Hugging Face's logo

xingjianleng
/
pretrained_vae

Model card Files
xet
Community
xingjianleng commited on
Commit
892ef8c
·
1 Parent(s): 62416d3

update: upload sdvae + vavae ckpts

Browse files
Files changed (4) hide show
  1. sit-xl-1-dinov2-b-enc8-vavae-freq1-lr1e-4-momentum0.1-l1-lpips-kl-gan-vaealign1.5-400k/args.json +54 -0
  2. sit-xl-1-dinov2-b-enc8-vavae-freq1-lr1e-4-momentum0.1-l1-lpips-kl-gan-vaealign1.5-400k/checkpoints/0400000.pt +3 -0
  3. sit-xl-dinov2-b-enc8-sdvae-freq1-lr1e-4-momentum0.1-l1-lpips-kl-gan-vaealign1.5-400k/args.json +60 -0
  4. sit-xl-dinov2-b-enc8-sdvae-freq1-lr1e-4-momentum0.1-l1-lpips-kl-gan-vaealign1.5-400k/checkpoints/0400000.pt +3 -0
sit-xl-1-dinov2-b-enc8-vavae-freq1-lr1e-4-momentum0.1-l1-lpips-kl-gan-vaealign1.5-400k/args.json ADDED
@@ -0,0 +1,54 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "output_dir": "exps",
3
+ "exp_name": "sit-xl-1-dinov2-b-enc8-vavae-freq1-lr1e-4-momentum0.1-l1-lpips-kl-gan-vaealign1.5-compile-400k",
4
+ "logging_dir": "logs",
5
+ "report_to": "wandb",
6
+ "sampling_steps": 10000,
7
+ "resume_step": 0,
8
+ "continue_train_exp_dir": null,
9
+ "wandb_history_path": null,
10
+ "model": "SiT-XL/1",
11
+ "num_classes": 1000,
12
+ "encoder_depth": 8,
13
+ "fused_attn": true,
14
+ "qk_norm": false,
15
+ "bn_momentum": 0.1,
16
+ "compile": true,
17
+ "data_dir": "data/imagenet-latents",
18
+ "resolution": 256,
19
+ "batch_size": 256,
20
+ "allow_tf32": true,
21
+ "mixed_precision": "fp16",
22
+ "epochs": 1400,
23
+ "max_train_steps": 400000,
24
+ "checkpointing_steps": 50000,
25
+ "gradient_accumulation_steps": 1,
26
+ "learning_rate": 0.0001,
27
+ "adam_beta1": 0.9,
28
+ "adam_beta2": 0.999,
29
+ "adam_weight_decay": 0.0,
30
+ "adam_epsilon": 1e-08,
31
+ "max_grad_norm": 1.0,
32
+ "seed": 0,
33
+ "num_workers": 4,
34
+ "path_type": "linear",
35
+ "prediction": "v",
36
+ "cfg_prob": 0.1,
37
+ "enc_type": "dinov2-vit-b",
38
+ "proj_coeff": 0.5,
39
+ "weighting": "uniform",
40
+ "legacy": false,
41
+ "vae": "f16d32",
42
+ "vae_ckpt": "pretrained_models/vavae-f16d32.pt",
43
+ "use_sd_vae_stats": false,
44
+ "vae_repa_align": false,
45
+ "freeze_bn_stats": false,
46
+ "imagenette_stats": 0,
47
+ "disc_pretrained_ckpt": "pretrained_models/vavae-discriminator-ckpt.pt",
48
+ "loss_cfg_path": "configs/loss_cfg/l1_lpips_kl_gan.yaml",
49
+ "vae_learning_rate": 0.0001,
50
+ "disc_learning_rate": 0.0001,
51
+ "vae_train_freq": 1,
52
+ "vae_align": true,
53
+ "vae_align_proj_coeff": 1.5
54
+ }
sit-xl-1-dinov2-b-enc8-vavae-freq1-lr1e-4-momentum0.1-l1-lpips-kl-gan-vaealign1.5-400k/checkpoints/0400000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7774bf4e6669ecf8bbb9e4aeddff9f43664edbcb8e1e3747f5045e782f4f7d57
3
+ size 11804655313
sit-xl-dinov2-b-enc8-sdvae-freq1-lr1e-4-momentum0.1-l1-lpips-kl-gan-vaealign1.5-400k/args.json ADDED
@@ -0,0 +1,60 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "output_dir": "exps",
3
+ "exp_name": "sit-xl-dinov2-b-enc8-sdvae-officialstats-freq1-lr1e-4-momentum0.1-l1-lpips-kl-gan-vaealign1.5-unfreezedecoder-compile-cont-300k-400k",
4
+ "logging_dir": "logs",
5
+ "report_to": "wandb",
6
+ "sampling_steps": 2000,
7
+ "resume_step": 300000,
8
+ "continue_train_exp_dir": "exps/sit-xl-dinov2-b-enc8-sdvae-officialstats-freq1-lr1e-4-momentum0.1-l1-lpips-kl-gan-vaealign1.5-unfreezedecoder-compile-cont-200k-300k",
9
+ "requeue": false,
10
+ "model": "SiT-XL/2",
11
+ "num_classes": 1000,
12
+ "encoder_depth": 8,
13
+ "fused_attn": true,
14
+ "qk_norm": false,
15
+ "bn_momentum": 0.1,
16
+ "bn_affine": false,
17
+ "compile": true,
18
+ "data_dir": "data/imagenet-latents",
19
+ "resolution": 256,
20
+ "batch_size": 256,
21
+ "allow_tf32": true,
22
+ "mixed_precision": "fp16",
23
+ "epochs": 1400,
24
+ "max_train_steps": 400000,
25
+ "checkpointing_steps": 5000,
26
+ "gradient_accumulation_steps": 1,
27
+ "learning_rate": 0.0001,
28
+ "adam_beta1": 0.9,
29
+ "adam_beta2": 0.999,
30
+ "adam_weight_decay": 0.0,
31
+ "adam_epsilon": 1e-08,
32
+ "max_grad_norm": 1.0,
33
+ "seed": 0,
34
+ "num_workers": 4,
35
+ "path_type": "linear",
36
+ "prediction": "v",
37
+ "cfg_prob": 0.1,
38
+ "enc_type": "dinov2-vit-b",
39
+ "proj_coeff": 0.5,
40
+ "weighting": "uniform",
41
+ "legacy": false,
42
+ "vae": "f8d4",
43
+ "vae_ckpt": "pretrained_models/sdvae-ft-mse-f8d4.pt",
44
+ "use_sd_vae_stats": true,
45
+ "vae_repa_align": false,
46
+ "freeze_bn_stats": false,
47
+ "vae_lora": false,
48
+ "vae_lora_rank": 64,
49
+ "vae_lora_alpha": 16,
50
+ "vae_lora_dropout": 0.0,
51
+ "vae_lora_bias": "none",
52
+ "disc_pretrained_ckpt": "pretrained_models/sdvae-original-discriminator-ckpt.pt",
53
+ "loss_cfg_path": "configs/loss_cfg/l1_lpips_kl_gan.yaml",
54
+ "vae_learning_rate": 0.0001,
55
+ "disc_learning_rate": 0.0001,
56
+ "vae_train_freq": 1,
57
+ "unfreeze_decoder": true,
58
+ "vae_align": true,
59
+ "vae_align_proj_coeff": 1.5
60
+ }
sit-xl-dinov2-b-enc8-sdvae-freq1-lr1e-4-momentum0.1-l1-lpips-kl-gan-vaealign1.5-400k/checkpoints/0400000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b67b62bb09159a1fd2455ce93edeb99f58ebead8252adbf0750f5ceef7dc53e8
3
+ size 11969765837