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

Bugfix: n_batch should always be <= n_ctx

Browse source
This commit is contained in:
Andrei Betlen 2023-04-04 13:08:21 -04:00
parent 248b0566fa
commit 5075c16fcc
1 changed files with 1 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
llama_cpp/llama.py
View file

@ -68,7 +68,7 @@ class Llama:
maxlen=self.last_n_tokens_size, maxlen=self.last_n_tokens_size,
) )
self.tokens_consumed = 0 self.tokens_consumed = 0
self.n_batch = n_batch self.n_batch = min(n_ctx, n_batch)
self.n_threads = n_threads or multiprocessing.cpu_count() self.n_threads = n_threads or multiprocessing.cpu_count()