mirrors/ComfyUI
1
0
Fork 0
mirror of https://github.com/comfyanonymous/ComfyUI.git synced 2025-09-10 19:46:38 +00:00
Code Issues Packages Projects Releases Wiki Activity

Basic hunyuan dit implementation. (#4102)

Browse Source 
* Let tokenizers return weights to be stored in the saved checkpoint.

* Basic hunyuan dit implementation.

* Fix some resolutions not working.

* Support hydit checkpoint save.

* Init with right dtype.

* Switch to optimized attention in pooler.

* Fix black images on hunyuan dit.
This commit is contained in:
comfyanonymous
2024-07-25 18:21:08 -04:00
committed by GitHub
parent f87810cd3e
commit a5f4292f9f
15 changed files with 48196 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

42
comfy/supported_models.py
View File

@@ -8,6 +8,7 @@ from . import sdxl_clip
import comfy.text_encoders.sd3_clip
import comfy.text_encoders.sa_t5
import comfy.text_encoders.aura_t5
import comfy.text_encoders.hydit
from . import supported_models_base
from . import latent_formats
@@ -580,6 +581,45 @@ class AuraFlow(supported_models_base.BASE):
def clip_target(self, state_dict={}):
return supported_models_base.ClipTarget(comfy.text_encoders.aura_t5.AuraT5Tokenizer, comfy.text_encoders.aura_t5.AuraT5Model)
models = [Stable_Zero123, SD15_instructpix2pix, SD15, SD20, SD21UnclipL, SD21UnclipH, SDXL_instructpix2pix, SDXLRefiner, SDXL, SSD1B, KOALA_700M, KOALA_1B, Segmind_Vega, SD_X4Upscaler, Stable_Cascade_C, Stable_Cascade_B, SV3D_u, SV3D_p, SD3, StableAudio, AuraFlow]
class HunyuanDiT(supported_models_base.BASE):
unet_config = {
"image_model": "hydit",
}
unet_extra_config = {
"attn_precision": torch.float32,
}
sampling_settings = {
"linear_start": 0.00085,
"linear_end": 0.018,
}
latent_format = latent_formats.SDXL
vae_key_prefix = ["vae."]
text_encoder_key_prefix = ["text_encoders."]
def get_model(self, state_dict, prefix="", device=None):
out = model_base.HunyuanDiT(self, device=device)
return out
def clip_target(self, state_dict={}):
return supported_models_base.ClipTarget(comfy.text_encoders.hydit.HyditTokenizer, comfy.text_encoders.hydit.HyditModel)
class HunyuanDiT1(HunyuanDiT):
unet_config = {
"image_model": "hydit1",
}
unet_extra_config = {}
sampling_settings = {
"linear_start" : 0.00085,
"linear_end" : 0.03,
}
models = [Stable_Zero123, SD15_instructpix2pix, SD15, SD20, SD21UnclipL, SD21UnclipH, SDXL_instructpix2pix, SDXLRefiner, SDXL, SSD1B, KOALA_700M, KOALA_1B, Segmind_Vega, SD_X4Upscaler, Stable_Cascade_C, Stable_Cascade_B, SV3D_u, SV3D_p, SD3, StableAudio, AuraFlow, HunyuanDiT, HunyuanDiT1]
models += [SVD_img2vid]