|
@@ -8,6 +8,7 @@
|
|
|
struct llm_build_granite_hybrid : public llm_graph_context_mamba {
|
|
struct llm_build_granite_hybrid : public llm_graph_context_mamba {
|
|
|
llm_build_granite_hybrid(const llama_model & model, const llm_graph_params & params);
|
|
llm_build_granite_hybrid(const llama_model & model, const llm_graph_params & params);
|
|
|
ggml_tensor * build_layer_ffn(ggml_tensor * cur, ggml_tensor * inpSA, const llama_model & model, const int il);
|
|
ggml_tensor * build_layer_ffn(ggml_tensor * cur, ggml_tensor * inpSA, const llama_model & model, const int il);
|
|
|
- ggml_tensor * build_attention_layer(ggml_tensor * cur, ggml_tensor * inp_pos, llm_graph_input_attn_kv * inp_attn,
|
|
|
|
|
|
|
+ ggml_tensor * build_attention_layer(ggml_tensor * cur, ggml_tensor * inp_pos, llm_graph_input_attn_kv * inp_attn,
|
|
|
const llama_model & model,const int64_t n_embd_head, const int il);
|
|
const llama_model & model,const int64_t n_embd_head, const int il);
|
|
|
-};
|
|
|
|
|
|
|
+};
|
|
|
|
|
+
|