Public/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-09-20 01:31:31 +02:00
Code Issues Packages Projects Releases Wiki Activity

mtmd: weave deepseek-ocr rows in one shot instead of per row (#26615)

Browse Source
This commit is contained in:
Saba Fallah
2026-08-05 13:07:53 +02:00
committed by GitHub
parent 6b0b1ea112
commit e77293d4cc
1 changed files with 19 additions and 48 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tools/mtmd/models/deepseekocr.cpp
+19 -48
View File
@@ -276,23 +276,11 @@ ggml_cgraph * clip_graph_deepseekocr::build() {
n_tiles_per_row = img.ny() / img.nx();
// each entry is one "row" image of shape [tile_size, tile_size * n_tiles_per_row, 3];
// reshape+concat all rows into one combined SAM input of shape
// merge the tile axis into the batch axis, giving a combined SAM input of shape
// [tile_size, tile_size, 3, n_tiles_per_row * n_rows_batch] (tile fast, row slow)
ggml_tensor * combined = nullptr;
for (int r = 0; r < n_rows_batch; r++) {
ggml_tensor * row = ggml_view_4d(ctx0, inp_raw, img.nx(), img.ny(), 3, 1,
inp_raw->nb[1], inp_raw->nb[2], inp_raw->nb[3],
r * inp_raw->nb[3]);
row = ggml_cont(ctx0, row);
// input shape: [tile_size, tile_size * n_tiles_per_row, 3, 1]
// we want to reshape it to [tile_size, tile_size, 3, n_tiles_per_row]
row = ggml_reshape_4d(ctx0, row, img.nx(), img.nx(), n_tiles_per_row, 3);
row = ggml_cont(ctx0, ggml_permute(ctx0, row, 0, 1, 3, 2));
combined = combined ? ggml_concat(ctx0, combined, row, 3) : row;
}
inp_raw = combined;
inp_raw = ggml_reshape_4d(ctx0, inp_raw, img.nx() * img.nx(), n_tiles_per_row, 3, n_rows_batch);
inp_raw = ggml_cont(ctx0, ggml_permute(ctx0, inp_raw, 0, 2, 1, 3));
inp_raw = ggml_reshape_4d(ctx0, inp_raw, img.nx(), img.nx(), 3, n_tiles_per_row * n_rows_batch);
}
ggml_tensor * sam_out = build_sam(inp_raw);
@@ -378,46 +366,29 @@ ggml_cgraph * clip_graph_deepseekocr::build() {
cur = ggml_concat(ctx0, cur, vs, 1); // (n_dim, h*(w+1) + 1, n_batch)
} else {
// tile row: interleave tiles within each row, add newline per row
// this weave is done per-row (cheap reshuffle ops) since a single row's
// interleave already uses all 4 ggml tensor axes, leaving no spare axis to
// also carry the n_rows_batch dimension through in one shot
const int grid_x = static_cast<int>(std::sqrt(static_cast<float>(clip_n_patches)));
const int grid_y = grid_x;
const auto n_dim = cur->ne[0];
// (n_dim, clip_n_patches, n_tiles_per_row * n_rows_batch) -> (n_dim, clip_n_patches, n_tiles_per_row, n_rows_batch)
ggml_tensor * cur4 = ggml_reshape_4d(ctx0, cur, n_dim, clip_n_patches, n_tiles_per_row, n_rows_batch);
// merge n_dim into the grid_x axis, freeing the 4th axis for n_rows_batch
// (n_dim, clip_n_patches, n_tiles_per_row * n_rows_batch) -> (n_dim*grid_x, grid_y, n_tiles_per_row, n_rows_batch)
cur = ggml_reshape_4d(ctx0, cur, n_dim * grid_x, grid_y, n_tiles_per_row, n_rows_batch);
ggml_tensor * rows_out = nullptr;
for (int r = 0; r < n_rows_batch; r++) {
ggml_tensor * row = ggml_view_3d(ctx0, cur4, n_dim, clip_n_patches, n_tiles_per_row,
cur4->nb[1], cur4->nb[2], r * cur4->nb[3]);
row = ggml_cont(ctx0, row);
// tiles: re-order from A.row0 A.row1 B.row0 B.row1 ...
// to A.row0 B.row0 A.row1 B.row1 ...
// then add nl: A.row0 B.row0 [nl] A.row1 B.row1 [nl] ...
// interleave tiles: -> (n_dim*grid_x, n_tiles_per_row, grid_y, n_rows_batch)
cur = ggml_cont(ctx0, ggml_permute(ctx0, cur, 0, 2, 1, 3));
// (n_dim, clip_n_patches, n_tiles_per_row) -> (n_dim, grid_x, grid_y, n_tiles_per_row)
row = ggml_reshape_4d(ctx0, row, n_dim, grid_x, grid_y, n_tiles_per_row);
// merge: -> (n_dim, grid_x*n_tiles_per_row, grid_y, n_rows_batch)
cur = ggml_reshape_4d(ctx0, cur, n_dim, grid_x * n_tiles_per_row, grid_y, n_rows_batch);
// tiles: re-order from A.row0 A.row1 B.row0 B.row1 ...
// to A.row0 B.row0 A.row1 B.row1 ...
// then add nl: A.row0 B.row0 [nl] A.row1 B.row1 [nl] ...
// interleave tiles: (n_dim, grid_x, grid_y, n_tiles_per_row) -> (n_dim, grid_x, n_tiles_per_row, grid_y)
row = ggml_cont(ctx0, ggml_permute(ctx0, row, 0, 1, 3, 2));
// append newline per row: (n_dim, grid_x*n_tiles_per_row+1, grid_y, n_rows_batch)
ggml_tensor * imgnl = ggml_repeat_4d(ctx0, model.image_newline, n_dim, 1, grid_y, n_rows_batch);
cur = ggml_concat(ctx0, cur, imgnl, 1);
// merge: (n_dim, grid_x, n_tiles_per_row, grid_y) -> (n_dim, grid_x*n_tiles_per_row, grid_y, 1)
row = ggml_reshape_4d(ctx0, row, n_dim, grid_x * n_tiles_per_row, grid_y, 1);
// append newline per row: (n_dim, grid_x*n_tiles_per_row+1, grid_y, 1)
ggml_tensor * imgnl = ggml_repeat_4d(ctx0, model.image_newline, n_dim, 1, grid_y, 1);
row = ggml_concat(ctx0, row, imgnl, 1);
// flatten: (n_dim, (grid_x*n_tiles_per_row+1)*grid_y, 1)
row = ggml_reshape_3d(ctx0, row, n_dim, (grid_x * n_tiles_per_row + 1) * grid_y, 1);
rows_out = rows_out ? ggml_concat(ctx0, rows_out, row, 2) : row;
}
// (n_dim, (grid_x*n_tiles_per_row+1)*grid_y, n_rows_batch)
cur = rows_out;
// flatten: (n_dim, (grid_x*n_tiles_per_row+1)*grid_y, n_rows_batch)
cur = ggml_reshape_3d(ctx0, cur, n_dim, (grid_x * n_tiles_per_row + 1) * grid_y, n_rows_batch);
}
cb(cur, "dsocr_output", -1);