vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

convert.py : fix of type and shorter code

Browse source
This commit is contained in:
Maximilian Markewitz 2023-07-27 20:03:43 +02:00
parent e15a67d6b2
commit 01d16e1a1e
1 changed files with 2 additions and 5 deletions
Download patch file Download diff file Expand all files Collapse all files

7
convert.py
View file

@ -146,7 +146,7 @@ class Params:
n_mult: int n_mult: int
n_head: int n_head: int
n_layer: int n_layer: int
n_kv_head: int # This parameter is only used for Llama 2 n_kv_head: Optional[int] # This parameter is only used for Llama 2
@staticmethod @staticmethod
def guessed(model: 'LazyModel') -> 'Params': def guessed(model: 'LazyModel') -> 'Params':
@ -185,10 +185,7 @@ class Params:
n_head = config["num_attention_heads"]; n_head = config["num_attention_heads"];
n_layer = config["num_hidden_layers"]; n_layer = config["num_hidden_layers"];
n_ff = config["intermediate_size"]; n_ff = config["intermediate_size"];
if "num_key_value_heads" in config: n_kv_head = config.get("num_key_value_heads")
n_kv_head = config["num_key_value_heads"]
else:
n_kv_head = None
n_mult = find_n_mult(n_ff, n_embd); n_mult = find_n_mult(n_ff, n_embd);