76 lines
2.8 KiB
Python
76 lines
2.8 KiB
Python
"""HF config for the Kohaku decoder. Ships inside an exported repository.
|
|
|
|
Standalone by construction: an exported repo is loaded with
|
|
``trust_remote_code=True`` on machines that do not have kohakuwullm installed, so
|
|
nothing here may import it. See docs/guides/hf-export.md.
|
|
"""
|
|
|
|
from transformers.configuration_utils import PretrainedConfig
|
|
|
|
|
|
class KohakuConfig(PretrainedConfig):
|
|
"""Kohaku: GQA + per-head QK-norm, SwiGLU, and DeepSeek-style sparse MLPs.
|
|
|
|
Layers below ``first_k_dense`` use a dense SwiGLU; the rest use one shared
|
|
expert plus ``num_experts_per_tok`` of ``n_routed_experts``, selected on
|
|
sigmoid scores offset by a selection-only bias and weighted by the unbiased
|
|
score.
|
|
"""
|
|
|
|
model_type = "kohaku"
|
|
keys_to_ignore_at_inference = ["past_key_values"]
|
|
|
|
def __init__(
|
|
self,
|
|
vocab_size: int = 65536,
|
|
hidden_size: int = 768,
|
|
num_hidden_layers: int = 16,
|
|
num_attention_heads: int = 12,
|
|
num_key_value_heads: int = 2,
|
|
head_dim: int = 64,
|
|
intermediate_size: int = 2048,
|
|
moe_intermediate_size: int = 384,
|
|
n_routed_experts: int = 64,
|
|
n_shared_experts: int = 1,
|
|
num_experts_per_tok: int = 8,
|
|
first_k_dense: int = 1,
|
|
norm_topk_prob: bool = True,
|
|
routed_scaling_factor: float = 1.0,
|
|
scoring_func: str = "sigmoid",
|
|
max_position_embeddings: int = 4096,
|
|
rope_theta: float = 100000.0,
|
|
rms_norm_eps: float = 1e-6,
|
|
qk_norm: bool = True,
|
|
tie_word_embeddings: bool = False,
|
|
bos_token_id: int | None = 64000,
|
|
eos_token_id: int | None = 64001,
|
|
pad_token_id: int | None = 64002,
|
|
**kwargs,
|
|
) -> None:
|
|
self.vocab_size = vocab_size
|
|
self.hidden_size = hidden_size
|
|
self.num_hidden_layers = num_hidden_layers
|
|
self.num_attention_heads = num_attention_heads
|
|
self.num_key_value_heads = num_key_value_heads
|
|
self.head_dim = head_dim
|
|
self.intermediate_size = intermediate_size
|
|
self.moe_intermediate_size = moe_intermediate_size
|
|
self.n_routed_experts = n_routed_experts
|
|
self.n_shared_experts = n_shared_experts
|
|
self.num_experts_per_tok = num_experts_per_tok
|
|
self.first_k_dense = first_k_dense
|
|
self.norm_topk_prob = norm_topk_prob
|
|
self.routed_scaling_factor = routed_scaling_factor
|
|
self.scoring_func = scoring_func
|
|
self.max_position_embeddings = max_position_embeddings
|
|
self.rope_theta = rope_theta
|
|
self.rms_norm_eps = rms_norm_eps
|
|
self.qk_norm = qk_norm
|
|
super().__init__(
|
|
bos_token_id=bos_token_id,
|
|
eos_token_id=eos_token_id,
|
|
pad_token_id=pad_token_id,
|
|
tie_word_embeddings=tie_word_embeddings,
|
|
**kwargs,
|
|
)
|