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

Context size by default should be 2048, not 512

Browse source 
Set `n_ctx` equal to `n_batch` in `Opt` class. Now context size is
a more reasonable 2048.

Signed-off-by: Eric Curtin <ecurtin@redhat.com>
This commit is contained in:
Eric Curtin 2025-01-05 23:03:06 +00:00
parent 0da5d86026
commit 62fe73d4b5
1 changed files with 1 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

1
examples/run/run.cpp
View file

@ -83,6 +83,7 @@ class Opt {
} }
ctx_params.n_batch = context_size >= 0 ? context_size : context_size_default; ctx_params.n_batch = context_size >= 0 ? context_size : context_size_default;
ctx_params.n_ctx = ctx_params.n_batch;
model_params.n_gpu_layers = ngl >= 0 ? ngl : ngl_default; model_params.n_gpu_layers = ngl >= 0 ? ngl : ngl_default;
temperature = temperature >= 0 ? temperature : temperature_default; temperature = temperature >= 0 ? temperature : temperature_default;