|
--- |
|
tags: |
|
- text-to-image |
|
- torch |
|
datasets: |
|
- laion/laion_100m_vqgan_f8 |
|
--- |
|
|
|
This model is collaboratively trained a part of the NeurIPS 2021 demonstration ["Training Transformers Together"](https://training-transformers-together.github.io/). |
|
|
|
# Model Description |
|
|
|
We train a model similar to [OpenAI DALL-E](https://openai.com/blog/dall-e/) — a Transformer model that generates images from text descriptions. Training happens collaboratively — volunteers from all over the Internet contribute to the training using hardware available to them. We use [LAION-400M](https://laion.ai/laion-400-open-dataset/), the world's largest openly available image-text-pair dataset with 400 million samples. Our model is based on the [dalle‑pytorch](https://github.com/lucidrains/DALLE-pytorch) implementation by [Phil Wang](https://github.com/lucidrains) with a few tweaks to make it communication-efficient. |
|
|
|
# Training |
|
|
|
You can check our [dashboard](https://huggingface.co/spaces/training-transformers-together/Dashboard) to see what is happening during the collaborative training (loss over time, number of active sessions over time, contribution of each participant, leaderboard, etc. ). |
|
|
|
# How to Use |
|
|
|
This section will be updated soon |
|
|
|
# Limitations |
|
|
|
This model is still being trained, so its generative capabilities will evolve with you! |
