Commit
•
ea14bb9
1
Parent(s):
bd73f2a
Update README.md
Browse files
README.md
CHANGED
@@ -238,11 +238,7 @@ Using the model to generate content that is cruel to individuals is a misuse of
|
|
238 |
- The model does not perform well on more difficult tasks which involve compositionality, such as rendering an image corresponding to “A red cube on top of a blue sphere”
|
239 |
- Faces and people in general may not be generated properly.
|
240 |
- The model was trained mainly with English captions and will not work as well in other languages.
|
241 |
-
- The autoencoding part of the model is lossy
|
242 |
-
- The model was trained on a large-scale dataset
|
243 |
-
[LAION-5B](https://laion.ai/blog/laion-5b/) which contains adult material
|
244 |
-
and is not fit for product use without additional safety mechanisms and
|
245 |
-
considerations.
|
246 |
- No additional measures were used to deduplicate the dataset. As a result, we observe some degree of memorization for images that are duplicated in the training data.
|
247 |
The training data can be searched at [https://rom1504.github.io/clip-retrieval/](https://rom1504.github.io/clip-retrieval/) to possibly assist in the detection of memorized images.
|
248 |
|
|
|
238 |
- The model does not perform well on more difficult tasks which involve compositionality, such as rendering an image corresponding to “A red cube on top of a blue sphere”
|
239 |
- Faces and people in general may not be generated properly.
|
240 |
- The model was trained mainly with English captions and will not work as well in other languages.
|
241 |
+
- The autoencoding part of the model is lossy.
|
|
|
|
|
|
|
|
|
242 |
- No additional measures were used to deduplicate the dataset. As a result, we observe some degree of memorization for images that are duplicated in the training data.
|
243 |
The training data can be searched at [https://rom1504.github.io/clip-retrieval/](https://rom1504.github.io/clip-retrieval/) to possibly assist in the detection of memorized images.
|
244 |
|