Upload xl_256_rot_asym_10_8_6_4/config.yaml with huggingface_hub
Browse files
xl_256_rot_asym_10_8_6_4/config.yaml
ADDED
|
@@ -0,0 +1,82 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
ckpt_path: '/workspace/LightningDiT/output/xl_256_rot_asym_10_8_6_4/checkpoints/0080000.pt'
|
| 2 |
+
data:
|
| 3 |
+
data_path: /workspace/vavae_f16d32/imagenet_train_256
|
| 4 |
+
fid_reference_file: /workspace/VIRTUAL_imagenet256_labeled.npz
|
| 5 |
+
image_size: 256
|
| 6 |
+
latent_multiplier: 1.0
|
| 7 |
+
latent_norm: true
|
| 8 |
+
num_classes: 1000
|
| 9 |
+
num_workers: 8
|
| 10 |
+
model:
|
| 11 |
+
degree_list:
|
| 12 |
+
- 270
|
| 13 |
+
- 270
|
| 14 |
+
- 0
|
| 15 |
+
- 90
|
| 16 |
+
- 0
|
| 17 |
+
- 90
|
| 18 |
+
- 90
|
| 19 |
+
- 180
|
| 20 |
+
- 270
|
| 21 |
+
- 180
|
| 22 |
+
- 90
|
| 23 |
+
- 0
|
| 24 |
+
- 270
|
| 25 |
+
- 0
|
| 26 |
+
- 0
|
| 27 |
+
- 180
|
| 28 |
+
- 90
|
| 29 |
+
- 180
|
| 30 |
+
- 0
|
| 31 |
+
- 0
|
| 32 |
+
- 180
|
| 33 |
+
- 90
|
| 34 |
+
- 0
|
| 35 |
+
- 90
|
| 36 |
+
- 180
|
| 37 |
+
- 0
|
| 38 |
+
- 90
|
| 39 |
+
- 0
|
| 40 |
+
in_chans: 32
|
| 41 |
+
model_type: LightningDiT-XL/1
|
| 42 |
+
use_qknorm: false
|
| 43 |
+
use_rmsnorm: true
|
| 44 |
+
use_rope: true
|
| 45 |
+
use_swiglu: true
|
| 46 |
+
wo_shift: false
|
| 47 |
+
optimizer:
|
| 48 |
+
beta2: 0.95
|
| 49 |
+
lr: 0.0002
|
| 50 |
+
sample:
|
| 51 |
+
atol: 1.0e-06
|
| 52 |
+
cfg_interval_start: 0.11
|
| 53 |
+
cfg_scale: 10.0
|
| 54 |
+
fid_num: 50000
|
| 55 |
+
likelihood: false
|
| 56 |
+
mode: ODE
|
| 57 |
+
num_sampling_steps: 250
|
| 58 |
+
per_proc_batch_size: 64
|
| 59 |
+
reverse: false
|
| 60 |
+
rtol: 0.001
|
| 61 |
+
sampling_method: euler
|
| 62 |
+
timestep_shift: 0.3
|
| 63 |
+
train:
|
| 64 |
+
ckpt: null
|
| 65 |
+
ckpt_every: 40000
|
| 66 |
+
exp_name: xl_256_rot_asym_10_8_6_4
|
| 67 |
+
global_batch_size: 1024
|
| 68 |
+
global_seed: 0
|
| 69 |
+
log_every: 100
|
| 70 |
+
max_steps: 80000
|
| 71 |
+
output_dir: output
|
| 72 |
+
transport:
|
| 73 |
+
loss_weight: null
|
| 74 |
+
path_type: Linear
|
| 75 |
+
prediction: velocity
|
| 76 |
+
sample_eps: null
|
| 77 |
+
train_eps: null
|
| 78 |
+
use_cosine_loss: true
|
| 79 |
+
use_lognorm: true
|
| 80 |
+
vae:
|
| 81 |
+
downsample_ratio: 16
|
| 82 |
+
model_name: vavae_f16d32
|