upstream-archive/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-10-02 10:57:33 -05:00
Code Issues Packages Projects Releases Wiki Activity

common : enable default speculative config

Browse Source 
Assisted-by: pi:llama.cpp/DeepSeek-V4-Flash-Vision-Exp
This commit is contained in:
Georgi Gerganov
2026-09-04 16:25:14 +03:00
parent d646c9d155
commit 7c2e0dea43
10 changed files with 196 additions and 25 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tests/test-model-resolution.cpp
+7
View File
@@ -453,6 +453,13 @@ static void test_task_assembly() {
// -hfd on a repo without sidecars keeps resolving a full model as draft
common_params params;
assemble({"server", "-hf", "test/main:Q8_0", "-hfd", "test/small"}, params);
REQUIRE(params.speculative.types == std::vector<enum common_speculative_type>{COMMON_SPECULATIVE_TYPE_NGRAM_MOD});
REQUIRE_EQ(params.speculative.draft.mparams.path, cached("test/small", "draft-model-Q4_K_M.gguf"));
}
{
// --no-spec-type removes the default speculative type and keeps the fallback draft resolution
common_params params;
assemble({"server", "-hf", "test/main:Q8_0", "-hfd", "test/small", "--no-spec-type", "ngram-mod"}, params);
REQUIRE(params.speculative.types == std::vector<enum common_speculative_type>{COMMON_SPECULATIVE_TYPE_NONE});
REQUIRE_EQ(params.speculative.draft.mparams.path, cached("test/small", "draft-model-Q4_K_M.gguf"));
}