Hugging Face's logo

xingjianleng
/
pretrained_vae

Model card Files
xet
Community
xingjianleng commited on
Commit
cea0304
·
1 Parent(s): 6f421a4

upload 4m intermediate checkpoints

Browse files
Files changed (6) hide show
  1. sit-xl-1-dinov2-b-enc8-invae-freq1-lr1e-4-momentum0.1-l1-lpips-kl-gan-vaealign1.5-4m/args.json +59 -0
  2. sit-xl-1-dinov2-b-enc8-invae-freq1-lr1e-4-momentum0.1-l1-lpips-kl-gan-vaealign1.5-4m/checkpoints/1040000.pt +3 -0
  3. sit-xl-1-dinov2-b-enc8-vavae-freq1-lr1e-4-momentum0.1-l1-lpips-kl-gan-vaealign1.5-4m/args.json +59 -0
  4. sit-xl-1-dinov2-b-enc8-vavae-freq1-lr1e-4-momentum0.1-l1-lpips-kl-gan-vaealign1.5-4m/checkpoints/1700000.pt +3 -0
  5. sit-xl-dinov2-b-enc8-sdvae-officialstats-freq1-lr1e-4-momentum0.1-l1-lpips-kl-gan-vaealign1.5-4m/args.json +59 -0
  6. sit-xl-dinov2-b-enc8-sdvae-officialstats-freq1-lr1e-4-momentum0.1-l1-lpips-kl-gan-vaealign1.5-4m/checkpoints/1320000.pt +3 -0
sit-xl-1-dinov2-b-enc8-invae-freq1-lr1e-4-momentum0.1-l1-lpips-kl-gan-vaealign1.5-4m/args.json ADDED
@@ -0,0 +1,59 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "output_dir": "exps",
3
+ "exp_name": "sit-xl-1-dinov2-b-enc8-invae-freq1-lr1e-4-momentum0.1-l1-lpips-kl-gan-vaealign1.5-4m",
4
+ "logging_dir": "logs",
5
+ "report_to": "wandb",
6
+ "sampling_steps": 9999999999,
7
+ "resume_step": 400000,
8
+ "continue_train_exp_dir": "pretrained_models/sit-xl-1-dinov2-b-enc8-invae-freq1-lr1e-4-momentum0.1-l1-lpips-kl-gan-vaealign1.5-unfreezedecoder-compile-400k",
9
+ "wandb_history_path": null,
10
+ "model": "SiT-XL/1",
11
+ "num_classes": 1000,
12
+ "encoder_depth": 8,
13
+ "qk_norm": false,
14
+ "bn_momentum": 0.1,
15
+ "compile": true,
16
+ "fused_attn": true,
17
+ "use_swiglu": true,
18
+ "use_rope": true,
19
+ "use_rmsnorm": true,
20
+ "wo_shift": false,
21
+ "data_dir": "data/imagenet-latents",
22
+ "resolution": 256,
23
+ "batch_size": 256,
24
+ "allow_tf32": true,
25
+ "mixed_precision": "fp16",
26
+ "epochs": 1400,
27
+ "max_train_steps": 4000000,
28
+ "checkpointing_steps": 20000,
29
+ "gradient_accumulation_steps": 1,
30
+ "learning_rate": 0.0001,
31
+ "adam_beta1": 0.9,
32
+ "adam_beta2": 0.999,
33
+ "adam_weight_decay": 0.0,
34
+ "adam_epsilon": 1e-08,
35
+ "max_grad_norm": 1.0,
36
+ "seed": 0,
37
+ "num_workers": 4,
38
+ "freeze_bn": false,
39
+ "path_type": "linear",
40
+ "prediction": "v",
41
+ "cfg_prob": 0.1,
42
+ "enc_type": "dinov2-vit-b",
43
+ "proj_coeff": 0.5,
44
+ "weighting": "uniform",
45
+ "cos_loss": false,
46
+ "legacy": false,
47
+ "vae": "f16d32",
48
+ "vae_ckpt": "pretrained_models/invae-f16d32.pt",
49
+ "use_sd_vae_stats": false,
50
+ "imagenette_stats": 0,
51
+ "disc_pretrained_ckpt": "pretrained_models/invae-discriminator-ckpt.pt",
52
+ "loss_cfg_path": "configs/loss_cfg/l1_lpips_kl_gan.yaml",
53
+ "vae_learning_rate": 0.0001,
54
+ "disc_learning_rate": 0.0001,
55
+ "vae_train_freq": 1,
56
+ "vae_align": true,
57
+ "vae_align_proj_coeff": 1.5,
58
+ "vae_align_weighting": "uniform"
59
+ }
sit-xl-1-dinov2-b-enc8-invae-freq1-lr1e-4-momentum0.1-l1-lpips-kl-gan-vaealign1.5-4m/checkpoints/1040000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:57fa074be2cce6b090622bd3b0dadbc593a3cfda63400fa6c2ed3e3f4d90824b
3
+ size 11804662737
sit-xl-1-dinov2-b-enc8-vavae-freq1-lr1e-4-momentum0.1-l1-lpips-kl-gan-vaealign1.5-4m/args.json ADDED
@@ -0,0 +1,59 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "output_dir": "exps",
3
+ "exp_name": "sit-xl-1-dinov2-b-enc8-vavae-freq1-lr1e-4-momentum0.1-l1-lpips-kl-gan-vaealign1.5-4m",
4
+ "logging_dir": "logs",
5
+ "report_to": "wandb",
6
+ "sampling_steps": 9999999999,
7
+ "resume_step": 400000,
8
+ "continue_train_exp_dir": "pretrained_models/sit-xl-1-dinov2-b-enc8-vavae-freq1-lr1e-4-momentum0.1-l1-lpips-kl-gan-vaealign1.5-400k",
9
+ "wandb_history_path": null,
10
+ "model": "SiT-XL/1",
11
+ "num_classes": 1000,
12
+ "encoder_depth": 8,
13
+ "qk_norm": false,
14
+ "bn_momentum": 0.1,
15
+ "compile": true,
16
+ "fused_attn": true,
17
+ "use_swiglu": true,
18
+ "use_rope": true,
19
+ "use_rmsnorm": true,
20
+ "wo_shift": false,
21
+ "data_dir": "data/imagenet-latents",
22
+ "resolution": 256,
23
+ "batch_size": 256,
24
+ "allow_tf32": true,
25
+ "mixed_precision": "fp16",
26
+ "epochs": 1400,
27
+ "max_train_steps": 4000000,
28
+ "checkpointing_steps": 20000,
29
+ "gradient_accumulation_steps": 1,
30
+ "learning_rate": 0.0001,
31
+ "adam_beta1": 0.9,
32
+ "adam_beta2": 0.999,
33
+ "adam_weight_decay": 0.0,
34
+ "adam_epsilon": 1e-08,
35
+ "max_grad_norm": 1.0,
36
+ "seed": 0,
37
+ "num_workers": 4,
38
+ "freeze_bn": false,
39
+ "path_type": "linear",
40
+ "prediction": "v",
41
+ "cfg_prob": 0.1,
42
+ "enc_type": "dinov2-vit-b",
43
+ "proj_coeff": 0.5,
44
+ "weighting": "uniform",
45
+ "cos_loss": false,
46
+ "legacy": false,
47
+ "vae": "f16d32",
48
+ "vae_ckpt": "pretrained_models/vavae-f16d32.pt",
49
+ "use_sd_vae_stats": false,
50
+ "imagenette_stats": 0,
51
+ "disc_pretrained_ckpt": "pretrained_models/vavae-discriminator-ckpt.pt",
52
+ "loss_cfg_path": "configs/loss_cfg/l1_lpips_kl_gan.yaml",
53
+ "vae_learning_rate": 0.0001,
54
+ "disc_learning_rate": 0.0001,
55
+ "vae_train_freq": 1,
56
+ "vae_align": true,
57
+ "vae_align_proj_coeff": 1.5,
58
+ "vae_align_weighting": "uniform"
59
+ }
sit-xl-1-dinov2-b-enc8-vavae-freq1-lr1e-4-momentum0.1-l1-lpips-kl-gan-vaealign1.5-4m/checkpoints/1700000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0daaeba42aeb36711991480427459f77aa7a3912e40977a9a1c5f9ced5bf2fad
3
+ size 11804662673
sit-xl-dinov2-b-enc8-sdvae-officialstats-freq1-lr1e-4-momentum0.1-l1-lpips-kl-gan-vaealign1.5-4m/args.json ADDED
@@ -0,0 +1,59 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "output_dir": "exps",
3
+ "exp_name": "sit-xl-dinov2-b-enc8-sdvae-officialstats-freq1-lr1e-4-momentum0.1-l1-lpips-kl-gan-vaealign1.5-4m",
4
+ "logging_dir": "logs",
5
+ "report_to": "wandb",
6
+ "sampling_steps": 9999999999,
7
+ "resume_step": 400000,
8
+ "continue_train_exp_dir": "pretrained_models/sit-xl-dinov2-b-enc8-sdvae-freq1-lr1e-4-momentum0.1-l1-lpips-kl-gan-vaealign1.5-400k",
9
+ "wandb_history_path": null,
10
+ "model": "SiT-XL/2",
11
+ "num_classes": 1000,
12
+ "encoder_depth": 8,
13
+ "qk_norm": false,
14
+ "bn_momentum": 0.1,
15
+ "compile": true,
16
+ "fused_attn": true,
17
+ "use_swiglu": true,
18
+ "use_rope": true,
19
+ "use_rmsnorm": true,
20
+ "wo_shift": false,
21
+ "data_dir": "data/imagenet-latents",
22
+ "resolution": 256,
23
+ "batch_size": 256,
24
+ "allow_tf32": true,
25
+ "mixed_precision": "fp16",
26
+ "epochs": 1400,
27
+ "max_train_steps": 4000000,
28
+ "checkpointing_steps": 20000,
29
+ "gradient_accumulation_steps": 1,
30
+ "learning_rate": 0.0001,
31
+ "adam_beta1": 0.9,
32
+ "adam_beta2": 0.999,
33
+ "adam_weight_decay": 0.0,
34
+ "adam_epsilon": 1e-08,
35
+ "max_grad_norm": 1.0,
36
+ "seed": 0,
37
+ "num_workers": 4,
38
+ "freeze_bn": false,
39
+ "path_type": "linear",
40
+ "prediction": "v",
41
+ "cfg_prob": 0.1,
42
+ "enc_type": "dinov2-vit-b",
43
+ "proj_coeff": 0.5,
44
+ "weighting": "uniform",
45
+ "cos_loss": false,
46
+ "legacy": false,
47
+ "vae": "f8d4",
48
+ "vae_ckpt": "pretrained_models/sdvae-ft-mse-f8d4.pt",
49
+ "use_sd_vae_stats": true,
50
+ "imagenette_stats": 0,
51
+ "disc_pretrained_ckpt": "pretrained_models/sdvae-original-discriminator-ckpt.pt",
52
+ "loss_cfg_path": "configs/loss_cfg/l1_lpips_kl_gan.yaml",
53
+ "vae_learning_rate": 0.0001,
54
+ "disc_learning_rate": 0.0001,
55
+ "vae_train_freq": 1,
56
+ "vae_align": true,
57
+ "vae_align_proj_coeff": 1.5,
58
+ "vae_align_weighting": "uniform"
59
+ }
sit-xl-dinov2-b-enc8-sdvae-officialstats-freq1-lr1e-4-momentum0.1-l1-lpips-kl-gan-vaealign1.5-4m/checkpoints/1320000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f4694265274047266f15728271821be0af91b0cf630fe4c7eee084a8e28501c7
3
+ size 11969765773