BingoG commited on
Commit
37e3d56
·
verified ·
1 Parent(s): e8b1ab6

Add E2 resolved training config

Browse files
Files changed (1) hide show
  1. E2/config.resolved.yaml +148 -0
E2/config.resolved.yaml ADDED
@@ -0,0 +1,148 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # resolver_provenance: version=spatialav2av_resolver_v2, profile=paper, counterfactual_interval=n/a
2
+ model:
3
+ model_path: /apdcephfs_zwfy4/share_302243908/helensliang/model_weights/LTX-2.3/ltx-2.3-22b-dev.safetensors
4
+ text_encoder_path: /apdcephfs_zwfy4/share_302243908/helensliang/model_weights/gemma-3-12b-it-qat-q4_0-unquantized
5
+ training_mode: full
6
+ load_checkpoint: null
7
+ lora: null
8
+ training_strategy:
9
+ name: spatialav2av
10
+ with_audio: true
11
+ use_refimg: true
12
+ use_ref_audio: false
13
+ masked_temporal_offset: 0.0
14
+ audio_mask_mode: zero
15
+ edit_margin_audio: 2
16
+ loss_video_weight: 0.85
17
+ loss_audio_weight: 0.15
18
+ spatial_aux:
19
+ enabled: true
20
+ method: dual
21
+ reference_mode: dual
22
+ dynamic_enabled: false
23
+ decoder_domain: log_mel
24
+ aux_interval: 1
25
+ warmup_fraction: 0.08
26
+ sigma_min: 0.05
27
+ sigma_max: 0.8
28
+ common_weight: 0.05
29
+ differential_weight: 0.1
30
+ position_weight: 0.0
31
+ velocity_weight: 0.0
32
+ windows: 24
33
+ bands: 4
34
+ mirror_probability: 0.0
35
+ counterfactual:
36
+ paired_forward_enabled: false
37
+ directional_consistency_enabled: false
38
+ required: false
39
+ pair_manifest_path: null
40
+ pair_manifest_sha256: null
41
+ pair_manifest_report_path: null
42
+ pair_manifest_report_sha256: null
43
+ expected_pair_split: train
44
+ expected_pair_schema: spatialav2av_cf_pair_v2
45
+ parent_data_list_sha256: null
46
+ crop_contract_version: av_temporal_crop_v1
47
+ sigma_cf: 1.0
48
+ interval: 16
49
+ pair_types:
50
+ - pan
51
+ - rotate
52
+ rank_weight: 0.0
53
+ min_gt_gap_db: 1.0
54
+ rank_margin_db: 1.0
55
+ score_type: endpoint_delta_rminusl_tild_db_v2_ltx23
56
+ gt_order_from_manifest: true
57
+ pair_branch_base_loss_enabled: false
58
+ same_source_required: true
59
+ same_crop_required: true
60
+ share_audio_noise_across_pair: true
61
+ share_video_noise_across_pair: true
62
+ independent_noise_between_modalities: true
63
+ stateless_pair_seed: true
64
+ isolate_pair_rng: true
65
+ pair_branches_require_grad: true
66
+ sync_trigger_across_ranks: true
67
+ on_missing_or_invalid_pair: error
68
+ optimization:
69
+ learning_rate: 1.0e-05
70
+ steps: 1000
71
+ batch_size: 1
72
+ gradient_accumulation_steps: 1
73
+ max_grad_norm: 1.0
74
+ optimizer_type: adamw
75
+ scheduler_type: constant
76
+ scheduler_params: {}
77
+ enable_gradient_checkpointing: true
78
+ acceleration:
79
+ mixed_precision_mode: bf16
80
+ quantization: null
81
+ load_text_encoder_in_8bit: false
82
+ offload_optimizer_during_validation: true
83
+ data:
84
+ online_encoding: true
85
+ online_offload: true
86
+ num_dataloader_workers: 4
87
+ json_list: /apdcephfs_sh3/share_300819555/helensliang/spatialav2av_data/split_v1/train.list
88
+ edit_min_size: 480
89
+ edit_sample_n_frames: 113
90
+ edit_spatial_mode: max_cap
91
+ edit_max_short_side: 480
92
+ mask_dilate: false
93
+ mask_dilate_range: 0
94
+ mask_bbox: false
95
+ edit_extend_prob: 0.5
96
+ edit_full_prob: 1.0
97
+ edit_zero_prob: 0.0
98
+ edit_margin_video: 0
99
+ resolution_buckets: ''
100
+ validation:
101
+ prompts: []
102
+ negative_prompt: ''
103
+ images: null
104
+ video_dims:
105
+ - 864
106
+ - 480
107
+ - 113
108
+ frame_rate: 25.0
109
+ seed: 42
110
+ inference_steps: 20
111
+ interval: null
112
+ guidance_scale: 4.0
113
+ stg_scale: 1.0
114
+ stg_blocks:
115
+ - 29
116
+ stg_mode: stg_av
117
+ generate_audio: true
118
+ skip_initial_validation: true
119
+ checkpoints:
120
+ interval: 200
121
+ keep_last_n: 5
122
+ precision: bfloat16
123
+ no_resume: true
124
+ save_training_state: 'off'
125
+ flow_matching:
126
+ timestep_sampling_mode: shifted_logit_normal
127
+ timestep_sampling_params: {}
128
+ diagnostics:
129
+ enabled: true
130
+ interval: 200
131
+ max_frames: 113
132
+ hub:
133
+ push_to_hub: false
134
+ hub_model_id: null
135
+ wandb:
136
+ enabled: false
137
+ project: ltx-2.3-trainer
138
+ entity: null
139
+ tags:
140
+ - ltx23
141
+ - spatialav2av
142
+ - E2
143
+ - dual
144
+ - 480p-cap
145
+ - 113f
146
+ log_validation_videos: false
147
+ seed: 42
148
+ output_dir: /apdcephfs_sh3/share_300819555/helensliang/spatialav2av_data/SpatialAV2AV_runs/dual/seed-42/dual_480pcap113f_4node_s42_20260726T195021Z_098d1dc