DarthAffe/stable-diffusion.cpp
1
0
Fork 0
mirror of https://github.com/leejet/stable-diffusion.cpp.git synced 2025-12-13 05:48:56 +00:00
Code Issues Packages Projects Releases Wiki Activity

to_add_out precision fix

Browse Source
This commit is contained in:
leejet 2025-10-12 17:27:51 +08:00
parent cc064a0530
commit 7519e2f11a
1 changed files with 4 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
qwen_image.hpp
View File

@ -97,7 +97,10 @@ namespace Qwen {
blocks["to_out.0"] = std::shared_ptr<GGMLBlock>(new Linear(inner_dim, out_dim, out_bias)); blocks["to_out.0"] = std::shared_ptr<GGMLBlock>(new Linear(inner_dim, out_dim, out_bias));
// to_out.1 is nn.Dropout // to_out.1 is nn.Dropout
blocks["to_add_out"] = std::shared_ptr<GGMLBlock>(new Linear(inner_dim, out_context_dim, out_bias)); float scale = 1.f / 32.f;
// The purpose of the scale here is to prevent NaN issues in certain situations.
// For example when using CUDA but the weights are k-quants (not all prompts).
blocks["to_add_out"] = std::shared_ptr<GGMLBlock>(new Linear(inner_dim, out_context_dim, out_bias, false, false, scale));
} }
std::pair<ggml_tensor*, ggml_tensor*> forward(struct ggml_context* ctx, std::pair<ggml_tensor*, ggml_tensor*> forward(struct ggml_context* ctx,