EveryDream-trainer/configs/stable-diffusion/v1-finetune_everydream.yaml

106 lines
3.0 KiB
YAML

model:
base_learning_rate: 1.0e-6
target: ldm.models.diffusion.ddpm.LatentDiffusion
params:
reg_weight: 1.0
linear_start: 0.00085
linear_end: 0.0120
num_timesteps_cond: 1
log_every_t: 200
timesteps: 1000
first_stage_key: image
cond_stage_key: caption
image_size: 64
channels: 4
cond_stage_trainable: true # Note: different from the one we trained before
conditioning_key: crossattn
monitor: val/loss_simple_ema
scale_factor: 0.18215
use_ema: False
embedding_reg_weight: 0.0
unfreeze_model: True
model_lr: 1.0e-6
unet_config:
target: ldm.modules.diffusionmodules.openaimodel.UNetModel
params:
image_size: 32 # unused
in_channels: 4
out_channels: 4
model_channels: 320
attention_resolutions: [ 4, 2, 1 ]
num_res_blocks: 2
channel_mult: [ 1, 2, 4, 4 ]
num_heads: 8
use_spatial_transformer: True
transformer_depth: 1
context_dim: 768
use_checkpoint: True
legacy: False
first_stage_config:
target: ldm.models.autoencoder.AutoencoderKL
params:
embed_dim: 4
monitor: val/rec_loss
ddconfig:
double_z: true
z_channels: 4
resolution: 512
in_channels: 3
out_ch: 3
ch: 128
ch_mult:
- 1
- 2
- 4
- 4
num_res_blocks: 2
attn_resolutions: []
dropout: 0.0
lossconfig:
target: torch.nn.Identity
cond_stage_config:
target: ldm.modules.encoders.modules.FrozenCLIPEmbedder
data:
target: main.DataModuleFromConfig
params:
batch_size: 6
num_workers: 8
wrap: falsegit
train:
target: ldm.data.every_dream.EveryDreamBatch
params:
size: 512
set: train
repeats: 5 # suggested setting: 5 with 5000 images, 10 for 1000 images, 50 for 500 images, 100 for <50 images
validation:
target: ldm.data.personalized.PersonalizedBase
params:
size: 512
set: val
repeats: 1
lightning:
modelcheckpoint:
params:
every_n_epochs: 1
#every_n_train_steps: 1400 # can only use epoch or train step checkpoints
save_on_train_epoch_end: False # avoid dupes because it is saved after validation
callbacks:
image_logger:
target: main.ImageLogger
params:
batch_frequency: 200
max_images: 16
increase_log_steps: False
trainer:
benchmark: True
max_epochs: 3 # epoch step count will be (total training images) / batch_size * repeats, suggest 1-4 epochs depending on dataset size and repeats
max_steps: 99000 # better to end on epochs not steps, especially with >500 images to ensure even distribution, but you can set this if you really want...
#check_val_every_n_epoch: 2 # can skip val every epoch if you want by increasing value
#precision: 16 # need lightning 1.6+ ?? *WIP*
#num_nodes: 2 # for multigpu *WIP*