[DLMED] update tutorial doc

Signed-off-by: Nic Ma <[email protected]>

Author: Nic-Ma

acceleration/fast_training_tutorial.ipynb

@@ -20,7 +20,7 @@
     "5. Use MONAI `DiceCE` loss instead of regular `Dice` loss.\n",
     "6. Use MONAI `Novograd` optimizer instead of regular `Adam` optimizer.\n",
     "\n",
-    "With a V100 GPU and the target validation mean dice of `0.95` on the `forground` channel only,  it's approximately `200x` speedup compared with the Pytorch regular implementation when achieving same metric. And every epoch is `20x` faster than regular training.\n",
+    "With a V100 GPU and the target validation `mean dice = 0.94` of the `forground` channel only,  it's approximately `40x` speedup compared with the Pytorch regular implementation when achieving the same metric. And every epoch is `20x` faster than regular training.\n",
     "\n",
     "It's modified from the Spleen 3D segmentation tutorial notebook, the Spleen dataset can be downloaded from http://medicaldecathlon.com/.\n",
     "\n",