feat: load scaled FP8 weights without upfront conversion (#1913)

This commit is contained in:
leejet
2026-08-27 20:50:47 +08:00
committed by GitHub
parent 50d6405683
commit be0e34480d
9 changed files with 73 additions and 101 deletions
+1 -1
View File
@@ -142,7 +142,7 @@ namespace Ideogram4 {
__STATIC_INLINE__ std::shared_ptr<Linear> make_linear(int64_t in_features,
int64_t out_features,
bool bias = true) {
return std::make_shared<Linear>(in_features, out_features, bias, false, false, 1.f, true);
return std::make_shared<Linear>(in_features, out_features, bias);
}
__STATIC_INLINE__ std::vector<float> gen_ideogram4_pe(int grid_h,