# LLAMIAFlux - Pretrained Model

This model was pretrained on the coyo-hd-11m-llavanext dataset to predict CLIP embeddings from text descriptions.

- Number of image generation heads: 4
- Training parameters:
  - Batch size: 32
  - Learning rate: 3e-05
  - Weight decay: 0.01
  - Epochs: 2
  - Model base: ssingh22/LLAMIAFlux-7b-unprojector-inverted
  - Trained on 9438926 image-caption pairs from coyo-hd-11m-llavanext
  - Trained on 9438926 image-caption pairs from coyo-hd-11m-llavanext