diff options
Diffstat (limited to 'modules/sd_hijack.py')
| -rw-r--r-- | modules/sd_hijack.py | 36 |
1 files changed, 27 insertions, 9 deletions
diff --git a/modules/sd_hijack.py b/modules/sd_hijack.py index c8fdd4f1..592f0055 100644 --- a/modules/sd_hijack.py +++ b/modules/sd_hijack.py @@ -2,7 +2,6 @@ import torch from torch.nn.functional import silu
from types import MethodType
-import modules.textual_inversion.textual_inversion
from modules import devices, sd_hijack_optimizations, shared, script_callbacks, errors, sd_unet
from modules.hypernetworks import hypernetwork
from modules.shared import cmd_opts
@@ -30,8 +29,10 @@ ldm.modules.attention.MemoryEfficientCrossAttention = ldm.modules.attention.Cros ldm.modules.attention.BasicTransformerBlock.ATTENTION_MODES["softmax-xformers"] = ldm.modules.attention.CrossAttention
# silence new console spam from SD2
-ldm.modules.attention.print = lambda *args: None
-ldm.modules.diffusionmodules.model.print = lambda *args: None
+ldm.modules.attention.print = shared.ldm_print
+ldm.modules.diffusionmodules.model.print = shared.ldm_print
+ldm.util.print = shared.ldm_print
+ldm.models.diffusion.ddpm.print = shared.ldm_print
optimizers = []
current_optimizer: sd_hijack_optimizations.SdOptimization = None
@@ -164,12 +165,13 @@ class StableDiffusionModelHijack: clip = None
optimization_method = None
- embedding_db = modules.textual_inversion.textual_inversion.EmbeddingDatabase()
-
def __init__(self):
+ import modules.textual_inversion.textual_inversion
+
self.extra_generation_params = {}
self.comments = []
+ self.embedding_db = modules.textual_inversion.textual_inversion.EmbeddingDatabase()
self.embedding_db.add_embedding_dir(cmd_opts.embeddings_dir)
def apply_optimizations(self, option=None):
@@ -197,7 +199,7 @@ class StableDiffusionModelHijack: conditioner.embedders[i] = sd_hijack_clip.FrozenCLIPEmbedderForSDXLWithCustomWords(embedder, self)
text_cond_models.append(conditioner.embedders[i])
if typename == 'FrozenOpenCLIPEmbedder2':
- embedder.model.token_embedding = EmbeddingsWithFixes(embedder.model.token_embedding, self)
+ embedder.model.token_embedding = EmbeddingsWithFixes(embedder.model.token_embedding, self, textual_inversion_key='clip_g')
conditioner.embedders[i] = sd_hijack_open_clip.FrozenOpenCLIPEmbedder2WithCustomWords(embedder, self)
text_cond_models.append(conditioner.embedders[i])
@@ -243,7 +245,21 @@ class StableDiffusionModelHijack: ldm.modules.diffusionmodules.openaimodel.UNetModel.forward = sd_unet.UNetModel_forward
def undo_hijack(self, m):
- if type(m.cond_stage_model) == sd_hijack_xlmr.FrozenXLMREmbedderWithCustomWords:
+ conditioner = getattr(m, 'conditioner', None)
+ if conditioner:
+ for i in range(len(conditioner.embedders)):
+ embedder = conditioner.embedders[i]
+ if isinstance(embedder, (sd_hijack_open_clip.FrozenOpenCLIPEmbedderWithCustomWords, sd_hijack_open_clip.FrozenOpenCLIPEmbedder2WithCustomWords)):
+ embedder.wrapped.model.token_embedding = embedder.wrapped.model.token_embedding.wrapped
+ conditioner.embedders[i] = embedder.wrapped
+ if isinstance(embedder, sd_hijack_clip.FrozenCLIPEmbedderForSDXLWithCustomWords):
+ embedder.wrapped.transformer.text_model.embeddings.token_embedding = embedder.wrapped.transformer.text_model.embeddings.token_embedding.wrapped
+ conditioner.embedders[i] = embedder.wrapped
+
+ if hasattr(m, 'cond_stage_model'):
+ delattr(m, 'cond_stage_model')
+
+ elif type(m.cond_stage_model) == sd_hijack_xlmr.FrozenXLMREmbedderWithCustomWords:
m.cond_stage_model = m.cond_stage_model.wrapped
elif type(m.cond_stage_model) == sd_hijack_clip.FrozenCLIPEmbedderWithCustomWords:
@@ -292,10 +308,11 @@ class StableDiffusionModelHijack: class EmbeddingsWithFixes(torch.nn.Module):
- def __init__(self, wrapped, embeddings):
+ def __init__(self, wrapped, embeddings, textual_inversion_key='clip_l'):
super().__init__()
self.wrapped = wrapped
self.embeddings = embeddings
+ self.textual_inversion_key = textual_inversion_key
def forward(self, input_ids):
batch_fixes = self.embeddings.fixes
@@ -309,7 +326,8 @@ class EmbeddingsWithFixes(torch.nn.Module): vecs = []
for fixes, tensor in zip(batch_fixes, inputs_embeds):
for offset, embedding in fixes:
- emb = devices.cond_cast_unet(embedding.vec)
+ vec = embedding.vec[self.textual_inversion_key] if isinstance(embedding.vec, dict) else embedding.vec
+ emb = devices.cond_cast_unet(vec)
emb_len = min(tensor.shape[0] - offset - 1, emb.shape[0])
tensor = torch.cat([tensor[0:offset + 1], emb[0:emb_len], tensor[offset + 1 + emb_len:]])
|