Fix imports.

comfy/cldm/cldm.py

@@ -5,17 +5,17 @@ import torch
 import torch as th
 import torch.nn as nn
 
-from ldm.modules.diffusionmodules.util import (
+from ..ldm.modules.diffusionmodules.util import (
     conv_nd,
     linear,
     zero_module,
     timestep_embedding,
 )
 
-from ldm.modules.attention import SpatialTransformer
-from ldm.modules.diffusionmodules.openaimodel import UNetModel, TimestepEmbedSequential, ResBlock, Downsample, AttentionBlock
-from ldm.models.diffusion.ddpm import LatentDiffusion
-from ldm.util import log_txt_as_img, exists, instantiate_from_config
+from ..ldm.modules.attention import SpatialTransformer
+from ..ldm.modules.diffusionmodules.openaimodel import UNetModel, TimestepEmbedSequential, ResBlock, Downsample, AttentionBlock
+from ..ldm.models.diffusion.ddpm import LatentDiffusion
+from ..ldm.util import log_txt_as_img, exists, instantiate_from_config
 
 
 class ControlledUnetModel(UNetModel):

comfy/gligen.py

 import torch
 from torch import nn, einsum
-from ldm.modules.attention import CrossAttention
+from .ldm.modules.attention import CrossAttention
 from inspect import isfunction
 
 

comfy/ldm/models/autoencoder.py

@@ -3,11 +3,11 @@ import torch
 import torch.nn.functional as F
 from contextlib import contextmanager
 
-from ldm.modules.diffusionmodules.model import Encoder, Decoder
-from ldm.modules.distributions.distributions import DiagonalGaussianDistribution
+from comfy.ldm.modules.diffusionmodules.model import Encoder, Decoder
+from comfy.ldm.modules.distributions.distributions import DiagonalGaussianDistribution
 
-from ldm.util import instantiate_from_config
-from ldm.modules.ema import LitEma
+from comfy.ldm.util import instantiate_from_config
+from comfy.ldm.modules.ema import LitEma
 
 # class AutoencoderKL(pl.LightningModule):
 class AutoencoderKL(torch.nn.Module):

comfy/ldm/models/diffusion/ddim.py

@@ -4,7 +4,7 @@ import torch
 import numpy as np
 from tqdm import tqdm
 
-from ldm.modules.diffusionmodules.util import make_ddim_sampling_parameters, make_ddim_timesteps, noise_like, extract_into_tensor
+from comfy.ldm.modules.diffusionmodules.util import make_ddim_sampling_parameters, make_ddim_timesteps, noise_like, extract_into_tensor
 
 
 class DDIMSampler(object):

comfy/ldm/models/diffusion/ddpm.py

@@ -19,12 +19,12 @@ from tqdm import tqdm
 from torchvision.utils import make_grid
 # from pytorch_lightning.utilities.distributed import rank_zero_only
 
-from ldm.util import log_txt_as_img, exists, default, ismap, isimage, mean_flat, count_params, instantiate_from_config
-from ldm.modules.ema import LitEma
-from ldm.modules.distributions.distributions import normal_kl, DiagonalGaussianDistribution
-from ldm.models.autoencoder import IdentityFirstStage, AutoencoderKL
-from ldm.modules.diffusionmodules.util import make_beta_schedule, extract_into_tensor, noise_like
-from ldm.models.diffusion.ddim import DDIMSampler
+from comfy.ldm.util import log_txt_as_img, exists, default, ismap, isimage, mean_flat, count_params, instantiate_from_config
+from comfy.ldm.modules.ema import LitEma
+from comfy.ldm.modules.distributions.distributions import normal_kl, DiagonalGaussianDistribution
+from ..autoencoder import IdentityFirstStage, AutoencoderKL
+from comfy.ldm.modules.diffusionmodules.util import make_beta_schedule, extract_into_tensor, noise_like
+from .ddim import DDIMSampler
 
 
 __conditioning_keys__ = {'concat': 'c_concat',

comfy/ldm/modules/attention.py

@@ -6,7 +6,7 @@ from torch import nn, einsum
 from einops import rearrange, repeat
 from typing import Optional, Any
 
-from ldm.modules.diffusionmodules.util import checkpoint
+from .diffusionmodules.util import checkpoint
 from .sub_quadratic_attention import efficient_dot_product_attention
 
 from comfy import model_management
@@ -21,7 +21,7 @@ if model_management.xformers_enabled():
 import os
 _ATTN_PRECISION = os.environ.get("ATTN_PRECISION", "fp32")
 
-from cli_args import args
+from comfy.cli_args import args
 
 def exists(val):
     return val is not None

comfy/ldm/modules/diffusionmodules/model.py

@@ -6,7 +6,7 @@ import numpy as np
 from einops import rearrange
 from typing import Optional, Any
 
-from ldm.modules.attention import MemoryEfficientCrossAttention
+from ..attention import MemoryEfficientCrossAttention
 from comfy import model_management
 
 if model_management.xformers_enabled_vae():

comfy/ldm/modules/diffusionmodules/openaimodel.py

@@ -6,7 +6,7 @@ import torch as th
 import torch.nn as nn
 import torch.nn.functional as F
 
-from ldm.modules.diffusionmodules.util import (
+from .util import (
     checkpoint,
     conv_nd,
     linear,
@@ -15,8 +15,8 @@ from ldm.modules.diffusionmodules.util import (
     normalization,
     timestep_embedding,
 )
-from ldm.modules.attention import SpatialTransformer
-from ldm.util import exists
+from ..attention import SpatialTransformer
+from comfy.ldm.util import exists
 
 
 # dummy replace

comfy/ldm/modules/diffusionmodules/upscaling.py

@@ -3,8 +3,8 @@ import torch.nn as nn
 import numpy as np
 from functools import partial
 
-from ldm.modules.diffusionmodules.util import extract_into_tensor, make_beta_schedule
-from ldm.util import default
+from .util import extract_into_tensor, make_beta_schedule
+from comfy.ldm.util import default
 
 
 class AbstractLowScaleModel(nn.Module):

comfy/ldm/modules/diffusionmodules/util.py

@@ -15,7 +15,7 @@ import torch.nn as nn
 import numpy as np
 from einops import repeat
 
-from ldm.util import instantiate_from_config
+from comfy.ldm.util import instantiate_from_config
 
 
 def make_beta_schedule(schedule, n_timestep, linear_start=1e-4, linear_end=2e-2, cosine_s=8e-3):

comfy/ldm/modules/encoders/noise_aug_modules.py

-from ldm.modules.diffusionmodules.upscaling import ImageConcatWithNoiseAugmentation
-from ldm.modules.diffusionmodules.openaimodel import Timestep
+from ..diffusionmodules.upscaling import ImageConcatWithNoiseAugmentation
+from ..diffusionmodules.openaimodel import Timestep
 import torch
 
 class CLIPEmbeddingNoiseAugmentation(ImageConcatWithNoiseAugmentation):

comfy/model_management.py

 import psutil
 from enum import Enum
-from cli_args import args
+from .cli_args import args
 
 class VRAMState(Enum):
     CPU = 0

comfy/sd.py

@@ -2,8 +2,8 @@ import torch
 import contextlib
 import copy
 
-import sd1_clip
-import sd2_clip
+from . import sd1_clip
+from . import sd2_clip
 from comfy import model_management
 from .ldm.util import instantiate_from_config
 from .ldm.models.autoencoder import AutoencoderKL
@@ -446,10 +446,10 @@ class CLIP:
         else:
             params = {}
 
-        if self.target_clip == "ldm.modules.encoders.modules.FrozenOpenCLIPEmbedder":
+        if self.target_clip.endswith("FrozenOpenCLIPEmbedder"):
             clip = sd2_clip.SD2ClipModel
             tokenizer = sd2_clip.SD2Tokenizer
-        elif self.target_clip == "ldm.modules.encoders.modules.FrozenCLIPEmbedder":
+        elif self.target_clip.endswith("FrozenCLIPEmbedder"):
             clip = sd1_clip.SD1ClipModel
             tokenizer = sd1_clip.SD1Tokenizer
 
@@ -896,9 +896,9 @@ def load_clip(ckpt_path, embedding_directory=None):
     clip_data = utils.load_torch_file(ckpt_path)
     config = {}
     if "text_model.encoder.layers.22.mlp.fc1.weight" in clip_data:
-        config['target'] = 'ldm.modules.encoders.modules.FrozenOpenCLIPEmbedder'
+        config['target'] = 'comfy.ldm.modules.encoders.modules.FrozenOpenCLIPEmbedder'
     else:
-        config['target'] = 'ldm.modules.encoders.modules.FrozenCLIPEmbedder'
+        config['target'] = 'comfy.ldm.modules.encoders.modules.FrozenCLIPEmbedder'
     clip = CLIP(config=config, embedding_directory=embedding_directory)
     clip.load_from_state_dict(clip_data)
     return clip
@@ -974,9 +974,9 @@ def load_checkpoint_guess_config(ckpt_path, output_vae=True, output_clip=True, o
     if output_clip:
         clip_config = {}
         if "cond_stage_model.model.transformer.resblocks.22.attn.out_proj.weight" in sd_keys:
-            clip_config['target'] = 'ldm.modules.encoders.modules.FrozenOpenCLIPEmbedder'
+            clip_config['target'] = 'comfy.ldm.modules.encoders.modules.FrozenOpenCLIPEmbedder'
         else:
-            clip_config['target'] = 'ldm.modules.encoders.modules.FrozenCLIPEmbedder'
+            clip_config['target'] = 'comfy.ldm.modules.encoders.modules.FrozenCLIPEmbedder'
         clip = CLIP(config=clip_config, embedding_directory=embedding_directory)
         w.cond_stage_model = clip.cond_stage_model
         load_state_dict_to = [w]
@@ -997,7 +997,7 @@ def load_checkpoint_guess_config(ckpt_path, output_vae=True, output_clip=True, o
             noise_schedule_config["timesteps"] = sd[noise_aug_key].shape[0]
             noise_schedule_config["beta_schedule"] = "squaredcos_cap_v2"
             params["noise_schedule_config"] = noise_schedule_config
-            noise_aug_config['target'] = "ldm.modules.encoders.noise_aug_modules.CLIPEmbeddingNoiseAugmentation"
+            noise_aug_config['target'] = "comfy.ldm.modules.encoders.noise_aug_modules.CLIPEmbeddingNoiseAugmentation"
             if size == 1280: #h
                 params["timestep_dim"] = 1024
             elif size == 1024: #l
@@ -1049,19 +1049,19 @@ def load_checkpoint_guess_config(ckpt_path, output_vae=True, output_clip=True, o
     unet_config["in_channels"] = sd['model.diffusion_model.input_blocks.0.0.weight'].shape[1]
     unet_config["context_dim"] = sd['model.diffusion_model.input_blocks.1.1.transformer_blocks.0.attn2.to_k.weight'].shape[1]
 
-    sd_config["unet_config"] = {"target": "ldm.modules.diffusionmodules.openaimodel.UNetModel", "params": unet_config}
-    model_config = {"target": "ldm.models.diffusion.ddpm.LatentDiffusion", "params": sd_config}
+    sd_config["unet_config"] = {"target": "comfy.ldm.modules.diffusionmodules.openaimodel.UNetModel", "params": unet_config}
+    model_config = {"target": "comfy.ldm.models.diffusion.ddpm.LatentDiffusion", "params": sd_config}
 
     if noise_aug_config is not None: #SD2.x unclip model
         sd_config["noise_aug_config"] = noise_aug_config
         sd_config["image_size"] = 96
         sd_config["embedding_dropout"] = 0.25
         sd_config["conditioning_key"] = 'crossattn-adm'
-        model_config["target"] = "ldm.models.diffusion.ddpm.ImageEmbeddingConditionedLatentDiffusion"
+        model_config["target"] = "comfy.ldm.models.diffusion.ddpm.ImageEmbeddingConditionedLatentDiffusion"
     elif unet_config["in_channels"] > 4: #inpainting model
         sd_config["conditioning_key"] = "hybrid"
         sd_config["finetune_keys"] = None
-        model_config["target"] = "ldm.models.diffusion.ddpm.LatentInpaintDiffusion"
+        model_config["target"] = "comfy.ldm.models.diffusion.ddpm.LatentInpaintDiffusion"
     else:
         sd_config["conditioning_key"] = "crossattn"