vrr/prima.cpp
2
0
Fork 0
mirror of https://github.com/Lizonghang/prima.cpp.git synced 2025-09-11 03:34:34 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

reset backend_embd and out_embd as input tensors

Browse source
This commit is contained in:
Lizonghang 2024-11-02 20:40:21 +04:00
parent 7aa771e701
commit a83f577c63
1 changed files with 2 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

4
src/llama.cpp
View file

@ -10367,7 +10367,7 @@ struct llm_build_context {
const llm_build_cb & cb) {
lctx.backend_embd = ggml_new_tensor_2d(ctx, GGML_TYPE_F32, hparams.n_embd, batch.n_tokens);
cb(lctx.backend_embd, "backend_embd", -1);
// ggml_set_input(lctx.backend_embd); // set it on the device of the adjacent node
ggml_set_input(lctx.backend_embd);
return lctx.backend_embd;
}
@ -10378,7 +10378,7 @@ struct llm_build_context {
const llm_build_cb & cb) {
lctx.out_embd = ggml_new_tensor_2d(ctx, GGML_TYPE_F32, hparams.n_embd, n_outputs);
cb(lctx.out_embd, "out_embd", -1);
// ggml_set_input(lctx.out_embd); // set it on the device of the adjacent node
ggml_set_input(lctx.out_embd);
return lctx.out_embd;
}