You're looking at a specific version of this model. Jump to the model overview.
zelenioncode /dreambooth_sdxl:748b6c0a
Input schema
The fields you can use to run this model with an API. If you don’t give a value for a field its default value will be used.
| Field | Type | Default value | Description |
|---|---|---|---|
| gender |
string
(enum)
|
woman
Options: woman, man |
Gender of person in training photo ( woman or man )
|
| name_model |
string
|
Dreambooth_sdxl
|
Give name for your .safetensors model
|
| send_to_huggingface |
boolean
|
False
|
Send folder have .safetensors model direct to your huggingface account
|
| token_huggingface |
string
|
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
|
If you use huggingface, enter your API TOKEN
|
| repo_id_huggingface |
string
|
WGlint/SafetensorsFromReplicate
|
If you use huggingface, enter repo_id to one of your project
|
| folder_huggingface |
string
|
|
Enter a path to download your .safetensors model. Default = ./
|
| input |
string
|
https://huggingface.co/WGlint/SafetensorsFromReplicate/resolve/main/input.zip
|
Direct link download with training picture (Only .zip file and picture in 1024px/1024px !)
|
| repeat_input |
integer
|
100
Min: 1 Max: 1000 |
Repeat of time GPU look training data ( e.g. 15 pictures * 100 repeat = 1500 steps )
|
| use_class_reg |
boolean
|
False
|
Use regulat classification picture
|
| repeat_class_reg |
integer
|
1
Min: 1 Max: 1000 |
Repeat of time GPU look class reg picture ( e.g. 5000 pictures * 2 repeat = 10000 steps cache latents )
|
| class_reg |
string
|
|
Direct link download for regular classification picture ( Default = class image of gender you use )
|
| model_sdxl |
string
(enum)
|
Stable Diffusion XL
Options: Stable Diffusion XL, RealVisXL_2, RealVisXL_3 |
Choice a model pretrained can run for SDXL training with dreambooth
|
| num_cpu_threads_per_process |
integer
|
4
Min: 1 Max: 10 |
Number CPU thread use with accelerate module
|
| resolution |
string
|
1024,1024
|
Resolution of your training picture data. WARNING ! Write in this formet : width,height ( e.g. 1024,1024 )
|
| vae |
string
|
stabilityai/sdxl-vae
|
VAE use for create model training
|
| lr_scheduler_num_cycles |
integer
|
1
Min: 1 Max: 1000 |
Num learning rate cycles for your training
|
| max_data_loader_n_workers |
integer
|
0
Max: 100 |
Maximun data loader for n workers you set
|
| learning_rate_te1 |
number
|
0.00001
|
Value for learning_rate te1
|
| learning_rate_te2 |
number
|
0.00001
|
Value for learning_rate te2
|
| learning_rate |
number
|
0.00001
|
Value for learning_rate
|
| lr_scheduler |
string
(enum)
|
constant
Options: constant, linear, cosine, cosine_with_restarts, polynomial, constant_with_warmup, adafactor |
Method use for learning rate scheduler
|
| train_batch_size |
integer
|
1
Min: 1 Max: 64 |
Select value for device max train step and speed the generation, WARINING ! High value = High value to have CUDA Memory
|
| max_train_steps |
integer
|
3000
Min: 1 Max: 25000 |
Number of step you want for your training, and in average 1000 steps = 10 minutes
|
| save_every_n_epochs |
integer
|
1
Min: 1 Max: 64 |
Number of epochs model you want
|
| mixed_precision |
string
(enum)
|
fp16
Options: no, fp16, bf16 |
Select if you want to use miwed precision
|
| save_precision |
string
(enum)
|
fp16
Options: no, fp16, bf16 |
Select if you want to use save precision
|
| optimizer_type |
string
(enum)
|
AdaFactor
Options: AdamW, AdamW8bit, PagedAdamW, PagedAdamW8bit, PagedAdamW32bit, Lion8bit, PagedLion8bit, Lion, SGDNesterov, SGDNesterov8bit, DAdaptation, DAdaptAdaGrad, DAdaptAdam, DAdaptAdan, DAdaptAdanIP, DAdaptLion, DAdaptSGD, AdaFactor |
Select a optimiser type
|
| scale_parameter |
boolean
|
False
|
Use scale parameter
|
| relative_step |
boolean
|
False
|
Use relative step
|
| warmup_init |
boolean
|
False
|
Use warmup init
|
| weight_decay |
number
|
0.01
|
Give a float value for weight decay
|
| bucket_reso_steps |
integer
|
64
Min: 1 Max: 1000 |
Give a int value for bucket reso steps
|
| save_every_n_steps |
integer
|
1
Min: 1 Max: 5 |
Number of .safetensors model you want, if you select 2 with 2000 max train steps, you well get 2 .safetensors. 1 with 1000 steps and 1 with 2000 steps
|
| noise_offset |
number
|
0
|
Give a float value for noise offset
|
| max_grad_norm |
number
|
0
|
Give a float value for max grad norm
|
| cache_latents_to_disk |
boolean
|
True
|
None
|
| cache_latents |
boolean
|
True
|
None
|
| mem_eff_attn |
boolean
|
True
|
None
|
| gradient_checkpointing |
boolean
|
True
|
None
|
| full_fp16 |
boolean
|
True
|
None
|
| xformers |
boolean
|
True
|
None
|
| bucket_no_upscale |
boolean
|
True
|
None
|
| no_half_vae |
boolean
|
True
|
None
|
| train_text_encoder |
boolean
|
True
|
None
|
| learning_rate_te1_bool |
number
|
0.000003
|
value for learning rate te1 bool
|
| learning_rate_te2_bool |
number
|
0
|
value for learning rate te2 bool
|
Output schema
The shape of the response you’ll get when you run this model with an API.
{'format': 'uri', 'title': 'Output', 'type': 'string'}