Upload subfolder: output/exp-vq
Browse files- output/exp-vq/best_ckpt.pt +3 -0
- output/exp-vq/config.yaml +82 -0
- output/exp-vq/last_ckpt.pt +3 -0
output/exp-vq/best_ckpt.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e9dff792df6684301e13de8239fe1a14aa846a6372ba5da52a051eb32a10f47b
|
| 3 |
+
size 3536901338
|
output/exp-vq/config.yaml
ADDED
|
@@ -0,0 +1,82 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
data_path: /mnt/localssd/ImageNet2012/train
|
| 2 |
+
data_face_path:
|
| 3 |
+
cloud_save_path: output/exp-vq
|
| 4 |
+
no_local_save: false
|
| 5 |
+
vq_model: VQ-16
|
| 6 |
+
vq_ckpt:
|
| 7 |
+
finetune: false
|
| 8 |
+
ema: true
|
| 9 |
+
codebook_size: 4096
|
| 10 |
+
codebook_embed_dim: 32
|
| 11 |
+
codebook_l2_norm: true
|
| 12 |
+
codebook_weight: 1.0
|
| 13 |
+
entropy_loss_ratio: 0.0
|
| 14 |
+
commit_loss_beta: 0.25
|
| 15 |
+
reconstruction_weight: 1.0
|
| 16 |
+
reconstruction_loss: l2
|
| 17 |
+
perceptual_weight: 1.0
|
| 18 |
+
disc_weight: 0.5
|
| 19 |
+
disc_epoch_start: 56
|
| 20 |
+
disc_start: 0
|
| 21 |
+
disc_type: dinodisc
|
| 22 |
+
disc_loss: hinge
|
| 23 |
+
gen_loss: hinge
|
| 24 |
+
compile: false
|
| 25 |
+
dropout_p: 0.0
|
| 26 |
+
results_dir: results_tokenizer_image
|
| 27 |
+
dataset: imagenet
|
| 28 |
+
image_size: 256
|
| 29 |
+
epochs: 200
|
| 30 |
+
lr: 3e-5
|
| 31 |
+
disc_lr: 0.0001
|
| 32 |
+
max_grad_norm: 0.0
|
| 33 |
+
lr_scheduler: cosine
|
| 34 |
+
weight_decay: 0.0
|
| 35 |
+
disc_weight_decay: 0.0005
|
| 36 |
+
beta1: 0.9
|
| 37 |
+
beta2: 0.95
|
| 38 |
+
global_batch_size: 1024
|
| 39 |
+
global_seed: 0
|
| 40 |
+
num_workers: 16
|
| 41 |
+
log_every: 100
|
| 42 |
+
vis_every: 5000
|
| 43 |
+
ckpt_every: 10000
|
| 44 |
+
gradient_accumulation_steps: 1
|
| 45 |
+
mixed_precision: bf16
|
| 46 |
+
save_best: true
|
| 47 |
+
val_data_path: /mnt/localssd/ImageNet2012/val
|
| 48 |
+
sample_folder_dir: samples
|
| 49 |
+
reconstruction_folder_dir: reconstruction
|
| 50 |
+
v_patch_nums:
|
| 51 |
+
- 16
|
| 52 |
+
enc_type: dinov2
|
| 53 |
+
dec_type: dinov2
|
| 54 |
+
semantic_guide: dinov2
|
| 55 |
+
num_latent_tokens: 256
|
| 56 |
+
encoder_model: vit_base_patch14_dinov2.lvd142m
|
| 57 |
+
decoder_model: vit_base_patch14_dinov2.lvd142m
|
| 58 |
+
disc_adaptive_weight: true
|
| 59 |
+
abs_pos_embed: true
|
| 60 |
+
product_quant: 2
|
| 61 |
+
share_quant_resi: 4
|
| 62 |
+
codebook_drop: 0.1
|
| 63 |
+
half_sem: true
|
| 64 |
+
start_drop: 3
|
| 65 |
+
lecam_loss_weight: 0.001
|
| 66 |
+
sem_loss_weight: 0.1
|
| 67 |
+
enc_tuning_method: full
|
| 68 |
+
dec_tuning_method: full
|
| 69 |
+
clip_norm: false
|
| 70 |
+
sem_loss_scale: 1.0
|
| 71 |
+
config: configs/tokenizer.yaml
|
| 72 |
+
norm_type: bn
|
| 73 |
+
aug_prob: 1.0
|
| 74 |
+
aug_fade_steps: 0
|
| 75 |
+
disc_reinit: 0
|
| 76 |
+
debug_disc: false
|
| 77 |
+
rank: 0
|
| 78 |
+
world_size: 32
|
| 79 |
+
gpu: 0
|
| 80 |
+
dist_url: env://
|
| 81 |
+
distributed: true
|
| 82 |
+
dist_backend: nccl
|
output/exp-vq/last_ckpt.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a2efb98708a66dfc0115ff732fb96a94a32b87fc1af0122b107725ffece71fa2
|
| 3 |
+
size 3536901338
|