File size: 3,266 Bytes
5e373a9 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 | # Copyright 2024 Adobe. All rights reserved.
model:
base_learning_rate: 1.0e-05
target: ldm.models.diffusion.ddpm.LatentDiffusion
params:
linear_start: 0.00085
linear_end: 0.0120
num_timesteps_cond: 1
log_every_t: 200
timesteps: 1000
first_stage_key: "inpaint"
cond_stage_key: "image"
image_size: 64
channels: 4
cond_stage_trainable: true # Note: different from the one we trained before
conditioning_key: "rewarp"
monitor: val/loss_simple_ema
u_cond_percent: 0.2
scale_factor: 0.18215
use_ema: False
context_embedding_dim: 768 # TODO embedding # 1024 clip, DINO: 'small': 384,'big': 768,'large': 1024,'huge': 1536
scheduler_config: # 10000 warmup steps
target: ldm.lr_scheduler.LambdaLinearScheduler
params:
warm_up_steps: [ 10000 ]
cycle_lengths: [ 10000000000000 ] # incredibly large number to prevent corner cases
f_start: [ 1.e-6 ]
f_max: [ 1. ]
f_min: [ 1. ]
unet_config:
target: ldm.modules.diffusionmodules.openaimodel.UNetModel
params:
image_size: 32 # unused
in_channels: 9
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
add_conv_in_front_of_unet: 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: 256
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.DINOEmbedder # TODO embedding
params:
dino_version: "big" # [small, big, large, huge]
data:
target: main.DataModuleFromConfig
params:
batch_size: 2
num_workers: 8
use_worker_init_fn: False
wrap: False
train:
target: ldm.data.collage_dataset.CollageDataset
params:
split_files: "<specify value train path>"
image_size: 512
embedding_type: 'dino' # TODO embedding
warping_type: 'collage'
validation:
target: ldm.data.collage_dataset.CollageDataset
params:
split_files: "<specify value val path>"
image_size: 512
embedding_type: 'dino' # TODO embedding
warping_type: 'mix'
test:
target: ldm.data.collage_dataset.CollageDataset
params:
split_files: "<specify value val path>"
image_size: 512
embedding_type: 'dino' # TODO embedding
warping_type: 'mix'
lightning:
trainer:
max_epochs: 500
num_nodes: 1
num_sanity_val_steps: 0
accelerator: 'gpu'
gpus: "0,1,2,3,4,5,6,7"
|