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

quantize: options for output and token embedding tensors qtype (#6239)

Browse source 
* quantize: be able to specify the output tensor type

* quantize: be able to specify the token embedding tensor type

---------

Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
This commit is contained in:
Kawrakow 2024-03-22 19:47:14 +01:00 • committed by GitHub
parent dba1af6129
commit 1d0331c12a
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
3 changed files with 61 additions and 26 deletions
Download patch file Download diff file Expand all files Collapse all files

24
examples/quantize/quantize.cpp
View file

@ -189,6 +189,18 @@ static void prepare_imatrix(const std::string& imatrix_file,
}
}
static ggml_type parse_ggml_type(const char * arg) {
ggml_type result = GGML_TYPE_COUNT;
for (int j = 0; j < GGML_TYPE_COUNT; ++j) {
auto type = ggml_type(j);
const auto * name = ggml_type_name(type);
if (name && strcmp(arg, name) == 0) {
result = type; break;
}
}
return result;
}
int main(int argc, char ** argv) {
if (argc < 3) {
usage(argv[0]);
@ -203,6 +215,18 @@ int main(int argc, char ** argv) {
for (; arg_idx < argc && strncmp(argv[arg_idx], "--", 2) == 0; arg_idx++) {
if (strcmp(argv[arg_idx], "--leave-output-tensor") == 0) {
params.quantize_output_tensor = false;
} else if (strcmp(argv[arg_idx], "--output-tensor-type") == 0) {
if (arg_idx < argc-1) {
params.output_tensor_type = parse_ggml_type(argv[++arg_idx]);
} else {
usage(argv[0]);
}
} else if (strcmp(argv[arg_idx], "--token-embedding-type") == 0) {
if (arg_idx < argc-1) {
params.token_embedding_type = parse_ggml_type(argv[++arg_idx]);
} else {
usage(argv[0]);
}
} else if (strcmp(argv[arg_idx], "--allow-requantize") == 0) {
params.allow_requantize = true;
} else if (strcmp(argv[arg_idx], "--pure") == 0) {