897c2521e9a4",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "dict_keys([frozenset({'t'}), frozenset({'v_latents'}), frozenset({'t', 'v_latents'})])\n",
      "{'t': tensor([[-0.9933, -1.0000, -1.0000,  ...,  1.0000,  1.0000, -0.5814],\n",
      "        [ 1.0000, -1.0000, -1.0000,  ...,  1.0000,  1.0000, -0.9995],\n",
      "        [ 0.8803, -1.0000, -1.0000,  ...,  1.0000,  1.0000, -0.5440],\n",
      "        ...,\n",
      "        [-1.0000, -1.0000, -1.0000,  ...,  1.0000,  1.0000,  1.0000],\n",
      "        [-0.2907, -1.0000, -1.0000,  ...,  1.0000, -1.0000,  1.0000],\n",
      "        [-0.1505, -1.0000, -1.0000,  ...,  1.0000, -1.0000, -0.4493]],\n",
      "       device='cuda:0', grad_fn=<TanhBackward0>)}\n"
     ]
    }
   ],
   "source": [
    "unimodal_latents = global_workspace.encode_domains(train_samples)\n",
    "print(unimodal_latents.keys())\n",
    "print(unimodal_latents[frozenset([\"t\"])])"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e9287c3f-9a52-4b1e-a399-2c71e073da69",
   "metadata": {},
   "source": [
    "We see that all domain groups were encoded into latent reprensentations.\n",
    "\n",
    "We now need a selection module (see [https://ruflab.github.io/shimmer/latest/shimmer/modules/selection.html](https://ruflab.github.io/shimmer/latest/shimmer/modules/selection.html)) that defines the policy on how to fuse the different domains in a group.\n",
    "\n",
    "We want to use a selection mechanism that only select the image domain and removes the text domain.\n",
    "\n",
    "All selection mechanism inherit from `SelectionBase` and define the forward which text the unimodal latent representations of the domains, and the pre-fusion encoded representation as parameters:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "7fa1c82f-ddc4-4a6c-893c-936e7bea6d86",
   "metadata": {},
   "outputs": [],
   "source": [
    "class OnlyVisionSelection(SelectionBase):\n",
    "    def forward(\n",
    "        self, domains: Mapping[str, torch.Tensor], encodings_pre_fusion: Mapping[str, torch.Tensor]\n",
    "    ) -> dict[str, torch.Tensor]:\n",
    "        # If only one domain, we keep it\n",
    "        if len(domains) == 1:\n",
    "            return {domain: torch.ones(32).to(device) for domain in domains}\n",
    "        else:\n",
    "            selection: dict[str, torch.Tensor] = {}\n",
    "            for domain in domains:\n",
    "                # We only keep the visual latents\n",
    "                if domain == \"v_latents\":\n",
    "                    selection[\"v_latents\"] = torch.ones(32).to(device)\n",
    "                else:\n",
    "                    # and set 0 to all other domain\n",
    "                    selection[domain] = torch.zeros(32).to(device)\n",
    "            return selection"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "f7f7db2d-6da8-4fba-adb2-7291951915f0",
   "metadata": {},
   "outputs": [],
   "source": [
    "selection_mod = OnlyVisionSelection()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "b245799a-e028-4d5b-af3f-54ecd679f871",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "dict_keys([frozenset({'t'}), frozenset({'v_latents'}), frozenset({'t', 'v_latents'})])\n"
     ]
    }
   ],
   "source": [
    "gw_latents = global_workspace.encode_and_fuse(unimodal_latents, selection_mod)\n",
    "print(gw_latents.keys())"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "58363ce7-c0aa-4bce-9fd1-a013bfaef738",
   "metadata": {},
   "source": [
    "Finally we can decode all latents and look at reconstructed domains."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "cf4b3fa6-0ffc-45e4-90ee-14fbfbc963b2",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "dict_keys([frozenset({'t'}), frozenset({'v_latents'}), frozenset({'t', 'v_latents'})])\n",
      "dict_keys([frozenset({'t'}), frozenset({'v_latents'}), frozenset({'t', 'v_latents'})])\n"
     ]
    }
   ],
   "source": [
    "decoded_latents = global_workspace.decode(gw_latents)\n",
    "print(decoded_latents.keys())\n",
    "reconstructed_domains = global_workspace.decode_domains(decoded_latents)\n",
    "print(reconstructed_domains.keys())\n",
    "# We keep the one that are paired. As the selection mechanism only keep the visual domains,\n",
    "# the text reconstructions are translations, and the images are demi-cycles.\n",
    "reconstructed_texts = reconstructed_domains[frozenset([\"t\", \"v_latents\"])][\"t\"]\n",
    "reconstructed_images = reconstructed_domains[frozenset([\"t\", \"v_latents\"])][\"v_latents\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "4e73a2f3-536c-400f-996f-f42e69063086",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "dict_keys(['token_dist', 'tokens'])"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "reconstructed_texts.keys()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "b566dfd3-f891-4a7e-9786-2459e86de857",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "torch.Size([32, 12])"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "reconstructed_images.size()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "cc14e8b5-f769-477a-91de-93dafe48e8e7",
   "metadata": {},
   "source": [
    "First let's look at the original images, and their reconstructions (demi-cycle):"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "8cafc661-29a6-444b-94bb-fa22580b3bf7",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "original_images = visual_module.decode_images(train_samples[frozenset([\"v_latents\", \"t\"])][\"v_latents\"])\n",
    "plot_image_grid(original_images)\n",
    "decoded_images = visual_module.decode_images(reconstructed_images)\n",
    "plot_image_grid(decoded_images)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4b2f4447-6a03-430b-8240-b3af3074a866",
   "metadata": {},
   "source": [
    "And now let's look at the predicted texts (translations):"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "c311bc41-ce50-4e12-9c16-aa790f903b6b",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1 The image is a thistle colored tiny kite, and is pointing towards the bottom-left corner, and is located at the center left side.\n",
      "2 The image represents a large shape in dark slate blue color pointing towards the right top-right corner, and is at the upper side right, on the very right side, and looks like a bullet.\n",
      "3 The image contains a tiny forest green colored shape pointing towards the top-right corner, it is located at the lower side left side, it looks like a kite.\n",
      "4 The image contains a medium sized silver colored object pointing towards the bottom-right corner, and it is located slightly bottom, at the center, and looks like a bullet.\n",
      "5 The image represents a tiny dark slate grey colored object pointing towards the top-left corner, and it is located at the bottom, on the left side, and looks like a bullet.\n",
      "6 The image contains a medium sized dark goldenrod colored triangle, and is pointing towards the left bottom-left corner, and is located slightly top, in the center.\n",
      "7 The image represents a medium sized dark grey colored object pointing towards the top-right corner, and it is located at the lower side left side, and looks like a deformed square shape.\n",
      "8 The image contains a medium sized object in wheat color pointing towards the bottom bottom-right corner, it is located slightly upper side, at the center, it looks like a triangle.\n",
      "9 The image is a small dark sea green colored object pointing towards the right top-right corner, it is located at the upper side right, on the very right side, and looks like a kite.\n",
      "10 The image represents a large indigo colored shape pointing to the bottom bottom-left corner, and is at the lower side center, and looks like a kite.\n",
      "11 The image represents a medium sized shape in indigo color pointing towards the right side, and is located in the middle right, on the very right side, and looks like a four-sided shape.\n",
      "12 The image represents a medium slate blue colored medium sized isosceles triangle, and is pointing to the left bottom-left, and is located at the bottom.\n",
      "13 The image contains a medium sized shape in dark slate grey color pointing towards the top-left corner, and is located slightly top, in the center, and looks like a bullet.\n",
      "14 The image contains a tiny shape in plum color pointing to the bottom, and is at the very top, on the left side, and it looks like an isosceles triangle.\n",
      "15 The image is a saddle brown colored isosceles triangle, and it is pointing towards the bottom-left corner, and it is at the upper side left, and it is small.\n",
      "16 The image represents a tiny dark grey colored shape pointing towards the top-right corner, it is at the very bottom, at the right side, and looks like a bullet.\n",
      "17 The image represents a large object in dark slate blue color pointing towards the top-right corner, and it is at the very bottom, in the middle, and looks like a bullet.\n",
      "18 The image contains a plum colored medium sized bullet, it is pointing towards the top-left corner, it is at the very bottom, at the right side.\n",
      "19 The image represents a little slate grey colored shape pointing towards the top top-left corner, it is located in the center left side, and it looks like a guitar pick.\n",
      "20 The image contains an average sized object in olive drab color pointing towards the top-left corner, and it is located at the very upper side, slightly right, and looks like a bullet.\n",
      "21 The image represents a medium sized grey colored shape pointing towards the bottom bottom-right corner, it is at the very upper side left, on the very left side, it looks like a deformed square shape.\n",
      "22 The image represents a tiny shape in indian red color pointing towards the top-right corner, and is located slightly upper side, at the left side, and it looks like a guitar pick.\n",
      "23 The image contains a pale green colored large deformed square shape, and is pointing towards the right top-right corner, and is at the very top, slightly left.\n",
      "24 The image is a large maroon colored object pointing towards the bottom, and is located in the center left side, and looks like a kite.\n",
      "25 The image represents a medium sized khaki colored shape pointing towards the bottom-left corner, and is located at the upper side right side, and looks like a bullet.\n",
      "26 The image represents a medium sized light blue colored isosceles triangle, and is pointing to the left, and is located at the top, slightly right.\n",
      "27 The image represents a dark blue colored little four-sided shape, and it is pointing towards the left bottom-left corner, and it is at the very bottom, slightly right.\n",
      "28 The image represents a tiny shape in linen color pointing towards the top top-left corner, it is at the very bottom, slightly right, and looks like a triangle.\n",
      "29 The image is a silver colored large deformed square shape, it is pointing to the bottom-left corner, it is at the bottom, on the left side.\n",
      "30 The image is a small old lace colored object pointing towards the top, it is at the bottom right, on the very right side, and looks like a bullet.\n",
      "31 The image contains a little medium orchid colored object pointing towards the bottom-right corner, it is at the very top, on the left side, and looks like a guitar pick.\n",
      "32 The image is a tiny dark slate grey colored object pointing towards the top-right corner, and it is located in the center, on the left side, it looks like an isosceles triangle.\n"
     ]
    }
   ],
   "source": [
    "reconstructed_text_strings = get_text_samples(reconstructed_texts)\n",
    "for k, text in enumerate(reconstructed_text_strings, 1):\n",
    "    print(k, text.replace(\"<pad>\", \"\"))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "545130d4-4798-47be-9a2a-f647e6122d45",
   "metadata": {},
   "source": [
    "## Interactive play\n",
    "\n",
    "Here, you can do a translation from an image to the text. You can create the image with different attribute, and see how the global workspace translates this image into a description."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "42c71c7a-9046-4d90-8a76-d94e684d7434",
   "metadata": {},
   "outputs": [],
   "source": [
    "import io\n",
    "import math\n",
    "\n",
    "import ipywidgets as widgets\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "from ipywidgets import interact, interact_manual\n",
    "from PIL import Image\n",
    "from shimmer_ssd.logging import attribute_image_grid\n",
    "from torch.nn.functional import one_hot\n",
    "from torchvision.transforms.functional import to_tensor\n",
    "\n",
    "from simple_shapes_dataset.cli import generate_image\n",
    "%matplotlib widget"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "id": "a0707356-68e1-414a-942e-f66107068a98",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "f5d056ad57e84d2ab2fa0c23055a3b6c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "interactive(children=(Dropdown(description='cat', options=('Triangle', 'Egg', 'Diamond'), value='Triangle'), F…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\n",
    "global_workspace.to(device)\n",
    "\n",
    "cat2idx = {\"Diamond\": 0, \"Egg\": 1, \"Triangle\": 2}\n",
    "\n",
    "\n",
    "def get_image(cat, x, y, size, rot, color_r, color_g, color_b):\n",
    "    fig, ax = plt.subplots(figsize=(32, 32), dpi=1)\n",
    "    # The dataset generatoion tool has function to generate a matplotlib shape\n",
    "    # from the attributes. \n",
    "    generate_image(\n",
    "        ax,\n",
    "        cat2idx[cat],\n",
    "        [int(x * 18 + 7), int(y * 18 + 7)],\n",
    "        size * 7 + 7,\n",
    "        rot * 2 * math.pi,\n",
    "        np.array([color_r * 255, color_g * 255, color_b * 255]),\n",
    "        imsize=32,\n",
    "    )\n",
    "    ax.set_facecolor(\"black\")\n",
    "    plt.tight_layout(pad=0)\n",
    "    # Return this as a PIL Image.\n",
    "    # This is to have the same dpi as saved images\n",
    "    # otherwise matplotlib will render this in very high quality\n",
    "    buf = io.BytesIO()\n",
    "    fig.savefig(buf)\n",
    "    buf.seek(0)\n",
    "    image = Image.open(buf)\n",
    "    plt.close(fig)\n",
    "    return image\n",
    "\n",
    "\n",
    "@interact(\n",
    "    cat=[\"Triangle\", \"Egg\", \"Diamond\"],\n",
    "    x=(0, 1, 0.1),\n",
    "    y=(0, 1, 0.1),\n",
    "    rot=(0, 1, 0.1),\n",
    "    size=(0, 1, 0.1),\n",
    "    color_r=(0, 1, 0.1),\n",
    "    color_g=(0, 1, 0.1),\n",
    "    color_b=(0, 1, 0.1),\n",
    ")\n",
    "def play_with_gw(\n",
    "    cat: str = \"Triangle\",\n",
    "    x: float = 0.5,\n",
    "    y: float = 0.5,\n",
    "    rot: float = 0.5,\n",
    "    size: float = 0.5,\n",
    "    color_r: float = 1,\n",
    "    color_g: float = 0,\n",
    "    color_b: float = 0,\n",
    "):\n",
    "    fig, ax = plt.subplots()\n",
    "    image = get_image(cat, x, y, size, rot, color_r, color_g, color_b)\n",
    "    ax.set_facecolor(\"black\")\n",
    "    ax.set_title(\"Original image from attributes\")\n",
    "    ax.set_xticks([])\n",
    "    ax.set_yticks([])\n",
    "    ax.imshow(image)\n",
    "    plt.show()\n",
    "\n",
    "    image_latent = visual_module.visual_module.encode(to_tensor(image)[:3].unsqueeze(0).to(device))\n",
    "    attr_gw_latent = global_workspace.gw_mod.encode({\"v_latents\": global_workspace.encode_domain(image_latent, \"v_latents\")})\n",
    "    gw_latent = global_workspace.gw_mod.fuse(\n",
    "        attr_gw_latent, {\"v_latents\": torch.ones(attr_gw_latent[\"v_latents\"].size(0)).to(device)}\n",
    "    )\n",
    "    decoded_latents = global_workspace.gw_mod.decode(gw_latent)[\"t\"]\n",
    "    decoded_texts = global_workspace.decode_domain(decoded_latents, \"t\")\n",
    "    reconstructed_text_strings = get_text_samples(decoded_texts)[0]\n",
    "    print(reconstructed_text_strings.replace(\"<pad>\", \"\"))"
   ]
  }
 ],
 "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": "3.11.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}