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:
#	.github/workflows/docker.yml
#	ggml/src/ggml-opencl/kernels/mul_mm_f16_f32_l4_lm.cl
#	ggml/src/ggml-opencl/kernels/mul_mm_f32_f32_l4_lm.cl
#	ggml/src/ggml-sycl/rope.cpp
#	ggml/src/ggml-webgpu/wgsl-shaders/rope.tmpl.wgsl
#	requirements/requirements-convert_legacy_llama.txt
#	tests/test-backend-ops.cpp
#	tests/test-rope.cpp
#	tools/server/README.md
This commit is contained in:
Concedo 2025-10-31 10:52:57 +08:00
commit 2b00e55356
35 changed files with 1905 additions and 290 deletions
Download patch file Download diff file Expand all files Collapse all files

2
src/llama-kv-cache.cpp
View file

@ -1375,7 +1375,7 @@ ggml_tensor * llama_kv_cache::build_rope_shift(
const auto & yarn_beta_slow = cparams.yarn_beta_slow;
const auto & n_rot = hparams.n_rot;
const auto & rope_type = hparams.rope_type == LLAMA_ROPE_TYPE_MROPE
const auto & rope_type = hparams.rope_type == LLAMA_ROPE_TYPE_MROPE || hparams.rope_type == LLAMA_ROPE_TYPE_IMROPE
// @ngxson : this is a workaround
// for M-RoPE, we want to rotate the whole vector when doing KV shift
// a normal RoPE should work, we just need to use the correct ordering