| { |
| "base_model_dir": "/mnt/beegfs/omniref-v4-training/models/pdd-checkpoint-1250-resume800", |
| "config": { |
| "callbacks": { |
| "grad_clip": { |
| "_target_": "fastvideo.train.callbacks.grad_clip.GradNormClipCallback", |
| "max_grad_norm": 1.0, |
| "measure_unclipped_grad_norms": true |
| }, |
| "latent_vis": { |
| "_target_": "fastvideo.train.callbacks.latent_vis.LatentVisCallback", |
| "every_steps": 100, |
| "fps": 24 |
| }, |
| "launch_observation": { |
| "_target_": "omniref_b300_observation.B300LaunchObservation" |
| }, |
| "validation": { |
| "_target_": "fastvideo.train.callbacks.minimax_h3_ref2va_pdd_validation.MiniMaxH3Ref2VADMPDDValidationCallback", |
| "dataset_file": "/mnt/beegfs/omniref-v5/validation/v5-b300-heldout-23-20260929/validation.json", |
| "evaluation_label": "v5_heldout23_768p_15s", |
| "every_steps": 100, |
| "expected_forwards": 8, |
| "expected_height": 768, |
| "expected_num_frames": 345, |
| "expected_requested_duration_seconds": 14.375, |
| "expected_validation_records": 23, |
| "expected_width": 1344, |
| "guidance_scale": 1.0, |
| "max_record_num_frames": 345, |
| "num_videos_per_prompt": 1, |
| "offload_training_state": true, |
| "pipeline_target": "fastvideo.pipelines.basic.minimax_h3.minimax_h3_pipeline.MiniMaxH3Ref2VAModularPipeline", |
| "require_reference_artifacts": true, |
| "require_reference_preview": false, |
| "run_at_start": true, |
| "sampling_steps": [ |
| 8 |
| ], |
| "text_encoder_cpu_offload": true, |
| "training_contract_profile": "native_v5_dmpdd_w03_cfg2_data_latent_b300_world64", |
| "use_record_dimensions": true, |
| "use_validation_media_conditioning": false, |
| "vae_cpu_offload": true |
| } |
| }, |
| "method": { |
| "_target_": "fastvideo.train.methods.distribution_matching.dmpdd.DMPDDMethod", |
| "block_size_max": 4, |
| "block_size_min": 4, |
| "cfg_uncond": { |
| "text": "zero" |
| }, |
| "fake_score_betas": [ |
| 0.9, |
| 0.999 |
| ], |
| "fake_score_cfg_dropout_rate": 0.1, |
| "fake_score_guidance_scale": 2.0, |
| "fake_score_learning_rate": 5e-06, |
| "fake_score_loss_space": "x0", |
| "fake_score_lr_scheduler": "constant", |
| "generator_update_interval": 5, |
| "max_timestep_ratio": 0.999, |
| "min_timestep_ratio": 0.001, |
| "pdd_loss_weight": 0.3, |
| "precision_decoding": "fp32", |
| "real_score_guidance_scale": 2.0, |
| "rollout_carry": false, |
| "rollout_mode": "data_latent", |
| "score_timestep_continuous": true, |
| "score_timestep_shift": 2.4, |
| "score_timestep_warp_max": 0.999, |
| "student_grid_indices": { |
| "8": [ |
| 0, |
| 4, |
| 8, |
| 12, |
| 16, |
| 20, |
| 24, |
| 28, |
| 32 |
| ] |
| }, |
| "student_sample_steps": 8, |
| "teacher_guidance_scale": 2.0, |
| "type_rk": "euler" |
| }, |
| "models": { |
| "critic": { |
| "_target_": "fastvideo.train.models.minimax_h3.MiniMaxH3Ref2VDMDModel", |
| "attention_backend": "FLASH_ATTN", |
| "disable_custom_init_weights": true, |
| "enable_gradient_checkpointing_type": "full", |
| "init_from": "/mnt/beegfs/omniref-v4-training/models/MiniMax-H3-ref2va", |
| "trainable": true |
| }, |
| "student": { |
| "_target_": "fastvideo.train.models.minimax_h3.MiniMaxH3Ref2VPDDModel", |
| "attention_backend": "VIDEO_SPARSE_ATTN_H3", |
| "enable_gradient_checkpointing_type": "full", |
| "init_from": "/mnt/beegfs/omniref-v4-training/models/pdd-checkpoint-1250-resume800", |
| "pdd_steps": 32, |
| "trainable": true |
| }, |
| "teacher": { |
| "_target_": "fastvideo.train.models.minimax_h3.MiniMaxH3Ref2VDMDModel", |
| "attention_backend": "FLASH_ATTN", |
| "disable_custom_init_weights": true, |
| "init_from": "/mnt/beegfs/omniref-v4-training/models/MiniMax-H3-ref2va", |
| "trainable": false |
| } |
| }, |
| "pipeline": { |
| "audio_scheduler_shift": 3.0, |
| "dit_config": { |
| "uniform_parameter_dtype": false |
| }, |
| "video_scheduler_shift": 12.0 |
| }, |
| "training": { |
| "checkpoint": { |
| "checkpointing_start_step": 100, |
| "checkpoints_total_limit": 2, |
| "inference_checkpoint_dtype": "bfloat16", |
| "inference_checkpoint_role": "student", |
| "inference_checkpoints_total_limit": 0, |
| "output_dir": "/mnt/beegfs/omniref-v4-training/outputs/omniref_v5_dmpdd_grid32_w03_cfg2_data_latent_b300_seed43_step1250_20261002T201404Z", |
| "require_complete_training_checkpoint": true, |
| "resume_from_checkpoint": "/mnt/beegfs/omniref-v4-training/outputs/omniref_v5_dmpdd_grid32_w1_data_latent_b300_seed43_step1250_20260930T063120Z/checkpoint-1800", |
| "save_inference_checkpoint_on_validation": true, |
| "training_state_checkpointing_steps": 100 |
| }, |
| "data": { |
| "data_path": "/mnt/beegfs/omniref-v5/encoded/full-b300-20260928T230836Z-r3/parquet", |
| "dataloader_num_workers": 0, |
| "fixed_sequence_capacities": [ |
| 44032, |
| 70144, |
| 102912, |
| 133888, |
| 221440 |
| ], |
| "max_presentation_tokens": 65536, |
| "native_shape_bucketing": true, |
| "preprocessed_data_type": "ref2va", |
| "require_existing_parquet_cache": true, |
| "seed": 43, |
| "train_batch_size": 1, |
| "training_cfg_rate": 0.0 |
| }, |
| "distributed": { |
| "hsdp_replicate_dim": 1, |
| "hsdp_shard_dim": 64, |
| "num_gpus": 64, |
| "sp_size": 4, |
| "tp_size": 1 |
| }, |
| "dit_precision": "fp32", |
| "loop": { |
| "gradient_accumulation_steps": 4, |
| "max_train_steps": 4000 |
| }, |
| "model": { |
| "enable_gradient_checkpointing_type": "full", |
| "enable_torch_compile": true, |
| "minimax_h3_accumulated_recompile_limit": 1024, |
| "minimax_h3_compile_limit_fallback": true, |
| "minimax_h3_student_compile": "grad_only_full_block", |
| "precondition_outputs": false, |
| "torch_compile_kwargs": { |
| "dynamic": true, |
| "fullgraph": true, |
| "recompile_limit": 128 |
| } |
| }, |
| "optimizer": { |
| "betas": [ |
| 0.9, |
| 0.999 |
| ], |
| "learning_rate": 5e-06, |
| "lr_scheduler": "constant", |
| "lr_warmup_steps": 0, |
| "weight_decay": 0.01 |
| }, |
| "tracker": { |
| "project_name": "h3-ref2va-dmd2", |
| "run_name": "omniref_v5_dmpdd_grid32_w03_cfg2_data_latent_b300_seed43_step1250_20261002T201404Z", |
| "trackers": [ |
| "wandb" |
| ] |
| }, |
| "vsa": { |
| "cache_tile_buf": false, |
| "ref_keep_rate": 0.1, |
| "sparsity": 0.9, |
| "tile_size": 128 |
| } |
| } |
| }, |
| "config_overrides": { |
| "pdd_steps": 32 |
| }, |
| "dtype": "bfloat16", |
| "format_version": 1, |
| "kind": "inference", |
| "lora": "none", |
| "max_shard_size_bytes": 5368709120, |
| "module": "transformer_ref", |
| "role": "student", |
| "shard_count": 14, |
| "shard_file_sizes": [ |
| 5296609992, |
| 5318721248, |
| 5318721240, |
| 4952636512, |
| 5318505992, |
| 5164580688, |
| 5318721240, |
| 5318721248, |
| 4875566056, |
| 5318505984, |
| 5164580688, |
| 5318721256, |
| 5318721208, |
| 2138941296 |
| ], |
| "shard_sizes": [ |
| 5296601856, |
| 5318715392, |
| 5318715392, |
| 4952630784, |
| 5318500352, |
| 5164574720, |
| 5318715392, |
| 5318715392, |
| 4875560448, |
| 5318500352, |
| 5164574720, |
| 5318715392, |
| 5318715392, |
| 2138938880 |
| ], |
| "step": 3500, |
| "tensor_count": 688, |
| "total_size": 70142174464 |
| } |
|
|
