xinyun/ComfyUI
1
0
Fork 0
mirror of https://git.datalinker.icu/comfyanonymous/ComfyUI synced 2026-09-17 15:17:52 +08:00
Code Issues Packages Projects Releases Wiki Activity

Formatting

Browse Source
This commit is contained in:
City 2024-12-14 21:12:39 +01:00
parent dc64dbfd3f
commit e270142f9a
2 changed files with 13 additions and 25 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
comfy/ldm/pixart/pixart.py
View File

@ -3,7 +3,6 @@
# https://github.com/PixArt-alpha/PixArt-sigma [Apache 2.0 license] # https://github.com/PixArt-alpha/PixArt-sigma [Apache 2.0 license]
import torch import torch
import torch.nn as nn import torch.nn as nn
import numpy as np
from .blocks import ( from .blocks import (
t2i_modulate, t2i_modulate,
@ -11,7 +10,6 @@ from .blocks import (
AttentionKVCompress, AttentionKVCompress,
MultiHeadCrossAttention, MultiHeadCrossAttention,
T2IFinalLayer, T2IFinalLayer,
TimestepEmbedder,
) )
from comfy.ldm.modules.diffusionmodules.mmdit import PatchEmbed, TimestepEmbedder, Mlp, get_1d_sincos_pos_embed_from_grid_torch from comfy.ldm.modules.diffusionmodules.mmdit import PatchEmbed, TimestepEmbedder, Mlp, get_1d_sincos_pos_embed_from_grid_torch

36
comfy/utils.py
View File

@ -385,52 +385,42 @@ def mmdit_to_diffusers(mmdit_config, output_prefix=""):
return key_map return key_map
PIXART_MAP_BASIC = { PIXART_MAP_BASIC = {
# Resolution
("csize_embedder.mlp.0.weight", "adaln_single.emb.resolution_embedder.linear_1.weight"), ("csize_embedder.mlp.0.weight", "adaln_single.emb.resolution_embedder.linear_1.weight"),
("csize_embedder.mlp.0.bias", "adaln_single.emb.resolution_embedder.linear_1.bias"), ("csize_embedder.mlp.0.bias", "adaln_single.emb.resolution_embedder.linear_1.bias"),
("csize_embedder.mlp.2.weight", "adaln_single.emb.resolution_embedder.linear_2.weight"), ("csize_embedder.mlp.2.weight", "adaln_single.emb.resolution_embedder.linear_2.weight"),
("csize_embedder.mlp.2.bias", "adaln_single.emb.resolution_embedder.linear_2.bias"), ("csize_embedder.mlp.2.bias", "adaln_single.emb.resolution_embedder.linear_2.bias"),
# Aspect ratio
("ar_embedder.mlp.0.weight", "adaln_single.emb.aspect_ratio_embedder.linear_1.weight"), ("ar_embedder.mlp.0.weight", "adaln_single.emb.aspect_ratio_embedder.linear_1.weight"),
("ar_embedder.mlp.0.bias", "adaln_single.emb.aspect_ratio_embedder.linear_1.bias"), ("ar_embedder.mlp.0.bias", "adaln_single.emb.aspect_ratio_embedder.linear_1.bias"),
("ar_embedder.mlp.2.weight", "adaln_single.emb.aspect_ratio_embedder.linear_2.weight"), ("ar_embedder.mlp.2.weight", "adaln_single.emb.aspect_ratio_embedder.linear_2.weight"),
("ar_embedder.mlp.2.bias", "adaln_single.emb.aspect_ratio_embedder.linear_2.bias"), ("ar_embedder.mlp.2.bias", "adaln_single.emb.aspect_ratio_embedder.linear_2.bias"),
# Patch embeddings
("x_embedder.proj.weight", "pos_embed.proj.weight"), ("x_embedder.proj.weight", "pos_embed.proj.weight"),
("x_embedder.proj.bias", "pos_embed.proj.bias"), ("x_embedder.proj.bias", "pos_embed.proj.bias"),
# Caption projection
("y_embedder.y_embedding", "caption_projection.y_embedding"), ("y_embedder.y_embedding", "caption_projection.y_embedding"),
("y_embedder.y_proj.fc1.weight", "caption_projection.linear_1.weight"), ("y_embedder.y_proj.fc1.weight", "caption_projection.linear_1.weight"),
("y_embedder.y_proj.fc1.bias", "caption_projection.linear_1.bias"), ("y_embedder.y_proj.fc1.bias", "caption_projection.linear_1.bias"),
("y_embedder.y_proj.fc2.weight", "caption_projection.linear_2.weight"), ("y_embedder.y_proj.fc2.weight", "caption_projection.linear_2.weight"),
("y_embedder.y_proj.fc2.bias", "caption_projection.linear_2.bias"), ("y_embedder.y_proj.fc2.bias", "caption_projection.linear_2.bias"),
# AdaLN-single LN
("t_embedder.mlp.0.weight", "adaln_single.emb.timestep_embedder.linear_1.weight"), ("t_embedder.mlp.0.weight", "adaln_single.emb.timestep_embedder.linear_1.weight"),
("t_embedder.mlp.0.bias", "adaln_single.emb.timestep_embedder.linear_1.bias"), ("t_embedder.mlp.0.bias", "adaln_single.emb.timestep_embedder.linear_1.bias"),
("t_embedder.mlp.2.weight", "adaln_single.emb.timestep_embedder.linear_2.weight"), ("t_embedder.mlp.2.weight", "adaln_single.emb.timestep_embedder.linear_2.weight"),
("t_embedder.mlp.2.bias", "adaln_single.emb.timestep_embedder.linear_2.bias"), ("t_embedder.mlp.2.bias", "adaln_single.emb.timestep_embedder.linear_2.bias"),
# Shared norm
("t_block.1.weight", "adaln_single.linear.weight"), ("t_block.1.weight", "adaln_single.linear.weight"),
("t_block.1.bias", "adaln_single.linear.bias"), ("t_block.1.bias", "adaln_single.linear.bias"),
# Final block
("final_layer.linear.weight", "proj_out.weight"), ("final_layer.linear.weight", "proj_out.weight"),
("final_layer.linear.bias", "proj_out.bias"), ("final_layer.linear.bias", "proj_out.bias"),
("final_layer.scale_shift_table", "scale_shift_table"), ("final_layer.scale_shift_table", "scale_shift_table"),
} }
PIXART_MAP_BLOCK = { PIXART_MAP_BLOCK = {
(f"scale_shift_table", f"scale_shift_table"), ("scale_shift_table", "scale_shift_table"),
# Projection ("attn.proj.weight", "attn1.to_out.0.weight"),
(f"attn.proj.weight", f"attn1.to_out.0.weight"), ("attn.proj.bias", "attn1.to_out.0.bias"),
(f"attn.proj.bias", f"attn1.to_out.0.bias"), ("mlp.fc1.weight", "ff.net.0.proj.weight"),
# Feed-forward ("mlp.fc1.bias", "ff.net.0.proj.bias"),
(f"mlp.fc1.weight", f"ff.net.0.proj.weight"), ("mlp.fc2.weight", "ff.net.2.weight"),
(f"mlp.fc1.bias", f"ff.net.0.proj.bias"), ("mlp.fc2.bias", "ff.net.2.bias"),
(f"mlp.fc2.weight", f"ff.net.2.weight"), ("cross_attn.proj.weight" ,"attn2.to_out.0.weight"),
(f"mlp.fc2.bias", f"ff.net.2.bias"), ("cross_attn.proj.bias" ,"attn2.to_out.0.bias"),
# Cross-attention (proj)
(f"cross_attn.proj.weight" ,f"attn2.to_out.0.weight"),
(f"cross_attn.proj.bias" ,f"attn2.to_out.0.bias"),
} }
def pixart_to_diffusers(mmdit_config, output_prefix=""): def pixart_to_diffusers(mmdit_config, output_prefix=""):