|
|
--- |
|
|
tags: |
|
|
- text-to-image |
|
|
- lora |
|
|
- diffusers |
|
|
- template:diffusion-lora |
|
|
- flux |
|
|
- flux.1/dev |
|
|
- flux.1/schnell |
|
|
- the_artist |
|
|
- transformers |
|
|
widget: |
|
|
- text: >- |
|
|
the_artist, the ai artist creating images inside the latent space, fourier |
|
|
waves, grids, reflective ground, |
|
|
output: |
|
|
url: images/PXeDW6vfKOjNhAcxzruRo.png |
|
|
- text: >- |
|
|
the_artist, concept design style as the artistic representation of a |
|
|
transformer (AI), the opening is the part where the user hits enter and the |
|
|
tokenizers start to create embedding vectors which are each straw. the small |
|
|
nodes are the attention mechanisms, the colours the attention heads. and |
|
|
they keep progressing as in a time continuum tunnel (the inference), until |
|
|
logits explode and the model feels confident enough for the EOS token, |
|
|
white background, |
|
|
output: |
|
|
url: images/tr02.png |
|
|
- text: >- |
|
|
the_artist, the ai artist creating images inside the latent space, fourier |
|
|
waves, grids, reflective ground, |
|
|
output: |
|
|
url: images/877361962999146847.png |
|
|
- text: >- |
|
|
the_artist, concept design style as the artistic representation of a |
|
|
transformer (AI), the opening is the part where the user hits enter and the |
|
|
tokenizers start to create embedding vectors which are each straw. the small |
|
|
nodes are the attention mechanisms, the colours the attention heads. and |
|
|
they keep progressing as in a time continuum tunnel (the inference), until |
|
|
logits explode and the model feels confident enough for the EOS token, |
|
|
white background,, |
|
|
output: |
|
|
url: images/tr01.png |
|
|
- text: >- |
|
|
the_artist, inside the latent space where the AI generate images, grids, |
|
|
geometry, waves |
|
|
output: |
|
|
url: images/877363701387119392.png |
|
|
- text: >- |
|
|
the_artist, inside the latent space where the AI generate images, grids, |
|
|
geometry, waves |
|
|
output: |
|
|
url: images/877363632667641856.png |
|
|
- text: >- |
|
|
the_artist, inside the latent space where the AI generate images, grids, |
|
|
geometry, waves |
|
|
output: |
|
|
url: images/877363756147997457.png |
|
|
- text: >- |
|
|
the_artist, inside the latent space where the AI generate images, grids, |
|
|
geometry, waves |
|
|
output: |
|
|
url: images/877363863522180708.png |
|
|
base_model: black-forest-labs/FLUX.1-dev |
|
|
instance_prompt: the_artist |
|
|
license: cc-by-4.0 |
|
|
language: |
|
|
- en |
|
|
datasets: |
|
|
- Hawkwind/the_artist_flux |
|
|
--- |
|
|
# The Artist | Flux Editition |
|
|
|
|
|
<Gallery /> |
|
|
|
|
|
|
|
|
|
|
|
--- |
|
|
|
|
|
# Model description |
|
|
|
|
|
Experimental version for Flux. |
|
|
|
|
|
With enough creativity and prompting, "the Artist" can help you generate images for diverse types and processes inside the Neural Network in an artistic/abstract way. |
|
|
|
|
|
Ever wanted to create artistic representation of Neural Networks such as Transformers to explain how they work in a fashion viewers can understand it? Now you can: |
|
|
|
|
|
# The Colours of Attention |
|
|
<audio controls src="https://cdn-uploads.huggingface.co/production/uploads/6740a691ddc2c8e208a41102/GfXJR3NyQ7f2OFkBpNbFm.mpga"></audio> |
|
|
|
|
|
``` |
|
|
the_artist, concept design style as the artistic representation of a |
|
|
transformer (AI), the opening is the part where the user hits enter and |
|
|
the tokenizers start to create embedding vectors which are each straw. |
|
|
the small nodes are the attention mechanisms, the colours the attention |
|
|
heads. and they keep progressing as in a time continuum tunnel |
|
|
(the inference), until logits explode and the model feels confident |
|
|
enough for the EOS token, white background, |
|
|
|
|
|
DPM++ 2S A Karras, Guiding Scale 3.5 CFG 6, Steps 5, Seed 2282625028, Clip Skip 1 |
|
|
<Lora:theartist.flux_.safetensors:1.0> |
|
|
Model: FusionV2 (Flux Schnell) |
|
|
``` |
|
|
|
|
|
 |
|
|
|
|
|
|
|
|
This image Lora model should mainly be used for research and educational needs. |
|
|
|
|
|
Although it is licensed under CC 4.0, which means that all generated images can be used for diverse ends, such as illustrations for articles, books, |
|
|
banners, posters. |
|
|
|
|
|
Derivative works are accepted, allowing an educator to edit the images to add captions or fix/change specific traits. |
|
|
|
|
|
The training model was presented with a dataset showcasing artistic interpretations of latent spaces, U-Nets, convolutions, diffusers and even transformers. |
|
|
|
|
|
Any feedback and ideas on how we could enhance and approach related process are very welcome. |
|
|
|
|
|
--- |
|
|
|
|
|
## Soundtrack |
|
|
|
|
|
# Song "LuvDiffuser" by Hawkwind |
|
|
|
|
|
<audio controls src="https://cdn-uploads.huggingface.co/production/uploads/6740a691ddc2c8e208a41102/FwKAg0dSCDviApjx_CMgt.mpga"></audio> |
|
|
|
|
|
He also took the lyrics and used them as image prompt with the_artist at 0.8, flux.1/schnell with dpm++ 2s A karras in 5 steps, guidance 3.5 and cfg 6: |
|
|
|
|
|
<div style="display: flex; justify-content: space-between;"> |
|
|
<img src="https://cdn-uploads.huggingface.co/production/uploads/6740a691ddc2c8e208a41102/QPNy-6qNnsIUWfFct5TzQ.png" alt="luvdiffuser.png" width="400px"> |
|
|
<img src="https://cdn-uploads.huggingface.co/production/uploads/6740a691ddc2c8e208a41102/dUdxh8l2x9jcI_reiax4d.png" alt="luvdiffuser2.png" width="400px"> |
|
|
</div> |
|
|
|
|
|
Caption for the first image: |
|
|
<audio controls src="https://cdn-uploads.huggingface.co/production/uploads/6740a691ddc2c8e208a41102/wJbug38XZWLMQjFS-NdKw.mpga"></audio> |
|
|
|
|
|
|
|
|
--- |
|
|
--- |
|
|
|
|
|
|
|
|
--- |
|
|
|
|
|
Disclaimer: |
|
|
|
|
|
This model is provided “as is” without any warranty. The creators are not responsible for any misuse or unintended consequences of using this model. |
|
|
|
|
|
--- |
|
|
|
|
|
|
|
|
|
|
|
--- |
|
|
|
|
|
For extra information, please proceed to Illustrious version: |
|
|
https://huggingface.co/robb-0/TheArtist-Style-IllustriousXL |
|
|
|
|
|
## Trigger words |
|
|
|
|
|
You should use `the_artist` to trigger the image generation. |
|
|
|
|
|
--- |
|
|
|
|
|
This is a collaborative work of a small group of community members. |
|
|
|
|
|
All songs by Hawkwind |
|
|
https://huggingface.co/Hawkwind |
|
|
|
|
|
--- |
|
|
|
|
|
## Download model |
|
|
|
|
|
Weights for this model are available in Safetensors format. |
|
|
|
|
|
[Download](/robb-0/the_artist_flux/tree/main) them in the Files & versions tab. |
|
|
|
|
|
--- |
|
|
|
|
|
Training settings |
|
|
|
|
|
``` |
|
|
General |
|
|
Batch |
|
|
1 |
|
|
Gradient Acc. Steps |
|
|
2 |
|
|
Resolution |
|
|
1024x1024 |
|
|
Clip Skip |
|
|
1 |
|
|
Epoch |
|
|
5 of 5 |
|
|
Steps |
|
|
465 of 465 |
|
|
|
|
|
Network |
|
|
Module |
|
|
networks.lora_flux |
|
|
Algorithm |
|
|
- |
|
|
Dim / Alpha |
|
|
32 / 16 |
|
|
Conv Dim / Alpha |
|
|
8 / 1 |
|
|
Network Dropout |
|
|
None |
|
|
IP Noise Gamma |
|
|
None |
|
|
Optimizer |
|
|
Type |
|
|
AdamW8bit |
|
|
Scheduler |
|
|
cosine |
|
|
Learning Rates |
|
|
LR: 0.000002 |
|
|
TE: |
|
|
[ |
|
|
0.00001 |
|
|
] |
|
|
UNET: 0.0005 |
|
|
Optional Args |
|
|
- |
|
|
SNR |
|
|
None |
|
|
Warmup Steps |
|
|
0 |
|
|
Noise Offset |
|
|
Noise Offset |
|
|
0.03 |
|
|
Pyramid Noise Iterations |
|
|
10 |
|
|
Discount |
|
|
0.1 |
|
|
Training Info |
|
|
Train Date |
|
|
Jun 21, 2025 |
|
|
Train Time |
|
|
0h 52m 26s |
|
|
Total Images |
|
|
37 |
|
|
Dataset |
|
|
{ |
|
|
"image_dir": { |
|
|
"n_repeats": 5, |
|
|
"img_count": 37 |
|
|
} |
|
|
} |
|
|
``` |
