Hugging Face's logo

Cccccz
/
HY

Diffusers
Safetensors
Model card Files
xet
 Community
Cccccz commited on
Commit
21cd5b1
·
verified ·
1 Parent(s): ac4da24

Upload batch 18: 13 files (7.20 GiB)

Browse files
Files changed (13) hide show
  1. checkpoints/predictor_v4_rollout_fppp_blocks1-52_overfit8_seed20260721_fp32master_8gpu_flr1e-4_blr1e-5_schedsteps2000_swarmup8_bwarmup8_swanlab/training_latest.pt +3 -0
  2. checkpoints/predictor_v4_rollout_smoke_8gpu_flr1e-5_blr1e-6_schedsteps2_swarmup50_bwarmup-none/predictor_step_00002.safetensors +3 -0
  3. checkpoints/predictor_v4_rollout_smoke_8gpu_flr1e-5_blr1e-6_schedsteps2_swarmup50_bwarmup-none/train_config.json +47 -0
  4. checkpoints/predictor_v4_rollout_smoke_8gpu_flr1e-5_blr1e-6_schedsteps2_swarmup50_bwarmup-none/train_log.jsonl +2 -0
  5. checkpoints/predictor_v4_rollout_smoke_8gpu_flr1e-5_blr1e-6_schedsteps2_swarmup50_bwarmup-none/training_latest.pt +3 -0
  6. checkpoints/predictor_v4_rollout_smoke_full_single_flr1e-5_blr1e-6_schedsteps1_swarmup50_bwarmup-none/train_config.json +47 -0
  7. checkpoints/predictor_v4_rollout_smoke_full_single_v2_flr1e-5_blr1e-6_schedsteps1_swarmup50_bwarmup-none/predictor_step_00001.safetensors +3 -0
  8. checkpoints/predictor_v4_rollout_smoke_full_single_v2_flr1e-5_blr1e-6_schedsteps1_swarmup50_bwarmup-none/train_config.json +47 -0
  9. checkpoints/predictor_v4_rollout_smoke_full_single_v2_flr1e-5_blr1e-6_schedsteps1_swarmup50_bwarmup-none/train_log.jsonl +1 -0
  10. checkpoints/predictor_v4_rollout_smoke_full_single_v2_flr1e-5_blr1e-6_schedsteps1_swarmup50_bwarmup-none/training_latest.pt +3 -0
  11. checkpoints/predictor_v4_rollout_smoke_latent-grad_flr1e-5_blr1e-6_schedsteps1_swarmup50_bwarmup-none/predictor_step_00001.safetensors +3 -0
  12. checkpoints/predictor_v4_rollout_smoke_latent-grad_flr1e-5_blr1e-6_schedsteps1_swarmup50_bwarmup-none/train_config.json +47 -0
  13. checkpoints/predictor_v4_rollout_smoke_latent-grad_flr1e-5_blr1e-6_schedsteps1_swarmup50_bwarmup-none/train_log.jsonl +1 -0
checkpoints/predictor_v4_rollout_fppp_blocks1-52_overfit8_seed20260721_fp32master_8gpu_flr1e-4_blr1e-5_schedsteps2000_swarmup8_bwarmup8_swanlab/training_latest.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:315e63c190b05e0cece254c932f37d4e766321a55c7ecfe1725036296db330ed
3
+ size 2668812046
checkpoints/predictor_v4_rollout_smoke_8gpu_flr1e-5_blr1e-6_schedsteps2_swarmup50_bwarmup-none/predictor_step_00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4c95bf406fd257900f83b5cc950dbc55f8ebf0504a99e21e42457832e05f6da8
3
+ size 696559936
checkpoints/predictor_v4_rollout_smoke_8gpu_flr1e-5_blr1e-6_schedsteps2_swarmup50_bwarmup-none/train_config.json ADDED
@@ -0,0 +1,47 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "base_model": "/mnt/s3files/s3-us-west2-default/zoubin/cz/checkpoints/hy_worldplay/huggingface/hub/models--tencent--HunyuanVideo-1.5/snapshots/9b49404b3f5df2a8f0b31df27a0c7ab872e7b038",
3
+ "blocks_lr": 1e-06,
4
+ "data_root": "/mnt/local_nvme/zoubin/cz/hyworldplay_predictor_prefeature_v4",
5
+ "dtype": "bfloat16",
6
+ "fusion_lr": 1e-05,
7
+ "global_trajectories_per_step": 8,
8
+ "gpu": "NVIDIA H100 80GB HBM3",
9
+ "grad_clip": 1.0,
10
+ "gradient_checkpointing": true,
11
+ "hidden_weight": 0.1,
12
+ "init_weights": "/mnt/s3files/s3-us-west2-default/zoubin/cz/projects/HY-WorldPlay-DEV-Predictor/checkpoints/predictor_v4_prefeature_blocks1-52_bf16_8gpu_flr1e-4_blr1e-5_schedsteps2000_swarmup100_bwarmup100_swanlab/predictor_step_02000.safetensors",
13
+ "keep_snapshots": 1,
14
+ "log_every": 1,
15
+ "max_steps": 2,
16
+ "max_trajectories": 16,
17
+ "num_chunks": 8,
18
+ "online_teacher_targets": true,
19
+ "online_text_kv": true,
20
+ "output_dir": "/mnt/s3files/s3-us-west2-default/zoubin/cz/projects/HY-WorldPlay-DEV-Predictor/checkpoints/predictor_v4_rollout_smoke_8gpu_flr1e-5_blr1e-6_schedsteps2_swarmup50_bwarmup-none",
21
+ "resume": null,
22
+ "reuse_image_condition_latent": true,
23
+ "save_every": 2,
24
+ "schedule": "chunk0=FFFF,chunks1-7=FPPP",
25
+ "scheduler_validation": {
26
+ "scheduler_max_abs_error": 0.0,
27
+ "scheduler_velocity_grad_l1": 0.14734959602355957
28
+ },
29
+ "scheduler_warmup_steps": 50,
30
+ "seed": 0,
31
+ "source_blocks": [
32
+ 1,
33
+ 52
34
+ ],
35
+ "swanlab_experiment": "predictor-v4-rollout-fppp-1-52-8gpu-2000steps",
36
+ "swanlab_mode": "disabled",
37
+ "swanlab_project": "HY-WorldPlay-Predictor-v4",
38
+ "swanlab_workspace": null,
39
+ "teacher_checkpoint": "/mnt/s3files/s3-us-west2-default/zoubin/cz/checkpoints/hy_worldplay/huggingface/hub/models--tencent--HY-WorldPlay/snapshots/f4c29235647707b571479a69b569e4166f9f5bf8/ar_distilled_action_model/diffusion_pytorch_model.safetensors",
40
+ "trainable_parameters": 197218816,
41
+ "trajectories": 16,
42
+ "trajectories_per_rank_per_epoch": 2,
43
+ "validate_scheduler_only": false,
44
+ "velocity_weight": 1.0,
45
+ "weight_decay": 0.01,
46
+ "world_size": 8
47
+ }
checkpoints/predictor_v4_rollout_smoke_8gpu_flr1e-5_blr1e-6_schedsteps2_swarmup50_bwarmup-none/train_log.jsonl ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ {"batch_in_epoch": 1, "epoch": 0, "global_step": 1, "grad_norm_max": 0.04052734375, "hidden_mse_step1": 0.011618375231461999, "hidden_mse_step2": 0.013987048005219549, "hidden_mse_step3": 0.01689611858455464, "loss": 0.013338339569080352, "lr_blocks": 4e-08, "lr_fusion": 4.0000000000000003e-07, "peak_memory_gib": 47.582953453063965, "trajectory_time_s": 81.97154285112265, "velocity_mse_step1": 0.009096657169201144, "velocity_mse_step2": 0.010728619466785208, "velocity_mse_step3": 0.01593958613479377}
2
+ {"batch_in_epoch": 2, "epoch": 0, "global_step": 2, "grad_norm_max": 0.0625, "hidden_mse_step1": 0.01917185743721867, "hidden_mse_step2": 0.021857682873295353, "hidden_mse_step3": 0.024719686214146868, "loss": 0.02573103164572136, "lr_blocks": 6e-08, "lr_fusion": 6.000000000000001e-07, "peak_memory_gib": 48.31588268280029, "trajectory_time_s": 79.01340394999352, "velocity_mse_step1": 0.01932535518946159, "velocity_mse_step2": 0.02160948164445082, "velocity_mse_step3": 0.029683332116941786}
checkpoints/predictor_v4_rollout_smoke_8gpu_flr1e-5_blr1e-6_schedsteps2_swarmup50_bwarmup-none/training_latest.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5b8ebf3714bc62be434d914eacce40dbc19275b231dec98e2575928776519ac9
3
+ size 1485498574
checkpoints/predictor_v4_rollout_smoke_full_single_flr1e-5_blr1e-6_schedsteps1_swarmup50_bwarmup-none/train_config.json ADDED
@@ -0,0 +1,47 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "base_model": "/mnt/s3files/s3-us-west2-default/zoubin/cz/checkpoints/hy_worldplay/huggingface/hub/models--tencent--HunyuanVideo-1.5/snapshots/9b49404b3f5df2a8f0b31df27a0c7ab872e7b038",
3
+ "blocks_lr": 1e-06,
4
+ "data_root": "/mnt/local_nvme/zoubin/cz/hyworldplay_predictor_prefeature_v4",
5
+ "dtype": "bfloat16",
6
+ "fusion_lr": 1e-05,
7
+ "global_trajectories_per_step": 1,
8
+ "gpu": "NVIDIA H100 80GB HBM3",
9
+ "grad_clip": 1.0,
10
+ "gradient_checkpointing": true,
11
+ "hidden_weight": 0.1,
12
+ "init_weights": "/mnt/s3files/s3-us-west2-default/zoubin/cz/projects/HY-WorldPlay-DEV-Predictor/checkpoints/predictor_v4_prefeature_blocks1-52_bf16_8gpu_flr1e-4_blr1e-5_schedsteps2000_swarmup100_bwarmup100_swanlab/predictor_step_02000.safetensors",
13
+ "keep_snapshots": 1,
14
+ "log_every": 1,
15
+ "max_steps": 1,
16
+ "max_trajectories": 1,
17
+ "num_chunks": 8,
18
+ "online_teacher_targets": true,
19
+ "online_text_kv": true,
20
+ "output_dir": "/mnt/s3files/s3-us-west2-default/zoubin/cz/projects/HY-WorldPlay-DEV-Predictor/checkpoints/predictor_v4_rollout_smoke_full_single_flr1e-5_blr1e-6_schedsteps1_swarmup50_bwarmup-none",
21
+ "resume": null,
22
+ "reuse_image_condition_latent": true,
23
+ "save_every": 1,
24
+ "schedule": "chunk0=FFFF,chunks1-7=FPPP",
25
+ "scheduler_validation": {
26
+ "scheduler_max_abs_error": 0.0,
27
+ "scheduler_velocity_grad_l1": 0.04113103449344635
28
+ },
29
+ "scheduler_warmup_steps": 50,
30
+ "seed": 0,
31
+ "source_blocks": [
32
+ 1,
33
+ 52
34
+ ],
35
+ "swanlab_experiment": "predictor-v4-rollout-fppp-1-52-8gpu-2000steps",
36
+ "swanlab_mode": "disabled",
37
+ "swanlab_project": "HY-WorldPlay-Predictor-v4",
38
+ "swanlab_workspace": null,
39
+ "teacher_checkpoint": "/mnt/s3files/s3-us-west2-default/zoubin/cz/checkpoints/hy_worldplay/huggingface/hub/models--tencent--HY-WorldPlay/snapshots/f4c29235647707b571479a69b569e4166f9f5bf8/ar_distilled_action_model/diffusion_pytorch_model.safetensors",
40
+ "trainable_parameters": 197218816,
41
+ "trajectories": 1,
42
+ "trajectories_per_rank_per_epoch": 1,
43
+ "validate_scheduler_only": false,
44
+ "velocity_weight": 1.0,
45
+ "weight_decay": 0.01,
46
+ "world_size": 1
47
+ }
checkpoints/predictor_v4_rollout_smoke_full_single_v2_flr1e-5_blr1e-6_schedsteps1_swarmup50_bwarmup-none/predictor_step_00001.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dde15b78fd9848d9a2bad9082eb2633c743a3c6a555d2d65d75dfdeb2cc90cbf
3
+ size 696559936
checkpoints/predictor_v4_rollout_smoke_full_single_v2_flr1e-5_blr1e-6_schedsteps1_swarmup50_bwarmup-none/train_config.json ADDED
@@ -0,0 +1,47 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "base_model": "/mnt/s3files/s3-us-west2-default/zoubin/cz/checkpoints/hy_worldplay/huggingface/hub/models--tencent--HunyuanVideo-1.5/snapshots/9b49404b3f5df2a8f0b31df27a0c7ab872e7b038",
3
+ "blocks_lr": 1e-06,
4
+ "data_root": "/mnt/local_nvme/zoubin/cz/hyworldplay_predictor_prefeature_v4",
5
+ "dtype": "bfloat16",
6
+ "fusion_lr": 1e-05,
7
+ "global_trajectories_per_step": 1,
8
+ "gpu": "NVIDIA H100 80GB HBM3",
9
+ "grad_clip": 1.0,
10
+ "gradient_checkpointing": true,
11
+ "hidden_weight": 0.1,
12
+ "init_weights": "/mnt/s3files/s3-us-west2-default/zoubin/cz/projects/HY-WorldPlay-DEV-Predictor/checkpoints/predictor_v4_prefeature_blocks1-52_bf16_8gpu_flr1e-4_blr1e-5_schedsteps2000_swarmup100_bwarmup100_swanlab/predictor_step_02000.safetensors",
13
+ "keep_snapshots": 1,
14
+ "log_every": 1,
15
+ "max_steps": 1,
16
+ "max_trajectories": 1,
17
+ "num_chunks": 8,
18
+ "online_teacher_targets": true,
19
+ "online_text_kv": true,
20
+ "output_dir": "/mnt/s3files/s3-us-west2-default/zoubin/cz/projects/HY-WorldPlay-DEV-Predictor/checkpoints/predictor_v4_rollout_smoke_full_single_v2_flr1e-5_blr1e-6_schedsteps1_swarmup50_bwarmup-none",
21
+ "resume": null,
22
+ "reuse_image_condition_latent": true,
23
+ "save_every": 1,
24
+ "schedule": "chunk0=FFFF,chunks1-7=FPPP",
25
+ "scheduler_validation": {
26
+ "scheduler_max_abs_error": 0.0,
27
+ "scheduler_velocity_grad_l1": 0.10987631976604462
28
+ },
29
+ "scheduler_warmup_steps": 50,
30
+ "seed": 0,
31
+ "source_blocks": [
32
+ 1,
33
+ 52
34
+ ],
35
+ "swanlab_experiment": "predictor-v4-rollout-fppp-1-52-8gpu-2000steps",
36
+ "swanlab_mode": "disabled",
37
+ "swanlab_project": "HY-WorldPlay-Predictor-v4",
38
+ "swanlab_workspace": null,
39
+ "teacher_checkpoint": "/mnt/s3files/s3-us-west2-default/zoubin/cz/checkpoints/hy_worldplay/huggingface/hub/models--tencent--HY-WorldPlay/snapshots/f4c29235647707b571479a69b569e4166f9f5bf8/ar_distilled_action_model/diffusion_pytorch_model.safetensors",
40
+ "trainable_parameters": 197218816,
41
+ "trajectories": 1,
42
+ "trajectories_per_rank_per_epoch": 1,
43
+ "validate_scheduler_only": false,
44
+ "velocity_weight": 1.0,
45
+ "weight_decay": 0.01,
46
+ "world_size": 1
47
+ }
checkpoints/predictor_v4_rollout_smoke_full_single_v2_flr1e-5_blr1e-6_schedsteps1_swarmup50_bwarmup-none/train_log.jsonl ADDED
@@ -0,0 +1 @@
 
 
1
+ {"batch_in_epoch": 1, "epoch": 0, "global_step": 1, "grad_norm_max": 0.0703125, "hidden_mse_step1": 0.02088463013725621, "hidden_mse_step2": 0.025278747880033082, "hidden_mse_step3": 0.02840455248951912, "loss": 0.025971859959619384, "lr_blocks": 4e-08, "lr_fusion": 4.0000000000000003e-07, "peak_memory_gib": 47.581183433532715, "trajectory_time_s": 62.63047818798805, "velocity_mse_step1": 0.017123487378869737, "velocity_mse_step2": 0.021688367639269148, "velocity_mse_step3": 0.03164692861693246}
checkpoints/predictor_v4_rollout_smoke_full_single_v2_flr1e-5_blr1e-6_schedsteps1_swarmup50_bwarmup-none/training_latest.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0bfab0bbfb2639f449c3ef7590045597a80501b31d0f0ec084828a7a1f86f34e
3
+ size 1485498574
checkpoints/predictor_v4_rollout_smoke_latent-grad_flr1e-5_blr1e-6_schedsteps1_swarmup50_bwarmup-none/predictor_step_00001.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d40d272b02b779b839e3168051a74c9a3adc154f8d139c919b8495a34a4ed40c
3
+ size 696559936
checkpoints/predictor_v4_rollout_smoke_latent-grad_flr1e-5_blr1e-6_schedsteps1_swarmup50_bwarmup-none/train_config.json ADDED
@@ -0,0 +1,47 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "base_model": "/mnt/s3files/s3-us-west2-default/zoubin/cz/checkpoints/hy_worldplay/huggingface/hub/models--tencent--HunyuanVideo-1.5/snapshots/9b49404b3f5df2a8f0b31df27a0c7ab872e7b038",
3
+ "blocks_lr": 1e-06,
4
+ "data_root": "/mnt/local_nvme/zoubin/cz/hyworldplay_predictor_prefeature_v4",
5
+ "dtype": "bfloat16",
6
+ "fusion_lr": 1e-05,
7
+ "global_trajectories_per_step": 1,
8
+ "gpu": "NVIDIA H100 80GB HBM3",
9
+ "grad_clip": 1.0,
10
+ "gradient_checkpointing": true,
11
+ "hidden_weight": 0.1,
12
+ "init_weights": "/mnt/s3files/s3-us-west2-default/zoubin/cz/projects/HY-WorldPlay-DEV-Predictor/checkpoints/predictor_v4_prefeature_blocks1-52_bf16_8gpu_flr1e-4_blr1e-5_schedsteps2000_swarmup100_bwarmup100_swanlab/predictor_step_02000.safetensors",
13
+ "keep_snapshots": 1,
14
+ "log_every": 1,
15
+ "max_steps": 1,
16
+ "max_trajectories": 1,
17
+ "num_chunks": 2,
18
+ "online_teacher_targets": true,
19
+ "online_text_kv": true,
20
+ "output_dir": "/mnt/s3files/s3-us-west2-default/zoubin/cz/projects/HY-WorldPlay-DEV-Predictor/checkpoints/predictor_v4_rollout_smoke_latent-grad_flr1e-5_blr1e-6_schedsteps1_swarmup50_bwarmup-none",
21
+ "resume": null,
22
+ "reuse_image_condition_latent": true,
23
+ "save_every": 1,
24
+ "schedule": "chunk0=FFFF,chunks1-7=FPPP",
25
+ "scheduler_validation": {
26
+ "scheduler_max_abs_error": 0.0,
27
+ "scheduler_velocity_grad_l1": 0.0557338148355484
28
+ },
29
+ "scheduler_warmup_steps": 50,
30
+ "seed": 0,
31
+ "source_blocks": [
32
+ 1,
33
+ 52
34
+ ],
35
+ "swanlab_experiment": "predictor-v4-rollout-fppp-1-52-8gpu-2000steps",
36
+ "swanlab_mode": "disabled",
37
+ "swanlab_project": "HY-WorldPlay-Predictor-v4",
38
+ "swanlab_workspace": null,
39
+ "teacher_checkpoint": "/mnt/s3files/s3-us-west2-default/zoubin/cz/checkpoints/hy_worldplay/huggingface/hub/models--tencent--HY-WorldPlay/snapshots/f4c29235647707b571479a69b569e4166f9f5bf8/ar_distilled_action_model/diffusion_pytorch_model.safetensors",
40
+ "trainable_parameters": 197218816,
41
+ "trajectories": 1,
42
+ "trajectories_per_rank_per_epoch": 1,
43
+ "validate_scheduler_only": false,
44
+ "velocity_weight": 1.0,
45
+ "weight_decay": 0.01,
46
+ "world_size": 1
47
+ }
checkpoints/predictor_v4_rollout_smoke_latent-grad_flr1e-5_blr1e-6_schedsteps1_swarmup50_bwarmup-none/train_log.jsonl ADDED
@@ -0,0 +1 @@
 
 
1
+ {"batch_in_epoch": 1, "epoch": 0, "global_step": 1, "grad_norm_max": 0.1533203125, "hidden_mse_step1": 0.04806334897875786, "hidden_mse_step2": 0.052682965993881226, "hidden_mse_step3": 0.053727686405181885, "latent_grad_l1_min": 0.003407946787774563, "loss": 0.04993429780006409, "lr_blocks": 4e-08, "lr_fusion": 4.0000000000000003e-07, "peak_memory_gib": 30.60678243637085, "trajectory_time_s": 19.161244955990696, "velocity_mse_step1": 0.03952038288116455, "velocity_mse_step2": 0.03904150426387787, "velocity_mse_step3": 0.05579359829425812}