llama : remove KV cache defragmentation logic (#15473)

ggml-ci
2025-08-22 12:22:13 +03:00 · 2025-08-22 12:22:13 +03:00 · 9ebebef62f
commit 9ebebef62f
parent ad5c975c2d
16 changed files with 32 additions and 440 deletions
--- a/common/common.h
+++ b/common/common.h
@ -288,7 +288,6 @@ struct common_params {
    float   yarn_beta_fast        = 32.0f; // YaRN low correction dim
    float   yarn_beta_slow        =  1.0f; // YaRN high correction dim
    int32_t yarn_orig_ctx         =     0; // YaRN original context length
-    float   defrag_thold          =  0.1f; // KV cache defragmentation threshold

    // offload params
    std::vector<ggml_backend_dev_t> devices; // devices to use for offloading