vrr/koboldcpp
2
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2026-05-12 22:32:18 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 3

Merge commit '2f966b8ed8' into concedo_experimental

Browse source 
# Conflicts:
#	.github/workflows/release.yml
#	docs/docker.md
#	ggml/src/CMakeLists.txt
#	ggml/src/ggml-cpu/CMakeLists.txt
#	tests/test-backend-ops.cpp
#	tests/test-thread-safety.cpp
#	tools/batched-bench/batched-bench.cpp
#	tools/mtmd/clip.cpp
This commit is contained in:
LostRuins Concedo 2025-11-08 10:34:17 +08:00
commit fdcb281a3a
31 changed files with 655 additions and 165 deletions
Download patch file Download diff file Expand all files Collapse all files

14
src/llama-model.cpp
View file

@ -6869,14 +6869,14 @@ float llama_model::get_rope_freq_scale(const llama_cparams & cparams, int il) co
}
ggml_tensor * llama_model::get_rope_factors(const llama_cparams & cparams, int il) const {
const uint32_t n_ctx_per_seq = cparams.n_ctx / cparams.n_seq_max;
const uint32_t n_ctx_seq = cparams.n_ctx_seq;
// choose long/short freq factors based on the context size
if (layers[il].rope_freqs != nullptr) {
return layers[il].rope_freqs;
}
if (n_ctx_per_seq > hparams.n_ctx_orig_yarn) {
if (n_ctx_seq > hparams.n_ctx_orig_yarn) {
return layers[il].rope_long;
}
@ -6952,12 +6952,6 @@ llama_memory_i * llama_model::create_memory(const llama_memory_params & params,
/* filter_attn */ std::move(filter_attn),
/* filter_recr */ std::move(filter_recr));
} else {
uint32_t n_ctx_per_stream = cparams.n_ctx;
if (!cparams.kv_unified) {
n_ctx_per_stream = (cparams.n_ctx + cparams.n_seq_max - 1)/cparams.n_seq_max;
}
llama_memory_i::layer_reuse_cb reuse = nullptr;
if (arch == LLM_ARCH_GEMMA3N) {
@ -6981,7 +6975,7 @@ llama_memory_i * llama_model::create_memory(const llama_memory_params & params,
cparams.offload_kqv,
params.swa_full,
cparams.kv_unified,
n_ctx_per_stream,
cparams.n_ctx_seq,
cparams.n_seq_max,
cparams.n_ubatch,
1,
@ -6997,7 +6991,7 @@ llama_memory_i * llama_model::create_memory(const llama_memory_params & params,
!cparams.flash_attn,
cparams.offload_kqv,
cparams.kv_unified,
n_ctx_per_stream,
cparams.n_ctx_seq,
cparams.n_seq_max,
1,
hparams.n_swa,