Remove incorrect links to zdevito/ATen (pytorch#50065)

Summary:
Similar to pytorch#49028, this PR removes a few more references to https://github.com/zdevito/ATen.

- The links for Functions.h, Tensor.h, and Type.h are simply broken, probably because they refer to `master` rather than a specific commit (cf. pytorch#47066)
- I'm unsure about the change to the `about` section of `aten/conda/meta.yaml`; can someone comment on whether I am understanding that field correctly?
- The reference to zdevito/ATen#163 remains [in `tools/autograd/derivatives.yaml`](https://github.com/pytorch/pytorch/blob/cd608fe59b70fa7cafb07110096b2e023a8b6e9c/tools/autograd/derivatives.yaml#L91), because the contents of that issue discussion don't seem to be mirrored anywhere else.

Pull Request resolved: pytorch#50065

Reviewed By: ezyang, walterddr

Differential Revision: D25767353

Pulled By: samestep

fbshipit-source-id: 265f46f058bc54ef6d1a77f112cdfa1f115b3247

Author: samestep

aten/conda/meta.yaml

@@ -24,7 +24,7 @@ requirements:
     - mkl # [not osx]
 
 about:
-  home: https://github.com/zdevito/ATen
+  home: https://github.com/pytorch/pytorch
   license: BSD
   summary: A TENsor library for C++14
 

caffe2/contrib/aten/README.md

@@ -1,15 +1,15 @@
 # An ATen operator for Caffe2
 
-[ATen](https://github.com/zdevito/aten) is a simple tensor library thats exposes the Tensor operations in Torch
+ATen is a simple tensor library thats exposes the Tensor operations in Torch
 and PyTorch directly in C++14. This library provides a generated wrapper around the ATen API
 that makes these functions available in Caffe2 as an operator. It also makes it accessible using the
 ToffeeIR.
 
 
 ### Example Usage in Caffe2
 
-First identify a function in ATen you want to call in [Functions.h](https://github.com/zdevito/ATen/blob/master/doc/Functions.h),
-[Tensor.h](https://github.com/zdevito/ATen/blob/master/doc/Tensor.h), or [Type.h](https://github.com/zdevito/ATen/blob/master/doc/Type.h).
+First identify a function in ATen you want to call in Functions.h,
+Tensor.h, or Type.h.
 
 We will call the `pow` operator:
 

caffe2/contrib/aten/docs/pytorch_to_caffe2.md

@@ -6,7 +6,7 @@ operators that haven't been standardized yet, or custom `torch.autograd.Function
 are specific to a network.
 
 To bridge this gap, we provide an experimental operator in ONNX that allows you to directly access PyTorch's tensor functions using the ATen library.
-[ATen](https://github.com/zdevito/aten) is the underlying C++ library that PyTorch uses to do tensor operations. Caffe2 has an [ATen operator](https://github.com/caffe2/caffe2/tree/master/caffe2/contrib/aten)
+[ATen](https://github.com/pytorch/pytorch/tree/master/aten) is the underlying C++ library that PyTorch uses to do tensor operations. Caffe2 has an [ATen operator](https://github.com/pytorch/pytorch/tree/master/caffe2/contrib/aten)
 that can run these tensor functions in a Caffe2 network after importing them through ONNX.
 
 This guide explains how to configure Caffe2 and modify your PyTorch program to use
@@ -61,8 +61,8 @@ We can add a `symbolic` method to it like so:
 
 The function `graph.at` adds a new ATen op the computation graph.
 You can call any ATen function using this facility. To do so,
-first identify a function in ATen you want to call in [Functions.h](https://github.com/zdevito/ATen/blob/master/doc/Functions.h),
-[Tensor.h](https://github.com/zdevito/ATen/blob/master/doc/Tensor.h), or [Type.h](https://github.com/zdevito/ATen/blob/master/doc/Type.h).
+first identify a function in ATen you want to call in Functions.h,
+Tensor.h, or Type.h.
 
 As an example, we might want to call the `pow` operator:
 
@@ -86,9 +86,9 @@ To call methods of ATen's `Type` objects, you provide an additional string attri
 that determines the type. For instance, `ones` creates a new constant tensor of all ones:
 ```
 class Type {
-	...
-	virtual Tensor ones(IntArrayRef size) const;
-	...
+  ...
+  virtual Tensor ones(IntArrayRef size) const;
+  ...
 };
 ```