mirrors/ollama-for-amd
1
0
Fork 0
mirror of https://github.com/likelovewant/ollama-for-amd.git synced 2025-12-21 22:33:56 +00:00
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

7
llama/llama.cpp/src/llama-hparams.h vendored
View File

@@ -2,8 +2,6 @@
#include "llama.h"
#include <algorithm>
#include <array>
// bump if necessary
@@ -44,7 +42,6 @@ struct llama_hparams {
uint32_t n_expert = 0;
uint32_t n_expert_used = 0;
uint32_t n_rel_attn_bkts = 0;
uint32_t n_vocab = 0;
// note: deepseek2 using MLA converts into MQA with larger heads, then decompresses to MHA
uint32_t n_embd_head_k_mla = 0;
@@ -59,7 +56,6 @@ struct llama_hparams {
std::array<uint32_t, LLAMA_MAX_LAYERS> n_ff_arr;
std::array<std::array<uint32_t, LLAMA_MAX_LAYERS>, 4> n_bskcn_arr = {};
std::array<uint32_t, LLAMA_MAX_LAYERS> cross_attn_layers;
uint32_t n_layer_dense_lead = 0;
uint32_t n_lora_q = 0;
@@ -163,9 +159,6 @@ struct llama_hparams {
// Block skip connection
bool n_bskcn(uint32_t n, uint32_t il) const;
// cross attention layers
bool cross_attention_layers(uint32_t il) const;
bool is_swa(uint32_t il) const;
};