Skip to content

avoid crash when using taesd for preview only #1141

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
leejet merged 1 commit into from
Dec 24, 2025
Merged

avoid crash when using taesd for preview only #1141

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
6 changes: 3 additions & 3 deletions examples/server/main.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -105,9 +105,9 @@ struct SDSvrParams {
std::string listen_ip = "127.0.0.1";
int listen_port = 1234;
std::string serve_html_path;
bool normal_exit = false;
bool verbose = false;
bool color = false;
bool normal_exit = false;
bool verbose = false;
bool color = false;

ArgOptions get_options() {
ArgOptions options;
Expand Down
2 changes: 1 addition & 1 deletion qwen_image.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -648,7 +648,7 @@ namespace Qwen {
modulate_index_vec.insert(modulate_index_vec.end(), num_ref_img_tokens, 1.f);
}

modulate_index = ggml_new_tensor_1d(compute_ctx, GGML_TYPE_F32, modulate_index_vec.size());
modulate_index = ggml_new_tensor_1d(compute_ctx, GGML_TYPE_F32, modulate_index_vec.size());
set_backend_tensor_data(modulate_index, modulate_index_vec.data());
}

Expand Down
79 changes: 39 additions & 40 deletions stable-diffusion.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -591,8 +591,8 @@ class StableDiffusionGGML {
vae_backend = backend;
}

if (sd_version_is_wan(version) || sd_version_is_qwen_image(version)) {
if (!use_tiny_autoencoder) {
if (!use_tiny_autoencoder || sd_ctx_params->tae_preview_only) {
if (sd_version_is_wan(version) || sd_version_is_qwen_image(version)) {
first_stage_model = std::make_shared<WAN::WanVAERunner>(vae_backend,
offload_params_to_cpu,
tensor_storage_map,
Expand All @@ -601,57 +601,56 @@ class StableDiffusionGGML {
version);
first_stage_model->alloc_params_buffer();
first_stage_model->get_param_tensors(tensors, "first_stage_model");
} else if (version == VERSION_CHROMA_RADIANCE) {
first_stage_model = std::make_shared<FakeVAE>(vae_backend,
offload_params_to_cpu);
} else {
first_stage_model = std::make_shared<AutoEncoderKL>(vae_backend,
offload_params_to_cpu,
tensor_storage_map,
"first_stage_model",
vae_decode_only,
false,
version);
if (sd_ctx_params->vae_conv_direct) {
LOG_INFO("Using Conv2d direct in the vae model");
first_stage_model->set_conv2d_direct_enabled(true);
}
if (version == VERSION_SDXL &&
(strlen(SAFE_STR(sd_ctx_params->vae_path)) == 0 || sd_ctx_params->force_sdxl_vae_conv_scale)) {
float vae_conv_2d_scale = 1.f / 32.f;
LOG_WARN(
"No VAE specified with --vae or --force-sdxl-vae-conv-scale flag set, "
"using Conv2D scale %.3f",
vae_conv_2d_scale);
first_stage_model->set_conv2d_scale(vae_conv_2d_scale);
}
first_stage_model->alloc_params_buffer();
first_stage_model->get_param_tensors(tensors, "first_stage_model");
}
}

if (use_tiny_autoencoder) {
if (sd_version_is_wan(version) || sd_version_is_qwen_image(version)) {
tae_first_stage = std::make_shared<TinyVideoAutoEncoder>(vae_backend,
offload_params_to_cpu,
tensor_storage_map,
"decoder",
vae_decode_only,
version);
if (sd_ctx_params->vae_conv_direct) {
LOG_INFO("Using Conv2d direct in the tae model");
tae_first_stage->set_conv2d_direct_enabled(true);
}
}
} else if (version == VERSION_CHROMA_RADIANCE) {
first_stage_model = std::make_shared<FakeVAE>(vae_backend,
offload_params_to_cpu);
} else if (!use_tiny_autoencoder || sd_ctx_params->tae_preview_only) {
first_stage_model = std::make_shared<AutoEncoderKL>(vae_backend,
offload_params_to_cpu,
tensor_storage_map,
"first_stage_model",
vae_decode_only,
false,
version);
if (sd_ctx_params->vae_conv_direct) {
LOG_INFO("Using Conv2d direct in the vae model");
first_stage_model->set_conv2d_direct_enabled(true);
}
if (version == VERSION_SDXL &&
(strlen(SAFE_STR(sd_ctx_params->vae_path)) == 0 || sd_ctx_params->force_sdxl_vae_conv_scale)) {
float vae_conv_2d_scale = 1.f / 32.f;
LOG_WARN(
"No VAE specified with --vae or --force-sdxl-vae-conv-scale flag set, "
"using Conv2D scale %.3f",
vae_conv_2d_scale);
first_stage_model->set_conv2d_scale(vae_conv_2d_scale);
} else {
tae_first_stage = std::make_shared<TinyImageAutoEncoder>(vae_backend,
offload_params_to_cpu,
tensor_storage_map,
"decoder.layers",
vae_decode_only,
version);
}
first_stage_model->alloc_params_buffer();
first_stage_model->get_param_tensors(tensors, "first_stage_model");
} else if (use_tiny_autoencoder) {
tae_first_stage = std::make_shared<TinyImageAutoEncoder>(vae_backend,
offload_params_to_cpu,
tensor_storage_map,
"decoder.layers",
vae_decode_only,
version);
if (sd_ctx_params->vae_conv_direct) {
LOG_INFO("Using Conv2d direct in the tae model");
tae_first_stage->set_conv2d_direct_enabled(true);
}
}
// first_stage_model->get_param_tensors(tensors, "first_stage_model.");

if (strlen(SAFE_STR(sd_ctx_params->control_net_path)) > 0) {
ggml_backend_t controlnet_backend = nullptr;
Expand Down
Loading