Merge branch 'upstream' into concedo_experimental

# Conflicts:
#	README.md
#	docs/build.md
#	ggml/src/ggml-cpu/CMakeLists.txt
#	ggml/src/ggml-cpu/kleidiai/kernels.cpp
#	ggml/src/ggml-cpu/kleidiai/kernels.h
#	ggml/src/ggml-cpu/kleidiai/kleidiai.cpp
#	tests/test-backend-ops.cpp
#	tools/server/README.md
This commit is contained in:
Concedo
2025-07-21 23:37:42 +08:00
4 changed files with 15 additions and 13 deletions
+10 -7
View File
@@ -786,14 +786,17 @@ int main(int argc, char ** argv) {
}
// check for reverse prompt using special tokens
llama_token last_token = common_sampler_last(smpl);
for (auto token : antiprompt_token) {
if (token == last_token) {
if (params.interactive) {
is_interacting = true;
// avoid calling common_sampler_last() if last_output is empty
if (!last_output.empty()) {
llama_token last_token = common_sampler_last(smpl);
for (auto token : antiprompt_token) {
if (token == last_token) {
if (params.interactive) {
is_interacting = true;
}
is_antiprompt = true;
break;
}
is_antiprompt = true;
break;
}
}
+2 -1
View File
@@ -4516,9 +4516,10 @@ int main(int argc, char ** argv) {
json tokens_response = json::array();
if (body.count("content") != 0) {
const bool add_special = json_value(body, "add_special", false);
const bool parse_special = json_value(body, "parse_special", true);
const bool with_pieces = json_value(body, "with_pieces", false);
llama_tokens tokens = tokenize_mixed(ctx_server.vocab, body.at("content"), add_special, true);
llama_tokens tokens = tokenize_mixed(ctx_server.vocab, body.at("content"), add_special, parse_special);
if (with_pieces) {
for (const auto& token : tokens) {