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

llama: rwkv6: Remove unused nodes

Browse source 
Signed-off-by: Molly Sophia <mollysophia379@gmail.com>
This commit is contained in:
Molly Sophia 2024-08-26 09:50:51 +08:00
parent e0ea51144e
commit 5f00c52be0
1 changed files with 0 additions and 9 deletions
Download patch file Download diff file Expand all files Collapse all files

9
src/llama.cpp
View file

@ -15135,15 +15135,6 @@ struct llm_build_context {
cur = ggml_add(ctx0, cur, llm_build_channel_mix_rwkv6(ctx0, layer, x_norm_ffn, x_prev)); cur = ggml_add(ctx0, cur, llm_build_channel_mix_rwkv6(ctx0, layer, x_norm_ffn, x_prev));
ggml_build_forward_expand(gf, cur); ggml_build_forward_expand(gf, cur);
token_shift = ggml_cont(
ctx0,
ggml_permute(
ctx0,
ggml_reshape_3d(ctx0, token_shift, n_embd, n_seqs, 2),
0, 2, 1, 3
)
);
struct ggml_tensor * last_norm_att = ggml_view_3d(ctx0, x_norm_att, n_embd, 1, n_seqs, x_norm_att->nb[1], x_norm_att->nb[2], (n_seq_tokens-1)*n_embd*ggml_element_size(x_norm_att)); struct ggml_tensor * last_norm_att = ggml_view_3d(ctx0, x_norm_att, n_embd, 1, n_seqs, x_norm_att->nb[1], x_norm_att->nb[2], (n_seq_tokens-1)*n_embd*ggml_element_size(x_norm_att));
struct ggml_tensor * last_norm_ffn = ggml_view_3d(ctx0, x_norm_ffn, n_embd, 1, n_seqs, x_norm_ffn->nb[1], x_norm_ffn->nb[2], (n_seq_tokens-1)*n_embd*ggml_element_size(x_norm_ffn)); struct ggml_tensor * last_norm_ffn = ggml_view_3d(ctx0, x_norm_ffn, n_embd, 1, n_seqs, x_norm_ffn->nb[1], x_norm_ffn->nb[2], (n_seq_tokens-1)*n_embd*ggml_element_size(x_norm_ffn));