Public/koboldcpp
1
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2026-09-18 16:55:14 +02:00
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'upstream' into concedo_experimental

Browse Source 
# Conflicts:
#	.github/workflows/server-self-hosted.yml
#	.github/workflows/ui-build-self-hosted.yml
#	.github/workflows/ui-build.yml
#	.github/workflows/ui.yml
#	ggml/src/ggml-sycl/mmvq.cpp
#	ggml/src/ggml-sycl/vecdotq.hpp
#	tests/test-backend-ops.cpp
#	tests/test-sampling.cpp
#	tools/cli/README.md
#	tools/server/README.md
This commit is contained in:
Concedo
2026-09-09 11:47:23 +08:00
parent b72d1de1ac 304665fe7a
commit ad1eee0c5f
18 changed files with 337 additions and 25 deletions
Download Patch File Download Diff File Expand all files Collapse all files
common/speculative.cpp
+12 -1
View File
@@ -2467,11 +2467,22 @@ common_params common_base_params_to_speculative(const common_params & params) {
result.pooling_type = LLAMA_POOLING_TYPE_UNSPECIFIED;
if (has_draft) {
result.devices = params_spec.devices;
// default to global devices value
if (!params_spec.devices.empty()) {
result.devices = params_spec.devices;
}
result.model = params_spec.mparams;
result.n_gpu_layers = params_spec.n_gpu_layers;
result.tensor_buft_overrides = params_spec.tensor_buft_overrides;
// a draft pinned to a single device doesn't need the meta wrapper an inherited -sm tensor would give it
// (the device list is null-terminated, so a single device means size 2)
const size_t n_devs = std::count_if(params_spec.devices.begin(), params_spec.devices.end(),
[](ggml_backend_dev_t d) { return d != nullptr; });
if (n_devs == 1) {
result.split_mode = LLAMA_SPLIT_MODE_LAYER;
}
if (params_spec.cpuparams.n_threads > 0) {
result.cpuparams.n_threads = params_spec.cpuparams.n_threads;
result.cpuparams_batch.n_threads = params_spec.cpuparams_batch.n_threads;