Merge pull request #62 from masa-su/add_log_prob

Add log_prob
masa-su · Apr 10, 2019 · e27a4e1 · e27a4e1
2 parents 796f0ac + 44319d3
commit e27a4e1
Show file tree

Hide file tree

Showing 33 changed files with 699 additions and 1,210 deletions.
diff --git a/examples/cvae.ipynb b/examples/cvae.ipynb
@@ -183,7 +183,7 @@
       "Distributions (for training): \n",
       "  q(z|x,y), p(x|z,y) \n",
       "Loss function: \n",
-      "  mean(-E_q(z|x,y)[log p(x|z,y)] + KL[q(z|x,y)||p_prior(z)]) \n",
+      "  mean(-(E_q(z|x,y)[log p(x|z,y)]) + KL[q(z|x,y)||p_prior(z)]) \n",
       "Optimizer: \n",
       "  Adam (\n",
       "  Parameter Group 0\n",
@@ -285,7 +285,7 @@
      "name": "stderr",
      "output_type": "stream",
      "text": [
-      "100%|██████████| 469/469 [00:04<00:00, 101.77it/s]"
+      "100%|██████████| 469/469 [00:04<00:00, 95.04it/s]"
      ]
     },
     {
@@ -314,7 +314,7 @@
      "name": "stderr",
      "output_type": "stream",
      "text": [
-      "100%|██████████| 469/469 [00:04<00:00, 110.71it/s]"
+      "100%|██████████| 469/469 [00:04<00:00, 100.45it/s]"
      ]
     },
     {
@@ -343,7 +343,7 @@
      "name": "stderr",
      "output_type": "stream",
      "text": [
-      "100%|██████████| 469/469 [00:04<00:00, 104.91it/s]"
+      "100%|██████████| 469/469 [00:04<00:00, 98.34it/s]"
      ]
     },
     {
@@ -372,7 +372,7 @@
      "name": "stderr",
      "output_type": "stream",
      "text": [
-      "100%|██████████| 469/469 [00:04<00:00, 97.32it/s]"
+      "100%|██████████| 469/469 [00:04<00:00, 101.22it/s]"
      ]
     },
     {
@@ -401,7 +401,7 @@
      "name": "stderr",
      "output_type": "stream",
      "text": [
-      "100%|██████████| 469/469 [00:04<00:00, 106.88it/s]"
+      "100%|██████████| 469/469 [00:04<00:00, 99.79it/s]"
      ]
     },
     {
@@ -430,7 +430,7 @@
      "name": "stderr",
      "output_type": "stream",
      "text": [
-      "100%|██████████| 469/469 [00:04<00:00, 109.72it/s]"
+      "100%|██████████| 469/469 [00:04<00:00, 109.00it/s]"
      ]
     },
     {
@@ -459,7 +459,7 @@
      "name": "stderr",
      "output_type": "stream",
      "text": [
-      "100%|██████████| 469/469 [00:04<00:00, 96.55it/s]"
+      "100%|██████████| 469/469 [00:04<00:00, 97.86it/s]"
      ]
     },
     {
@@ -488,7 +488,7 @@
      "name": "stderr",
      "output_type": "stream",
      "text": [
-      "100%|██████████| 469/469 [00:04<00:00, 101.35it/s]"
+      "100%|██████████| 469/469 [00:04<00:00, 96.30it/s]"
      ]
     },
     {
@@ -517,7 +517,7 @@
      "name": "stderr",
      "output_type": "stream",
      "text": [
-      "100%|██████████| 469/469 [00:04<00:00, 99.68it/s] "
+      "100%|██████████| 469/469 [00:03<00:00, 124.22it/s]"
      ]
     },
     {
@@ -546,14 +546,27 @@
      "name": "stderr",
      "output_type": "stream",
      "text": [
-      "100%|██████████| 469/469 [00:04<00:00, 112.04it/s]\n"
+      "100%|██████████| 469/469 [00:03<00:00, 126.78it/s]"
+     ]
+    },
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Epoch: 10 Train loss: 99.1070\n"
+     ]
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "\n"
      ]
     },
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "Epoch: 10 Train loss: 99.1070\n",
       "Test loss: 99.9338\n"
      ]
     }

diff --git a/examples/distributions.ipynb b/examples/distributions.ipynb
@@ -15,7 +15,7 @@
     {
      "data": {
       "text/plain": [
-       "<torch._C.Generator at 0x7fac2fc9b8b0>"
+       "<torch._C.Generator at 0x7fdf8eaad8d0>"
       ]
      },
      "execution_count": 1,
@@ -315,12 +315,14 @@
     {
      "data": {
       "text/plain": [
-       "{'x': tensor([[-1.1235, -1.1559,  0.4218,  0.8778, -0.1497,  0.2739,  1.1814, -0.7278,\n",
-       "           0.2572,  0.1075, -0.7142, -0.7021,  0.6641, -1.1700, -1.8278, -0.9027,\n",
-       "           0.6691,  0.2645,  0.2566, -0.1142],\n",
-       "         [-0.2431, -0.5863, -0.0452,  2.1263,  0.9091,  0.5982, -0.9394,  0.3520,\n",
-       "          -0.7051,  1.8862,  0.4602, -0.2422, -0.6304,  0.8388,  0.8246,  1.1748,\n",
-       "           0.3473, -0.8007,  0.2327,  0.3098]])}"
+       "{'x': tensor([[-1.1551e+00, -1.2686e+00,  4.2959e-01,  8.6341e-01, -1.2102e-02,\n",
+       "           3.1782e-01,  1.2648e+00, -8.4481e-01,  4.3645e-01,  1.8594e-01,\n",
+       "          -6.2100e-01, -5.1982e-01,  6.1928e-01, -1.1163e+00, -1.8660e+00,\n",
+       "          -9.2985e-01,  5.8562e-01,  2.9533e-01,  2.9360e-01,  1.1179e-03],\n",
+       "         [-2.5069e-01, -6.9296e-01,  1.7690e-02,  2.1784e+00,  8.5724e-01,\n",
+       "           5.4175e-01, -8.0418e-01,  3.2449e-01, -7.0768e-01,  1.9246e+00,\n",
+       "           6.1470e-01, -2.2973e-01, -6.1405e-01,  7.4305e-01,  6.9249e-01,\n",
+       "           1.1372e+00,  3.4499e-01, -6.6509e-01,  1.5735e-01,  2.6749e-01]])}"
       ]
      },
      "execution_count": 12,
@@ -347,26 +349,26 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "tensor([-26.2268, -23.8622], grad_fn=<SumBackward2>)\n",
-      "tensor([-18.9722, -19.5073], grad_fn=<SumBackward2>)\n",
-      "tensor([-45.1698, -41.0407], grad_fn=<AddBackward0>)\n",
+      "tensor([-26.2297, -23.6749], grad_fn=<SumBackward2>)\n",
+      "tensor([-19.4051, -19.5314], grad_fn=<SumBackward2>)\n",
+      "tensor([-45.6209, -40.9298], grad_fn=<AddBackward0>)\n",
       "tensor([-155.4684, -163.4326, -150.2627, -150.2103, -159.1462, -163.7559,\n",
       "        -168.1021, -162.1275, -160.1595, -142.4833], grad_fn=<AddBackward0>)\n"
      ]
     }
    ],
    "source": [
     "outputs = p1.sample({\"y\":y, \"a\":a})\n",
-    "print(p1.log_likelihood(outputs))\n",
+    "print(p1.log_prob().eval(outputs))\n",
     "\n",
     "outputs = p2.sample({\"x\":x, \"y\":y})\n",
-    "print(p2.log_likelihood(outputs))\n",
+    "print(p2.log_prob().eval(outputs))\n",
     "\n",
     "outputs = p3.sample({\"y\":y, \"a\":a})\n",
-    "print(p3.log_likelihood(outputs))\n",
+    "print(p3.log_prob().eval(outputs))\n",
     "\n",
     "outputs = p_all.sample(batch_size=10)\n",
-    "print(p_all.log_likelihood(outputs))"
+    "print(p_all.log_prob().eval(outputs))"
    ]
   },
   {