Remove extra overrides for non-leaf nodes.

[ailzhang]

Note that when removing extra overrides, we need to make sure their leaf part is added if not present. For example, arange should be removed but we should add arange_out.
For reference here's source of truth of all leaf nodes as of 10/21/2019:
https://gist.github.com/ailzhang/905aa41a5be8c5315ba0e3d8f0d0e326

• (Ailing)Tensor bitwise_not(const Tensor& self); // bitwise_not(Tensor)->Tensor
• (Ailing)Tensor& bitwise_not_(Tensor& self); // bitwise_not_(Tensor)->Tensor
• (Ailing) Tensor _cast_Byte(const Tensor& self, bool non_blocking); // _cast_Byte(Tensor,bool)->Tensor
• (Ailing) Tensor _cast_Char(const Tensor& self, bool non_blocking); // _cast_Char(Tensor,bool)->Tensor
• (Ailing) Tensor _cast_Float(const Tensor& self, bool non_blocking); // _cast_Float(Tensor,bool)->Tensor
• (Ailing) Tensor _cast_Int(const Tensor& self, bool non_blocking); // _cast_Int(Tensor,bool)->Tensor
• (Ailing) Tensor _cast_Long(const Tensor& self, bool non_blocking); // _cast_Long(Tensor,bool)->Tensor
• (Ailing) Tensor _cast_Short(const Tensor& self, bool non_blocking); // _cast_Short(Tensor,bool)->Tensor
• (Ailing)Tensor _dim_arange(const Tensor& like, int64_t dim); // _dim_arange(Tensor,int64_t)->Tensor
• (Ailing) Tensor argmax(const Tensor& self, c10::optional<int64_t> dim, bool keepdim); // argmax(Tensor,c10::optional<int64_t>,bool)->Tensor
• (Ailing) Tensor argmin(const Tensor& self, c10::optional<int64_t> dim, bool keepdim); // argmin(Tensor,c10::optional<int64_t>,bool)->Tensor
• (Ailing) Tensor argsort(const Tensor& self, int64_t dim, bool descending); // argsort(Tensor,int64_t,bool)->Tensor
• (Ailing) Tensor avg_pool1d(const Tensor& self, IntArrayRef kernel_size, IntArrayRef stride, IntArrayRef padding, bool ceil_mode, bool count_include_pad); // avg_pool1d(Tensor,IntArrayRef,IntArrayRef,IntArrayRef,bool,bool)->Tensor
• (Ailing) Tensor batch_norm(const Tensor& input, const Tensor& weight, const Tensor& bias, const Tensor& running_mean, const Tensor& running_var, bool training, double momentum, double eps, bool cudnn_enabled); // batch_norm(Tensor,Tensor,Tensor,Tensor,Tensor,bool,double,double,bool)->Tensor
• (Ailing) Tensor bernoulli(const Tensor& self, double p, Generator* generator); // bernoulli(Tensor,double,Generator)->Tensor
• (Ailing) Tensor bilinear(const Tensor& input1, const Tensor& input2, const Tensor& weight, const Tensor& bias); // bilinear(Tensor,Tensor,Tensor,Tensor)->Tensor
• (Ailing) Tensor binary_cross_entropy_with_logits_backward( const Tensor& grad_output, const Tensor& self, const Tensor& target, const Tensor& weight, const Tensor& pos_weight, int64_t reduction); // binary_cross_entropy_with_logits_backward(Tensor,Tensor,Tensor,Tensor,Tensor,int64_t)->Tensor
• (Ailing) std::vector broadcast_tensors(TensorList tensors); // broadcast_tensors(TensorList)->std::vector
• (Ailing) Tensor celu(const Tensor& self, Scalar alpha); // celu(Tensor,Scalar)->Tensor
• (Ailing) Tensor& celu_(Tensor& self, Scalar alpha); // celu_(Tensor,Scalar)->Tensor
• (Ailing) Tensor chain_matmul(TensorList matrices); // chain_matmul(TensorList)->Tensor
• (Ailing) Tensor contiguous(const Tensor& self, MemoryFormat memory_format); // contiguous(Tensor,MemoryFormat)->Tensor
• (Ailing) Tensor& copy_(Tensor& self, const Tensor& src, bool non_blocking); // copy_(Tensor,Tensor,bool)->Tensor
• (Ailing) Tensor cosine_embedding_loss(const Tensor& input1, const Tensor& input2, const Tensor& target, double margin, int64_t reduction); // cosine_embedding_loss(Tensor,Tensor,Tensor,double,int64_t)->Tensor
• (Ailing) Tensor cosine_similarity(const Tensor& x1, const Tensor& x2, int64_t dim, double eps); // cosine_similarity(Tensor,Tensor,int64_t,double)->Tensor
• (Ailing) Tensor diagflat(const Tensor& self, int64_t offset); // diagflat(Tensor,int64_t)->Tensor
• (Ailing) Tensor dropout(const Tensor& input, double p, bool train); // dropout(Tensor,double,bool)->Tensor
• (Ailing) Tensor& dropout_(Tensor& self, double p, bool train); // dropout_(Tensor,double,bool)->Tensor
• Tensor einsum(std::string equation, TensorList tensors); // einsum(std::string,TensorList)->Tensor
• Tensor empty_like(const Tensor& self); // empty_like(Tensor)->Tensor
• Tensor expand_as(const Tensor& self, const Tensor& other); // expand_as(Tensor,Tensor)->Tensor
• Tensor flatten(const Tensor& self, int64_t start_dim, int64_t end_dim); // flatten(Tensor,int64_t,int64_t)->Tensor
• Tensor frobenius_norm(const Tensor& self); // frobenius_norm(Tensor)->Tensor
• Tensor frobenius_norm(const Tensor& self, IntArrayRef dim, bool keepdim); // frobenius_norm(Tensor,IntArrayRef,bool)->Tensor
• Tensor full_like(const Tensor& self, Scalar fill_value); // full_like(Tensor,Scalar)->Tensor
• Tensor group_norm(const Tensor& input, int64_t num_groups, const Tensor& weight, const Tensor& bias, double eps, bool cudnn_enabled); // group_norm(Tensor,int64_t,Tensor,Tensor,double,bool)->Tensor
• Tensor hinge_embedding_loss(const Tensor& self, const Tensor& target, double margin, int64_t reduction); // hinge_embedding_loss(Tensor,Tensor,double,int64_t)->Tensor
• Tensor index_add(const Tensor& self, int64_t dim, const Tensor& index, const Tensor& source); // index_add(Tensor,int64_t,Tensor,Tensor)->Tensor
• Tensor index_copy(const Tensor& self, int64_t dim, const Tensor& index, const Tensor& source); // index_copy(Tensor,int64_t,Tensor,Tensor)->Tensor
• Tensor index_fill(const Tensor& self, int64_t dim, const Tensor& index, Scalar value); // index_fill(Tensor,int64_t,Tensor,Scalar)->Tensor
• Tensor index_fill(const Tensor& self, int64_t dim, const Tensor& index, const Tensor& value); // index_fill(Tensor,int64_t,Tensor,Tensor)->Tensor
• Tensor index_put(const Tensor& self, TensorList indices, const Tensor& values, bool accumulate); // index_put(Tensor,TensorList,Tensor,bool)->Tensor
• Tensor instance_norm( const Tensor& input, const Tensor& weight, const Tensor& bias, const Tensor& running_mean, const Tensor& running_var, bool use_input_stats, double momentum, double eps, bool cudnn_enabled); // instance_norm(Tensor,Tensor,Tensor,Tensor,Tensor,bool,double,double,bool)->Tensor
• bool is_floating_point(const Tensor& self); // is_floating_point(Tensor)->bool
• bool is_signed(const Tensor& self); // is_signed(Tensor)->bool
• Tensor layer_norm(const Tensor& input, IntArrayRef normalized_shape, const Tensor& weight, const Tensor& bias, double eps, bool cudnn_enable); // layer_norm(Tensor,IntArrayRef,Tensor,Tensor,double,bool)->Tensor
• Tensor linear(const Tensor& input, const Tensor& weight, const Tensor& bias); // linear(Tensor,Tensor,Tensor)->Tensor
• Tensor log_sigmoid(const Tensor& self); // log_sigmoid(Tensor)->Tensor
• Tensor log_softmax(const Tensor& self, int64_t dim, c10::optional dtype); // log_softmax(Tensor,int64_t,c10::optional)->Tensor
• Tensor margin_ranking_loss(const Tensor& input1, const Tensor& input2, const Tensor& target, double margin, int64_t reduction); // margin_ranking_loss(Tensor,Tensor,Tensor,double,int64_t)->Tensor
• Tensor masked_fill(const Tensor& self, const Tensor& mask, Scalar value); // masked_fill(Tensor,Tensor,Scalar)->Tensor
• Tensor masked_fill(const Tensor& self, const Tensor& mask, const Tensor& value); // masked_fill(Tensor,Tensor,Tensor)->Tensor
• Tensor matmul(const Tensor& self, const Tensor& other); // matmul(Tensor,Tensor)->Tensor
• Tensor max_pool1d(const Tensor& self, IntArrayRef kernel_size, IntArrayRef stride, IntArrayRef padding, IntArrayRef dilation, bool ceil_mode); // max_pool1d(Tensor,IntArrayRef,IntArrayRef,IntArrayRef,IntArrayRef,bool)->Tensor
• Tensor max_pool2d(const Tensor& self, IntArrayRef kernel_size, IntArrayRef stride, IntArrayRef padding, IntArrayRef dilation, bool ceil_mode); // max_pool2d(Tensor,IntArrayRef,IntArrayRef,IntArrayRef,IntArrayRef,bool)->Tensor
• Tensor max_pool3d(const Tensor& self, IntArrayRef kernel_size, IntArrayRef stride, IntArrayRef padding, IntArrayRef dilation, bool ceil_mode); // max_pool3d(Tensor,IntArrayRef,IntArrayRef,IntArrayRef,IntArrayRef,bool)->Tensor
• std::vector meshgrid(TensorList tensors); // meshgrid(TensorList)->std::vector
• Tensor narrow(const Tensor& self, int64_t dim, int64_t start, int64_t length); // narrow(Tensor,int64_t,int64_t,int64_t)->Tensor
• Tensor nll_loss(const Tensor& self, const Tensor& target, const Tensor& weight, int64_t reduction, int64_t ignore_index); // nll_loss(Tensor,Tensor,Tensor,int64_t,int64_t)->Tensor
• Tensor nuclear_norm(const Tensor& self, bool keepdim); // nuclear_norm(Tensor,bool)->Tensor
• Tensor one_hot(const Tensor& self, int64_t num_classes); // one_hot(Tensor,int64_t)->Tensor
• Tensor ones_like(const Tensor& self); // ones_like(Tensor)->Tensor
• Tensor pairwise_distance(const Tensor& x1, const Tensor& x2, double p, double eps, bool keepdim); // pairwise_distance(Tensor,Tensor,double,double,bool)->Tensor
• Tensor pixel_shuffle(const Tensor& self, int64_t upscale_factor); // pixel_shuffle(Tensor,int64_t)->Tensor
• Tensor pinverse(const Tensor& self, double rcond); // pinverse(Tensor,double)->Tensor
• Tensor reshape(const Tensor& self, IntArrayRef shape); // reshape(Tensor,IntArrayRef)->Tensor
• Tensor scatter(const Tensor& self, int64_t dim, const Tensor& index, const Tensor& src); // scatter(Tensor,int64_t,Tensor,Tensor)->Tensor
• Tensor scatter(const Tensor& self, int64_t dim, const Tensor& index, Scalar value); // scatter(Tensor,int64_t,Tensor,Scalar)->Tensor
• Tensor scatter_add(const Tensor& self, int64_t dim, const Tensor& index, const Tensor& src); // scatter_add(Tensor,int64_t,Tensor,Tensor)->Tensor
• Tensor selu(const Tensor& self); // selu(Tensor)->Tensor
• Tensor& selu_(Tensor& self); // selu_(Tensor)->Tensor
• int64_t size(const Tensor& self, int64_t dim); // size(Tensor,int64_t)->int64_t
• Tensor softmax(const Tensor& self, int64_t dim, c10::optional dtype); // softmax(Tensor,int64_t,c10::optional)->Tensor
• Tensor sum_to_size(const Tensor& self, IntArrayRef size); // sum_to_size(Tensor,IntArrayRef)->Tensor
• Tensor tensordot(const Tensor& self, const Tensor& other, IntArrayRef dims_self, IntArrayRef dims_other); // tensordot(Tensor,Tensor,IntArrayRef,IntArrayRef)->Tensor
• Tensor to(const Tensor& self, const TensorOptions& options, bool non_blocking, bool copy, c10::optional memory_format); // to(Tensor,TensorOptions,bool,bool,c10::optional)->Tensor
• Tensor to(const Tensor& self, Device device, ScalarType dtype, bool non_blocking, bool copy, c10::optional memory_format); // to(Tensor,Device,ScalarType,bool,bool,c10::optional)->Tensor
• Tensor to(const Tensor& self, ScalarType dtype, bool non_blocking, bool copy, c10::optional memory_format); // to(Tensor,ScalarType,bool,bool,c10::optional)->Tensor
• Tensor to(const Tensor& self, const Tensor& other, bool non_blocking, bool copy, c10::optional memory_format); // to(Tensor,Tensor,bool,bool,c10::optional)->Tensor
• Tensor triplet_margin_loss(const Tensor& anchor, const Tensor& positive, const Tensor& negative, double margin, double p, double eps, bool swap, int64_t reduction); // triplet_margin_loss(Tensor,Tensor,Tensor,double,double,double,bool,int64_t)->Tensor
• Tensor view_as(const Tensor& self, const Tensor& other); // view_as(Tensor,Tensor)->Tensor
• Tensor where(const Tensor& condition, const Tensor& self, const Tensor& other); // where(Tensor,Tensor,Tensor)->Tensor
• Tensor zeros_like(const Tensor& self); // zeros_like(Tensor)->Tensor

And a few more (non-leaf factory functions)

• Tensor arange(Scalar end, const TensorOptions& options); // arange(Scalar,TensorOptions)->Tensor
• Tensor arange(Scalar start, Scalar end, const TensorOptions& options); // arange(Scalar,Scalar,TensorOptions)->Tensor
• Tensor arange(Scalar start, Scalar end, Scalar step, const TensorOptions& options); // arange(Scalar,Scalar,Scalar,TensorOptions)->Tensor
• Tensor bartlett_window(int64_t window_length, const TensorOptions& options); // bartlett_window(int64_t,TensorOptions)->Tensor
• Tensor bartlett_window(int64_t window_length, bool periodic, const TensorOptions& options); // bartlett_window(int64_t,bool,TensorOptions)->Tensor
• Tensor blackman_window(int64_t window_length, const TensorOptions& options); // blackman_window(int64_t,TensorOptions)->Tensor
• Tensor blackman_window(int64_t window_length, bool periodic, const TensorOptions& options); // blackman_window(int64_t,bool,TensorOptions)->Tensor
• Tensor empty_like(const Tensor& self, const TensorOptions& options, c10::optional memory_format); // empty_like(Tensor,TensorOptions,c10::optional)->Tensor
• Tensor eye(int64_t n, const TensorOptions& options); // eye(int64_t,TensorOptions)->Tensor
• Tensor eye(int64_t n, int64_t m, const TensorOptions& options); // eye(int64_t,int64_t,TensorOptions)->Tensor
• Tensor full(IntArrayRef size, Scalar fill_value, const TensorOptions& options); // full(IntArrayRef,Scalar,TensorOptions)->Tensor
• Tensor full_like(const Tensor& self, Scalar fill_value, const TensorOptions& options, c10::optional memory_format); // full_like(Tensor,Scalar,TensorOptions,c10::optional)->Tensor
• Tensor hamming_window(int64_t window_length, const TensorOptions& options); // hamming_window(int64_t,TensorOptions)->Tensor
• Tensor hamming_window(int64_t window_length, bool periodic, const TensorOptions& options); // hamming_window(int64_t,bool,TensorOptions)->Tensor
• Tensor hamming_window(int64_t window_length, bool periodic, double alpha, const TensorOptions& options); // hamming_window(int64_t,bool,double,TensorOptions)->Tensor
• Tensor hamming_window(int64_t window_length, bool periodic, double alpha, double beta, const TensorOptions& options); // hamming_window(int64_t,bool,double,double,TensorOptions)->Tensor
• Tensor hann_window(int64_t window_length, const TensorOptions& options); // hann_window(int64_t,TensorOptions)->Tensor
• Tensor hann_window(int64_t window_length, bool periodic, const TensorOptions& options); // hann_window(int64_t,bool,TensorOptions)->Tensor
• Tensor ones(IntArrayRef size, const TensorOptions& options); // ones(IntArrayRef,TensorOptions)->Tensor
• Tensor ones_like(const Tensor& self, const TensorOptions& options, c10::optional memory_format); // ones_like(Tensor,TensorOptions,c10::optional)->Tensor
• Tensor randperm(int64_t n, const TensorOptions& options); // randperm(int64_t,TensorOptions)->Tensor
• Tensor randperm(int64_t n, Generator* generator, const TensorOptions& options); // randperm(int64_t,Generator,TensorOptions)->Tensor
• Tensor zeros_like(const Tensor& self, c10::optional memory_format); // zeros_like(Tensor,c10::optional)->Tensor
• Tensor zeros_like(const Tensor& self, const TensorOptions& options, c10::optional memory_format); // zeros_like(Tensor,TensorOptions,c10::optional)->Tensor

[dlibenzi]

The list above are all the non-leaf we override?

[ailzhang]

The list above are all the non-leaf we override?

Yep that's all of them