Any-to-Any
Transformers
Safetensors
gestalt
multimodal
discrete-diffusion
vision-language
unified-model
custom_code
Instructions to use GeWu-Lab/Gestalt with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use GeWu-Lab/Gestalt with Transformers:
# pip install -U transformers accelerate # Load model directly from transformers import GestaltModelLM model = GestaltModelLM.from_pretrained("GeWu-Lab/Gestalt", trust_remote_code=True, device_map="auto") - Notebooks
- Google Colab
- Kaggle
Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -4,7 +4,7 @@ tags:
|
|
| 4 |
- multimodal
|
| 5 |
- discrete-diffusion
|
| 6 |
- vision-language
|
| 7 |
-
-
|
| 8 |
---
|
| 9 |
|
| 10 |
<p align="center">
|
|
@@ -47,7 +47,7 @@ tags:
|
|
| 47 |
|
| 48 |
## Model Description
|
| 49 |
|
| 50 |
-
This is the
|
| 51 |
|
| 52 |
| Property | Value |
|
| 53 |
|:--|:--|
|
|
|
|
| 4 |
- multimodal
|
| 5 |
- discrete-diffusion
|
| 6 |
- vision-language
|
| 7 |
+
- unified-model
|
| 8 |
---
|
| 9 |
|
| 10 |
<p align="center">
|
|
|
|
| 47 |
|
| 48 |
## Model Description
|
| 49 |
|
| 50 |
+
This is the official checkpoint of **Gestalt**, a large multimodal interplay model trained through three progressive stages: multimodal pretraining (70M), continual pretraining (8M), and interplay-oriented supervised fine-tuning (13.7M + 2.5M T2I). Gestalt supports both multimodal understanding and image generation within a unified discrete diffusion framework.
|
| 51 |
|
| 52 |
| Property | Value |
|
| 53 |
|:--|:--|
|