Struct tgn::TGNConfig

ClassList > tgn > TGNConfig

Configuration parameters for the TGN model architecture.

• #include <tgn.h>

Public Attributes

Type	Name
torch::Device	device = torch::kCPU Default to CPU.
float	dropout = 0.1 TransformerConv dropout.
std::size_t	embedding_dim = 100 TransformerConv embedding size.
std::size_t	memory_dim = 100 TGNMemory embedding size.
std::size_t	num_heads = 2 TransformerConv multi-head attention heads.
std::size_t	num_nbrs = 10 RecencySampler neighbor buffer size.
std::size_t	time_dim = 100 TimeEncoder embedding size.

Public Attributes Documentation

variable device

Default to CPU.

torch::Device tgn::TGNConfig::device;

---

variable dropout

TransformerConv dropout.

float tgn::TGNConfig::dropout;

---

variable embedding_dim

TransformerConv embedding size.

std::size_t tgn::TGNConfig::embedding_dim;

---

variable memory_dim

TGNMemory embedding size.

std::size_t tgn::TGNConfig::memory_dim;

---

variable num_heads

TransformerConv multi-head attention heads.

std::size_t tgn::TGNConfig::num_heads;

---

variable num_nbrs

RecencySampler neighbor buffer size.

std::size_t tgn::TGNConfig::num_nbrs;

---

variable time_dim

TimeEncoder embedding size.

std::size_t tgn::TGNConfig::time_dim;

---

---

The documentation for this class was generated from the following file include/tgn.h