Commite248568

committed

cleanup: do not force VAE type to f32 on SDXL

This seems to be a leftover from the initial SDXL support: it'snot enough to avoid NaN issues, and it's not not needed for thefixed sdxl-vae-fp16-fix .

1 parent7a8ff2e commite248568Copy full SHA for e248568

File tree

1 file changed

-6

lines changed

stable-diffusion.cpp

1 file changed

-6

lines changed

`‎stable-diffusion.cpp‎`

Lines changed: 1 addition & 6 deletions

Original file line number	Diff line number	Diff line change
`@@ -282,11 +282,6 @@ class StableDiffusionGGML {`
`282`	`282`	`model_loader.set_wtype_override(wtype);`
`283`	`283`	`}`
`284`	`284`
`285`		`-if (sd_version_is_sdxl(version)) {`
`286`		`- vae_wtype = GGML_TYPE_F32;`
`287`		`- model_loader.set_wtype_override(GGML_TYPE_F32,"vae.");`
`288`		`- }`
`289`		`-`
`290`	`285`	`LOG_INFO("Weight type: %s", model_wtype != GGML_TYPE_COUNT ?ggml_type_name(model_wtype) :"??");`
`291`	`286`	`LOG_INFO("Conditioner weight type: %s", conditioner_wtype != GGML_TYPE_COUNT ?ggml_type_name(conditioner_wtype) :"??");`
`292`	`287`	`LOG_INFO("Diffusion model weight type: %s", diffusion_model_wtype != GGML_TYPE_COUNT ?ggml_type_name(diffusion_model_wtype) :"??");`
`@@ -2110,4 +2105,4 @@ sd_image_t* edit(sd_ctx_t* sd_ctx,`
`2110`	`2105`	`LOG_INFO("edit completed in %.2fs", (t2 - t0) *1.0f /1000);`
`2111`	`2106`
`2112`	`2107`	`return result_images;`
`2113`		`-}`
	`2108`	`+}`

0 commit comments

Comments

(0)

Movatterモバイル変換

Navigation Menu

Search code, repositories, users, issues, pull requests...

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Commite248568

File tree

1 file changed

1 file changed

`‎stable-diffusion.cpp‎`

0 commit comments