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

gguf : fix resource leaks (#6061)

Browse source 
There several places where a gguf context is allocated. A call to gguf_free
is missing in some error paths. Also on linux, llama-bench was missing a
fclose.
This commit is contained in:
Steve Grubb 2024-03-14 14:29:32 -04:00 • committed by GitHub
parent 727107707a
commit 6e0438da3c
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
4 changed files with 7 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

4
examples/llava/clip.cpp
View file

@ -995,6 +995,7 @@ struct clip_ctx * clip_model_load(const char * fname, const int verbosity = 1) {
if (!new_clip->ctx_data) {
fprintf(stderr, "%s: ggml_init() failed\n", __func__);
clip_free(new_clip);
gguf_free(ctx);
return nullptr;
}
@ -1002,6 +1003,7 @@ struct clip_ctx * clip_model_load(const char * fname, const int verbosity = 1) {
if (!fin) {
printf("cannot open model file for loading tensors\n");
clip_free(new_clip);
gguf_free(ctx);
return nullptr;
}
@ -1023,6 +1025,7 @@ struct clip_ctx * clip_model_load(const char * fname, const int verbosity = 1) {
if (!fin) {
printf("%s: failed to seek for tensor %s\n", __func__, name);
clip_free(new_clip);
gguf_free(ctx);
return nullptr;
}
int num_bytes = ggml_nbytes(cur);
@ -1908,6 +1911,7 @@ bool clip_model_quantize(const char * fname_inp, const char * fname_out, const i
break;
default:
printf("Please use an input file in f32 or f16\n");
gguf_free(ctx_out);
return false;
}