vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

separate vision ctx and llm ctx

Browse source
This commit is contained in:
Xuan Son Nguyen 2025-02-06 20:32:09 +01:00
parent ff77b15845
commit fa55281759
7 changed files with 139 additions and 35 deletions
Download patch file Download diff file Expand all files Collapse all files

4
src/llama-arch.cpp
View file

@ -1576,8 +1576,8 @@ static const std::map<llm_tensor, llm_tensor_info> LLM_TENSOR_INFOS = {
{LLM_TENSOR_V_ENC_OUTPUT_NORM, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
{LLM_TENSOR_V_ENC_FFN_UP, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
{LLM_TENSOR_V_ENC_FFN_DOWN, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
{LLM_TENSOR_V_PRE_NORM, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
{LLM_TENSOR_V_POST_NORM, {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
{LLM_TENSOR_V_PRE_NORM, {LLM_TENSOR_LAYER_INPUT, GGML_OP_MUL}},
{LLM_TENSOR_V_POST_NORM, {LLM_TENSOR_LAYER_OUTPUT, GGML_OP_MUL}},
{LLM_TENSOR_V_RESMPL_POS_EMBD_K, {LLM_TENSOR_LAYER_PROJECTION, GGML_OP_ADD}},
{LLM_TENSOR_V_RESMPL_ATTN_Q, {LLM_TENSOR_LAYER_PROJECTION, GGML_OP_MUL_MAT}},
{LLM_TENSOR_V_RESMPL_ATTN_K, {LLM_TENSOR_LAYER_PROJECTION, GGML_OP_MUL_MAT}},