Banaxi-Tech commited on
Commit
ee57398
·
verified ·
1 Parent(s): 4c8a7d2

Publish BananaMind 2 Pico architecture for training

Browse files
Files changed (1) hide show
  1. configuration_bananamind2pico.py +1 -1
configuration_bananamind2pico.py CHANGED
@@ -20,7 +20,7 @@ class BananaMind2PicoConfig(PretrainedConfig):
20
  refresh_layers: tuple[int, ...] | list[int] = (3, 5),
21
  refresh_kernel_size: int = 9,
22
  refresh_alpha_init: float = 0.0,
23
- max_position_embeddings: int = 8192,
24
  rope_theta: float = 100000.0,
25
  rms_norm_eps: float = 1e-6,
26
  initializer_range: float = 0.02,
 
20
  refresh_layers: tuple[int, ...] | list[int] = (3, 5),
21
  refresh_kernel_size: int = 9,
22
  refresh_alpha_init: float = 0.0,
23
+ max_position_embeddings: int = 4096,
24
  rope_theta: float = 100000.0,
25
  rms_norm_eps: float = 1e-6,
26
  initializer_range: float = 0.02,