vrr/koboldcpp
2
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2026-05-21 18:52:02 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

Merge branch 'upstream' into concedo_experimental

Browse source 
# Conflicts:
#	README.md
#	docs/build.md
#	ggml/src/ggml-cpu/CMakeLists.txt
#	ggml/src/ggml-cpu/kleidiai/kernels.cpp
#	ggml/src/ggml-cpu/kleidiai/kernels.h
#	ggml/src/ggml-cpu/kleidiai/kleidiai.cpp
#	tests/test-backend-ops.cpp
#	tools/server/README.md
This commit is contained in:
Concedo 2025-07-21 23:37:42 +08:00
commit 4abea4b5c9
4 changed files with 15 additions and 13 deletions
Download patch file Download diff file Expand all files Collapse all files

2
ggml/src/ggml-cuda/im2col.cu
View file

@ -10,7 +10,7 @@ static __global__ void im2col_kernel(
return;
}
const int64_t ksize = OW * (KH > 1 ? KW : 1);
const int64_t ksize = OW * KH;
const int64_t kx = i / ksize;
const int64_t kd = kx * ksize;
const int64_t ky = (i - kd) / OW;

6
ggml/src/ggml-vulkan/vulkan-shaders/im2col.comp
View file

@ -40,12 +40,10 @@ void main() {
const uint src_base = ic * p.offset_delta + batch * p.batch_offset;
const uint dst_base = ((batch * p.OH + oh) * p.OW) * p.CHW + ic * (p.KW * p.KH);
const int oh_s1 = int(oh) * p.s1;
const uint ksize = p.OW * (p.KH > 1 ? p.KW : 1);
const uint ksize = p.OW * p.KH;
const uint base_linear_idx = gidx * NUM_ITER;
const uint max_ky = ksize / p.OW;
uint current_kx = base_linear_idx / ksize;
const uint rem = base_linear_idx - (current_kx * ksize);
uint current_ky = rem / p.OW;
@ -76,7 +74,7 @@ void main() {
if (++current_ix == p.OW) {
current_ix = 0;
if (++current_ky == max_ky) {
if (++current_ky == p.KH) {
current_ky = 0;
current_kx++;
}

17
tools/main/main.cpp
View file

@ -786,14 +786,17 @@ int main(int argc, char ** argv) {
}
// check for reverse prompt using special tokens
llama_token last_token = common_sampler_last(smpl);
for (auto token : antiprompt_token) {
if (token == last_token) {
if (params.interactive) {
is_interacting = true;
// avoid calling common_sampler_last() if last_output is empty
if (!last_output.empty()) {
llama_token last_token = common_sampler_last(smpl);
for (auto token : antiprompt_token) {
if (token == last_token) {
if (params.interactive) {
is_interacting = true;
}
is_antiprompt = true;
break;
}
is_antiprompt = true;
break;
}
}

3
tools/server/server.cpp
View file

@ -4516,9 +4516,10 @@ int main(int argc, char ** argv) {
json tokens_response = json::array();
if (body.count("content") != 0) {
const bool add_special = json_value(body, "add_special", false);
const bool parse_special = json_value(body, "parse_special", true);
const bool with_pieces = json_value(body, "with_pieces", false);
llama_tokens tokens = tokenize_mixed(ctx_server.vocab, body.at("content"), add_special, true);
llama_tokens tokens = tokenize_mixed(ctx_server.vocab, body.at("content"), add_special, parse_special);
if (with_pieces) {
for (const auto& token : tokens) {