Skip to content

model: fix multimodal padding token for gemma3n/gemma4 #21625

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

Merged
ngxson merged 3 commits into from
Apr 9, 2026
Merged

model: fix multimodal padding token for gemma3n/gemma4 #21625

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
4a6366b
model: fix multimodal padding token for gemma3n/gemma4
ngxson Apr 8, 2026
de54699
Merge branch 'master' into xsn/scale_gemma4_3n_pad
ngxson Apr 8, 2026
3a6ee0c
nits
ngxson Apr 8, 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
12 changes: 7 additions & 5 deletions src/models/gemma3n-iswa.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -250,27 +250,29 @@ ggml_tensor * llm_build_gemma3n_iswa::calc_magnitude(ggml_tensor * x) {
ggml_tensor * llm_build_gemma3n_iswa::build_inp_per_layer() {
auto inp = std::make_unique<llm_graph_input_embd>(n_embd);
ggml_tensor * inp_per_layer;
float tok_embd_scale = sqrtf((float) n_embd_altup);
if (ubatch.token) {
inp->tokens = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, ubatch.n_tokens);
ggml_set_input(inp->tokens);
res->t_inp_tokens = inp->tokens;
inp_per_layer = ggml_get_rows(ctx0, model.per_layer_tok_embd, inp->tokens);
inp_per_layer = ggml_get_rows (ctx0, model.per_layer_tok_embd, inp->tokens);
inp_per_layer = ggml_reshape_3d(ctx0, inp_per_layer, n_embd_altup, n_layer, n_tokens);
inp_per_layer = ggml_scale(ctx0, inp_per_layer, sqrtf((float) n_embd_altup));
inp_per_layer = ggml_scale (ctx0, inp_per_layer, tok_embd_scale);
cb(inp_per_layer, "inp_per_layer_selected", -1);
res->add_input(std::move(inp));
} else {
// Vision embedding path: use padding token (ID=0) embedding
// Multimodal embedding path: use padding token (ID=0) embedding
// TODO: verify if this is the correct behavior in transformers implementation
const int64_t embd_size = model.per_layer_tok_embd->ne[0]; // n_embd_altup * n_layer

// Extract and dequantize padding token embedding (row 0)
ggml_tensor * padding = ggml_view_1d(ctx0, model.per_layer_tok_embd, embd_size, 0);
inp_per_layer = ggml_cast(ctx0, padding, GGML_TYPE_F32);
inp_per_layer = ggml_cast (ctx0, padding, GGML_TYPE_F32);
inp_per_layer = ggml_scale(ctx0, inp_per_layer, tok_embd_scale);

// Reshape to [n_embd_altup, n_layer, 1]
inp_per_layer = ggml_reshape_3d(ctx0, inp_per_layer, n_embd_altup, n_layer, 1);
cb(inp_per_layer, "inp_per_layer_vision", -1);
cb(inp_per_layer, "inp_per_layer_multimodal", -1);
}
return inp_per_layer;
}
Expand Down
10 changes: 6 additions & 4 deletions src/models/gemma4-iswa.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -265,29 +265,31 @@ ggml_tensor * llm_build_gemma4_iswa::build_inp_per_layer() {
auto inp = std::make_unique<llm_graph_input_embd>(n_embd);

ggml_tensor * inp_per_layer;
float tok_embd_scale = sqrtf((float) n_embd_per_layer);
if (ubatch.token) {
inp->tokens = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, ubatch.n_tokens);
ggml_set_input(inp->tokens);
res->t_inp_tokens = inp->tokens;

inp_per_layer = ggml_get_rows (ctx0, model.per_layer_tok_embd, inp->tokens);
inp_per_layer = ggml_reshape_3d(ctx0, inp_per_layer, n_embd_per_layer, n_layer, n_tokens);
inp_per_layer = ggml_scale (ctx0, inp_per_layer, sqrtf((float) n_embd_per_layer));
inp_per_layer = ggml_scale (ctx0, inp_per_layer, tok_embd_scale);
cb(inp_per_layer, "inp_per_layer_selected", -1);

res->add_input(std::move(inp));
} else {
// Vision embedding path: use padding token (ID=0) embedding
// Multimodal embedding path: use padding token (ID=0) embedding
// TODO: verify if this is the correct behavior in transformers implementation
const int64_t embd_size = model.per_layer_tok_embd->ne[0]; // n_embd_per_layer * n_layer

// Extract and dequantize padding token embedding (row 0)
ggml_tensor * padding = ggml_view_1d(ctx0, model.per_layer_tok_embd, embd_size, 0);
inp_per_layer = ggml_cast(ctx0, padding, GGML_TYPE_F32);
inp_per_layer = ggml_cast (ctx0, padding, GGML_TYPE_F32);
inp_per_layer = ggml_scale(ctx0, inp_per_layer, tok_embd_scale);

// Reshape to [n_embd_per_layer, n_layer, 1]
inp_per_layer = ggml_reshape_3d(ctx0, inp_per_layer, n_embd_per_layer, n_layer, 1);
cb(inp_per_layer, "inp_per_layer_vision", -1);
cb(inp_per_layer, "inp_per_layer_multimodal", -1);
}
return inp_per_layer;
}
Expand Down
Loading