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

ggml : move FP16 <-> FP32 code to ggml-impl.h (#3861)

Browse source 
* ggml : move FP16 <-> FP32 stuff to ggml-impl.h

ggml-ci

* tests : fix ARM build

* ggml : explicitly initialize deprecated type traits

* ggml : add math.h to ggml-impl.h

* ggml : remove duplicate static assert macros

* ggml : prefix lookup tables with ggml_

ggml-ci

* ggml-impl : move extern "C" to start of file
This commit is contained in:
Georgi Gerganov 2023-10-30 19:19:15 +02:00 • committed by GitHub
parent 6e08281e58
commit 207b51900e
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
7 changed files with 470 additions and 424 deletions
Download patch file Download diff file Expand all files Collapse all files

7
tests/test-quantize-fns.cpp
View file

@ -129,6 +129,13 @@ int main(int argc, char * argv[]) {
ggml_type type = (ggml_type) i;
ggml_type_traits_t qfns = ggml_internal_get_type_traits(type);
// deprecated - skip
if (qfns.blck_size == 0) {
continue;
}
printf("Testing %s\n", ggml_type_name((ggml_type) i));
if (qfns.from_float && qfns.to_float) {
const float total_error = total_quantization_error(qfns, test_size, test_data.data());
const float max_quantization_error =