mikhail/llama.cpp-mtp-turboquant
1
0
Fork 0
Code Issues Pull Requests Actions 33 Packages Projects Releases Wiki Activity

common/autoparser: fixes for newline handling / forced tool calls (#22654)

Browse Source 
* chat/autoparser: the fixes

* Move optspace() to chat-peg-parser, comment out server tests invalidated due to content now allowed with forced tool calls.

* Trim whitespace on apply instead
This commit is contained in:
Piotr Wilkin (ilintar)
2026-05-04 13:18:11 +02:00
committed by GitHub
parent 994118a183
commit a4701c98f7
10 changed files with 392 additions and 97 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tests/test-reasoning-budget.cpp
+7 -7
View File
@@ -70,20 +70,20 @@ static void test_reasoning_budget(
llama_sampler_apply(sampler, &cur_p);
// Check if forcing is active (all logits except one should be -INFINITY)
size_t finite_count = 0;
llama_token finite_token = -1;
size_t not_neg_inf = 0;
llama_token not_neg_inf_token = -1;
for (size_t j = 0; j < cur.size(); j++) {
if (std::isfinite(cur[j].logit)) {
finite_count++;
finite_token = cur[j].id;
if (std::isfinite(cur[j].logit) || cur[j].logit > 0) { // +INFINITY
not_neg_inf++;
not_neg_inf_token = cur[j].id;
}
}
llama_sampler_accept(sampler, sequence[i]);
fprintf(stderr, " i=%zu: token=%d, finite_count=%zu, finite_token=%d\n", i, (int)sequence[i], finite_count, (int)finite_token);
fprintf(stderr, " i=%zu: token=%d, not_neg_inf_count=%zu, not_neg_inf_token=%d\n", i, (int)sequence[i], not_neg_inf, (int)not_neg_inf_token);
if (finite_count == 1) {
if (not_neg_inf == 1) {
if (actual_force_start == SIZE_MAX) {
actual_force_start = i;
}