CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-11-03 09:22:01 +00:00
Code Issues Packages Projects Releases Wiki Activity

llama : update per-seq context computation

Browse Source
This commit is contained in:
Georgi Gerganov
2025-10-23 17:54:53 +03:00
parent f3d1607579
commit 2ca720c859
8 changed files with 40 additions and 36 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
tools/server/server.cpp
View File

@@ -2379,10 +2379,6 @@ struct server_context {
llama_batch_free(batch);
}
int32_t n_ctx_slot() const {
return params_base.kv_unified ? n_ctx : n_ctx / params_base.n_parallel;
}
bool load_model(const common_params & params) {
SRV_INF("loading model '%s'\n", params.model.path.c_str());
@@ -2411,7 +2407,7 @@ struct server_context {
params_dft.devices = params_base.speculative.devices;
params_dft.model = params_base.speculative.model;
params_dft.n_ctx = params_base.speculative.n_ctx == 0 ? n_ctx_slot() : params_base.speculative.n_ctx;
params_dft.n_ctx = params_base.speculative.n_ctx == 0 ? llama_n_ctx_seq(ctx) : params_base.speculative.n_ctx;
params_dft.n_gpu_layers = params_base.speculative.n_gpu_layers;
params_dft.n_parallel = 1;
params_dft.cache_type_k = params_base.speculative.cache_type_k;
@@ -2506,7 +2502,7 @@ struct server_context {
slot.id = i;
slot.ctx = ctx;
slot.n_ctx = n_ctx_slot();
slot.n_ctx = llama_n_ctx_seq(ctx);
slot.mctx = mctx;
slot.prompt.tokens.has_mtmd = mctx != nullptr;