1-layer-addition-v2 / model_config.py
melephant's picture
Publish addition-transformer run rmb41c76
f8ccbd6 verified
Raw
History Blame Contribute Delete
258 Bytes
from __future__ import annotations
from typing import Protocol
class AdditionModelConfig(Protocol):
d_model: int
n_heads: int
d_mlp: int
n_layers: int
max_seq_len: int
rope_theta: float
init_mode: str
residual_alpha: float