PyTorch now requires float Tensor for gradient computation

Author: Atcold

03-autograd_tutorial.ipynb

@@ -221,7 +221,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "x = torch.arange(1, n + 1, requires_grad=True)\n",
+    "x = torch.arange(1., n + 1, requires_grad=True)\n",
     "w = torch.ones(n, requires_grad=True)\n",
     "z = w @ x\n",
     "z.backward()\n",
@@ -234,7 +234,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "x = torch.arange(1, n + 1)\n",
+    "x = torch.arange(1., n + 1)\n",
     "w = torch.ones(n, requires_grad=True)\n",
     "z = w @ x\n",
     "z.backward()\n",
@@ -248,7 +248,7 @@
    "outputs": [],
    "source": [
     "with torch.no_grad():\n",
-    "    x = torch.arange(1, n + 1)\n",
+    "    x = torch.arange(1., n + 1)\n",
     "    w = torch.ones(n, requires_grad=True)\n",
     "    z = w @ x\n",
     "    z.backward()\n",
@@ -283,7 +283,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.6.5"
+   "version": "3.6.6"
   }
  },
  "nbformat": 4,