vrr/koboldcpp
2
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2026-05-12 05:52:26 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Merge branch 'upstream' into concedo_experimental

Browse source 
# Conflicts:
#	CODEOWNERS
#	docs/ops.md
#	docs/ops/SYCL.csv
#	examples/embedding/README.md
#	ggml/src/ggml-cann/aclnn_ops.cpp
#	ggml/src/ggml-cann/ggml-cann.cpp
#	ggml/src/ggml-sycl/backend.hpp
#	ggml/src/ggml-sycl/ggml-sycl.cpp
#	ggml/src/ggml-sycl/norm.cpp
#	ggml/src/ggml-sycl/norm.hpp
#	scripts/snapdragon/adb/run-bench.sh
#	scripts/snapdragon/adb/run-cli.sh
#	src/llama-batch.cpp
#	tests/test-backend-ops.cpp
#	tests/test-chat.cpp
#	tests/test-json-schema-to-grammar.cpp
#	tools/llama-bench/README.md
This commit is contained in:
Concedo 2025-10-30 13:44:46 +08:00
commit 16cbe9f24e
55 changed files with 2408 additions and 1222 deletions
Download patch file Download diff file Expand all files Collapse all files

25
src/llama-model.cpp
View file

@ -2243,7 +2243,7 @@ bool llama_model::load_tensors(llama_model_loader & ml) {
// define a comparator for the buft -> ctx map to ensure that the order is well-defined:
struct ggml_backend_buft_comparator {
bool operator()(const ggml_backend_buffer_type_t & lhs, const ggml_backend_buffer_type_t & rhs) const {
return ggml_backend_buft_name(lhs) < ggml_backend_buft_name(rhs);
return strcmp(ggml_backend_buft_name(lhs), ggml_backend_buft_name(rhs)) < 0;
}
};
std::map<ggml_backend_buffer_type_t, ggml_context_ptr, ggml_backend_buft_comparator> ctx_map;
@ -19703,7 +19703,7 @@ struct llm_build_apertus : public llm_graph_context {
}
};
llama_memory_i * llama_model::create_memory(const llama_memory_params & params, llama_cparams & cparams) const {
llama_memory_i * llama_model::create_memory(const llama_memory_params & params, const llama_cparams & cparams) const {
llama_memory_i * res;
switch (arch) {
@ -19754,17 +19754,13 @@ llama_memory_i * llama_model::create_memory(const llama_memory_params & params,
};
}
const auto padding = llama_kv_cache::get_padding(cparams);
cparams.n_ctx = GGML_PAD(cparams.n_ctx, padding);
res = new llama_memory_hybrid(
/* model */ *this,
/* attn_type_k */ params.type_k,
/* attn_type_v */ params.type_v,
/* attn_v_trans */ !cparams.flash_attn,
/* attn_kv_size */ cparams.n_ctx,
/* attn_n_pad */ padding,
/* attn_n_pad */ 1,
/* attn_n_swa */ hparams.n_swa,
/* attn_swa_type */ hparams.swa_type,
/* recurrent_type_k */ GGML_TYPE_F32,
@ -19776,23 +19772,12 @@ 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 {
const auto padding = llama_kv_cache::get_padding(cparams);
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;
n_ctx_per_stream = GGML_PAD(n_ctx_per_stream, padding);
cparams.n_ctx = n_ctx_per_stream*cparams.n_seq_max;
} else {
n_ctx_per_stream = GGML_PAD(n_ctx_per_stream, padding);
cparams.n_ctx = n_ctx_per_stream;
}
LLAMA_LOG_DEBUG("%s: n_ctx = %u (padded)\n", __func__, cparams.n_ctx);
llama_memory_i::layer_reuse_cb reuse = nullptr;
if (arch == LLM_ARCH_GEMMA3N) {
@ -19819,7 +19804,7 @@ llama_memory_i * llama_model::create_memory(const llama_memory_params & params,
n_ctx_per_stream,
cparams.n_seq_max,
cparams.n_ubatch,
padding,
1,
nullptr,
reuse);
} else {
@ -19834,7 +19819,7 @@ llama_memory_i * llama_model::create_memory(const llama_memory_params & params,
cparams.kv_unified,
n_ctx_per_stream,
cparams.n_seq_max,
padding,
1,
hparams.n_swa,
hparams.swa_type,
nullptr,