A ZIP file containing your training images (JPG, PNG, etc. size not restricted). These images contain your 'subject' that you want the trained model to embed in the output domain for later generating customized scenes beyond the training images. For best results, use images without noise or unrelated objects in the background.

A seed for reproducible training

Default: 1337

The resolution for input images. All the images in the train/validation dataset will be resized to this resolution.

Default: 512

Whether to train the text encoder

Default: true

Batch size (per device) for the training dataloader.

Default: 1

Number of updates steps to accumulate before performing a backward/update pass.

Default: 4

Whether or not to use gradient checkpointing to save memory at the expense of slower backward pass.

Default: false

Scale the learning rate by the number of GPUs, gradient accumulation steps, and batch size.

Default: true

The scheduler type to use

Default: "constant"

Number of steps for the warmup in the lr scheduler.

Default: 0

Whether or not to perform Bayesian Learning Rule on norm of the CLIP latent.

Default: true

Whether or not to use color jitter at augmentation.

Default: true

Whether or not to continue inversion.

Default: false

The learning rate for continuing an inversion.

Default: 0.0001

The tokens to use for the initializer. If not provided, will randomly initialize from gaussian N(0,0.017^2)

The learning rate for the text encoder.

Default: 0.00001

The learning rate for the TI.

Default: 0.0005

The learning rate for the unet.

Default: 0.0001

Rank of the LoRA. Larger it is, more likely to capture fidelity but less likely to be editable. Larger rank will make the end result larger.

Default: 4

Dropout for the LoRA layer. Reference LoRA paper for more details.

Default: 0.1

Scaling parameter at the end of the LoRA layer.

Default: 1

The scheduler type to use

Default: "constant"

Number of steps for the warmup in the lr scheduler.

Default: 0

The maximum number of training steps for the TI.

Default: 500

The maximum number of training steps for the tuning.

Default: 1000

If this value is provided as 'X|Y', it will transform target word X into Y at caption. You are required to provide caption as filename (not regarding extension), and Y has to contain placeholder token below. You are also required to set `None` for `use_template` argument to use this feature.

The placeholder tokens to use for the initializer. If not provided, will use the first tokens of the data.

Default: "<s1>|<s2>"

Whether or not to use the face segmentation condition.

Default: false

The template to use for the inversion.

Default: "object"

The weight decay for the LORA loss.

Default: 0.001

The weight decay for the TI.

Default: 0