vrr/koboldcpp
2
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2026-08-26 00:34:30 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Merge commit '11b068d066' into concedo_experimental

Browse source 
# Conflicts:
#	CONTRIBUTING.md
#	docs/backend/SYCL.md
#	docs/install.md
#	docs/speculative.md
#	ggml/src/ggml-hip/CMakeLists.txt
#	ggml/src/ggml-opencl/ggml-opencl.cpp
#	ggml/src/ggml-sycl/common.hpp
#	ggml/src/ggml-sycl/element_wise.cpp
#	ggml/src/ggml-sycl/fattn-onednn.cpp
#	ggml/src/ggml-sycl/ggml-sycl.cpp
#	ggml/src/ggml-webgpu/ggml-webgpu-shader-lib.hpp
#	ggml/src/ggml-webgpu/ggml-webgpu.cpp
#	ggml/src/ggml-webgpu/wgsl-shaders/glu.wgsl
#	ggml/src/ggml-webgpu/wgsl-shaders/ssm_scan.wgsl
#	tests/test-backend-ops.cpp
#	tests/test-chat.cpp
#	tests/test-llama-archs.cpp
#	tools/cli/README.md
#	tools/llama-bench/llama-bench.cpp
#	tools/mtmd/CMakeLists.txt
#	tools/server/README.md
This commit is contained in:
Concedo 2026-07-31 17:06:19 +08:00
commit 98926f27c1
81 changed files with 5314 additions and 518 deletions
Download patch file Download diff file Expand all files Collapse all files

58
src/llama-model.cpp
View file

@ -126,6 +126,7 @@
#include "models/mistral4.cpp"
#include "models/modern-bert.cpp"
#include "models/mpt.cpp"
#include "models/nanbeige.cpp"
#include "models/nemotron-h-moe.cpp"
#include "models/nemotron-h.cpp"
#include "models/nemotron.cpp"
@ -226,6 +227,8 @@ static llama_model * llama_model_mapping(llm_arch arch, const llama_model_params
return new llama_model_stablelm(params);
case LLM_ARCH_MELLUM:
return new llama_model_mellum(params);
case LLM_ARCH_NANBEIGE:
return new llama_model_nanbeige(params);
case LLM_ARCH_QWEN:
return new llama_model_qwen(params);
case LLM_ARCH_QWEN2:
@ -957,6 +960,7 @@ const char * llm_type_name(llm_type type) {
case LLM_TYPE_100B_A6B: return "100B.A6B";
case LLM_TYPE_102B_A12B: return "102B.A12B";
case LLM_TYPE_106B_A12B: return "106B.A12B";
case LLM_TYPE_118B_A8B: return "118B.A8B";
case LLM_TYPE_120B_A12B: return "120B.A12B";
case LLM_TYPE_122B_A10B: return "122B.A10B";
case LLM_TYPE_196B_A11B: return "196B.A11B";
@ -2210,7 +2214,6 @@ llama_memory_i * llama_model::create_memory(const llama_memory_params & params,
res = nullptr;
} break;
case LLM_ARCH_DEEPSEEK32:
case LLM_ARCH_GLM_DSA:
{
res = new llama_kv_cache_dsa(
*this,
@ -2227,6 +2230,56 @@ llama_memory_i * llama_model::create_memory(const llama_memory_params & params,
nullptr,
nullptr);
} break;
case LLM_ARCH_GLM_DSA:
{
if (params.ctx_type == LLAMA_CONTEXT_TYPE_MTP && hparams.n_layer_nextn > 0) {
// The NextN/MTP draft head runs dense MLA (no DSA indexer), so the
// MTP context uses a plain attention KV cache holding only the
// nextn layer(s) - same pattern as the hybrid Qwen3.5 MTP context.
llama_kv_cache::layer_filter_cb filter =
[&](uint32_t il) { return il >= hparams.n_layer(); };
res = new llama_kv_cache(
*this,
hparams,
params.type_k,
params.type_v,
!cparams.flash_attn,
cparams.offload_kqv,
cparams.kv_unified,
cparams.n_ctx_seq,
cparams.n_seq_max,
1,
hparams.n_swa,
hparams.swa_type,
nullptr,
filter,
nullptr,
nullptr);
} else {
// Main context: DSA cache for the trunk layers only - the nextn
// layer(s) are never attended by the trunk graph.
llama_kv_cache::layer_filter_cb filter = nullptr;
if (hparams.n_layer_nextn > 0) {
filter = [&](uint32_t il) { return il < hparams.n_layer(); };
}
res = new llama_kv_cache_dsa(
*this,
params.type_k,
params.type_v,
!cparams.flash_attn,
cparams.offload_kqv,
cparams.kv_unified,
cparams.n_ctx_seq,
cparams.n_seq_max,
1,
hparams.n_swa,
hparams.swa_type,
filter,
nullptr);
}
} break;
// Models that need standard caching should rely on recurrent/hybrid
// checks
default:
@ -2332,7 +2385,7 @@ llama_memory_i * llama_model::create_memory(const llama_memory_params & params,
filter = [&](uint32_t il) { return il >= hparams.n_layer(); };
}
if ((arch == LLM_ARCH_STEP35 || arch == LLM_ARCH_HY_V3) && hparams.n_layer_nextn > 0) {
if ((arch == LLM_ARCH_STEP35 || arch == LLM_ARCH_HY_V3 || arch == LLM_ARCH_GLM_DSA) && hparams.n_layer_nextn > 0) {
if (params.ctx_type == LLAMA_CONTEXT_TYPE_MTP) {
filter = [&](uint32_t il) { return il >= hparams.n_layer(); };
} else {
@ -2632,6 +2685,7 @@ llama_rope_type llama_model_rope_type(const llama_model * model) {
case LLM_ARCH_LLAMA_EMBED:
case LLM_ARCH_MAINCODER:
case LLM_ARCH_GLM_DSA:
case LLM_ARCH_NANBEIGE:
return LLAMA_ROPE_TYPE_NORM;
// the pairs of head values are offset by n_rot/2