llama : add Gemma4 MTP (#23398)

2026-06-07 20:50:54 +08:00
parent 8a091c47ab
commit 04eb4c446d
31 changed files with 644 additions and 142 deletions
@@ -23,6 +23,8 @@ struct llama_memory_params {
    bool swa_full;

    llama_context_type ctx_type;
+
+    llama_memory_t mem_other;
 };

 enum llama_memory_status {
@@ -76,6 +78,8 @@ struct llama_memory_i {
    // return negative value to indicate that the layer il should not reuse memory
    using layer_reuse_cb = std::function<int32_t(int32_t il)>;

+    using layer_share_cb = std::function<int32_t(int32_t il)>;
+
    virtual ~llama_memory_i() = default;

    // split the input batch into a set of ubatches and verify that they can fit into the cache