chore: update mllama to use ollama engine (#10637)

This commit is contained in:
Michael Yang
2025-05-13 17:36:02 -07:00
committed by GitHub
parent 0478d440f0
commit 23125648b8
67 changed files with 785 additions and 4354 deletions

View File

@@ -11,7 +11,6 @@
#include <string>
#include <unordered_map>
#include <vector>
#include <stdexcept>
struct llama_cparams;
struct llama_ubatch;
@@ -75,7 +74,6 @@ enum llm_type {
LLM_TYPE_40B,
LLM_TYPE_65B,
LLM_TYPE_70B,
LLM_TYPE_90B,
LLM_TYPE_236B,
LLM_TYPE_290B,
LLM_TYPE_314B,
@@ -320,16 +318,6 @@ struct llama_layer {
struct ggml_tensor * bskcn_tv = nullptr;
// cross attention
struct ggml_tensor * cross_attn_k_norm = nullptr;
struct ggml_tensor * cross_attn_k_proj = nullptr;
struct ggml_tensor * cross_attn_o_proj = nullptr;
struct ggml_tensor * cross_attn_q_norm = nullptr;
struct ggml_tensor * cross_attn_q_proj = nullptr;
struct ggml_tensor * cross_attn_v_proj = nullptr;
struct ggml_tensor * cross_attn_attn_gate = nullptr;
struct ggml_tensor * cross_attn_mlp_gate = nullptr;
struct llama_layer_posnet posnet;
struct llama_layer_convnext convnext;