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

build_rwkv6: Simplify graph

Browse source 
Signed-off-by: Molly Sophia <mollysophia379@gmail.com>
This commit is contained in:
Molly Sophia 2024-08-12 14:47:26 +08:00
parent 12fbe1ade2
commit 276d53b18f
1 changed files with 11 additions and 33 deletions
Download patch file Download diff file Expand all files Collapse all files

44
src/llama.cpp
View file

@ -9398,40 +9398,18 @@ static struct ggml_tensor * llm_build_time_mix_rwkv6(
xxx xxx
); );
struct ggml_tensor *mw = ggml_new_tensor_1d(ctx, GGML_TYPE_F32, n_embed * n_tokens); // struct ggml_tensor *mw = ggml_new_tensor_1d(ctx, GGML_TYPE_F32, n_embed * n_tokens);
mw = ggml_reshape_2d( // mw = ggml_reshape_2d(
ctx, // ctx,
ggml_set_1d(ctx, mw, ggml_view_1d(ctx, xxx, n_embed * n_tokens, 0), 0), // ggml_set_1d(ctx, mw, ggml_view_1d(ctx, xxx, n_embed * n_tokens, 0), 0),
n_embed, n_tokens // n_embed, n_tokens
); // );
struct ggml_tensor *mk = ggml_new_tensor_1d(ctx, GGML_TYPE_F32, n_embed * n_tokens); struct ggml_tensor *mw = ggml_view_2d(ctx, xxx, n_embed, n_tokens, xxx->nb[1], 0);
mk = ggml_reshape_2d( struct ggml_tensor *mk = ggml_view_2d(ctx, xxx, n_embed, n_tokens, xxx->nb[1], n_embed * n_tokens * sizeof(float));
ctx, struct ggml_tensor *mv = ggml_view_2d(ctx, xxx, n_embed, n_tokens, xxx->nb[1], n_embed * n_tokens * 2 * sizeof(float));
ggml_set_1d(ctx, mk, ggml_view_1d(ctx, xxx, n_embed * n_tokens, n_embed * n_tokens * sizeof(float)), 0), struct ggml_tensor *mr = ggml_view_2d(ctx, xxx, n_embed, n_tokens, xxx->nb[1], n_embed * n_tokens * 3 * sizeof(float));
n_embed, n_tokens struct ggml_tensor *mg = ggml_view_2d(ctx, xxx, n_embed, n_tokens, xxx->nb[1], n_embed * n_tokens * 4 * sizeof(float));
);
struct ggml_tensor *mv = ggml_new_tensor_1d(ctx, GGML_TYPE_F32, n_embed * n_tokens);
mv = ggml_reshape_2d(
ctx,
ggml_set_1d(ctx, mv, ggml_view_1d(ctx, xxx, n_embed * n_tokens, n_embed * n_tokens * 2 * sizeof(float)), 0),
n_embed, n_tokens
);
struct ggml_tensor *mr = ggml_new_tensor_1d(ctx, GGML_TYPE_F32, n_embed * n_tokens);
mr = ggml_reshape_2d(
ctx,
ggml_set_1d(ctx, mr, ggml_view_1d(ctx, xxx, n_embed * n_tokens, n_embed * n_tokens * 3 * sizeof(float)), 0),
n_embed, n_tokens
);
struct ggml_tensor *mg = ggml_new_tensor_1d(ctx, GGML_TYPE_F32, n_embed * n_tokens);
mg = ggml_reshape_2d(
ctx,
ggml_set_1d(ctx, mg, ggml_view_1d(ctx, xxx, n_embed * n_tokens, n_embed * n_tokens * 4 * sizeof(float)), 0),
n_embed, n_tokens
);
struct ggml_tensor * xw = ggml_add( struct ggml_tensor * xw = ggml_add(
ctx, ctx,