Update description in fast_training_tutorial (Project-MONAI#1509)

Fixes Project-MONAI#1404.

### Description
Update description in `fast_training_tutorial`

### Checks
<!--- Put an `x` in all the boxes that apply, and remove the not
applicable items -->
- [ ] Avoid including large-size files in the PR.
- [ ] Clean up long text outputs from code cells in the notebook.
- [ ] For security purposes, please check the contents and remove any
sensitive info such as user names and private key.
- [ ] Ensure (1) hyperlinks and markdown anchors are working (2) use
relative paths for tutorial repo files (3) put figure and graphs in the
`./figure` folder
- [ ] Notebook runs automatically `./runner.sh -t <path to .ipynb file>`

Signed-off-by: KumoLiu <[email protected]>

Author: KumoLiu

acceleration/fast_model_training_guide.md

@@ -281,7 +281,7 @@ train_transforms = [
 ]
 dataset = CacheDataset(..., transform=train_trans)
 ```
-Here we convert to PyTorch `Tensor` with `EnsureTyped` transform and move data to GPU with `ToDeviced` transform. `CacheDataset` caches the transform results until `ToDeviced`, so it is in GPU memory. Then in every epoch, the program fetches cached data from GPU memory and only execute the random transform `RandCropByPosNegLabeld` on GPU directly.
+Here we convert to PyTorch `Tensor` and move data to GPU with `EnsureTyped` transform. `CacheDataset` caches the transform results until `EnsureTyped`, so it is in GPU memory. Then in every epoch, the program fetches cached data from GPU memory and only execute the random transform `RandCropByPosNegLabeld` on GPU directly.
 GPU caching example is available at [Spleen fast training tutorial](fast_training_tutorial.ipynb).
 
 ## Leveraging multi-GPU distributed training

acceleration/fast_training_tutorial.ipynb

@@ -385,7 +385,7 @@
     "For MONAI fast training progress, we mainly introduce the following features:\n",
     "1. `AMP` (auto mixed precision): AMP is an important feature released in PyTorch v1.6, NVIDIA CUDA 11 added strong support for AMP and significantly improved training speed.\n",
     "2. `CacheDataset`: Dataset with the cache mechanism that can load data and cache deterministic transforms' result during training.\n",
-    "3. `ToDeviced` transform: to move data to GPU and cache with `CacheDataset`, then execute random transforms on GPU directly, avoid CPU -> GPU sync in every epoch. Please note that not all the MONAI transforms support GPU operation so far, still working in progress.\n",
+    "3. `EnsureTyped` transform: to move data to GPU and cache with `CacheDataset`, then execute random transforms on GPU directly, avoid CPU -> GPU sync in every epoch. Please note that not all the MONAI transforms support GPU operation so far, still working in progress.\n",
     "4. `set_track_meta(False)`: to disable meta tracking in the random transforms to avoid unnecessary computation.\n",
     "5. `ThreadDataLoader`: uses multi-threads instead of multi-processing, faster than `DataLoader` in light-weight task as we already cached the results of most computation.\n",
     "6. `DiceCE` loss function: computes Dice loss and Cross Entropy Loss, returns the weighted sum of these two losses.\n",