Merge commit 'd646c9d15500a702425e8a90c19d2400deecbd0c' into concedo_experimental

# Conflicts:
#	.github/actions/windows-setup-rocm/action.yml
#	.github/workflows/build-apple.yml
#	.github/workflows/release.yml
#	.github/workflows/server-self-hosted.yml
#	examples/training/README.md
#	ggml/src/ggml-hexagon/ggml-hexagon.cpp
#	ggml/src/ggml-hexagon/htp/hvx-arith.h
#	ggml/src/ggml-hexagon/htp/hvx-log.h
#	ggml/src/ggml-hexagon/htp/hvx-norm.h
#	ggml/src/ggml-hexagon/htp/hvx-scale.h
#	ggml/src/ggml-hexagon/htp/hvx-sqrt.h
#	ggml/src/ggml-hexagon/htp/unary-ops.c
#	ggml/src/ggml-hexagon/htp/unary-ops.h
#	ggml/src/ggml-sycl/mmvq.cpp
#	ggml/src/ggml-sycl/vecdotq.hpp
#	tests/test-backend-ops.cpp
#	tests/test-mtmd-c-api.c
#	tests/test-mtmd-impl.cpp
This commit is contained in:
Concedo
2026-09-04 16:05:40 +08:00
78 changed files with 693 additions and 221 deletions
+11 -1
View File
@@ -961,6 +961,11 @@ static bool common_params_parse_ex(int argc, char ** argv, common_params_context
));
}
// if the preserve_reasoning kwarg was not specified explicitly, enable it by default
if (!params.default_template_kwargs.count("preserve_reasoning")) {
params.default_template_kwargs["preserve_reasoning"] = "true";
}
return true;
}
@@ -3554,6 +3559,10 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
LOG_WRN("Setting 'enable_thinking' via --chat-template-kwargs is deprecated. "
"Use --reasoning on / --reasoning off instead.\n");
}
if (item.key() == "preserve_reasoning") {
LOG_WRN("Setting 'preserve_reasoning' via --chat-template-kwargs is deprecated. "
"Use --reasoning-preserve / --no-reasoning-preserve instead.\n");
}
params.default_template_kwargs[item.key()] = item.value().dump();
}
}
@@ -3744,7 +3753,7 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
add_opt(common_arg(
{"--reasoning-preserve"},
{"--no-reasoning-preserve"},
"preserve reasoning trace in the full history, not just the last assistant message (default: template default)\n"
"preserve reasoning trace in the full history, not just the last assistant message (default: enabled)\n"
"compatible with certain templates having 'supports_preserve_reasoning' capability\n"
"example: https://docs.z.ai/guides/capabilities/thinking-mode#preserved-thinking",
[](common_params & params, bool value) {
@@ -3753,6 +3762,7 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
} else {
params.default_template_kwargs["preserve_reasoning"] = "false";
}
params.preserve_reasoning_specified = true;
}
).set_examples({LLAMA_EXAMPLE_SERVER, LLAMA_EXAMPLE_COMPLETION, LLAMA_EXAMPLE_CLI}).set_env("LLAMA_ARG_REASONING_PRESERVE"));
add_opt(common_arg(
+2 -1
View File
@@ -271,7 +271,7 @@ struct common_params_sampling {
COMMON_SAMPLER_TYPE_TEMPERATURE,
};
common_grammar grammar; // optional grammar constraint (user / output-format / tool-calls)
common_grammar grammar; // optional grammar constraint (user / output-format / tool-calls)
bool grammar_lazy = false;
std::vector<common_grammar_trigger> grammar_triggers; // optional triggers (for lazy grammars)
std::set<llama_token> preserved_tokens;
@@ -658,6 +658,7 @@ struct common_params {
std::string ssl_file_cert = ""; // NOLINT
std::map<std::string, std::string> default_template_kwargs;
bool preserve_reasoning_specified = false;
// CLI params
std::string server_base; // if set, connect to this server instead of starting a new one