mstar.model.ripple.config#

Ripple’s config, read off the HF config.json.

Classes

RippleConfig(vocab_size, hidden_size, ...)

class mstar.model.ripple.config.RippleConfig(vocab_size: 'int', hidden_size: 'int', intermediate_size: 'int', num_hidden_layers: 'int', num_attention_heads: 'int', num_key_value_heads: 'int', head_dim: 'int', layer_types: 'tuple[str, ...]', sliding_window: 'int', max_position_embeddings: 'int', rope_theta: 'float', rms_norm_eps: 'float', embed_scale: 'float', ffn_clamp_threshold: 'float', stop_token_ids: 'tuple[int, ...]')[source]#

Bases: object

Parameters:
  • vocab_size (int)

  • hidden_size (int)

  • intermediate_size (int)

  • num_hidden_layers (int)

  • num_attention_heads (int)

  • num_key_value_heads (int)

  • head_dim (int)

  • layer_types (tuple[str, ...])

  • sliding_window (int)

  • max_position_embeddings (int)

  • rope_theta (float)

  • rms_norm_eps (float)

  • embed_scale (float)

  • ffn_clamp_threshold (float)

  • stop_token_ids (tuple[int, ...])

embed_scale: float#
ffn_clamp_threshold: float#
classmethod from_dict(data)[source]#
Parameters:

data (dict)

Return type:

RippleConfig

classmethod from_hf(path)[source]#
Parameters:

path (str | Path)

Return type:

RippleConfig

head_dim: int#
hidden_size: int#
intermediate_size: int#
property is_sliding: tuple[bool, ...]#
layer_types: tuple[str, ...]#
max_position_embeddings: int#
num_attention_heads: int#
num_hidden_layers: int#
num_key_value_heads: int#
rms_norm_eps: float#
rope_theta: float#
sliding_window: int#
stop_token_ids: tuple[int, ...]#
vocab_size: int#