Skip to content

graph : constant topology for tokens/embeddings inputs #18549

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
ggerganov wants to merge 2 commits into from
Closed

graph : constant topology for tokens/embeddings inputs #18549

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
4ed59dc
graph : reduce topology branching
ggerganov Jan 2, 2026
bf3f12d
graph : constant topology for tokens/embeddings inputs
ggerganov Jan 2, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
28 changes: 15 additions & 13 deletions src/llama-graph.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,7 +21,8 @@ void llm_graph_input_embd::set_input(const llama_ubatch * ubatch) {
}

if (ubatch->embd) {
const int64_t n_embd = embd->ne[0];
GGML_ASSERT(n_embd == embd->ne[0]);

const int64_t n_tokens = ubatch->n_tokens;

ggml_backend_tensor_set(embd, ubatch->embd, 0, n_tokens*n_embd*ggml_element_size(embd));
Expand Down Expand Up @@ -1206,17 +1207,21 @@ ggml_tensor * llm_graph_context::build_moe_ffn(
ggml_tensor * llm_graph_context::build_inp_embd(ggml_tensor * tok_embd) const {
const int64_t n_embd = hparams.n_embd_inp();

auto inp = std::make_unique<llm_graph_input_embd>();
auto inp = std::make_unique<llm_graph_input_embd>(n_embd);

ggml_tensor * cur = nullptr;
inp->tokens = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, ubatch.n_tokens);
cb(inp->tokens, "inp_tokens", -1);
ggml_set_input(inp->tokens);
res->t_tokens = inp->tokens;

if (ubatch.token) {
inp->tokens = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, ubatch.n_tokens);
//cb(inp->tokens, "inp_tokens", -1);
ggml_set_input(inp->tokens);
res->t_tokens = inp->tokens;
inp->embd = ggml_new_tensor_2d(ctx0, GGML_TYPE_F32, n_embd, ubatch.n_tokens);
ggml_set_input(inp->embd);

ggml_tensor * cur = nullptr;

{
cur = ggml_get_rows(ctx0, tok_embd, inp->tokens);
cur = ggml_scale(ctx0, cur, ubatch.token ? 1.0f : 0.0f);

// apply lora for embedding tokens if needed
for (const auto & lora : *loras) {
Expand All @@ -1235,13 +1240,10 @@ ggml_tensor * llm_graph_context::build_inp_embd(ggml_tensor * tok_embd) const {

cur = ggml_add(ctx0, cur, inpL_delta);
}
} else {
inp->embd = ggml_new_tensor_2d(ctx0, GGML_TYPE_F32, n_embd, ubatch.n_tokens);
ggml_set_input(inp->embd);

cur = inp->embd;
}

cur = ggml_add(ctx0, cur, ggml_scale(ctx0, inp->embd, ubatch.embd ? 1.0f : 0.0f));

// For Granite architecture
if (hparams.f_embedding_scale != 0.0f) {
cur = ggml_scale(ctx0, cur, hparams.f_embedding_scale);
Expand Down
4 changes: 3 additions & 1 deletion src/llama-graph.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -104,7 +104,7 @@ using llm_graph_input_ptr = std::unique_ptr<llm_graph_input_i>;

class llm_graph_input_embd : public llm_graph_input_i {
public:
llm_graph_input_embd() = default;
llm_graph_input_embd(int64_t n_embd) : n_embd(n_embd) {}
virtual ~llm_graph_input_embd() = default;

void set_input(const llama_ubatch * ubatch) override;
Expand All @@ -113,6 +113,8 @@ class llm_graph_input_embd : public llm_graph_input_i {

ggml_tensor * tokens = nullptr; // I32 [n_batch]
ggml_tensor * embd = nullptr; // F32 [n_embd, n_batch]

const int64_t n_embd = 0;
};

class llm_graph_input_pos : public llm_graph_input_i {
Expand Down
8 changes: 5 additions & 3 deletions src/models/cogvlm.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3,12 +3,14 @@
llm_build_cogvlm::llm_build_cogvlm(const llama_model & model, const llm_graph_params & params) :
llm_graph_context(params) {
const int64_t n_embd_head = hparams.n_embd_head_v;
float kq_scale = 1.0f / sqrtf(float(n_embd_head));
const float kq_scale = 1.0f / sqrtf(float(n_embd_head));

GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
GGML_ASSERT(n_embd_head == hparams.n_rot);

ggml_tensor *inpL, *cur;
ggml_tensor * inpL;
ggml_tensor * cur;

inpL = build_inp_embd(model.tok_embd);

ggml_tensor * inp_pos = build_inp_pos();
Expand Down Expand Up @@ -44,7 +46,7 @@ llm_build_cogvlm::llm_build_cogvlm(const llama_model & model, const llm_graph_pa
}

ggml_tensor * inpSA = inpL;
cur = build_norm(inpSA, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);
cur = build_norm(inpSA, model.layers[il].attn_norm, NULL, LLM_NORM_RMS, il);

// build self attention
{
Expand Down
8 changes: 2 additions & 6 deletions src/models/gemma-embedding.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,7 +1,5 @@
#include "models.h"



llm_build_gemma_embedding::llm_build_gemma_embedding(const llama_model & model, const llm_graph_params & params) :
llm_graph_context(params) {
const int64_t n_embd_head = hparams.n_embd_head_k;
Expand All @@ -12,10 +10,8 @@ llm_build_gemma_embedding::llm_build_gemma_embedding(const llama_model & model,
inpL = build_inp_embd(model.tok_embd);

// important: do not normalize weights for raw embeddings input (i.e. encoded image emdeddings)
if (ubatch.token) {
inpL = ggml_scale(ctx0, inpL, sqrtf(n_embd));
cb(inpL, "inp_scaled", -1);
}
inpL = ggml_scale(ctx0, inpL, ubatch.token ? sqrtf(n_embd) : 1.0f);
cb(inpL, "inp_scaled", -1);

// inp_pos - contains the positions
ggml_tensor * inp_pos = build_inp_pos();
Expand Down
7 changes: 3 additions & 4 deletions src/models/gemma3.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10,10 +10,9 @@ llm_build_gemma3<iswa>::llm_build_gemma3(const llama_model & model, const llm_gr
inpL = build_inp_embd(model.tok_embd);

// important: do not normalize weights for raw embeddings input (i.e. encoded image emdeddings)
if (ubatch.token) {
inpL = ggml_scale(ctx0, inpL, sqrtf(n_embd));
cb(inpL, "inp_scaled", -1);
}
inpL = ggml_scale(ctx0, inpL, ubatch.token ? sqrtf(n_embd) : 1.0f);
cb(inpL, "inp_scaled", -1);

// inp_pos - contains the positions
ggml_tensor * inp_pos = build_inp_pos();

Expand Down
11 changes: 4 additions & 7 deletions src/models/gemma3n-iswa.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,7 +1,5 @@
#include "models.h"



llm_build_gemma3n_iswa::llm_build_gemma3n_iswa(const llama_model & model, const llm_graph_params & params) :
llm_graph_context(params),
model(model),
Expand All @@ -15,10 +13,9 @@ llm_build_gemma3n_iswa::llm_build_gemma3n_iswa(const llama_model & model, const
inpL = build_inp_embd(model.tok_embd);

// important: do not normalize weights for raw embeddings input (i.e. encoded image emdeddings)
if (ubatch.token) {
inpL = ggml_scale(ctx0, inpL, sqrtf(n_embd));
cb(inpL, "inp_scaled", -1);
}
inpL = ggml_scale(ctx0, inpL, ubatch.token ? sqrtf(n_embd) : 1.0f);
cb(inpL, "inp_scaled", -1);

// inp_pos - contains the positions
ggml_tensor * inp_pos = build_inp_pos();

Expand Down Expand Up @@ -248,7 +245,7 @@ ggml_tensor * llm_build_gemma3n_iswa::view_2d_slice(ggml_tensor * x, int idx) {
// equivalent to get_per_layer_inputs() in python code
// output shape: [n_embd_altup, n_layer, n_tokens]
ggml_tensor * llm_build_gemma3n_iswa::get_per_layer_inputs() {
auto inp = std::make_unique<llm_graph_input_embd>();
auto inp = std::make_unique<llm_graph_input_embd>(n_embd);
ggml_tensor * inp_per_layer;
if (ubatch.token) {
inp->tokens = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, ubatch.n_tokens);
Expand Down
Loading