Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,35 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
inference: false
|
| 3 |
+
---
|
| 4 |
+
|
| 5 |
+
<br>
|
| 6 |
+
<br>
|
| 7 |
+
|
| 8 |
+
# LWM-32K-Jax Model Card
|
| 9 |
+
|
| 10 |
+
## Model details
|
| 11 |
+
|
| 12 |
+
**Model type:**
|
| 13 |
+
LWM-32K-Jax is an open-source model trained from LLaMA-2 on a subset of Books3 filtered data, along with a large collection if image and video data. It is an auto-regressive vision-language model, based on the transformer architecture. These are the Jax / Flax version of the parameters.
|
| 14 |
+
|
| 15 |
+
**Model date:**
|
| 16 |
+
LWM-32K-Jax was trained in January 2024.
|
| 17 |
+
|
| 18 |
+
**Paper or resources for more information:**
|
| 19 |
+
https://largeworldmodel.github.io/
|
| 20 |
+
|
| 21 |
+
## License
|
| 22 |
+
Llama 2 is licensed under the LLAMA 2 Community License,
|
| 23 |
+
Copyright (c) Meta Platforms, Inc. All Rights Reserved.
|
| 24 |
+
|
| 25 |
+
**Where to send questions or comments about the model:**
|
| 26 |
+
https://github.com/LargeWorldModel/lwm/issues
|
| 27 |
+
|
| 28 |
+
## Training dataset
|
| 29 |
+
- Books3 dataset
|
| 30 |
+
- 700B text-image pairs from Laion-2B-en, filtered to only keep images with at least 256 resolution
|
| 31 |
+
- 400M text-image pairs from COYO-700M, filtered to only keep images with at least 256 resolution
|
| 32 |
+
- 10M text-video pairs from WebVid10M
|
| 33 |
+
- 3M text-video pairs from a subset of InternVid10M
|
| 34 |
+
- 73K text-video chat pairs from Valley-Instruct-73K
|
| 35 |
+
- 100K text-video chat pairs from Video-ChatGPT
|