[Docker] Add patches for one-hot, expand_as, and etc ops (#21)

In this PR, we update the dockerfile and image to include:
1. support one hot op.
2. support expand as.
3. some minor updates.

As the result, it is highly recommended to install the patched PyTorch.

README.md

@@ -12,7 +12,7 @@ We are refactoring our [internal LLM training system](https://arxiv.org/abs/2402
 
 ### From Source
 
-#### Install a Patched Version of PyTorch (optional)
+#### Install a Patched Version of PyTorch
 
 ```bash
 bash patches/build_pytorch_w_patch.sh
@@ -60,4 +60,4 @@ The veScale Project is under the Apache License v2.0.
 ## Acknowledgement
 
 veScale team would like to sincerely acknowledge the assistance of and collaboration with
-the [PyTorch DTensor team](https://github.com/pytorch/pytorch/tree/main/torch/distributed/_tensor).
+the [PyTorch DTensor team](https://github.com/pytorch/pytorch/tree/main/torch/distributed/_tensor).