Skip to content

server: do not propagate --kv-mean-center bias to draft model contexts #81

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
cnndabbler wants to merge 1 commit into
base: prism
Choose a base branch
from
+8 −0
Open

server: do not propagate --kv-mean-center bias to draft model contexts #81

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 8 additions & 0 deletions tools/server/server-context.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -901,6 +901,10 @@ struct server_context_impl {
params_dft.cache_type_k = params_spec.cache_type_k;
params_dft.cache_type_v = params_spec.cache_type_v;
params_dft.tensor_buft_overrides = params_spec.tensor_buft_overrides;
// the K-cache mean-center bias is calibrated for the target model
// (per-head/channel K layout); the draft model has a different K
// geometry, so it must not inherit the bias
params_dft.kv_mean_center_path.clear();
} else {
// MTP draft context lives on the target model, only context+compute are new
measure_model_bytes = false;
Expand Down Expand Up @@ -991,6 +995,10 @@ struct server_context_impl {
params_dft.n_gpu_layers = params_spec.n_gpu_layers;
params_dft.cache_type_k = params_spec.cache_type_k;
params_dft.cache_type_v = params_spec.cache_type_v;
// the K-cache mean-center bias is calibrated for the target model
// (per-head/channel K layout); the draft model has a different K
// geometry, so it must not inherit the bias
params_dft.kv_mean_center_path.clear();

if (params_spec.cpuparams.n_threads > 0) {
params_dft.cpuparams.n_threads = params_spec.cpuparams.n_threads;
Expand Down
Loading