fix param_bug

Author: xyliao

chapter3_NN/param_initialize.ipynb

@@ -25,7 +25,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 6,
+   "execution_count": 1,
    "metadata": {
     "collapsed": true
    },
@@ -38,7 +38,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 13,
+   "execution_count": 2,
    "metadata": {
     "collapsed": true
    },
@@ -56,7 +56,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 14,
+   "execution_count": 3,
    "metadata": {
     "collapsed": true
    },
@@ -69,23 +69,21 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 15,
-   "metadata": {
-    "collapsed": false
-   },
+   "execution_count": 4,
+   "metadata": {},
    "outputs": [
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
       "Parameter containing:\n",
-      "-0.0221 -0.0171  0.1221  ...  -0.0452 -0.1715 -0.0637\n",
-      "-0.0922 -0.1111  0.0822  ...   0.0316  0.1020 -0.0585\n",
-      "-0.0830  0.1037 -0.0572  ...  -0.1465 -0.1049 -0.0566\n",
+      " 0.1236 -0.1731 -0.0479  ...   0.0031  0.0784  0.1239\n",
+      " 0.0713  0.1615  0.0500  ...  -0.1757 -0.1274 -0.1625\n",
+      " 0.0638 -0.1543 -0.0362  ...   0.0316 -0.1774 -0.1242\n",
       "          ...             ⋱             ...          \n",
-      " 0.1485  0.1137  0.1745  ...   0.0073  0.0887  0.1143\n",
-      " 0.1634 -0.1478  0.0930  ...   0.1418 -0.0501  0.1266\n",
-      " 0.0943 -0.1595 -0.1742  ...  -0.1531  0.0786 -0.1594\n",
+      " 0.1551  0.1772  0.1537  ...   0.0730  0.0950  0.0627\n",
+      " 0.0495  0.0896  0.0243  ...  -0.1302 -0.0256 -0.0326\n",
+      "-0.1193 -0.0989 -0.1795  ...   0.0939  0.0774 -0.0751\n",
       "[torch.FloatTensor of size 40x30]\n",
       "\n"
      ]
@@ -104,7 +102,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 16,
+   "execution_count": 5,
    "metadata": {
     "collapsed": true
    },
@@ -116,23 +114,21 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 17,
-   "metadata": {
-    "collapsed": false
-   },
+   "execution_count": 6,
+   "metadata": {},
    "outputs": [
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
       "Parameter containing:\n",
-      " 4.6904  3.5478  4.0254  ...   3.6078  4.6897  4.8285\n",
-      " 4.6349  3.4475  4.8485  ...   3.8712  3.9396  4.7797\n",
-      " 3.0177  3.4870  4.5741  ...   4.6718  4.2548  4.6343\n",
+      " 4.5768  3.6175  3.3098  ...   4.7374  4.0164  3.3037\n",
+      " 4.1809  3.5624  3.1452  ...   3.0305  4.4444  4.1058\n",
+      " 3.5277  4.3712  3.7859  ...   3.5760  4.8559  4.3252\n",
       "          ...             ⋱             ...          \n",
-      " 3.3116  3.2907  4.5550  ...   3.5882  4.4668  3.6532\n",
-      " 4.2998  4.6337  3.8836  ...   3.1220  4.0567  4.3605\n",
-      " 4.3862  4.5433  4.1909  ...   4.2792  4.7513  3.7076\n",
+      " 4.8983  3.9855  3.2842  ...   4.7683  4.7590  3.3498\n",
+      " 4.9168  4.5723  3.5870  ...   3.2032  3.9842  3.2484\n",
+      " 4.2532  4.6352  4.4857  ...   3.7543  3.9885  4.4211\n",
       "[torch.DoubleTensor of size 40x30]\n",
       "\n"
      ]
@@ -151,10 +147,8 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 18,
-   "metadata": {
-    "collapsed": false
-   },
+   "execution_count": 7,
+   "metadata": {},
    "outputs": [],
    "source": [
     "for layer in net1:\n",
@@ -188,7 +182,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 28,
+   "execution_count": 8,
    "metadata": {
     "collapsed": true
    },
@@ -202,7 +196,7 @@
     "            nn.ReLU()\n",
     "        )\n",
     "        \n",
-    "        self.l1[0].weight.data = torch.randn(30, 40) # 直接对某一层初始化\n",
+    "        self.l1[0].weight.data = torch.randn(40, 30) # 直接对某一层初始化\n",
     "        \n",
     "        self.l2 = nn.Sequential(\n",
     "            nn.Linear(40, 50),\n",
@@ -223,7 +217,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 29,
+   "execution_count": 9,
    "metadata": {
     "collapsed": true
    },
@@ -234,10 +228,8 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 30,
-   "metadata": {
-    "collapsed": false
-   },
+   "execution_count": 10,
+   "metadata": {},
    "outputs": [
     {
      "name": "stdout",
@@ -266,10 +258,8 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 31,
-   "metadata": {
-    "collapsed": false
-   },
+   "execution_count": 11,
+   "metadata": {},
    "outputs": [
     {
      "name": "stdout",
@@ -327,10 +317,8 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 33,
-   "metadata": {
-    "collapsed": false
-   },
+   "execution_count": 12,
+   "metadata": {},
    "outputs": [],
    "source": [
     "for layer in net2.modules():\n",
@@ -356,7 +344,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 34,
+   "execution_count": 13,
    "metadata": {
     "collapsed": true
    },
@@ -367,23 +355,21 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 38,
-   "metadata": {
-    "collapsed": false
-   },
+   "execution_count": 14,
+   "metadata": {},
    "outputs": [
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
       "Parameter containing:\n",
-      " 0.2051 -0.4551  0.7049  ...   0.5223 -0.7658 -0.1899\n",
-      " 0.2562  0.2797  0.0012  ...  -0.5278 -0.4887 -0.8263\n",
-      " 0.4582 -0.1433  0.5009  ...   0.1000 -0.5663  0.1605\n",
+      " 0.8453  0.2891 -0.5276  ...  -0.1530 -0.4474 -0.5470\n",
+      "-0.1983 -0.4530 -0.1950  ...   0.4107 -0.4889  0.3654\n",
+      " 0.9149 -0.5641 -0.6594  ...   0.0734  0.1354 -0.4152\n",
       "          ...             ⋱             ...          \n",
-      " 0.8715  0.4053  0.3679  ...  -0.4733 -0.6270 -0.3325\n",
-      "-0.1898  0.6608  0.1111  ...   0.2294  0.2603 -0.0200\n",
-      "-0.3035  0.1876 -0.5422  ...   0.0505  0.6244 -0.2368\n",
+      "-0.4718 -0.5125 -0.5572  ...   0.0824 -0.6551  0.0840\n",
+      "-0.2374 -0.0036  0.6497  ...   0.7856 -0.1367 -0.8795\n",
+      " 0.0774  0.2609 -0.2358  ...  -0.8196  0.1696  0.5976\n",
       "[torch.DoubleTensor of size 40x30]\n",
       "\n"
      ]
@@ -395,26 +381,24 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 39,
-   "metadata": {
-    "collapsed": false
-   },
+   "execution_count": 15,
+   "metadata": {},
    "outputs": [
     {
      "data": {
       "text/plain": [
        "Parameter containing:\n",
-       "-0.0449 -0.2140  0.2820  ...  -0.2266  0.0365 -0.1897\n",
-       "-0.0313  0.1128  0.1789  ...  -0.1731  0.0590 -0.1085\n",
-       "-0.0347 -0.1429 -0.1646  ...   0.0212  0.1731 -0.0251\n",
+       "-0.2114  0.2704 -0.2186  ...   0.1727  0.2158  0.0775\n",
+       "-0.0736 -0.0565  0.0844  ...   0.1793  0.2520 -0.0047\n",
+       " 0.1331 -0.1843  0.2426  ...  -0.2199 -0.0689  0.1756\n",
        "          ...             ⋱             ...          \n",
-       " 0.0902 -0.1555  0.0562  ...  -0.0109 -0.2192 -0.1540\n",
-       "-0.1491 -0.2610 -0.2453  ...   0.2201  0.2257  0.1047\n",
-       " 0.0297  0.1414 -0.0139  ...  -0.1209 -0.0193 -0.1731\n",
+       " 0.2751 -0.1404  0.1225  ...   0.1926  0.0175 -0.2099\n",
+       " 0.0970 -0.0733 -0.2461  ...   0.0605  0.1915 -0.1220\n",
+       " 0.0199  0.1283 -0.1384  ...  -0.0344 -0.0560  0.2285\n",
        "[torch.DoubleTensor of size 40x30]"
       ]
      },
-     "execution_count": 39,
+     "execution_count": 15,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -425,23 +409,21 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 40,
-   "metadata": {
-    "collapsed": false
-   },
+   "execution_count": 16,
+   "metadata": {},
    "outputs": [
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
       "Parameter containing:\n",
-      "-0.0449 -0.2140  0.2820  ...  -0.2266  0.0365 -0.1897\n",
-      "-0.0313  0.1128  0.1789  ...  -0.1731  0.0590 -0.1085\n",
-      "-0.0347 -0.1429 -0.1646  ...   0.0212  0.1731 -0.0251\n",
+      "-0.2114  0.2704 -0.2186  ...   0.1727  0.2158  0.0775\n",
+      "-0.0736 -0.0565  0.0844  ...   0.1793  0.2520 -0.0047\n",
+      " 0.1331 -0.1843  0.2426  ...  -0.2199 -0.0689  0.1756\n",
       "          ...             ⋱             ...          \n",
-      " 0.0902 -0.1555  0.0562  ...  -0.0109 -0.2192 -0.1540\n",
-      "-0.1491 -0.2610 -0.2453  ...   0.2201  0.2257  0.1047\n",
-      " 0.0297  0.1414 -0.0139  ...  -0.1209 -0.0193 -0.1731\n",
+      " 0.2751 -0.1404  0.1225  ...   0.1926  0.0175 -0.2099\n",
+      " 0.0970 -0.0733 -0.2461  ...   0.0605  0.1915 -0.1220\n",
+      " 0.0199  0.1283 -0.1384  ...  -0.0344 -0.0560  0.2285\n",
       "[torch.DoubleTensor of size 40x30]\n",
       "\n"
      ]
@@ -472,9 +454,9 @@
  ],
  "metadata": {
   "kernelspec": {
-   "display_name": "mx",
+   "display_name": "Python 3",
    "language": "python",
-   "name": "mx"
+   "name": "python3"
   },
   "language_info": {
    "codemirror_mode": {
@@ -486,7 +468,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.6.0"
+   "version": "3.6.3"
   }
  },
  "nbformat": 4,