File size: 1,807 Bytes
83bd6f7 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 | """HuggingFace-compatible config for the flaxchat YatNMN-Softplus GPT."""
from transformers import PretrainedConfig
class YatGPTHfConfig(PretrainedConfig):
model_type = "yatnmn_gpt"
def __init__(
self,
sequence_len: int = 1024,
vocab_size: int = 32768,
n_layer: int = 12,
n_head: int = 12,
n_kv_head: int = 12,
n_embd: int = 768,
window_pattern: str = "SSSL",
tie_embeddings: bool = True,
rope_base: float = 100000.0,
pad_vocab_size_to: int = 64,
mlp_type: str = "yatnmn-softplus",
scalar_bias: bool = False,
softplus_bias: bool = True,
learnable_epsilon: bool = True,
epsilon_init: float = 1e-3,
constant_alpha: bool = False,
tie_word_embeddings: bool = True,
**kwargs,
):
self.sequence_len = sequence_len
self.vocab_size = vocab_size
self.n_layer = n_layer
self.n_head = n_head
self.n_kv_head = n_kv_head
self.n_embd = n_embd
self.window_pattern = window_pattern
self.tie_embeddings = tie_embeddings
self.rope_base = rope_base
self.pad_vocab_size_to = pad_vocab_size_to
self.mlp_type = mlp_type
self.scalar_bias = scalar_bias
self.softplus_bias = softplus_bias
self.learnable_epsilon = learnable_epsilon
self.epsilon_init = epsilon_init
self.constant_alpha = constant_alpha
self.max_position_embeddings = sequence_len * 10
# HF probes these standard names
self.num_hidden_layers = n_layer
self.num_attention_heads = n_head
self.num_key_value_heads = n_kv_head
self.hidden_size = n_embd
super().__init__(tie_word_embeddings=tie_word_embeddings, **kwargs)
|