Hugging Face's logo

BingoG
/
LTX-2-SpatialAV2AV-Checkpoints

PyTorch
ltx-video
spatial-audio
audio-video
Model card Files
xet
 Community
LTX-2-SpatialAV2AV-Checkpoints
File size: 3,893 Bytes
96759bd
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
# resolver_provenance: version=spatialav2av_resolver_v2, profile=paper, counterfactual_interval=n/a
model:
  model_path: /apdcephfs_zwfy4/share_302243908/helensliang/model_weights/LTX-2.3/ltx-2.3-22b-dev.safetensors
  text_encoder_path: /apdcephfs_zwfy4/share_302243908/helensliang/model_weights/gemma-3-12b-it-qat-q4_0-unquantized
  training_mode: full
  load_checkpoint: null
lora: null
training_strategy:
  name: spatialav2av
  with_audio: true
  use_refimg: true
  use_ref_audio: false
  masked_temporal_offset: 0.0
  audio_mask_mode: zero
  edit_margin_audio: 2
  loss_video_weight: 0.85
  loss_audio_weight: 0.15
  spatial_aux:
    enabled: true
    method: core
    reference_mode: dual
    dynamic_enabled: true
    decoder_domain: log_mel
    aux_interval: 1
    warmup_fraction: 0.08
    sigma_min: 0.05
    sigma_max: 0.8
    common_weight: 0.05
    differential_weight: 0.1
    position_weight: 0.05
    velocity_weight: 0.02
    windows: 24
    bands: 4
    mirror_probability: 0.0
    counterfactual:
      paired_forward_enabled: false
      directional_consistency_enabled: false
      required: false
      pair_manifest_path: null
      pair_manifest_sha256: null
      pair_manifest_report_path: null
      pair_manifest_report_sha256: null
      expected_pair_split: train
      expected_pair_schema: spatialav2av_cf_pair_v2
      parent_data_list_sha256: null
      crop_contract_version: av_temporal_crop_v1
      sigma_cf: 1.0
      interval: 16
      pair_types:
      - pan
      - rotate
      rank_weight: 0.0
      min_gt_gap_db: 1.0
      rank_margin_db: 1.0
      score_type: endpoint_delta_rminusl_tild_db_v2_ltx23
      gt_order_from_manifest: true
      pair_branch_base_loss_enabled: false
      same_source_required: true
      same_crop_required: true
      share_audio_noise_across_pair: true
      share_video_noise_across_pair: true
      independent_noise_between_modalities: true
      stateless_pair_seed: true
      isolate_pair_rng: true
      pair_branches_require_grad: true
      sync_trigger_across_ranks: true
      on_missing_or_invalid_pair: error
optimization:
  learning_rate: 1.0e-05
  steps: 1000
  batch_size: 1
  gradient_accumulation_steps: 1
  max_grad_norm: 1.0
  optimizer_type: adamw
  scheduler_type: constant
  scheduler_params: {}
  enable_gradient_checkpointing: true
acceleration:
  mixed_precision_mode: bf16
  quantization: null
  load_text_encoder_in_8bit: false
  offload_optimizer_during_validation: true
data:
  online_encoding: true
  online_offload: true
  num_dataloader_workers: 4
  json_list: /apdcephfs_sh3/share_300819555/helensliang/spatialav2av_data/split_v1/train.list
  edit_min_size: 480
  edit_sample_n_frames: 113
  edit_spatial_mode: max_cap
  edit_max_short_side: 480
  mask_dilate: false
  mask_dilate_range: 0
  mask_bbox: false
  edit_extend_prob: 0.5
  edit_full_prob: 1.0
  edit_zero_prob: 0.0
  edit_margin_video: 0
  resolution_buckets: ''
validation:
  prompts: []
  negative_prompt: ''
  images: null
  video_dims:
  - 864
  - 480
  - 113
  frame_rate: 25.0
  seed: 42
  inference_steps: 20
  interval: null
  guidance_scale: 4.0
  stg_scale: 1.0
  stg_blocks:
  - 29
  stg_mode: stg_av
  generate_audio: true
  skip_initial_validation: true
checkpoints:
  interval: 200
  keep_last_n: 5
  precision: bfloat16
  no_resume: true
  save_training_state: 'off'
flow_matching:
  timestep_sampling_mode: shifted_logit_normal
  timestep_sampling_params: {}
diagnostics:
  enabled: true
  interval: 200
  max_frames: 113
hub:
  push_to_hub: false
  hub_model_id: null
wandb:
  enabled: false
  project: ltx-2.3-trainer
  entity: null
  tags:
  - ltx23
  - spatialav2av
  - E4
  - core
  - 480p-cap
  - 113f
  log_validation_videos: false
seed: 42
output_dir: /apdcephfs_sh3/share_300819555/helensliang/spatialav2av_data/SpatialAV2AV_runs/core/seed-42/core_480pcap113f_4node_s42_20260724T185147Z_098d1dc