vrr/koboldcpp
2
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2025-09-10 09:04:36 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Merge branch 'master' into concedo_experimental

Browse source 
# Conflicts:
#	CMakeLists.txt
#	Makefile
#	README.md
#	docs/BLIS.md
This commit is contained in:
Concedo 2023-10-15 15:50:20 +08:00
commit 5cfabaee25
21 changed files with 10952 additions and 278 deletions
Download patch file Download diff file Expand all files Collapse all files

7
llama.cpp
View file

@ -2845,8 +2845,8 @@ static void llm_load_tensors(
auto & layer = model.layers[i];
layer.attn_norm = ml.create_tensor(ctx, tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}, backend);
layer.wqkv = ml.create_tensor(ctx, tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, 3*n_embd}, backend_split);
layer.wo = ml.create_tensor(ctx, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, backend_split);
layer.wqkv = ml.create_tensor(ctx, tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa}, backend_split);
layer.wo = ml.create_tensor(ctx, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}, backend_split);
layer.ffn_norm = ml.create_tensor(ctx, tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, backend);
@ -5374,7 +5374,7 @@ static struct ggml_cgraph * llm_build_mpt(
const int64_t n_layer = hparams.n_layer;
const int64_t n_ctx = cparams.n_ctx;
const int64_t n_head = hparams.n_head;
const int64_t n_head_kv = hparams.n_head_kv; // == n_head for MPT, as there's no MQA/GQA
const int64_t n_head_kv = hparams.n_head_kv;
const int64_t n_embd_head = hparams.n_embd_head();
const int64_t n_embd_gqa = hparams.n_embd_gqa();
@ -5727,7 +5727,6 @@ static struct ggml_cgraph * llama_build_graph(
//
// - lctx: llama context
// - batch: batch to evaluate
// - n_threads: number of threads to use
//
// return 0 on success
// return positive int on warning