wylab/llama.cpp
0
0
Fork 1
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-09-15 18:13:29 +02:00
Code Issues Packages Projects Releases Wiki Activity

wip

Browse Source
This commit is contained in:
Georgi Gerganov
2026-09-15 15:23:22 +03:00
parent 2efe7c9e5c
commit 59bf83ccd2
1 changed files with 9 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/llama-graph.cpp
+9 -1
View File
@@ -2451,6 +2451,7 @@ ggml_tensor * llm_graph_context::build_inp_pos() const {
cur = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, (int64_t)n_tokens*hparams.n_pos_per_embd());
ggml_set_input(cur);
cb(cur, "inp_pos", -1);
res->add_input(std::move(inp));
@@ -2465,7 +2466,7 @@ ggml_tensor * llm_graph_context::build_inp_attn_scale() const {
// this need to be 1x1xN for broadcasting
cur = ggml_new_tensor_3d(ctx0, GGML_TYPE_F32, 1, 1, n_tokens);
ggml_set_input(cur);
ggml_set_name(cur, "attn_scale");
cb(cur, "inp_attn_scale", -1);
res->add_input(std::move(inp));
@@ -2487,6 +2488,7 @@ ggml_tensor * llm_graph_context::build_inp_out_ids() const {
cur = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, n_outputs);
ggml_set_input(cur);
ggml_set_name(cur, "out_ids");
res->add_input(std::move(inp));
@@ -2500,6 +2502,7 @@ ggml_tensor * llm_graph_context::build_inp_mean() const {
cur = ggml_new_tensor_2d(ctx0, GGML_TYPE_F32, n_tokens, ubatch.n_seqs_unq);
ggml_set_input(cur);
ggml_set_name(cur, "mean");
res->add_input(std::move(inp));
@@ -2513,6 +2516,7 @@ ggml_tensor * llm_graph_context::build_inp_cls() const {
cur = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, ubatch.n_seqs_unq);
ggml_set_input(cur);
ggml_set_name(cur, "cls");
res->add_input(std::move(inp));
@@ -2537,6 +2541,7 @@ ggml_tensor * llm_graph_context::build_inp_cross_embd() const {
cur = ggml_new_tensor_2d(ctx0, GGML_TYPE_F32, n_embd, n_enc);
ggml_set_input(cur);
ggml_set_name(cur, "cross_embd");
res->add_input(std::move(inp));
@@ -2550,6 +2555,7 @@ ggml_tensor * llm_graph_context::build_inp_pos_bucket_enc() const {
cur = ggml_new_tensor_2d(ctx0, GGML_TYPE_I32, n_tokens, n_tokens);
ggml_set_input(cur);
ggml_set_name(cur, "pos_bucket_enc");
res->add_input(std::move(inp));
@@ -2567,6 +2573,7 @@ ggml_tensor * llm_graph_context::build_inp_pos_bucket_dec() const {
cur = ggml_new_tensor_2d(ctx0, GGML_TYPE_I32, n_kv, n_tokens);
ggml_set_input(cur);
ggml_set_name(cur, "pos_bucket_dec");
res->add_input(std::move(inp));
@@ -2735,6 +2742,7 @@ llm_graph_input_attn_no_cache * llm_graph_context::build_attn_inp_no_cache() con
// note: there is no KV cache, so the number of KV values is equal to the number of tokens in the batch
inp->self_kq_mask = ggml_new_tensor_4d(ctx0, type_mask, n_tokens, n_tokens, 1, 1);
ggml_set_input(inp->self_kq_mask);
cb(inp->self_kq_mask, "self_kq_mask", -1);
inp->self_kq_mask_cnv = inp->self_kq_mask;