求大佬指教在用秋叶炼丹炉训练lora模型时失败怎么回事
12:06:01-953368 INFO Training started with config file / 训练开始，使用配置文件:E:\comfyuiCHAJIAN\loraXUNLIAN\config\autosave\20251216-120601.toml
12:06:01-955358 INFO Task efb22f10-546b-4cfb-a283-5da34db8ad08 created
2025-12-16 12:06:08 INFO Loading settings from train_util.py:4012
E:\comfyuiCHAJIAN\loraXUNLIAN\config\autosave\20251216-120601.toml...
INFO E:\comfyuiCHAJIAN\loraXUNLIAN\config\autosave\20251216-120601 train_util.py:4031
2025-12-16 12:06:08 INFO prepare tokenizer train_util.py:4558
2025-12-16 12:06:12 INFO update token length: 255 train_util.py:4575
INFO Using DreamBooth method. train_network.py:172
INFO prepare images. train_util.py:1686
INFO found directory E:\comfyuiCHAJIAN\loraXUNLIAN\train\ysy\30_ysy containstrain_util.py:1633
35 image files
INFO 1050 train images with repeating. train_util.py:1727
INFO 0 reg images. train_util.py:1730
WARNINGno regularization images / 正則化画像が見つかりませんでした train_util.py:1735
INFO config_util.py:572
batch_size: 1
resolution: (512, 512)
enable_bucket: True
network_multiplier: 1.0
min_bucket_reso: 256
max_bucket_reso: 1024
bucket_reso_steps: 64
bucket_no_upscale: True
image_dir: "E:\comfyuiCHAJIAN\loraXUNLIAN\train\ysy\30_ysy"
image_count: 35
num_repeats: 30
shuffle_caption: False
keep_tokens: 0
keep_tokens_separator:
caption_separator: ,
secondary_separator: None
enable_wildcard: False
caption_dropout_rate: 0.0
caption_dropout_every_n_epoches: 0
caption_tag_dropout_rate: 0.0
caption_prefix: None
caption_suffix: None
color_aug: False
flip_aug: False
face_crop_aug_range: None
random_crop: False
token_warmup_min: 1,
token_warmup_step: 0,
alpha_mask: False,
is_reg: False
class_tokens: ysy
caption_extension: .txt
INFO config_util.py:578
INFO loading image sizes. train_util.py:901
100%|███████████████████████████████████████████████████████████████████████████████| 35/35
INFO make buckets train_util.py:907
WARNINGmin_bucket_reso and max_bucket_reso are ignored if bucket_no_upscale is train_util.py:924
set, because bucket reso is defined by image size automatically /
bucket_no_upscaleが指定された場合は、bucketの解像度は画像サイズから自動計
算されるため、min_bucket_resoとmax_bucket_resoは無視されます
INFO number of images (including repeats) / train_util.py:953
各bucketの画像枚数（繰り返し回数を含む）
INFO bucket 0: resolution (512, 512), count: 1050 train_util.py:958
INFO mean ar error (without repeats): 0.0 train_util.py:963
INFO preparing accelerator train_network.py:225
accelerator device: cuda
INFO loading model for process 0/1 train_util.py:4716
INFO load StableDiffusion checkpoint: train_util.py:4672
E:/秋叶/ComfyUI-aki-v2/ComfyUI/models/diffusion_models/majicflus_v10.saf
etensors
Traceback (most recent call last):
File "E:\comfyuiCHAJIAN\loraXUNLIAN\scripts\stable\train_network.py", line 1242, in <module>
trainer.train(args)
File "E:\comfyuiCHAJIAN\loraXUNLIAN\scripts\stable\train_network.py", line 234, in train
model_version, text_encoder, vae, unet = self.load_target_model(args, weight_dtype, accelerator)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "E:\comfyuiCHAJIAN\loraXUNLIAN\scripts\stable\train_network.py", line 101, in load_target_model
text_encoder, vae, unet, _ = train_util.load_target_model(args, weight_dtype, accelerator)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "E:\comfyuiCHAJIAN\loraXUNLIAN\scripts\stable\library\train_util.py", line 4718, in load_target_model
text_encoder, vae, unet, load_stable_diffusion_format = _load_target_model(
^^^^^^^^^^^^^^^^^^^
File "E:\comfyuiCHAJIAN\loraXUNLIAN\scripts\stable\library\train_util.py", line 4673, in _load_target_model
text_encoder, vae, unet = model_util.load_models_from_stable_diffusion_checkpoint(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "E:\comfyuiCHAJIAN\loraXUNLIAN\scripts\stable\library\model_util.py", line 1005, in load_models_from_stable_diffusion_checkpoint
converted_unet_checkpoint = convert_ldm_unet_checkpoint(v2, state_dict, unet_config)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "E:\comfyuiCHAJIAN\loraXUNLIAN\scripts\stable\library\model_util.py", line 267, in convert_ldm_unet_checkpoint
new_checkpoint["time_embedding.linear_1.weight"] = unet_state_dict["time_embed.0.weight"]
~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^
KeyError: 'time_embed.0.weight'
12:06:13-397552 ERROR Training failed / 训练失败
建议遇到报错直接找AI问一下。
Gemini给出的回复是：
核心原因： 你正在尝试用 SD1.5 (Stable Diffusion 1.x) 的训练配置/脚本去加载一个 Flux 架构的大模型。
我看了一下Gemmini的完整回复，它猜测你训练使用的底模是flux架构的，这不一定对，但是可以说明问题。
各种模型之间的架构不同，因此不能使用相同的训练方法，你可以去下载这个底模的网站看一下这个模型的底模（Base Model），然后在训练的时候选择适合的脚本。 ICRS 发表于 2025-12-22 16:15
建议遇到报错直接找AI问一下。
Gemini给出的回复是：
核心原因： 你正在尝试用 SD1.5 (Stable Diffusion 1. ...
谢谢谢谢
大部分问题可以直接复制下来，直接扔给AI，他会处理，kimi、DeepSeek、Grok等等都行
页:
[1]