DarthAffe/stable-diffusion.cpp
1
0 Fork 0
mirror of https://github.com/leejet/stable-diffusion.cpp.git synced 2026-11-11 19:36:33 +00:00
Code Issues Packages Projects Releases Wiki Activity

refactor: unify ggml operation helpers around runner context (#2314)

Browse Source
This commit is contained in:
leejet authored and GitHub committed 2026-11-10 01:58:04 +08:00
1 parent 73fef08224
commit a3c49b13be
76 files changed
+1983 -1831
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

src/model/diffusion/ideogram4.hpp
+17 -12
View File
@@ -73,12 +73,13 @@ namespace Ideogram4 {
}
};
__STATIC_INLINE__ ggml_tensor* timestep_embedding_sin_cos(ggml_context* ctx,
__STATIC_INLINE__ ggml_tensor* timestep_embedding_sin_cos(GGMLRunnerContext* runner_ctx,
ggml_tensor* timesteps,
int dim) {
auto ctx = runner_ctx->ggml_ctx;
GGML_ASSERT(dim % 2 == 0);
auto embedding = ggml_ext_timestep_embedding(ctx, timesteps, dim, TIMESTEP_MAX_PERIOD, 10.f);
auto chunks = ggml_ext_chunk(ctx, embedding, 2, 0);
auto embedding = ggml_ext_timestep_embedding(runner_ctx, timesteps, dim, TIMESTEP_MAX_PERIOD, 10.f);
auto chunks = ggml_ext_chunk(runner_ctx, embedding, 2, 0);
return ggml_concat(ctx, chunks[1], chunks[0], 0);
}
@@ -108,7 +109,10 @@ namespace Ideogram4 {
return ggml_ext_modulate(ctx, x, nullptr, scale);
}
__STATIC_INLINE__ ggml_tensor* patchify(ggml_context* ctx, ggml_tensor* x, const Ideogram4Config& config) {
__STATIC_INLINE__ ggml_tensor* patchify(GGMLRunnerContext* runner_ctx,
ggml_tensor* x,
const Ideogram4Config& config) {
auto ctx = runner_ctx->ggml_ctx;
// x: [N, 128, H, W] with channel order [ae, ph, pw].
// return: [N, H*W, 128] with token channel order [ph, pw, ae].
const int64_t W = x->ne[0];
@@ -126,11 +130,12 @@ namespace Ideogram4 {
return x;
}
__STATIC_INLINE__ ggml_tensor* unpatchify(ggml_context* ctx,
__STATIC_INLINE__ ggml_tensor* unpatchify(GGMLRunnerContext* runner_ctx,
ggml_tensor* x,
int64_t H,
int64_t W,
const Ideogram4Config& config) {
auto ctx = runner_ctx->ggml_ctx;
const int64_t C = x->ne[0];
const int64_t N = x->ne[2];
@@ -217,7 +222,7 @@ namespace Ideogram4 {
auto out_proj = std::dynamic_pointer_cast<Linear>(blocks["o"]);
auto qkv = qkv_proj->forward(ctx, x);
auto qkv_vec = split_qkv(ctx->ggml_ctx, qkv);
auto qkv_vec = ggml_ext_split_qkv(ctx, qkv);
auto q = ggml_reshape_4d(ctx->ggml_ctx, qkv_vec[0], head_dim, num_heads, n_token, N);
auto k = ggml_reshape_4d(ctx->ggml_ctx, qkv_vec[1], head_dim, num_heads, n_token, N);
auto v = ggml_reshape_4d(ctx->ggml_ctx, qkv_vec[2], head_dim, num_heads, n_token, N);
@@ -278,7 +283,7 @@ namespace Ideogram4 {
auto adaln_modulation = std::dynamic_pointer_cast<Linear>(blocks["adaln_modulation"]);
auto mod = adaln_modulation->forward(ctx, adaln_input);
auto mods = ggml_ext_chunk(ctx->ggml_ctx, mod, 4, 0);
auto mods = ggml_ext_chunk(ctx, mod, 4, 0);
auto scale_msa = mods[0];
auto gate_msa = to_token_modulation(ctx->ggml_ctx, ggml_tanh(ctx->ggml_ctx, mods[1]));
auto scale_mlp = mods[2];
@@ -315,7 +320,7 @@ namespace Ideogram4 {
auto mlp_in = std::dynamic_pointer_cast<Linear>(blocks["mlp_in"]);
auto mlp_out = std::dynamic_pointer_cast<Linear>(blocks["mlp_out"]);
x = timestep_embedding_sin_cos(ctx->ggml_ctx, x, static_cast<int>(dim));
x = timestep_embedding_sin_cos(ctx, x, static_cast<int>(dim));
x = ggml_silu(ctx->ggml_ctx, mlp_in->forward(ctx, x));
x = mlp_out->forward(ctx, x);
return x;
@@ -383,7 +388,7 @@ namespace Ideogram4 {
auto embed_image_indicator = std::dynamic_pointer_cast<Embedding>(blocks["embed_image_indicator"]);
auto final_layer = std::dynamic_pointer_cast<Ideogram4FinalLayer>(blocks["final_layer"]);
auto img = patchify(ctx->ggml_ctx, x, config);
auto img = patchify(ctx, x, config);
img = input_proj->forward(ctx, img);
ggml_tensor* h = img;
@@ -413,11 +418,11 @@ namespace Ideogram4 {
h = final_layer->forward(ctx, h, adaln_input);
if (context_len > 0) {
h = ggml_ext_slice(ctx->ggml_ctx, h, 1, context_len, h->ne[1]);
h = ggml_ext_slice(ctx, h, 1, context_len, h->ne[1]);
}
h = unpatchify(ctx->ggml_ctx, h, H, W, config);
h = ggml_ext_scale(ctx->ggml_ctx, h, -1.f);
h = unpatchify(ctx, h, H, W, config);
h = ggml_ext_scale(ctx, h, -1.f);
return h;
}
};