[Fig05]-lin+sftm-introspection.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Fig 05"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "feeds/mem-introspection-analysis/decent-galaxy-15-model:v0 | val_acc: 99.969\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\u001b[34m\u001b[1mwandb\u001b[0m:   1 of 1 files downloaded.  \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "T: 64 | p: 128 | seq_len: 10 | model_dim: 128 | no_softmax: False | dataset_type: backward | attention_input: linear\n"
     ]
    }
   ],
   "source": [
    "from data_import import *\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn.decomposition import PCA\n",
    "from plotting import *\n",
    "\n",
    "device = 'cuda:0'\n",
    "\n",
    "df = load_result_table('feeds/mem-introspection-analysis')\n",
    "figure_dir = FIGURE_DIR\n",
    "\n",
    "config = df.loc['decent-galaxy-15']\n",
    "model, all_embbedings = load_model(config)\n",
    "print_specs(config)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/tmp/ipykernel_3825541/3383846626.py:46: UserWarning: *c* argument looks like a single numeric RGB or RGBA sequence, which should be avoided as value-mapping will have precedence in case its length matches with *x* & *y*.  Please use the *color* keyword-argument or provide a 2D array with a single row if you intend to specify the same RGB or RGBA value for all points.\n",
      "  axes[-1].scatter([],[],c=color, label=f\"${i}$\")\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 900x300 with 3 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "n_points = 300\n",
    "x = np.linspace(0,1.0,n_points)\n",
    "y = np.linspace(0,1.0,n_points)\n",
    "xx, yy = np.meshgrid(x,y)\n",
    "xx = xx.flatten()\n",
    "yy = yy.flatten()\n",
    "mask = xx+yy <= 1\n",
    "#xx = xx[mask]\n",
    "#yy = yy[mask]\n",
    "grid = np.stack([xx,yy,1-xx-yy], axis=1) \n",
    "# todo: changed\n",
    "xx = grid[:,0] \n",
    "yy = grid[:,1] \n",
    "zz = grid[:,2] \n",
    "# compute the sum of the embeddings for each point in the grid\n",
    "grid = torch.tensor(grid, dtype=torch.float32)\n",
    "embeddings = (grid[:,0] *all_embbedings[0].reshape(-1,1)\n",
    "              + grid[:,1] * all_embbedings[1].reshape(-1,1) \n",
    "             + grid[:,2] * all_embbedings[2].reshape(-1,1) \n",
    "            ) # to permute randomly [torch.randperm(64)]\n",
    "k=3\n",
    "max_items_per_row = min(7,k)\n",
    "fig, axes = plt.subplots(k//max_items_per_row,max_items_per_row, figsize=(3*max_items_per_row,3*k//max_items_per_row),sharex=True, sharey=True)\n",
    "axes=axes.flatten()\n",
    "token_names = ['B','C','D']\n",
    "# get the tab10 colors in a list\n",
    "colors = plt.get_cmap('tab20b').colors\n",
    "from matplotlib.colors import ListedColormap\n",
    "cmap = ListedColormap(colors[:config['seq_len']])\n",
    "for i in range(k):\n",
    "    enc_output = all_embbedings[i]\n",
    "    enc_output = enc_output.repeat(len(xx),1)\n",
    "    c = model.get_output(embeddings.T.to(device), enc_output.to(device)).argmax(dim=1).cpu().detach().numpy()\n",
    "    #im = axes[i].scatter(xx,yy,c=[colors[v] for v in c])\n",
    "    axes[i].pcolor(x, y, np.ma.masked_where(~mask.reshape(n_points,n_points),c.reshape(n_points,n_points)), cmap=cmap)\n",
    "    axes[i].set_xlabel(f'$\\\\alpha_{token_names[0]}$')\n",
    "    \n",
    "    # add text that says hi in the top right corner\n",
    "    if i == 0:\n",
    "        axes[i].text(0.45, 0.85, f\"$\\\\mathcal{{T}} = \\\\{{ {','.join(token_names)} \\\\}}$\")\n",
    "        axes[i].text(0.45, 0.75, f\"$\\\\alpha_{token_names[2]} = 1 - \\\\alpha_{token_names[0]} - \\\\alpha_{token_names[1]}$\")\n",
    "    axes[i].set_title(f\"$f\\left(\\sum_{{\\mathcal{{T}}}}\\\\alpha_t e_t + e_{token_names[i]}\\\\right)$\")\n",
    "for i, color in enumerate(colors[:config['seq_len']]):\n",
    "    if i ==0:\n",
    "        continue\n",
    "    axes[-1].scatter([],[],c=color, label=f\"${i}$\")\n",
    "axes[0].set_ylabel(f'$\\\\alpha_{token_names[1]}$')\n",
    "plt.legend(title=r'$\\operatorname{argmax}_c f_c$')\n",
    "plt.savefig(figure_dir / 'FF_lincomb_linear_model.png', bbox_inches='tight')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "def highlight_cell(x,y,color,ax):\n",
    "    # Given a coordinate (x,y), highlight the corresponding cell using a colored frame in the ac\n",
    "    # after having called imshow already\n",
    "    rect = plt.Rectangle((x-.5, y-.5), 1,1, fill=False,color=color,lw=2)\n",
    "    ax.add_patch(rect)\n",
    "    return rect\n",
    "\n",
    "def adapt_to(x,ax,data,has_BOS=False):\n",
    "    for i, x_i in enumerate(x):\n",
    "        for j, x_j in enumerate(x):\n",
    "            if x_i == x_j:\n",
    "                highlight_cell(i,j, color='red',ax=ax)\n",
    "    for k in range(data.shape[0]):\n",
    "            for j in range(data.shape[1]):\n",
    "                ax.text(j, k, f'{int(np.round(data[k, j]*100))}', ha='center', va='center', color='white')\n",
    "    if has_BOS:\n",
    "        rect = plt.Rectangle((0-.5, 0-.5), data.shape[0],1, fill=False,color='cyan',lw=3)\n",
    "        ax.add_patch(rect)\n",
    "        rect = plt.Rectangle((0-.5, 0-.5), 1,data.shape[1], fill=False,color='cyan',lw=3)\n",
    "        ax.add_patch(rect)\n",
    "        \n",
    "    alpha = 'ABCDEFGHIJKLMNOPQRSTUVW'\n",
    "    alpha = alpha + (config['T']-len(alpha))*' ' + '$'\n",
    "    ax.set_xticks(np.arange(len(x)), [alpha[a] for a in x],fontsize=13)\n",
    "    ax.set_yticks(np.arange(len(x)), [alpha[a] for a in x],fontsize=13)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[1.        1.        1.        0.9999999 1.        1.        1.0000001\n",
      " 1.0000001 1.        1.       ]\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 520x500 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, axes = plt.subplots(1, 2, figsize=(5.2, 5),width_ratios=[5,0.2])\n",
    "\n",
    "vmax = 1.0\n",
    "cmap = ATTENTION_SCORE_CMAP\n",
    "\n",
    "x_orig = [1,1,2,1,1,1,3,4,4,4]\n",
    "\n",
    "# LINEAR\n",
    "ax = axes[0]\n",
    "ax.set_title('linear+sftm')\n",
    "X = torch.tensor(x_orig).unsqueeze(0).to(device)\n",
    "model(X)\n",
    "data = model.attn_probs.detach().cpu().numpy()[0]\n",
    "print(data.sum(axis=1))\n",
    "im = ax.imshow(data, cmap=cmap,vmin=0.0,vmax=vmax)\n",
    "adapt_to(x_orig,ax,data)\n",
    "\n",
    "fig.colorbar(im, cax=axes[-1], label=r'$A_{ij}$')\n",
    "\n",
    "plt.savefig(figure_dir / 'linear_matrix.pdf', bbox_inches='tight')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "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": "3.8.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}