汇聚嵌入运算符

本节包含用于各种汇聚嵌入操作的 CUDA 和 CPU 运算符，包括合并和排列运算符。

合并运算符

std::vector<at::Tensor> all_to_one_device(std::vector<at::Tensor> input_tensors, at::Device target_device)

排列运算符

at::Tensor permute_pooled_embs_split_gpu(const at::Tensor &pooled_embs, const at::Tensor &offset_dim_list, const at::Tensor &permute_list, const at::Tensor &inv_offset_dim_list, const at::Tensor &inv_permute_list)

at::Tensor permute_pooled_embs_auto_grad_split_gpu(const at::Tensor &pooled_embs, const at::Tensor &offset_dim_list, const at::Tensor &permute_list, const at::Tensor &inv_offset_dim_list, const at::Tensor &inv_permute_list)

Tensor permute_pooled_embs_auto_grad_gpu(const Tensor &pooled_embs, const Tensor &offset_dim_list, const Tensor &permute_list, const Tensor &inv_offset_dim_list, const Tensor &inv_permute_list)

at::Tensor permute_pooled_embs_cpu_impl(const at::Tensor &pooled_embs, const at::Tensor &offset_dim_list, const at::Tensor &permute_list, const at::Tensor &inv_offset_dim_list, const at::Tensor &inv_permute_list, const bool &allow_duplicates)

at::Tensor permute_pooled_embs_split_cpu(const at::Tensor &pooled_embs, const at::Tensor &offset_dim_list, const at::Tensor &permute_list, const at::Tensor &inv_offset_dim_list, const at::Tensor &inv_permute_list)

at::Tensor permute_pooled_embs_auto_grad_split_cpu(const at::Tensor &pooled_embs, const at::Tensor &offset_dim_list, const at::Tensor &permute_list, const at::Tensor &inv_offset_dim_list, const at::Tensor &inv_permute_list)

at::Tensor permute_pooled_embs_auto_grad(const Tensor &pooled_embs, const Tensor &offset_dim_list, const Tensor &permute_list, const Tensor &inv_offset_dim_list, const Tensor &inv_permute_list)

at::Tensor permute_pooled_embs_auto_grad_cpu(const Tensor &pooled_embs, const Tensor &offset_dim_list, const Tensor &permute_list, const Tensor &inv_offset_dim_list, const Tensor &inv_permute_list)