Public/automatic
1
0
Fork 0
mirror of https://github.com/vladmandic/automatic synced 2026-09-19 17:24:32 +02:00
Code Issues Packages Projects Releases Wiki Activity

lint formatting

Browse Source 
Signed-off-by: Vladimir Mandic <mandic00@live.com>
This commit is contained in:
Vladimir Mandic
2025-07-05 17:13:31 -04:00
parent a5b77b8ee2
commit 8c676d0a61
130 changed files with 398 additions and 446 deletions
Download Patch File Download Diff File Expand all files Collapse all files
scripts/consistory/consistory_pipeline.py
+8 -8
View File
@@ -82,7 +82,7 @@ class ConsistoryExtendAttnSDXLPipeline(
clip_skip: Optional[int] = None,
callback_on_step_end: Optional[Callable[[int, int, Dict], None]] = None,
callback_on_step_end_tensor_inputs: List[str] = ["latents"],
attention_store_kwargs: Optional[Dict] = None,
extended_attn_kwargs: Optional[Dict] = None,
share_queries: bool = False,
@@ -422,8 +422,8 @@ class ConsistoryExtendAttnSDXLPipeline(
t,
encoder_hidden_states=prompt_embeds,
timestep_cond=timestep_cond,
cross_attention_kwargs={'query_store': query_store,
'perform_extend_attn': False,
cross_attention_kwargs={'query_store': query_store,
'perform_extend_attn': False,
'record_attention': False},
added_cond_kwargs=added_cond_kwargs,
return_dict=False,
@@ -436,9 +436,9 @@ class ConsistoryExtendAttnSDXLPipeline(
t,
encoder_hidden_states=prompt_embeds,
timestep_cond=timestep_cond,
cross_attention_kwargs={'query_store': query_store,
'perform_extend_attn': True,
'record_attention': True,
cross_attention_kwargs={'query_store': query_store,
'perform_extend_attn': True,
'record_attention': True,
'feature_injector': feature_injector,
'anchors_cache': anchors_cache},
added_cond_kwargs=added_cond_kwargs,
@@ -483,7 +483,7 @@ class ConsistoryExtendAttnSDXLPipeline(
if XLA_AVAILABLE:
# xm.mark_step()
pass
# Update attention store mask
self.attention_store.aggregate_last_steps_attention()
@@ -516,4 +516,4 @@ class ConsistoryExtendAttnSDXLPipeline(
if not return_dict:
return (image,)
return StableDiffusionXLPipelineOutput(images=image)
return StableDiffusionXLPipelineOutput(images=image)
scripts/consistory/consistory_unet_sdxl.py
+1 -1
View File
@@ -1153,7 +1153,7 @@ class ConsistorySDXLUNet2DConditionModel(ModelMixin, ConfigMixin, UNet2DConditio
upsample_size=upsample_size,
scale=lora_scale,
)
self.latent_store(sample.detach(), t=timestep, layer_index=i)
# 6. post-process
scripts/consistory/utils/ptp_utils.py
+5 -5
View File
@@ -13,9 +13,9 @@
# limitations under the License.
# MIT License
#
#
# Copyright (c) 2023 AttendAndExcite
#
#
# Permission is hereby granted, free of charge, to any person obtaining a copy
# of this software and associated documentation files (the "Software"), to deal
# in the Software without restriction, including without limitation the rights
@@ -175,9 +175,9 @@ class AttentionStore:
attn_mask = self.last_mask_dropout[width]
if attn_mask is None:
return None
n_patches = width**2
output_attn_mask = torch.zeros((attn_mask.shape[0] * attn_mask.shape[1],), device=attn_mask.device, dtype=torch.bool)
for j in range(attn_mask.shape[0]):
@@ -191,4 +191,4 @@ class AttentionStore:
raise NotImplementedError('mask_background_query is not supported anymore')
output_attn_mask[0, attn_mask[i], k*n_patches:(k+1)*n_patches] = attn_mask[j].unsqueeze(0).expand(attn_mask[i].sum(), -1)
return output_attn_mask
return output_attn_mask
scripts/instantir/aggregator.py
+2 -2
View File
@@ -68,7 +68,7 @@ class SFT(nn.Module):
self.add = nn.Conv2d(nhidden, norm_nc, kernel_size=ks, padding=pw)
def forward(self, hidden_states, mask=False):
c, h = hidden_states
mask = mask or self.mask
assert mask is False
@@ -750,7 +750,7 @@ class Aggregator(ModelMixin, ConfigMixin, FromOriginalModelMixin):
image_embeds = self.encoder_hid_proj(image_embeds)
encoder_hidden_states = (encoder_hidden_states, image_embeds)
return encoder_hidden_states
def _set_gradient_checkpointing(self, module, value: bool = False) -> None:
if isinstance(module, (CrossAttnDownBlock2D, DownBlock2D)):
module.gradient_checkpointing = value
scripts/instantir/ip_adapter/attention_processor.py
+3 -3
View File
@@ -723,7 +723,7 @@ class AdditiveKV_AttnProcessor2_0(torch.nn.Module):
temb=None,
):
assert temb is not None, "Timestep embedding is needed for a time-aware attention processor."
residual = hidden_states
if attn.spatial_norm is not None:
@@ -834,7 +834,7 @@ class TA_AdditiveKV_AttnProcessor2_0(torch.nn.Module):
temb=None,
):
assert temb is not None, "Timestep embedding is needed for a time-aware attention processor."
residual = hidden_states
if attn.spatial_norm is not None:
@@ -1163,7 +1163,7 @@ class TA_IPAttnProcessor2_0(torch.nn.Module):
# for ip-adapter
ip_key = self.to_k_ip(ip_hidden_states)
ip_value = self.to_v_ip(ip_hidden_states)
# time-dependent adaLN
ip_key = self.ln_k_ip(ip_key, temb)
ip_value = self.ln_v_ip(ip_value, temb)
scripts/instantir/ip_adapter/ip_adapter.py
+2 -2
View File
@@ -47,14 +47,14 @@ class MLPProjModel(torch.nn.Module):
"""SD model with image prompt"""
def __init__(self, cross_attention_dim=2048, clip_embeddings_dim=1280):
super().__init__()
self.proj = torch.nn.Sequential(
torch.nn.Linear(clip_embeddings_dim, clip_embeddings_dim),
torch.nn.GELU(),
torch.nn.Linear(clip_embeddings_dim, cross_attention_dim),
torch.nn.LayerNorm(cross_attention_dim)
)
def forward(self, image_embeds):
clip_extra_context_tokens = self.proj(image_embeds)
return clip_extra_context_tokens
scripts/instantir/ip_adapter/utils.py
+1 -1
View File
@@ -245,4 +245,4 @@ def prepare_training_image_embeds(
)
image_embeds.append(single_image_embeds)
return image_embeds
return image_embeds
scripts/instantir/sdxl_instantir.py
+1 -1
View File
@@ -396,7 +396,7 @@ class InstantIRPipeline(
self.unet.disable_adapters()
return lora_alpha
# Copied from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline.encode_prompt
def encode_prompt(
self,
scripts/pulid/eva_clip/__init__.py
+1 -1
View File
@@ -8,4 +8,4 @@ from .openai import load_openai_model, list_openai_models
from .pretrained import list_pretrained, list_pretrained_models_by_tag, list_pretrained_tags_by_model,\
get_pretrained_url, download_pretrained_from_url, is_pretrained_cfg, get_pretrained_cfg, download_pretrained
from .tokenizer import SimpleTokenizer, tokenize
from .transform import image_transform
from .transform import image_transform
scripts/pulid/eva_clip/eva_vit_model.py
+24 -24
View File
@@ -11,7 +11,7 @@ try:
from timm.models.layers import drop_path, to_2tuple, trunc_normal_
except:
from timm.layers import drop_path, to_2tuple, trunc_normal_
from .transformer import PatchDropout
from .rope import VisionRotaryEmbedding, VisionRotaryEmbeddingFast
@@ -39,19 +39,19 @@ class DropPath(nn.Module):
def forward(self, x):
return drop_path(x, self.drop_prob, self.training)
def extra_repr(self) -> str:
return 'p={}'.format(self.drop_prob)
class Mlp(nn.Module):
def __init__(
self,
in_features,
hidden_features=None,
out_features=None,
act_layer=nn.GELU,
norm_layer=nn.LayerNorm,
self,
in_features,
hidden_features=None,
out_features=None,
act_layer=nn.GELU,
norm_layer=nn.LayerNorm,
drop=0.,
subln=False,
@@ -71,7 +71,7 @@ class Mlp(nn.Module):
x = self.fc1(x)
x = self.act(x)
# x = self.drop(x)
# commit this for the orignal BERT implement
# commit this for the orignal BERT implement
x = self.ffn_ln(x)
x = self.fc2(x)
@@ -79,7 +79,7 @@ class Mlp(nn.Module):
return x
class SwiGLU(nn.Module):
def __init__(self, in_features, hidden_features=None, out_features=None, act_layer=nn.SiLU, drop=0.,
def __init__(self, in_features, hidden_features=None, out_features=None, act_layer=nn.SiLU, drop=0.,
norm_layer=nn.LayerNorm, subln=False):
super().__init__()
out_features = out_features or in_features
@@ -91,7 +91,7 @@ class SwiGLU(nn.Module):
self.act = act_layer()
self.ffn_ln = norm_layer(hidden_features) if subln else nn.Identity()
self.w3 = nn.Linear(hidden_features, out_features)
self.drop = nn.Dropout(drop)
def forward(self, x):
@@ -172,20 +172,20 @@ class Attention(nn.Module):
def forward(self, x, rel_pos_bias=None, attn_mask=None):
B, N, C = x.shape
if self.subln:
if self.subln:
q = F.linear(input=x, weight=self.q_proj.weight, bias=self.q_bias)
k = F.linear(input=x, weight=self.k_proj.weight, bias=None)
v = F.linear(input=x, weight=self.v_proj.weight, bias=self.v_bias)
q = q.reshape(B, N, self.num_heads, -1).permute(0, 2, 1, 3) # B, num_heads, N, C
k = k.reshape(B, N, self.num_heads, -1).permute(0, 2, 1, 3)
v = v.reshape(B, N, self.num_heads, -1).permute(0, 2, 1, 3)
else:
k = k.reshape(B, N, self.num_heads, -1).permute(0, 2, 1, 3)
v = v.reshape(B, N, self.num_heads, -1).permute(0, 2, 1, 3)
else:
qkv_bias = None
if self.q_bias is not None:
qkv_bias = torch.cat((self.q_bias, torch.zeros_like(self.v_bias, requires_grad=False), self.v_bias))
qkv = F.linear(input=x, weight=self.qkv.weight, bias=qkv_bias)
qkv = qkv.reshape(B, N, 3, self.num_heads, -1).permute(2, 0, 3, 1, 4) # 3, B, num_heads, N, C
q, k, v = qkv[0], qkv[1], qkv[2]
@@ -232,7 +232,7 @@ class Attention(nn.Module):
if attn_mask is not None:
attn_mask = attn_mask.bool()
attn = attn.masked_fill(~attn_mask[:, None, None, :], float("-inf"))
attn = attn.softmax(dim=-1)
attn = self.attn_drop(attn)
@@ -262,15 +262,15 @@ class Block(nn.Module):
if naiveswiglu:
self.mlp = SwiGLU(
in_features=dim,
hidden_features=mlp_hidden_dim,
in_features=dim,
hidden_features=mlp_hidden_dim,
subln=subln,
norm_layer=norm_layer,
)
else:
self.mlp = Mlp(
in_features=dim,
hidden_features=mlp_hidden_dim,
in_features=dim,
hidden_features=mlp_hidden_dim,
act_layer=act_layer,
subln=subln,
drop=drop
@@ -407,7 +407,7 @@ class EVAVisionTransformer(nn.Module):
ft_seq_len=hw_seq_len if intp_freq else None,
# patch_dropout=patch_dropout
)
else:
else:
self.rope = None
self.naiveswiglu = naiveswiglu
@@ -469,7 +469,7 @@ class EVAVisionTransformer(nn.Module):
def get_num_layers(self):
return len(self.blocks)
def lock(self, unlocked_groups=0, freeze_bn_stats=False):
assert unlocked_groups == 0, 'partial locking not currently supported for this model'
for param in self.parameters():
@@ -491,7 +491,7 @@ class EVAVisionTransformer(nn.Module):
self.head = nn.Linear(self.embed_dim, num_classes) if num_classes > 0 else nn.Identity()
def forward_features(self, x, return_all_features=False, return_hidden=False, shuffle=False):
x = self.patch_embed(x)
batch_size, seq_len, _ = x.size()
scripts/pulid/eva_clip/factory.py
+8 -8
View File
@@ -93,7 +93,7 @@ def load_state_dict(checkpoint_path: str, map_location: str='cpu', model_key: st
state_dict = checkpoint
if next(iter(state_dict.items()))[0].startswith('module'):
state_dict = {k[7:]: v for k, v in state_dict.items()}
for k in skip_list:
if k in list(state_dict.keys()):
logging.info(f"Removing key {k} from pretrained checkpoint")
@@ -181,7 +181,7 @@ def load_pretrained_checkpoint(
visual_state_dict = load_clip_visual_state_dict(visual_checkpoint_path, is_openai=True, skip_list=skip_list)
else:
visual_state_dict = load_state_dict(visual_checkpoint_path, model_key=model_key, is_openai=False, skip_list=skip_list)
# resize_clip_pos_embed for CLIP and open CLIP
if 'positional_embedding' in visual_state_dict:
resize_visual_pos_embed(visual_state_dict, model)
@@ -202,7 +202,7 @@ def load_pretrained_checkpoint(
text_state_dict = load_state_dict(visual_checkpoint_path, model_key=model_key, is_openai=False, skip_list=skip_list)
text_incompatible_keys = model.text.load_state_dict(text_state_dict, strict=strict)
logging.info(f"num of loaded text_state_dict keys: {len(text_state_dict.keys())}")
logging.info(f"text_incompatible_keys.missing_keys: {text_incompatible_keys.missing_keys}")
@@ -255,7 +255,7 @@ def create_model(
if force_quick_gelu:
# override for use of QuickGELU on non-OpenAI transformer models
model_cfg["quick_gelu"] = True
if force_patch_dropout is not None:
# override the default patch dropout value
model_cfg['vision_cfg']["patch_dropout"] = force_patch_dropout
@@ -286,7 +286,7 @@ def create_model(
checkpoint_path,
model_key="model|module|state_dict",
strict=False
)
)
else:
error_str = (
f'Pretrained weights ({pretrained}) not found for model {model_name}.'
@@ -296,7 +296,7 @@ def create_model(
else:
visual_checkpoint_path = ''
text_checkpoint_path = ''
if pretrained_image:
pretrained_visual_model = pretrained_visual_model.replace('/', '-') # for callers using old naming with / in ViT names
pretrained_image_cfg = get_pretrained_cfg(pretrained_visual_model, pretrained_image)
@@ -321,7 +321,7 @@ def create_model(
else:
logging.warning(f'Pretrained weights ({text_checkpoint_path}) not found for model {model_name}.text.')
raise RuntimeError(f'Pretrained weights ({text_checkpoint_path}) not found for model {model_name}.text.')
if visual_checkpoint_path:
logging.info(f'Loading pretrained {model_name}.visual weights ({visual_checkpoint_path}).')
if text_checkpoint_path:
@@ -338,7 +338,7 @@ def create_model(
model_key="model|module|state_dict",
skip_list=skip_list
)
if "fp16" in precision or "bf16" in precision:
logging.info(f'convert precision to {precision}')
model = model.to(torch.bfloat16) if 'bf16' in precision else model.to(torch.float16)
scripts/pulid/eva_clip/hf_model.py
+14 -14
View File
@@ -62,19 +62,19 @@ class ClsPooler(nn.Module):
self.use_pooler_output = use_pooler_output
def forward(self, x:BaseModelOutput, attention_mask:TensorType):
if (self.use_pooler_output and
if (self.use_pooler_output and
isinstance(x, (BaseModelOutputWithPooling, BaseModelOutputWithPoolingAndCrossAttentions)) and
(x.pooler_output is not None)
):
return x.pooler_output
return x.last_hidden_state[:, self.cls_token_position, :]
class HFTextEncoder(nn.Module):
"""HuggingFace model adapter"""
def __init__(
self,
self,
model_name_or_path: str,
output_dim: int,
tokenizer_name: str = None,
@@ -134,10 +134,10 @@ class HFTextEncoder(nn.Module):
self.tokenizer = AutoTokenizer.from_pretrained(tokenizer_name)
# def forward_itm(self, x:TensorType, image_embeds:TensorType) -> TensorType:
# image_atts = torch.ones(image_embeds.size()[:-1],dtype=torch.long).to(x.device)
# image_atts = torch.ones(image_embeds.size()[:-1],dtype=torch.long).to(x.device)
# attn_mask = (x != self.config.pad_token_id).long()
# out = self.transformer(
# input_ids=x,
# input_ids=x,
# attention_mask=attn_mask,
# encoder_hidden_states = image_embeds,
# encoder_attention_mask = image_atts,
@@ -147,14 +147,14 @@ class HFTextEncoder(nn.Module):
# return self.itm_proj(pooled_out)
def mask(self, input_ids, vocab_size, device, targets=None, masked_indices=None, probability_matrix=None):
if masked_indices is None:
if masked_indices is None:
masked_indices = torch.bernoulli(probability_matrix).bool()
masked_indices[input_ids == self.tokenizer.pad_token_id] = False
masked_indices[input_ids == self.tokenizer.cls_token_id] = False
if targets is not None:
targets[~masked_indices] = -100 # We only compute loss on masked tokens
targets[~masked_indices] = -100 # We only compute loss on masked tokens
# 80% of the time, we replace masked input tokens with tokenizer.mask_token ([MASK])
indices_replaced = torch.bernoulli(torch.full(input_ids.shape, 0.8)).bool() & masked_indices
@@ -163,9 +163,9 @@ class HFTextEncoder(nn.Module):
# 10% of the time, we replace masked input tokens with random word
indices_random = torch.bernoulli(torch.full(input_ids.shape, 0.5)).bool() & masked_indices & ~indices_replaced
random_words = torch.randint(vocab_size, input_ids.shape, dtype=torch.long).to(device)
input_ids[indices_random] = random_words[indices_random]
# The rest of the time (10% of the time) we keep the masked input tokens unchanged
input_ids[indices_random] = random_words[indices_random]
# The rest of the time (10% of the time) we keep the masked input tokens unchanged
if targets is not None:
return input_ids, targets
else:
@@ -174,7 +174,7 @@ class HFTextEncoder(nn.Module):
def forward_mlm(self, input_ids, image_embeds, mlm_probability=0.25):
labels = input_ids.clone()
attn_mask = (input_ids != self.config.pad_token_id).long()
image_atts = torch.ones(image_embeds.size()[:-1],dtype=torch.long).to(input_ids.device)
image_atts = torch.ones(image_embeds.size()[:-1],dtype=torch.long).to(input_ids.device)
vocab_size = getattr(self.config, arch_dict[self.config.model_type]["config_names"]["vocab_size"])
probability_matrix = torch.full(labels.shape, mlm_probability)
input_ids, labels = self.mask(input_ids, vocab_size, input_ids.device, targets=labels,
scripts/pulid/eva_clip/loss.py
+3 -3
View File
@@ -119,7 +119,7 @@ class ClipLoss(nn.Module):
self.prev_num_logits = num_logits
else:
labels = self.labels[device]
if self.label_smoothing_cross_entropy:
total_loss = (
self.label_smoothing_cross_entropy(logits_per_image, labels) +
@@ -130,9 +130,9 @@ class ClipLoss(nn.Module):
F.cross_entropy(logits_per_image, labels) +
F.cross_entropy(logits_per_text, labels)
) / 2
acc = None
i2t_acc = (logits_per_image.argmax(-1) == labels).sum() / len(logits_per_image)
t2i_acc = (logits_per_text.argmax(-1) == labels).sum() / len(logits_per_text)
acc = {"i2t": i2t_acc, "t2i": t2i_acc}
return total_loss, acc
return total_loss, acc
scripts/pulid/eva_clip/model.py
+2 -2
View File
@@ -231,7 +231,7 @@ class CLIP(nn.Module):
def set_grad_checkpointing(self, enable=True):
self.visual.set_grad_checkpointing(enable)
self.transformer.grad_checkpointing = enable
@torch.jit.ignore
def no_weight_decay(self):
return {'logit_scale'}
@@ -309,7 +309,7 @@ def convert_weights_to_lp(model: nn.Module, dtype=torch.float16):
"""Convert applicable model parameters to low-precision (bf16 or fp16)"""
def _convert_weights(l):
if isinstance(l, (nn.Conv1d, nn.Conv2d, nn.Linear)):
l.weight.data = l.weight.data.to(dtype)
if l.bias is not None:
scripts/pulid/eva_clip/model_configs/EVA01-CLIP-B-16.json
+1 -1
View File
@@ -16,4 +16,4 @@
"heads": 8,
"layers": 12
}
}
}
scripts/pulid/eva_clip/model_configs/EVA01-CLIP-g-14-plus.json
+1 -1
View File
@@ -21,4 +21,4 @@
"xattn": false,
"fusedLN": true
}
}
}
scripts/pulid/eva_clip/model_configs/EVA01-CLIP-g-14.json
+1 -1
View File
@@ -21,4 +21,4 @@
"xattn": false,
"fusedLN": true
}
}
}
scripts/pulid/eva_clip/model_configs/EVA02-CLIP-B-16.json
+1 -1
View File
@@ -26,4 +26,4 @@
"xattn": true,
"fusedLN": true
}
}
}
scripts/pulid/eva_clip/model_configs/EVA02-CLIP-L-14-336.json
+1 -1
View File
@@ -26,4 +26,4 @@
"xattn": false,
"fusedLN": true
}
}
}
scripts/pulid/eva_clip/model_configs/EVA02-CLIP-L-14.json
+1 -1
View File
@@ -26,4 +26,4 @@
"xattn": false,
"fusedLN": true
}
}
}
scripts/pulid/eva_clip/model_configs/EVA02-CLIP-bigE-14.json
+1 -1
View File
@@ -22,4 +22,4 @@
"xattn": false,
"fusedLN": true
}
}
}
scripts/pulid/eva_clip/rope.py
+2 -2
View File
@@ -60,7 +60,7 @@ class VisionRotaryEmbedding(nn.Module):
freqs_w = torch.einsum('..., f -> ... f', t, freqs)
freqs_w = repeat(freqs_w, '... n -> ... (n r)', r = 2)
freqs = broadcat((freqs_h[:, None, :], freqs_w[None, :, :]), dim = -1)
freqs = broadcat((freqs_h[:, None, :], freqs_w[None, :, :]), dim = -1)
self.register_buffer("freqs_cos", freqs.cos())
self.register_buffer("freqs_sin", freqs.sin())
@@ -134,4 +134,4 @@ class VisionRotaryEmbeddingFast(nn.Module):
return t * freqs_cos + rotate_half(t) * freqs_sin
return t * self.freqs_cos + rotate_half(t) * self.freqs_sin
return t * self.freqs_cos + rotate_half(t) * self.freqs_sin
scripts/pulid/eva_clip/transformer.py
+7 -7
View File
@@ -12,7 +12,7 @@ try:
from timm.models.layers import trunc_normal_
except:
from timm.layers import trunc_normal_
from .rope import VisionRotaryEmbedding, VisionRotaryEmbeddingFast
from .utils import to_2tuple
@@ -311,7 +311,7 @@ class CustomAttention(nn.Module):
attn = self.attn_drop(attn)
x = torch.bmm(attn, v)
if self.head_scale is not None:
x = x.view(B_q, self.num_heads, N_q, C_q) * self.head_scale
x = x.view(-1, N_q, C_q)
@@ -411,7 +411,7 @@ class CustomTransformer(nn.Module):
])
def get_cast_dtype(self) -> torch.dtype:
return self.resblocks[0].mlp.c_fc.weight.dtype
return self.resblocks[0].mlp.c_fc.weight.dtype
def forward(self, q: torch.Tensor, k: torch.Tensor = None, v: torch.Tensor = None, attn_mask: Optional[torch.Tensor] = None):
if k is None and v is None:
@@ -532,7 +532,7 @@ class VisionTransformer(nn.Module):
# setting a patch_dropout of 0. would mean it is disabled and this function would be the identity fn
self.patch_dropout = PatchDropout(patch_dropout) if patch_dropout > 0. else nn.Identity()
self.ln_pre = norm_layer(width)
self.transformer = Transformer(
width,
layers,
@@ -551,7 +551,7 @@ class VisionTransformer(nn.Module):
def lock(self, unlocked_groups=0, freeze_bn_stats=False):
for param in self.parameters():
param.requires_grad = False
if unlocked_groups != 0:
groups = [
[
@@ -655,7 +655,7 @@ class TextTransformer(nn.Module):
norm_layer=norm_layer,
xattn=xattn
)
self.xattn = xattn
self.ln_final = norm_layer(width)
self.text_projection = nn.Parameter(torch.empty(width, output_dim))
@@ -686,7 +686,7 @@ class TextTransformer(nn.Module):
@torch.jit.ignore
def set_grad_checkpointing(self, enable=True):
self.transformer.grad_checkpointing = enable
@torch.jit.ignore
def no_weight_decay(self):
# return {'positional_embedding', 'token_embedding'}
scripts/pulid/eva_clip/utils.py
+2 -2
View File
@@ -135,7 +135,7 @@ def resize_eva_pos_embed(state_dict, model, interpolation: str = 'bicubic', seq_
patch_size = model.visual.patch_embed.patch_size
state_dict['patch_embed.proj.weight'] = torch.nn.functional.interpolate(
patch_embed_proj.float(), size=patch_size, mode='bicubic', align_corners=False)
def resize_rel_pos_embed(state_dict, model, interpolation: str = 'bicubic', seq_dim=1):
all_keys = list(state_dict.keys())
@@ -323,4 +323,4 @@ class AllGather(torch.autograd.Function):
None
)
allgather = AllGather.apply
allgather = AllGather.apply
scripts/xadapter/pipeline_sd_xl_adapter_controlnet.py
-1
View File
@@ -1824,4 +1824,3 @@ class StableDiffusionXLAdapterControlnetPipeline(DiffusionPipeline, FromSingleFi
init_latents = self.scheduler.add_noise(init_latents, noise, timestep)
return init_latents
scripts/xadapter/pipeline_sd_xl_adapter_controlnet_img2img.py
-1
View File
@@ -1910,4 +1910,3 @@ class StableDiffusionXLAdapterControlnetI2IPipeline(DiffusionPipeline, FromSingl
# image.save(f'./test_img/noisy_image_sd1_5_{int(timestep)}.jpg')
return init_latents
scripts/xadapter/utils.py
+1 -1
View File
@@ -11,4 +11,4 @@ from safetensors import safe_open
from tqdm import tqdm
from einops import rearrange
from model.convert_from_ckpt import convert_ldm_unet_checkpoint, convert_ldm_clip_checkpoint, convert_ldm_vae_checkpoint
# from animatediff.utils.convert_lora_safetensor_to_diffusers import convert_lora, convert_motion_lora_ckpt_to_diffusers
# from animatediff.utils.convert_lora_safetensor_to_diffusers import convert_lora, convert_motion_lora_ckpt_to_diffusers