llama.cpp
833dfb54 - fix: Use per-layer n_embd_k/v_s calls for mamba (1) layers

Commit

114 days ago

fix: Use per-layer n_embd_k/v_s calls for mamba (1) layers Branch: HybridRecurrentCache Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

References

Author

gabe-l-hart

Committer

gabe-l-hart

Parents