Commit
·
35ece41
1
Parent(s):
af7bd31
Update README.md
Browse files
README.md
CHANGED
|
@@ -29,6 +29,22 @@ configs:
|
|
| 29 |
- split: train
|
| 30 |
path: data/train-*
|
| 31 |
---
|
| 32 |
-
#
|
| 33 |
|
| 34 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 29 |
- split: train
|
| 30 |
path: data/train-*
|
| 31 |
---
|
| 32 |
+
# Karlo
|
| 33 |
|
| 34 |
+
All images included in this dataset were voted as "Not solved" by the community in https://huggingface.co/spaces/OpenGenAI/open-parti-prompts.
|
| 35 |
+
This means that according to the community the model did not generate an image that corresponds sufficiently enough to the prompt.
|
| 36 |
+
|
| 37 |
+
The following script was used to generate the images:
|
| 38 |
+
|
| 39 |
+
```py
|
| 40 |
+
from diffusers import DiffusionPipeline
|
| 41 |
+
import torch
|
| 42 |
+
|
| 43 |
+
pipe = DiffusionPipeline.from_pretrained("kakaobrain/karlo-v1-alpha", torch_dtype=torch.float16)
|
| 44 |
+
pipe.to("cuda")
|
| 45 |
+
|
| 46 |
+
prompt = "" # a parti prompt
|
| 47 |
+
generator = torch.Generator("cuda").manual_seed(0)
|
| 48 |
+
|
| 49 |
+
image = pipe(prompt, prior_num_inference_steps=50, decoder_num_inference_steps=100, generator=generator).images[0]
|
| 50 |
+
```
|