chore: update image links and weight (#66)

Author: ydcjeff

tutorials/advanced/01-collective-communication.ipynb

@@ -43,7 +43,7 @@
       "source": [
         "<!-- ---\n",
         "title: Collective Communication with Ignite\n",
-        "weight: 5\n",
+        "weight: 1\n",
         "date: 2021-10-5\n",
         "downloads: true\n",
         "sidebar: true\n",
@@ -122,7 +122,7 @@
       "source": [
         "## All Reduce\n",
         "\n",
-        "![All Reduce Diagram](https://github.com/pytorch-ignite/examples/blob/add-collective-comm-nb/tutorials/assets/all-reduce.png?raw=1)\n",
+        "![All Reduce Diagram](https://github.com/pytorch-ignite/examples/blob/main/tutorials/assets/all-reduce.png?raw=1)\n",
         "\n",
         "The [`all_reduce()`](https://pytorch.org/ignite/distributed.html#ignite.distributed.utils.all_reduce) method is used to collect specified tensors from each process and make them available on every node then perform a specified operation (sum, product, min, max, etc) on them. Let's spawn 3 processes with ranks 0, 1 and 2 and define a `tensor` on all of them. If we performed `all_reduce` with the operation SUM on `tensor` then `tensor` on all ranks will be gathered, added and stored in `tensor` as shown below:"
       ]
@@ -224,7 +224,7 @@
       "source": [
         "## All Gather\n",
         "\n",
-        "![All Gather Diagram](https://github.com/pytorch-ignite/examples/blob/add-collective-comm-nb/tutorials/assets/all-gather.png?raw=1)\n",
+        "![All Gather Diagram](https://github.com/pytorch-ignite/examples/blob/main/tutorials/assets/all-gather.png?raw=1)\n",
         "\n",
         "The [`all_gather()`](https://pytorch.org/ignite/distributed.html#ignite.distributed.utils.all_gather) method is used when you just want to collect a tensor, number or string across all participating processes. As a basic example, suppose you have to collect all the different values stored in `num` on all ranks. You can achieve this by using `all_gather` as below:"
       ]
@@ -315,7 +315,7 @@
       "source": [
         "## Broadcast\n",
         "\n",
-        "![Broadcast Diagram](https://github.com/pytorch-ignite/examples/blob/add-collective-comm-nb/tutorials/assets/broadcast.png?raw=1)\n",
+        "![Broadcast Diagram](https://github.com/pytorch-ignite/examples/blob/main/tutorials/assets/broadcast.png?raw=1)\n",
         "\n",
         "The [`broadcast()`](https://pytorch.org/ignite/distributed.html#ignite.distributed.utils.broadcast) method copies a tensor, float or string from a source process to all the other processes. For example, you need to send a message from rank 0 to all other ranks. You can do this by creating the actual message on rank 0 and a placeholder on all other ranks, then broadcast the message mentioning a source rank. You can also use `safe_mode=True` in case the placeholder is not defined on all ranks. "
       ]
@@ -427,4 +427,4 @@
       "outputs": []
     }
   ]
-}
+}