LouisCastricato commited on
Commit
ef135cd
·
verified ·
1 Parent(s): f392710

Upload folder using huggingface_hub

Browse files
Files changed (5) hide show
  1. ae/ckpt.pt +3 -0
  2. ae/config.yaml +79 -0
  3. config.yaml +40 -0
  4. model.safetensors +3 -0
  5. owl_wms.commit +1 -0
ae/ckpt.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9a1b558adab097e601598e7bb781e0f26035f5ac4a8fcfbe6eaa2b0f1eee3a07
3
+ size 6811962067
ae/config.yaml ADDED
@@ -0,0 +1,79 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ model:
2
+ model_id: dcae
3
+ sample_size: [360, 640]
4
+ channels: 3
5
+ latent_size: 32
6
+ latent_channels: 16
7
+
8
+ ch_0: 256
9
+ ch_max: 2048
10
+
11
+ encoder_blocks_per_stage: [4, 4, 4, 8]
12
+ decoder_blocks_per_stage: [4, 4, 4, 8]
13
+
14
+ use_middle_block: false
15
+ do_channel_mask: false
16
+
17
+ train:
18
+ trainer_id: rec
19
+ data_id: video_dir_loader
20
+ data_kwargs:
21
+ source:
22
+ - /mnt/data/datasets/extracted_tars/kbm/fps/*/*.mp4
23
+ - /mnt/data/datasets/extracted_tars/kbm/3ps/*/*.mp4
24
+ - /mnt/data/datasets/extracted_tars/kbm/other/*/*.mp4
25
+ target_size: [360, 640]
26
+
27
+ target_batch_size: 32
28
+ batch_size: 4
29
+
30
+ epochs: 200
31
+
32
+ #opt: AdamW
33
+ #opt_kwargs:
34
+ # lr: 3.0e-5
35
+ # weight_decay: 1.0e-4
36
+ # betas: [0.9, 0.95]
37
+ # eps: 1.0e-15
38
+
39
+ opt: Muon
40
+ opt_kwargs:
41
+ lr: 1.0e-3
42
+ momentum: 0.95
43
+ adamw_lr: 1.0e-5
44
+ adamw_wd: 1.0e-2
45
+ adamw_eps: 1.0e-6
46
+ adamw_betas: [0.9, 0.95]
47
+ adamw_keys:
48
+ - encoder.conv_in
49
+ - encoder.conv_out
50
+ - encoder.conv_out_logvar
51
+ - decoder.conv_in
52
+ - decoder.conv_out
53
+ - .up.
54
+ - .down.
55
+ - .residuals.
56
+
57
+ lpips_type: convnext
58
+ loss_weights:
59
+ kl: 3.0e-7
60
+ lpips: 12.0
61
+ l2: 1.0
62
+ dwt: 0.25
63
+
64
+
65
+ scheduler: LinearWarmup
66
+ scheduler_kwargs:
67
+ warmup_steps: 3000
68
+ min_lr: 1.0e-5
69
+
70
+ checkpoint_dir: checkpoints/waypoint_1_vae_owlc_f16_c16
71
+ resume_ckpt: checkpoints/waypoint_1_vae_owlc_f16_c16/step_200000.pt
72
+
73
+ sample_interval: 1000
74
+ save_interval: 5000
75
+
76
+ wandb:
77
+ name: shahbuland
78
+ project: new_vaes_v2
79
+ run_name: waypoint_1_vae_owlc_f16_c16_highres
config.yaml ADDED
@@ -0,0 +1,40 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ channels: 16
2
+ n_layers: 22
3
+ n_heads: 40
4
+ n_kv_heads: 20
5
+ d_model: 2560
6
+ mlp_ratio: 5
7
+ mlp_gradient_checkpointing: true
8
+ causal: true
9
+ n_buttons: 256
10
+ tokens_per_frame: 256
11
+ height: 16
12
+ width: 16
13
+ patch:
14
+ - 2
15
+ - 2
16
+ base_fps: 60
17
+ local_window: 16
18
+ global_window: 128
19
+ global_pinned_dilation: 8
20
+ global_attn_period: 4
21
+ global_attn_offset: 0
22
+ value_residual: false
23
+ n_frames: 800
24
+ rope_impl: ortho
25
+ gated_attn: true
26
+ noise_conditioning: wan
27
+ ctrl_conditioning: mlp_fusion
28
+ ctrl_cond_dropout: 0.0
29
+ ctrl_conditioning_period: 3
30
+ prompt_conditioning: cross_attention
31
+ prompt_cond_dropout: 0.0
32
+ prompt_embedding_dim: 2048
33
+ prompt_conditioning_period: 3
34
+ block_gradient_checkpointing: true
35
+ scheduler_sigmas:
36
+ - 1.0
37
+ - 0.8609585762023926
38
+ - 0.729332447052002
39
+ - 0.3205108940601349
40
+ - 0.0
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:14356db9229453850f9ad650f31c3e1c4744066abd43562f6fbee161fb36c9e6
3
+ size 12515075376
owl_wms.commit ADDED
@@ -0,0 +1 @@
 
 
1
+ da4ed5ea709dd8418ebb2ac1082083758a458bf4