zeke / dreambooth

A copy of replicate/dreambooth for testing

  • Public
  • 1 run
  • T4

Input

*string
Shift + Return to add a new line

The prompt with identifier specifying the instance

*string
Shift + Return to add a new line

The prompt to specify images in the same class as provided instance images.

*file

A ZIP file containing the training data of instance images

file

A ZIP file containing the training data of class images. Images will be generated if you do not provide.

integer

Minimal class images for prior preservation loss. If not enough images are provided in class_data, additional images will be sampled with class_prompt.

Default: 50

string
Shift + Return to add a new line

The prompt used to generate sample outputs to save.

string
Shift + Return to add a new line

The negative prompt used to generate sample outputs to save.

integer

The number of samples to save.

Default: 4

number

CFG for save sample.

Default: 7.5

integer

The number of inference steps for save sample.

Default: 50

boolean

Flag to pad tokens to length 77.

Default: false

boolean

Flag to add prior preservation loss.

Default: true

number

Weight of prior preservation loss.

Default: 1

integer

A seed for reproducible training

Default: 1337

integer

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

Default: 512

boolean

Whether to center crop images before resizing to resolution

Default: false

boolean

Whether to train the text encoder

Default: true

integer

Batch size (per device) for the training dataloader.

Default: 1

integer

Batch size (per device) for sampling images.

Default: 4

integer

Default: 1

integer

Total number of training steps to perform. If provided, overrides num_train_epochs.

Default: 2000

integer

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

Default: 1

boolean

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

Default: false

number

Initial learning rate (after the potential warmup period) to use.

Default: 0.000001

boolean

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

Default: false

string

The scheduler type to use

Default: "constant"

integer

Number of steps for the warmup in the lr scheduler.

Default: 0

boolean

Whether or not to use 8-bit Adam from bitsandbytes.

Default: false

number

The beta1 parameter for the Adam optimizer.

Default: 0.9

number

The beta2 parameter for the Adam optimizer.

Default: 0.999

number

Weight decay to use

Default: 0.01

number

Epsilon value for the Adam optimizer

Default: 1e-8

number

Max gradient norm.

Default: 1

Output

No output yet! Press "Submit" to start a prediction.

Run time and cost

This model runs on Nvidia T4 GPU hardware. We don't yet have enough runs of this model to provide performance information.

Readme

Model description

Intended use

Ethical considerations

Caveats and recommendations