TinyChatEngine
|
Public Member Functions | |
Fp32llamaDecoder (std::string param_path, const struct model_config config) | |
Matrix3D< float > | prepare_decoder_attention_mask (int length, int past_length) |
struct Fp32llamaDecoder_output | forward (const struct Fp32llamaDecoder_input &input) |
Public Attributes | |
Embedding | embed_tokens |
LlamaRMSNorm | norm |
float | rms_norm_eps |
int | voc_size |
int | embed_dim |
int | padding_idx |
int | hidden_dim |
int | num_heads |
std::vector< Fp32llamaDecoderLayer > | layers |
std::string | profile_name = "Fp32llamaDecoder" |