patrickvonplaten
commited on
Commit
•
ac7275a
1
Parent(s):
597ad97
Update README.md
Browse files
README.md
CHANGED
@@ -8,7 +8,7 @@ license: openrail++
|
|
8 |
inference: false
|
9 |
---
|
10 |
|
11 |
-
# Latent Consistency Model (LCM) LoRA:
|
12 |
|
13 |
Latent Consistency Model (LCM) LoRA was proposed in [LCM-LoRA: A universal Stable-Diffusion Acceleration Module](TODO:)
|
14 |
by *Simian Luo, Yiqin Tan, Suraj Patil, Daniel Gu et al.*
|
@@ -35,15 +35,15 @@ pip install --upgrade diffusers transformers accelerate peft
|
|
35 |
|
36 |
### Text-to-Image
|
37 |
|
38 |
-
The adapter can be loaded with it's base model `
|
39 |
Please make sure to either disable `guidance_scale` or use values between 1.0 and 2.0.
|
40 |
|
41 |
```python
|
42 |
import torch
|
43 |
from diffusers import LCMScheduler, AutoPipelineForText2Image
|
44 |
|
45 |
-
model_id = "
|
46 |
-
adapter_id = "latent-consistency/lcm-lora-
|
47 |
|
48 |
pipe = AutoPipelineForText2Image.from_pretrained(model_id, torch_dtype=torch.float16, variant="fp16")
|
49 |
pipe.scheduler = LCMScheduler.from_config(pipe.scheduler.config)
|
|
|
8 |
inference: false
|
9 |
---
|
10 |
|
11 |
+
# Latent Consistency Model (LCM) LoRA: SDv1-5
|
12 |
|
13 |
Latent Consistency Model (LCM) LoRA was proposed in [LCM-LoRA: A universal Stable-Diffusion Acceleration Module](TODO:)
|
14 |
by *Simian Luo, Yiqin Tan, Suraj Patil, Daniel Gu et al.*
|
|
|
35 |
|
36 |
### Text-to-Image
|
37 |
|
38 |
+
The adapter can be loaded with it's base model `runwayml/stable-diffusion-v1-5`. Next, the scheduler needs to be changed to [`LCMScheduler`](https://huggingface.co/docs/diffusers/v0.22.3/en/api/schedulers/lcm#diffusers.LCMScheduler) and we can reduce the number of inference steps to just 2 to 8 steps.
|
39 |
Please make sure to either disable `guidance_scale` or use values between 1.0 and 2.0.
|
40 |
|
41 |
```python
|
42 |
import torch
|
43 |
from diffusers import LCMScheduler, AutoPipelineForText2Image
|
44 |
|
45 |
+
model_id = "runwayml/stable-diffusion-v1-5"
|
46 |
+
adapter_id = "latent-consistency/lcm-lora-sdv1-5"
|
47 |
|
48 |
pipe = AutoPipelineForText2Image.from_pretrained(model_id, torch_dtype=torch.float16, variant="fp16")
|
49 |
pipe.scheduler = LCMScheduler.from_config(pipe.scheduler.config)
|