Add detail branch v1b best39500 results
Browse files
configs/detail_branch_v1b_aug_photo130k_lsdir.yaml
ADDED
|
@@ -0,0 +1,90 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
project:
|
| 2 |
+
name: detail_branch_v1b_aug_photo130k_lsdir
|
| 3 |
+
output_dir: /home/ubuntu/scratch/sr-diffusion/runs/detail_branch_v1b_aug_photo130k_lsdir
|
| 4 |
+
|
| 5 |
+
logging:
|
| 6 |
+
wandb:
|
| 7 |
+
enabled: true
|
| 8 |
+
project: sr-diffusion
|
| 9 |
+
entity: jwheo
|
| 10 |
+
name: detail_branch_v1b_aug_photo130k_lsdir
|
| 11 |
+
mode: online
|
| 12 |
+
dir: /home/ubuntu/scratch/sr-diffusion/wandb
|
| 13 |
+
group: stage-detail
|
| 14 |
+
job_type: detail-branch
|
| 15 |
+
tags: [detail-branch, stage2-base, high-frequency, photo130k-lsdir, photo-detail-mix, hflip, texture-crop, hr-color-jitter]
|
| 16 |
+
watch: false
|
| 17 |
+
|
| 18 |
+
seed: 1337
|
| 19 |
+
|
| 20 |
+
autoencoder:
|
| 21 |
+
config: configs/autoencoder_photo10k.yaml
|
| 22 |
+
checkpoint: checkpoints/stage1_autoencoder_best_eval_recon.pt
|
| 23 |
+
|
| 24 |
+
condition_encoder:
|
| 25 |
+
config: configs/latent_pretrain_photo130k_lsdir_dual_multiscale_long.yaml
|
| 26 |
+
checkpoint: checkpoints/stage2_photo130k_lsdir_dual_multiscale_best98000.pt
|
| 27 |
+
trainable: false
|
| 28 |
+
|
| 29 |
+
data:
|
| 30 |
+
manifest: /home/ubuntu/scratch/sr-diffusion/data/manifest_photo130k_lsdir.csv
|
| 31 |
+
split: train
|
| 32 |
+
hr_size: 512
|
| 33 |
+
scale: 4
|
| 34 |
+
degradation_preset: photo_detail_mix
|
| 35 |
+
domains:
|
| 36 |
+
photo: 0
|
| 37 |
+
anime: 1
|
| 38 |
+
hflip_prob: 0.5
|
| 39 |
+
texture_crop_retries: 4
|
| 40 |
+
texture_crop_downsample: 128
|
| 41 |
+
hr_color_jitter_prob: 0.25
|
| 42 |
+
hr_color_jitter: [0.97, 1.03]
|
| 43 |
+
num_workers: 4
|
| 44 |
+
|
| 45 |
+
model:
|
| 46 |
+
image_channels: 3
|
| 47 |
+
latent_channels: 16
|
| 48 |
+
hidden_channels: 96
|
| 49 |
+
num_blocks: 8
|
| 50 |
+
norm_groups: 32
|
| 51 |
+
num_domains: 2
|
| 52 |
+
residual_scale: 0.18
|
| 53 |
+
gate_bias: -2.0
|
| 54 |
+
highpass_kernel: 15
|
| 55 |
+
use_condition_latent: false
|
| 56 |
+
|
| 57 |
+
loss:
|
| 58 |
+
image_weight: 1.0
|
| 59 |
+
residual_target_weight: 0.75
|
| 60 |
+
highpass_weight: 1.25
|
| 61 |
+
laplacian_weight: 0.75
|
| 62 |
+
lowpass_anchor_weight: 1.0
|
| 63 |
+
gate_l1_weight: 0.001
|
| 64 |
+
residual_l1_weight: 0.001
|
| 65 |
+
highpass_kernel: 15
|
| 66 |
+
lowpass_kernel: 31
|
| 67 |
+
charbonnier_eps: 0.001
|
| 68 |
+
|
| 69 |
+
train:
|
| 70 |
+
device: auto
|
| 71 |
+
dtype: bf16
|
| 72 |
+
batch_size: 4
|
| 73 |
+
max_steps: 40000
|
| 74 |
+
lr: 0.0001
|
| 75 |
+
weight_decay: 0.0
|
| 76 |
+
grad_accum_steps: 4
|
| 77 |
+
log_every: 25
|
| 78 |
+
save_every: 2000
|
| 79 |
+
|
| 80 |
+
eval:
|
| 81 |
+
enabled: true
|
| 82 |
+
split: val
|
| 83 |
+
limit: 100
|
| 84 |
+
batch_size: 4
|
| 85 |
+
num_workers: 4
|
| 86 |
+
every: 500
|
| 87 |
+
run_at_start: true
|
| 88 |
+
best_metric: eval/detail_score
|
| 89 |
+
best_mode: max
|
| 90 |
+
sample_count: 8
|