theluyuan/DiffSynth-Studio
1
0
Fork 0
mirror of https://github.com/modelscope/DiffSynth-Studio.git synced 2026-03-23 17:38:10 +00:00
Code Issues Packages Projects Releases Wiki Activity

Merge pull request #260 from mi804/sd3.5

Browse Source 
update default t5_sequence_length to 77
This commit is contained in:
Zhongjie Duan
2024-11-11 16:39:31 +08:00
committed by GitHub
parent 39ddb7c3e3 4378b540cf
commit d6f4fb67cc
2 changed files with 10 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
diffsynth/pipelines/sd3_image.py
View File

@@ -59,9 +59,9 @@ class SD3ImagePipeline(BasePipeline):
return image return image
def encode_prompt(self, prompt, positive=True): def encode_prompt(self, prompt, positive=True, t5_sequence_length=77):
prompt_emb, pooled_prompt_emb = self.prompter.encode_prompt( prompt_emb, pooled_prompt_emb = self.prompter.encode_prompt(
prompt, device=self.device, positive=positive prompt, device=self.device, positive=positive, t5_sequence_length=t5_sequence_length
) )
return {"prompt_emb": prompt_emb, "pooled_prompt_emb": pooled_prompt_emb} return {"prompt_emb": prompt_emb, "pooled_prompt_emb": pooled_prompt_emb}
@@ -84,6 +84,7 @@ class SD3ImagePipeline(BasePipeline):
height=1024, height=1024,
width=1024, width=1024,
num_inference_steps=20, num_inference_steps=20,
t5_sequence_length=77,
tiled=False, tiled=False,
tile_size=128, tile_size=128,
tile_stride=64, tile_stride=64,
@@ -109,9 +110,9 @@ class SD3ImagePipeline(BasePipeline):
# Encode prompts # Encode prompts
self.load_models_to_device(['text_encoder_1', 'text_encoder_2', 'text_encoder_3']) self.load_models_to_device(['text_encoder_1', 'text_encoder_2', 'text_encoder_3'])
prompt_emb_posi = self.encode_prompt(prompt, positive=True) prompt_emb_posi = self.encode_prompt(prompt, positive=True, t5_sequence_length=t5_sequence_length)
prompt_emb_nega = self.encode_prompt(negative_prompt, positive=False) prompt_emb_nega = self.encode_prompt(negative_prompt, positive=False, t5_sequence_length=t5_sequence_length)
prompt_emb_locals = [self.encode_prompt(prompt_local) for prompt_local in local_prompts] prompt_emb_locals = [self.encode_prompt(prompt_local, t5_sequence_length=t5_sequence_length) for prompt_local in local_prompts]
# Denoise # Denoise
self.load_models_to_device(['dit']) self.load_models_to_device(['dit'])

7
diffsynth/prompters/sd3_prompter.py
View File

@@ -67,7 +67,8 @@ class SD3Prompter(BasePrompter):
self, self,
prompt, prompt,
positive=True, positive=True,
device="cuda" device="cuda",
t5_sequence_length=77,
): ):
prompt = self.process_prompt(prompt, positive=positive) prompt = self.process_prompt(prompt, positive=positive)
@@ -77,9 +78,9 @@ class SD3Prompter(BasePrompter):
# T5 # T5
if self.text_encoder_3 is None: if self.text_encoder_3 is None:
prompt_emb_3 = torch.zeros((prompt_emb_1.shape[0], 256, 4096), dtype=prompt_emb_1.dtype, device=device) prompt_emb_3 = torch.zeros((prompt_emb_1.shape[0], t5_sequence_length, 4096), dtype=prompt_emb_1.dtype, device=device)
else: else:
prompt_emb_3 = self.encode_prompt_using_t5(prompt, self.text_encoder_3, self.tokenizer_3, 256, device) prompt_emb_3 = self.encode_prompt_using_t5(prompt, self.text_encoder_3, self.tokenizer_3, t5_sequence_length, device)
prompt_emb_3 = prompt_emb_3.to(prompt_emb_1.dtype) # float32 -> float16 prompt_emb_3 = prompt_emb_3.to(prompt_emb_1.dtype) # float32 -> float16
# Merge # Merge