File size: 3,991 Bytes
e062867
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
---
frameworks:
- Pytorch
license: Apache License 2.0
tasks:
- text-to-image-synthesis

#model-type:
##如 gpt、phi、llama、chatglm、baichuan 等
#- gpt

#domain:
##如 nlp、cv、audio、multi-modal
#- nlp

#language:
##语言代码列表 https://help.aliyun.com/document_detail/215387.html?spm=a2c4g.11186623.0.0.9f8d7467kni6Aa
#- cn 

#metrics:
##如 CIDEr、Blue、ROUGE 等
#- CIDEr

#tags:
##各种自定义,包括 pretrained、fine-tuned、instruction-tuned、RL-tuned 等训练方法和其他
#- pretrained

#tools:
##如 vllm、fastchat、llamacpp、AdaSeq 等
#- vllm
---
# LoRA 编码器(FLUX.1-Dev)

本模型可以将 FLUX 模型的 LoRA 模型编码为 Embedding 向量,激发出 LoRA 模型的能力。

以 LoRA 模型 [VoidOc/F.1_动物森友会LoRA](https://www.modelscope.cn/models/VoidOc/flux_animal_forest1) 为例,LoRA 编码器有以下几种使用方法。

## 使用方法1:LoRA 用途推断

给定一个 LoRA 模型,在没有任何额外信息的条件下,使用空提示词可以直接激发 LoRA 模型的能力,进而推断出 LoRA 的用途。

提示词:`""`

|不使用 LoRA 编码器|使用 LoRA 编码器|
|-|-|
|![](./assets/image_1_origin.jpg)|![](./assets/image_1.jpg)|

## 使用方法2:免触发词激发 LoRA 能力

无需填写触发词,即可自动激发 LoRA 的能力。

提示词:`"a car"`

|不使用 LoRA 编码器|使用 LoRA 编码器|
|-|-|
|![](./assets/image_2_origin.jpg)|![](./assets/image_2.jpg)|

## 使用方法3:LoRA 强度控制

我们预留了一个额外的参数 `scale`,控制 LoRA 对模型生成图像的影响大小。

在下面的例子中,提示词为“a cat”,当 `scale=1` 时,LoRA 强度为最大,画面中生成了动物森友会中的角色和一只猫;当 `scale=0.5` 时,LoRA 强度被减弱,画面中生成了动物森友会中的猫猫角色。`scale` 的最优数值与 LoRA 模型本身有关,我们建议在角色 LoRA 上使用较大的数值,在风格 LoRA 上使用较小的数值。

提示词:`"a cat"`

|`scale=1`|`scale=0.5`|
|-|-|
|![](./assets/image_3.jpg)|![](./assets/image_3_scale.jpg)|

## 推理代码

```
git clone https://github.com/modelscope/DiffSynth-Studio.git  
cd DiffSynth-Studio
pip install -e .
```

```python
import torch
from diffsynth.pipelines.flux_image_new import FluxImagePipeline, ModelConfig


pipe = FluxImagePipeline.from_pretrained(
    torch_dtype=torch.bfloat16,
    device="cuda",
    model_configs=[
        ModelConfig(model_id="black-forest-labs/FLUX.1-dev", origin_file_pattern="flux1-dev.safetensors"),
        ModelConfig(model_id="black-forest-labs/FLUX.1-dev", origin_file_pattern="text_encoder/model.safetensors"),
        ModelConfig(model_id="black-forest-labs/FLUX.1-dev", origin_file_pattern="text_encoder_2/"),
        ModelConfig(model_id="black-forest-labs/FLUX.1-dev", origin_file_pattern="ae.safetensors"),
        ModelConfig(model_id="DiffSynth-Studio/LoRA-Encoder-FLUX.1-Dev", origin_file_pattern="model.safetensors"),
    ],
)
pipe.enable_lora_magic()

lora = ModelConfig(model_id="VoidOc/flux_animal_forest1", origin_file_pattern="20.safetensors")
pipe.load_lora(pipe.dit, lora, hotload=True) # Use `pipe.clear_lora()` to drop the loaded LoRA.

# Empty prompt can automatically activate LoRA capabilities.
image = pipe(prompt="", seed=0, lora_encoder_inputs=lora)
image.save("image_1.jpg")

image = pipe(prompt="", seed=0)
image.save("image_1_origin.jpg")

# Prompt without trigger words can also activate LoRA capabilities.
image = pipe(prompt="a car", seed=0, lora_encoder_inputs=lora)
image.save("image_2.jpg")

image = pipe(prompt="a car", seed=0,)
image.save("image_2_origin.jpg")

# Adjust the activation intensity through the scale parameter.
image = pipe(prompt="a cat", seed=0, lora_encoder_inputs=lora, lora_encoder_scale=1.0)
image.save("image_3.jpg")

image = pipe(prompt="a cat", seed=0, lora_encoder_inputs=lora, lora_encoder_scale=0.5)
image.save("image_3_scale.jpg")
```