Int4OPTDecoder Class Reference

Public Member Functions
	Int4OPTDecoder (std::string param_path, const struct model_config config)
Matrix3D< float >	prepare_decoder_attention_mask (int length, int past_length)
Matrix3D< float >	get_position_embed (int sql_length, int past_length)
struct Int4OPTDecoder_output	forward (const struct Int4OPTDecoder_input &input)

Public Attributes
Embedding	embed_tokens
Embedding	embed_positions
int	voc_size
int	embed_dim
int	padding_idx
int	hidden_dim
int	num_heads
std::vector< Int4OPTDecoderLayer >	layers
LayerNorm	final_layer_norm
std::string	profile_name = "Int4OPTDecoder"

---

The documentation for this class was generated from the following files:

• llm/include/nn_modules/Int4OPTDecoder.h
• llm/src/nn_modules/Int4OPTDecoder.cc