Skip to content

Add --parse-special for enabling parsing of special tokens in imatrix calculation #13389

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
May 9, 2025
Merged

Add --parse-special for enabling parsing of special tokens in imatrix calculation #13389

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
cbf7e25
Add --parse-special for enabling parsing of special tokens in imatrix…
bartowski1182 May 8, 2025
e4bd553
whitespace
bartowski1182 May 8, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
7 changes: 7 additions & 0 deletions common/arg.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2627,6 +2627,13 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
params.i_chunk = value;
}
).set_examples({LLAMA_EXAMPLE_IMATRIX}));
add_opt(common_arg(
{"--parse-special"},
string_format("prase special tokens (chat, tool, etc) (default: %s)", params.parse_special ? "true" : "false"),
[](common_params & params) {
params.parse_special = true;
}
).set_examples({LLAMA_EXAMPLE_IMATRIX}));
add_opt(common_arg(
{"-pps"},
string_format("is the prompt shared across parallel sequences (default: %s)", params.is_pp_shared ? "true" : "false"),
Expand Down
1 change: 1 addition & 0 deletions common/common.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -409,6 +409,7 @@ struct common_params {

bool process_output = false; // collect data for the output tensor
bool compute_ppl = true; // whether to compute perplexity
bool parse_special = false; // whether to parse special tokens during imatrix tokenization

// cvector-generator params
int n_pca_batch = 100;
Expand Down
5 changes: 3 additions & 2 deletions tools/imatrix/imatrix.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -24,7 +24,8 @@ static void print_usage(int, char ** argv) {
LOG("\n %s \\\n"
" -m model.gguf -f some-text.txt [-o imatrix.dat] [--process-output] \\\n"
" [--no-ppl] [--chunk 123] [--output-frequency 10] [--save-frequency 0] \\\n"
" [--in-file imatrix-prev-0.dat --in-file imatrix-prev-1.dat ...]\n" , argv[0]);
" [--in-file imatrix-prev-0.dat --in-file imatrix-prev-1.dat ...] \\\n"
" [--parse-special]\n" , argv[0]);
LOG("\n");
}

Expand Down Expand Up @@ -439,7 +440,7 @@ static bool compute_imatrix(llama_context * ctx, const common_params & params) {
auto tim1 = std::chrono::high_resolution_clock::now();
LOG_INF("%s: tokenizing the input ..\n", __func__);

std::vector<llama_token> tokens = common_tokenize(ctx, params.prompt, true);
std::vector<llama_token> tokens = common_tokenize(ctx, params.prompt, true, params.parse_special);

auto tim2 = std::chrono::high_resolution_clock::now();
LOG_INF("%s: tokenization took %g ms\n",__func__,1e-3*std::chrono::duration_cast<std::chrono::microseconds>(tim2-tim1).count());
Expand Down
Loading