GroupQueryAttentionΒΆ
Grouped Query Attention layer.
Abstract Signature:
GroupQueryAttention(head_dim: int, num_query_heads: int, num_key_value_heads: int, dropout: float = 0.0)
Grouped Query Attention layer.
Abstract Signature:
GroupQueryAttention(head_dim: int, num_query_heads: int, num_key_value_heads: int, dropout: float = 0.0)