File size: 375 Bytes
0f10080
 
 
 
 
 
 
 
 
78c5b19
 
e799e08
 
78c5b19
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
"""
helpers for lora embeddings
"""


def get_linear_embedding_layers(model_type):
    """
    returns the linear embedding layers needed for loras, dependent on the model arch
    """
    if model_type == "gpt_neox":
        return ["embed_in", "embed_out"]
    if model_type == "falcon":
        return ["word_embeddings", "lm_head"]
    return ["embed_tokens", "lm_head"]