{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "vision_models-01",
   "metadata": {},
   "source": "# Configure vision models and backends\n\n`VisionModel` provides one interface over `transformers`, `timm`, and `thingsvision` for layer-level activation extraction.\n\n| Backend | Install extra | Example model ID |\n|---|---|---|\n| `transformers` | `vneurotk[vision]` | `facebook/dinov2-base` |\n| `timm` | `vneurotk[timm]` | `resnet50.a1_in1k` |\n| `thingsvision` | `vneurotk[thingsvision]` | backend-specific |\n\nModel construction can access the network unless weights and processor/config files are already cached. The cells below require the selected optional extra and model assets; the documentation build does not execute them."
  },
  {
   "cell_type": "markdown",
   "id": "vision_models-02",
   "metadata": {},
   "source": [
    "## Load a model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "vision_models-03",
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "\n",
    "import vneurotk as vtk\n",
    "\n",
    "device = \"cuda\" if torch.cuda.is_available() else \"cpu\"\n",
    "model = vtk.VisionModel(\n",
    "    \"facebook/dinov2-base\",\n",
    "    backend=\"transformers\",\n",
    "    device=device,\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "vision_models-04",
   "metadata": {},
   "source": [
    "VneuroTK passes model identifiers to the selected backend unchanged and does not silently fall back to another backend or model."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "vision_models-05",
   "metadata": {},
   "source": [
    "## Inspect and select layers"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "vision_models-06",
   "metadata": {},
   "outputs": [],
   "source": [
    "model.print_modules(max_depth=3)\n",
    "model.set_selector(module_type=\"Dinov2Layer\")\n",
    "model.module_names"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "vision_models-07",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Select by name, or take a union of type and name matches.\n",
    "model.set_selector(module_type=\"Dinov2Layer\", module_name=\"layernorm\")\n",
    "\n",
    "# A custom list can be assembled from model.module_list.\n",
    "last_norm = [item for item in model.module_list if item.module_type == \"LayerNorm\"][-1]\n",
    "model.set_selector([last_norm])"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "vision_models-08",
   "metadata": {},
   "source": "Use `vneurotk.print_cached_models()` when you need to inspect locally discoverable assets. Its result depends on the machine cache and is intentionally not stored in this notebook.\n\n## Next steps\n\n- [Extract from an image mapping](vision_alone)\n- [Extract with VneuroTK neural data](vision_union)\n- [Vision API](../api/vision.md)"
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "name": "python",
   "version": "3.11"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}