mirrors/ollama
1
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2025-05-12 02:46:36 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

remove mllama patch

Browse source
This commit is contained in:
Michael Yang 2025-04-25 15:54:42 -07:00
parent 0d6e35d3c6
commit f8586c6b2b
35 changed files with 37 additions and 2444 deletions
Download patch file Download diff file Expand all files Collapse all files

6
llama/llama.cpp/include/llama.h vendored
View file

@ -256,7 +256,6 @@ extern "C" {
llama_token * token;
float * embd;
int32_t n_embd;
llama_pos * pos;
int32_t * n_seq_id;
llama_seq_id ** seq_id;
@ -359,7 +358,6 @@ extern "C" {
bool offload_kqv; // whether to offload the KQV ops (including the KV cache) to GPU
bool flash_attn; // whether to use flash attention [EXPERIMENTAL]
bool no_perf; // whether to measure performance timings
bool cross_attn; // whether to use cross attention
// Abort callback
// if it returns true, execution of llama_decode() will be aborted
@ -461,10 +459,6 @@ extern "C" {
struct llama_context_params params),
"use llama_init_from_model instead");
// TODO (jmorganca): this should most likely be passed in as part of a batch
// and not set on the context for all batches.
LLAMA_API void llama_set_cross_attention(struct llama_context * ctx, bool cross_attn_state);
// Frees all allocated memory
LLAMA_API void llama_free(struct llama_context * ctx);