highperfocused/ollama
1
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2025-11-12 10:28:00 +01:00
Code Issues Packages Projects Releases Wiki Activity

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

Browse Source
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
Download Patch File Download Diff File Expand all files Collapse all files

12
llama/llama.cpp/src/llama-graph.h vendored
View File

@@ -87,7 +87,6 @@ public:
ggml_tensor * tokens = nullptr; // I32 [n_batch]
ggml_tensor * embd = nullptr; // F32 [n_embd, n_batch]
ggml_tensor * cross_attn_state; // F32 [4, n_embd, 1061]
};
class llm_graph_input_pos : public llm_graph_input_i {
@@ -285,16 +284,6 @@ public:
const llama_cross * cross = nullptr;
};
class llm_graph_input_cross_attn_state : public llm_graph_input_i {
public:
llm_graph_input_cross_attn_state() = default;
virtual ~llm_graph_input_cross_attn_state() = default;
void set_input(const llama_ubatch * ubatch) override;
ggml_tensor * cross_attn_state; // F32 [4, n_embd, 1061]
};
//
// llm_graph_result
//
@@ -506,7 +495,6 @@ struct llm_graph_context {
ggml_tensor * build_inp_cls() const;
ggml_tensor * build_inp_s_copy() const;
ggml_tensor * build_inp_s_mask() const;
ggml_tensor * build_inp_cross_attn_state() const;
ggml_tensor * build_inp_cross_embd() const;
ggml_tensor * build_inp_pos_bucket_enc() const;