theluyuan/DiffSynth-Studio
1
0
Fork 0
mirror of https://github.com/modelscope/DiffSynth-Studio.git synced 2026-03-18 22:08:13 +00:00
Code Issues Packages Projects Releases Wiki Activity

bugfix

Browse Source
This commit is contained in:
Artiprocher
2025-12-20 14:00:22 +08:00
parent c6722b3f56
commit 20e1aaf908
4 changed files with 10 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
diffsynth/models/general_modules.py
View File

@@ -87,6 +87,7 @@ class TimestepEmbeddings(torch.nn.Module):
self.timestep_embedder = torch.nn.Sequential( self.timestep_embedder = torch.nn.Sequential(
torch.nn.Linear(dim_in, dim_out), torch.nn.SiLU(), torch.nn.Linear(dim_out, dim_out) torch.nn.Linear(dim_in, dim_out), torch.nn.SiLU(), torch.nn.Linear(dim_out, dim_out)
) )
self.use_additional_t_cond = use_additional_t_cond
if use_additional_t_cond: if use_additional_t_cond:
self.addition_t_embedding = torch.nn.Embedding(2, dim_out) self.addition_t_embedding = torch.nn.Embedding(2, dim_out)

2
diffsynth/pipelines/qwen_image.py
View File

@@ -762,7 +762,7 @@ def model_fn_qwen_image(
conditioning = dit.time_text_embed( conditioning = dit.time_text_embed(
timestep, timestep,
image.dtype, image.dtype,
addition_t_cond=None if layer_num is None else torch.tensor([0]).to(device=image.device, dtype=torch.long) addition_t_cond=None if not dit.time_text_embed.use_additional_t_cond else torch.tensor([0]).to(device=image.device, dtype=torch.long)
) )
if entity_prompt_emb is not None: if entity_prompt_emb is not None:

6
examples/qwen_image/model_training/validate_full/Qwen-Image-Layered.py
View File

@@ -18,9 +18,11 @@ state_dict = load_state_dict("models/train/Qwen-Image-Layered_full/epoch-1.safet
pipe.dit.load_state_dict(state_dict) pipe.dit.load_state_dict(state_dict)
prompt = "a poster" prompt = "a poster"
input_image = Image.open("data/example_image_dataset/layer/image.png").convert("RGBA").resize((864, 480)) input_image = Image.open("data/example_image_dataset/layer/image.png").convert("RGBA").resize((864, 480))
image = pipe( images = pipe(
prompt, seed=0, prompt, seed=0,
height=480, width=864, height=480, width=864,
layer_input_image=input_image, layer_num=3, layer_input_image=input_image, layer_num=3,
) )
image.save("image.jpg") for i, image in enumerate(images):
if i == 0: continue # The first image is the input image.
image.save(f"image_{i}.png")

6
examples/qwen_image/model_training/validate_lora/Qwen-Image-Layered.py
View File

@@ -17,9 +17,11 @@ pipe = QwenImagePipeline.from_pretrained(
pipe.load_lora(pipe.dit, "models/train/Qwen-Image-Layered_lora/epoch-4.safetensors") pipe.load_lora(pipe.dit, "models/train/Qwen-Image-Layered_lora/epoch-4.safetensors")
prompt = "a poster" prompt = "a poster"
input_image = Image.open("data/example_image_dataset/layer/image.png").convert("RGBA").resize((864, 480)) input_image = Image.open("data/example_image_dataset/layer/image.png").convert("RGBA").resize((864, 480))
image = pipe( images = pipe(
prompt, seed=0, prompt, seed=0,
height=480, width=864, height=480, width=864,
layer_input_image=input_image, layer_num=3, layer_input_image=input_image, layer_num=3,
) )
image.save("image.jpg") for i, image in enumerate(images):
if i == 0: continue # The first image is the input image.
image.save(f"image_{i}.png")