Hugging Face's logo

Efficient-Large-Model
/
LongLive-Plug-MiniMax-H3-few-step

Text-to-Video
PEFT
lora
distillation
few-step
Model card Files
xet
 Community
2
LongLive-Plug-MiniMax-H3-few-step / training_config.json
Andyson's picture
Andyson
Duplicate from Perflow-Shuai/MiniMax-H3-LongLive-LoRA-4Step-Audio2-PromptMix-iter1000
d651476
Raw History Blame Contribute Delete
36.1 kB
{
"schema_version": 1,
"model_id": "/lustre/fs1/portfolios/nvr/projects/nvr_elm_llm/users/shyang/LongLive-LoRA/Minimax-H3/training/LongLive-LoRA/models/MiniMax-H3-Base-T2VA-fa9c8ab1",
"requested_revision": "fa9c8ab1eaa21c8ae25e7e40b83b2e6002f340af",
"resolved_revision": "fa9c8ab1eaa21c8ae25e7e40b83b2e6002f340af",
"workflow": "t2va",
"ir": false,
"regenerator": false,
"cfg": false,
"shared_base_roles": [
"student",
"critic",
"teacher"
],
"teacher_adapter_state": "disabled",
"student_adapter": "student",
"critic_adapter": "critic",
"lora_rank": 128,
"lora_alpha": 128,
"lora_targets": [
"token_refiner.refiner_blocks.0.attn.to_k",
"token_refiner.refiner_blocks.0.attn.to_out.0",
"token_refiner.refiner_blocks.0.attn.to_q",
"token_refiner.refiner_blocks.0.attn.to_v",
"token_refiner.refiner_blocks.0.ff.net.0.proj",
"token_refiner.refiner_blocks.0.ff.net.2",
"token_refiner.refiner_blocks.1.attn.to_k",
"token_refiner.refiner_blocks.1.attn.to_out.0",
"token_refiner.refiner_blocks.1.attn.to_q",
"token_refiner.refiner_blocks.1.attn.to_v",
"token_refiner.refiner_blocks.1.ff.net.0.proj",
"token_refiner.refiner_blocks.1.ff.net.2",
"transformer_blocks.0.attn.to_k",
"transformer_blocks.0.attn.to_out.0",
"transformer_blocks.0.attn.to_q",
"transformer_blocks.0.attn.to_v",
"transformer_blocks.0.ff.net.0.proj",
"transformer_blocks.0.ff.net.2",
"transformer_blocks.1.attn.to_k",
"transformer_blocks.1.attn.to_out.0",
"transformer_blocks.1.attn.to_q",
"transformer_blocks.1.attn.to_v",
"transformer_blocks.1.ff.net.0.proj",
"transformer_blocks.1.ff.net.2",
"transformer_blocks.10.attn.to_k",
"transformer_blocks.10.attn.to_out.0",
"transformer_blocks.10.attn.to_q",
"transformer_blocks.10.attn.to_v",
"transformer_blocks.10.ff.net.0.proj",
"transformer_blocks.10.ff.net.2",
"transformer_blocks.11.attn.to_k",
"transformer_blocks.11.attn.to_out.0",
"transformer_blocks.11.attn.to_q",
"transformer_blocks.11.attn.to_v",
"transformer_blocks.11.ff.net.0.proj",
"transformer_blocks.11.ff.net.2",
"transformer_blocks.12.attn.to_k",
"transformer_blocks.12.attn.to_out.0",
"transformer_blocks.12.attn.to_q",
"transformer_blocks.12.attn.to_v",
"transformer_blocks.12.ff.net.0.proj",
"transformer_blocks.12.ff.net.2",
"transformer_blocks.13.attn.to_k",
"transformer_blocks.13.attn.to_out.0",
"transformer_blocks.13.attn.to_q",
"transformer_blocks.13.attn.to_v",
"transformer_blocks.13.ff.net.0.proj",
"transformer_blocks.13.ff.net.2",
"transformer_blocks.14.attn.to_k",
"transformer_blocks.14.attn.to_out.0",
"transformer_blocks.14.attn.to_q",
"transformer_blocks.14.attn.to_v",
"transformer_blocks.14.ff.net.0.proj",
"transformer_blocks.14.ff.net.2",
"transformer_blocks.15.attn.to_k",
"transformer_blocks.15.attn.to_out.0",
"transformer_blocks.15.attn.to_q",
"transformer_blocks.15.attn.to_v",
"transformer_blocks.15.ff.net.0.proj",
"transformer_blocks.15.ff.net.2",
"transformer_blocks.16.attn.to_k",
"transformer_blocks.16.attn.to_out.0",
"transformer_blocks.16.attn.to_q",
"transformer_blocks.16.attn.to_v",
"transformer_blocks.16.ff.net.0.proj",
"transformer_blocks.16.ff.net.2",
"transformer_blocks.17.attn.to_k",
"transformer_blocks.17.attn.to_out.0",
"transformer_blocks.17.attn.to_q",
"transformer_blocks.17.attn.to_v",
"transformer_blocks.17.ff.net.0.proj",
"transformer_blocks.17.ff.net.2",
"transformer_blocks.18.attn.to_k",
"transformer_blocks.18.attn.to_out.0",
"transformer_blocks.18.attn.to_q",
"transformer_blocks.18.attn.to_v",
"transformer_blocks.18.ff.net.0.proj",
"transformer_blocks.18.ff.net.2",
"transformer_blocks.19.attn.to_k",
"transformer_blocks.19.attn.to_out.0",
"transformer_blocks.19.attn.to_q",
"transformer_blocks.19.attn.to_v",
"transformer_blocks.19.ff.net.0.proj",
"transformer_blocks.19.ff.net.2",
"transformer_blocks.2.attn.to_k",
"transformer_blocks.2.attn.to_out.0",
"transformer_blocks.2.attn.to_q",
"transformer_blocks.2.attn.to_v",
"transformer_blocks.2.ff.net.0.proj",
"transformer_blocks.2.ff.net.2",
"transformer_blocks.20.attn.to_k",
"transformer_blocks.20.attn.to_out.0",
"transformer_blocks.20.attn.to_q",
"transformer_blocks.20.attn.to_v",
"transformer_blocks.20.ff.net.0.proj",
"transformer_blocks.20.ff.net.2",
"transformer_blocks.21.attn.to_k",
"transformer_blocks.21.attn.to_out.0",
"transformer_blocks.21.attn.to_q",
"transformer_blocks.21.attn.to_v",
"transformer_blocks.21.ff.net.0.proj",
"transformer_blocks.21.ff.net.2",
"transformer_blocks.22.attn.to_k",
"transformer_blocks.22.attn.to_out.0",
"transformer_blocks.22.attn.to_q",
"transformer_blocks.22.attn.to_v",
"transformer_blocks.22.ff.net.0.proj",
"transformer_blocks.22.ff.net.2",
"transformer_blocks.23.attn.to_k",
"transformer_blocks.23.attn.to_out.0",
"transformer_blocks.23.attn.to_q",
"transformer_blocks.23.attn.to_v",
"transformer_blocks.23.ff.net.0.proj",
"transformer_blocks.23.ff.net.2",
"transformer_blocks.24.attn.to_k",
"transformer_blocks.24.attn.to_out.0",
"transformer_blocks.24.attn.to_q",
"transformer_blocks.24.attn.to_v",
"transformer_blocks.24.ff.net.0.proj",
"transformer_blocks.24.ff.net.2",
"transformer_blocks.25.attn.to_k",
"transformer_blocks.25.attn.to_out.0",
"transformer_blocks.25.attn.to_q",
"transformer_blocks.25.attn.to_v",
"transformer_blocks.25.ff.net.0.proj",
"transformer_blocks.25.ff.net.2",
"transformer_blocks.26.attn.to_k",
"transformer_blocks.26.attn.to_out.0",
"transformer_blocks.26.attn.to_q",
"transformer_blocks.26.attn.to_v",
"transformer_blocks.26.ff.net.0.proj",
"transformer_blocks.26.ff.net.2",
"transformer_blocks.27.attn.to_k",
"transformer_blocks.27.attn.to_out.0",
"transformer_blocks.27.attn.to_q",
"transformer_blocks.27.attn.to_v",
"transformer_blocks.27.ff.net.0.proj",
"transformer_blocks.27.ff.net.2",
"transformer_blocks.28.attn.to_k",
"transformer_blocks.28.attn.to_out.0",
"transformer_blocks.28.attn.to_q",
"transformer_blocks.28.attn.to_v",
"transformer_blocks.28.ff.net.0.proj",
"transformer_blocks.28.ff.net.2",
"transformer_blocks.29.attn.to_k",
"transformer_blocks.29.attn.to_out.0",
"transformer_blocks.29.attn.to_q",
"transformer_blocks.29.attn.to_v",
"transformer_blocks.29.ff.net.0.proj",
"transformer_blocks.29.ff.net.2",
"transformer_blocks.3.attn.to_k",
"transformer_blocks.3.attn.to_out.0",
"transformer_blocks.3.attn.to_q",
"transformer_blocks.3.attn.to_v",
"transformer_blocks.3.ff.net.0.proj",
"transformer_blocks.3.ff.net.2",
"transformer_blocks.30.attn.to_k",
"transformer_blocks.30.attn.to_out.0",
"transformer_blocks.30.attn.to_q",
"transformer_blocks.30.attn.to_v",
"transformer_blocks.30.ff.net.0.proj",
"transformer_blocks.30.ff.net.2",
"transformer_blocks.31.attn.to_k",
"transformer_blocks.31.attn.to_out.0",
"transformer_blocks.31.attn.to_q",
"transformer_blocks.31.attn.to_v",
"transformer_blocks.31.ff.net.0.proj",
"transformer_blocks.31.ff.net.2",
"transformer_blocks.32.attn.to_k",
"transformer_blocks.32.attn.to_out.0",
"transformer_blocks.32.attn.to_q",
"transformer_blocks.32.attn.to_v",
"transformer_blocks.32.ff.net.0.proj",
"transformer_blocks.32.ff.net.2",
"transformer_blocks.33.attn.to_k",
"transformer_blocks.33.attn.to_out.0",
"transformer_blocks.33.attn.to_q",
"transformer_blocks.33.attn.to_v",
"transformer_blocks.33.ff.net.0.proj",
"transformer_blocks.33.ff.net.2",
"transformer_blocks.34.attn.to_k",
"transformer_blocks.34.attn.to_out.0",
"transformer_blocks.34.attn.to_q",
"transformer_blocks.34.attn.to_v",
"transformer_blocks.34.ff.net.0.proj",
"transformer_blocks.34.ff.net.2",
"transformer_blocks.35.attn.to_k",
"transformer_blocks.35.attn.to_out.0",
"transformer_blocks.35.attn.to_q",
"transformer_blocks.35.attn.to_v",
"transformer_blocks.35.ff.net.0.proj",
"transformer_blocks.35.ff.net.2",
"transformer_blocks.36.attn.to_k",
"transformer_blocks.36.attn.to_out.0",
"transformer_blocks.36.attn.to_q",
"transformer_blocks.36.attn.to_v",
"transformer_blocks.36.ff.net.0.proj",
"transformer_blocks.36.ff.net.2",
"transformer_blocks.37.attn.to_k",
"transformer_blocks.37.attn.to_out.0",
"transformer_blocks.37.attn.to_q",
"transformer_blocks.37.attn.to_v",
"transformer_blocks.37.ff.net.0.proj",
"transformer_blocks.37.ff.net.2",
"transformer_blocks.38.attn.to_k",
"transformer_blocks.38.attn.to_out.0",
"transformer_blocks.38.attn.to_q",
"transformer_blocks.38.attn.to_v",
"transformer_blocks.38.ff.net.0.proj",
"transformer_blocks.38.ff.net.2",
"transformer_blocks.39.attn.to_k",
"transformer_blocks.39.attn.to_out.0",
"transformer_blocks.39.attn.to_q",
"transformer_blocks.39.attn.to_v",
"transformer_blocks.39.ff.net.0.proj",
"transformer_blocks.39.ff.net.2",
"transformer_blocks.4.attn.to_k",
"transformer_blocks.4.attn.to_out.0",
"transformer_blocks.4.attn.to_q",
"transformer_blocks.4.attn.to_v",
"transformer_blocks.4.ff.net.0.proj",
"transformer_blocks.4.ff.net.2",
"transformer_blocks.40.attn.to_k",
"transformer_blocks.40.attn.to_out.0",
"transformer_blocks.40.attn.to_q",
"transformer_blocks.40.attn.to_v",
"transformer_blocks.40.ff.net.0.proj",
"transformer_blocks.40.ff.net.2",
"transformer_blocks.41.attn.to_k",
"transformer_blocks.41.attn.to_out.0",
"transformer_blocks.41.attn.to_q",
"transformer_blocks.41.attn.to_v",
"transformer_blocks.41.ff.net.0.proj",
"transformer_blocks.41.ff.net.2",
"transformer_blocks.42.attn.to_k",
"transformer_blocks.42.attn.to_out.0",
"transformer_blocks.42.attn.to_q",
"transformer_blocks.42.attn.to_v",
"transformer_blocks.42.ff.net.0.proj",
"transformer_blocks.42.ff.net.2",
"transformer_blocks.43.attn.to_k",
"transformer_blocks.43.attn.to_out.0",
"transformer_blocks.43.attn.to_q",
"transformer_blocks.43.attn.to_v",
"transformer_blocks.43.ff.net.0.proj",
"transformer_blocks.43.ff.net.2",
"transformer_blocks.44.attn.to_k",
"transformer_blocks.44.attn.to_out.0",
"transformer_blocks.44.attn.to_q",
"transformer_blocks.44.attn.to_v",
"transformer_blocks.44.ff.net.0.proj",
"transformer_blocks.44.ff.net.2",
"transformer_blocks.45.attn.to_k",
"transformer_blocks.45.attn.to_out.0",
"transformer_blocks.45.attn.to_q",
"transformer_blocks.45.attn.to_v",
"transformer_blocks.45.ff.net.0.proj",
"transformer_blocks.45.ff.net.2",
"transformer_blocks.46.attn.to_k",
"transformer_blocks.46.attn.to_out.0",
"transformer_blocks.46.attn.to_q",
"transformer_blocks.46.attn.to_v",
"transformer_blocks.46.ff.net.0.proj",
"transformer_blocks.46.ff.net.2",
"transformer_blocks.47.attn.to_k",
"transformer_blocks.47.attn.to_out.0",
"transformer_blocks.47.attn.to_q",
"transformer_blocks.47.attn.to_v",
"transformer_blocks.47.ff.net.0.proj",
"transformer_blocks.47.ff.net.2",
"transformer_blocks.48.attn.to_k",
"transformer_blocks.48.attn.to_out.0",
"transformer_blocks.48.attn.to_q",
"transformer_blocks.48.attn.to_v",
"transformer_blocks.48.ff.net.0.proj",
"transformer_blocks.48.ff.net.2",
"transformer_blocks.49.attn.to_k",
"transformer_blocks.49.attn.to_out.0",
"transformer_blocks.49.attn.to_q",
"transformer_blocks.49.attn.to_v",
"transformer_blocks.49.ff.net.0.proj",
"transformer_blocks.49.ff.net.2",
"transformer_blocks.5.attn.to_k",
"transformer_blocks.5.attn.to_out.0",
"transformer_blocks.5.attn.to_q",
"transformer_blocks.5.attn.to_v",
"transformer_blocks.5.ff.net.0.proj",
"transformer_blocks.5.ff.net.2",
"transformer_blocks.6.attn.to_k",
"transformer_blocks.6.attn.to_out.0",
"transformer_blocks.6.attn.to_q",
"transformer_blocks.6.attn.to_v",
"transformer_blocks.6.ff.net.0.proj",
"transformer_blocks.6.ff.net.2",
"transformer_blocks.7.attn.to_k",
"transformer_blocks.7.attn.to_out.0",
"transformer_blocks.7.attn.to_q",
"transformer_blocks.7.attn.to_v",
"transformer_blocks.7.ff.net.0.proj",
"transformer_blocks.7.ff.net.2",
"transformer_blocks.8.attn.to_k",
"transformer_blocks.8.attn.to_out.0",
"transformer_blocks.8.attn.to_q",
"transformer_blocks.8.attn.to_v",
"transformer_blocks.8.ff.net.0.proj",
"transformer_blocks.8.ff.net.2",
"transformer_blocks.9.attn.to_k",
"transformer_blocks.9.attn.to_out.0",
"transformer_blocks.9.attn.to_q",
"transformer_blocks.9.attn.to_v",
"transformer_blocks.9.ff.net.0.proj",
"transformer_blocks.9.ff.net.2"
],
"rollout_mode": "predict_x0_fresh_renoise",
"attention_mode": "native_non_causal",
"attention_is_causal": false,
"temporal_ar_blocks": 0,
"model_evaluations_per_rollout": 4,
"sigma_endpoints_per_rollout": 5,
"student_positive_forwards_per_rollout": 4,
"transformer_block_count": 50,
"token_refiner_block_count": 2,
"normalization": "per_modality_full_sequence",
"cuda_cache_policy": "empty_cache_before_generator_backward",
"cuda_empty_cache_before_generator_backward": true,
"cuda_allocator_config": "expandable_segments:True",
"saved_tensor_offload_policy": "student_rollout_save_on_cpu",
"saved_tensor_offload_pin_memory": false,
"video_shift": 12,
"audio_shift": 3,
"height": 768,
"width": 1344,
"num_frames": 124,
"strategy": "fsdp",
"gradient_checkpointing": true,
"seed": 20260812,
"student_lr": 0.00002,
"critic_lr": 0.00005,
"max_steps": 1500,
"prompt_start_index": 0,
"prompt_cache": "/lustre/fs1/portfolios/nvr/projects/nvr_elm_llm/users/shyang/LongLive-LoRA/Minimax-H3/training/LongLive-LoRA/artifacts/minimax_h3/prompt-cache-v5-w16-s2000/collection.json",
"prompt_cache_sha256": "590c072c951049c660ad8cbe3399746d0224029eb1ed4083119a9b50e4eb8e46",
"fsdp_sharding": "FULL_SHARD",
"fsdp_use_orig_params": true,
"fsdp_auto_wrap_classes": [
"MiniMaxH3TransformerBlock",
"MiniMaxH3TokenRefinerBlock"
],
"fsdp_auto_wrap_policy": "h3_peft_attention_ffn_base_projection_v8",
"fsdp_top_level_wrap_modules": [
"proj_in",
"audio_proj_in",
"context_embedder",
"time_embedder",
"token_refiner",
"norm_out",
"proj_out",
"audio_proj_out"
],
"fsdp_inner_wrap_modules": [
"attn"
],
"fsdp_attention_projection_wrap_modules": [
"to_q",
"to_k",
"to_v",
"to_out.0"
],
"fsdp_attention_base_wrap_module": "base_layer",
"fsdp_ffn_projection_wrap_modules": [
"net.0.proj",
"net.2"
],
"fsdp_handle_uniform_dtype_required": true,
"fsdp_root_parameter_tensors_expected": 0,
"fsdp_cpu_offload_policy": "frozen_h3_base_handles_cpu_lora_cuda",
"fsdp_cpu_offload_params": true,
"fsdp_limit_all_gathers": true,
"fsdp_backward_prefetch": "disabled",
"flex_attention_backend": "native",
"fsdp_sync_module_states": false,
"longrun_schema_version": 12,
"total_optimizer_steps": 1500,
"checkpoint_every": 100,
"critic_update_every": 1,
"student_update_ratio": 5,
"student_update_rule": "(global_step-1)%5==0",
"adam_betas": [
0,
0.9
],
"student_betas": [
0,
0.9
],
"critic_betas": [
0.9,
0.999
],
"weight_decay": 0.01,
"lora_dropout": 0,
"max_grad_norm": 1,
"sample_rng_scheme": "named_rank_local_outer_step_substreams_v1",
"joint_optimizer": "AdamW",
"optimizer_layout": "joint_role_groups_per_role_adamw_v2",
"optimizer_param_group_roles": [
"student",
"critic"
],
"longrun_config_version": 12,
"checkpoint_schema_version": 2,
"lora_target_policy": "h3_full_50_main_plus_2_token_refiner_attention_ffn_v1",
"lora_target_scope": "50_main_plus_2_token_refiner_blocks",
"lora_target_branches": [
"attn",
"ff"
],
"lora_target_module_count": 312,
"lora_excluded_scopes": [
"adaln_and_timestep",
"normalization",
"top_level_input_context_output_projections"
],
"prompt_runtime_selection": {
"policy": "rank_balanced_50pct_authored_audio_v1",
"rank_cache_loader": "minimax_h3.longlive_prompt_distributed.load_rank_prompt_cache",
"index_formula": "global_step-1",
"absolute_global_step_resume_guard": true,
"dataset_rows": 248221,
"sampler_seed_shared_across_ranks": 20260806,
"sampler_epoch": 0,
"sampler_shuffle": true,
"sampler_drop_last": true,
"sampler_rank_epoch_rows": 15513,
"sampler_epoch_dropped_rows": 13,
"world_size": 16,
"conditions_per_rank": 2000,
"scheduled_conditions": 32000,
"shared_prompt_across_ranks": false,
"rank_distinct_noise": true,
"critic_conditions_per_rank": 1500,
"student_update_rule": "(global_step-1)%5==0",
"student_conditions_per_rank": 300,
"student_uses_same_rank_condition_as_critic": true,
"source_original_cache_conditions": 32000,
"used_outer_condition_slots": 24000,
"used_authored_outer_slots": 12000,
"used_authored_student_slots": 2400,
"original_condition_seed_retained": true,
"text_selection_uses_rng": false,
"distinct_prompt_text_per_rank_verified": false,
"rank_distinct_condition_ids": true
},
"technical_document": {
"path": "/lustre/fs1/portfolios/nvr/projects/nvr_elm_llm/users/shyang/LongLive-LoRA/Minimax-H3/MINIMAX_H3_DMD_TECHNICAL_DOCUMENT_ZH.md",
"sha256": "e293cd2045f26ee6da47d32b440ee10db2a540eeaa4626e5e8b73a7f0318392b",
"documented_steps": 4,
"requested_steps": 4,
"grid_adaptation": "none_exact_document_four_step_grid"
},
"semantic_role_map": {
"generator": "student",
"fake_score": "critic",
"real_score_teacher": "adapters_disabled"
},
"model_variant": "standard_full_adaln",
"adaln_curve_rank": null,
"student_weight_decay": 0.01,
"critic_weight_decay": 0,
"lr_scheduler": "constant",
"lora_scale": 1,
"lora_master_dtype": "float32",
"fresh_noise_transitions_per_rollout": 3,
"target_stage_distribution": "uniform_0_through_3",
"outer_step_shared_random_variables": [
"target_stage",
"score_base_sigma"
],
"score_base_sigma_min": 0.02,
"score_base_sigma_max": 0.98,
"video_sigma_grid": [
1,
0.972972972972973,
0.9230769230769231,
0.8,
0
],
"audio_sigma_grid": [
1,
0.9,
0.75,
0.5,
0
],
"fresh_noise_draw_order": [
"video",
"audio"
],
"video_audio_transition_noise_independent": true,
"sample_rng_substreams": [
"detached_initial_latents",
"generator_initial_latents",
"shared_stage_sigma",
"detached_rollout",
"fake_dsm",
"generator_rollout",
"generator_dmd"
],
"sampler_matches_training_and_inference": true,
"teacher_trajectory_regression": true,
"real_latent_regression": false,
"gan_loss": false,
"ema": false,
"loss_objective": "audio_quality_audio_mixed_endpoint_latent",
"video_loss_weight": 1,
"audio_loss_weight": 2,
"audio_tokens_retained": true,
"audio_forward_retained": true,
"audio_raw_loss_metric_retained": true,
"audio_output_retained": true,
"rng_and_forward_contract_unchanged": false,
"student_rng_and_forward_contract_unchanged": true,
"gradient_checkpointing_scope": "main_transformer_blocks",
"gradient_checkpointing_policy": "main_transformer_blocks_reentrant",
"gradient_checkpointing_use_reentrant": true,
"token_refiner_gradient_checkpointing": false,
"attention_backend": "pytorch_native_flash_sdpa_forced",
"attention_backend_dispatch_name": "_native_flash",
"attention_requires_padless_training_sequence": true,
"layout_cache_policy": "per_step_rebuild_release_no_gpu_cache",
"distributed_backend": "FSDP_FULL_SHARD",
"documented_distributed_backend": "ZeRO-3",
"distributed_backend_deviation_reason": "The validated local 16-rank launcher uses equivalent full parameter, gradient, and optimizer sharding through FSDP FULL_SHARD.",
"data_deviation": "Half of each global batch retains original LongLive text; half uses separately authored audio-conditioned text. Other endpoint-latent settings are unchanged.",
"distributed_contract": {
"nodes": 4,
"gpus_per_node": 4,
"world_size": 16,
"per_rank_batch_size": 1,
"global_batch_size": 16,
"global_steps": 1500,
"scheduled_conditions": 24000
},
"experiment_id": "audio-data-mixed-fresh-20260914",
"teacher_video_cfg": 1,
"teacher_audio_cfg": 1,
"teacher_reference_cache": "/lustre/fs1/portfolios/nvr/projects/nvr_elm_llm/users/shyang/LongLive-LoRA/Minimax-H3/experiments/audio_cfg_1600_20260911/pilot_outputs/reference-rank0.pt",
"teacher_reference_cache_sha256": "902fd97f6ced7146ebaeb790bd2d5ef89173594ef1f9a650099a0bbeeca342d3",
"teacher_reference_kind": "whitespace_conditioned_not_calibrated_unconditional",
"extra_guidance_target": "real_score_teacher_only",
"cfg_enabled": false,
"cfg_scope": "teacher_only_student_inference_is_single_pass",
"student_extra_cfg": false,
"critic_extra_cfg": false,
"used_conditions_per_rank": 1500,
"used_total_conditions": 24000,
"source_cache_conditions_per_rank": 2000,
"training_source_sha256": {
"__init__.py": "7928c5364f079a90f05d14d9ef8c3949173f1e5caf9010f08779793b3fa50291",
"build_prompt_cache.py": "5ed06cb54ca28bc776d92f6da13b37aaa5baccb9e7e537fbcd37282f871cae98",
"cfg_guidance.py": "ee796ccfc5a03aef1e98054eaf30585d74b190e54308ba1ca192e2373e06df24",
"dmd_4step.py": "3d72c4dd3b3793f91fdcce3d37a809e18be805889fd3cc4bf57d96426eef8d0f",
"dmd_8step.py": "77ab7e2c7ce40cad12b141b1b24467406ff66a7fa420b5a752bdc5d5b74dc2b1",
"dmd_core.py": "4c2cc6909d41356d27c95275a0e32a5a3d96038634cb83e42a270424f6a86164",
"encode_longlive_prompt_cache.py": "7bdd4004bf4973158bac9866975c08b825579ee279c45916bd7789eacd57dc42",
"encode_longlive_prompt_cache_v4.py": "b39b37977235320b541e54d78833088cabd3199c8fe77059399ea8438cabcb29",
"extract_dmd8_inline5.py": "9f43ef1bad719a328f7ece6d14657534ec9cf76ebe91405760fed15d5bed821e",
"fresh_noise_scheduler.py": "404cbb1e6eb05e5a11222b95f86ba5eec88fb09718d32094a04dd4ac834fbcdb",
"fresh_noise_scheduler_4step.py": "ae6d4887471edb94096be0d65dfcd850131f819620d35450a26289dd90a5dc8e",
"full_h3_lora.py": "ad4b08ff2aae784452f5dd3468fc7e571cecb9665ada80ef5f60f04bf8cf4343",
"infer_base.py": "dd6a5ef8aad12ba76c65d74048726fd0f9623a116f216b4a300ef42272d09b93",
"infer_student.py": "c3dfa6456fcc1d49500c5a6445431e0254181f0d5342fa24353dc57e8c18c5d8",
"infer_student_4step.py": "3bbb180764ef744791065e172acf759b65f8ff7b8e85a3ac8cdc9ee59c738db3",
"infer_student_8step.py": "867252e18cb1c53c266d2a3a658453d63e7b9a38711a94ec36bcf88d7f944f5b",
"longlive_lora.py": "4d061332055b2b59dfa7632ca7502f98ecdace283333b28f872b1f74f2c1cb4e",
"longlive_prompt_distributed.py": "eff9f6e68cc17f4b0237990ac79478a45d5a6e846a4b9b812a508cc03faca48d",
"longlive_prompt_set.py": "ed80195faf3f1614c2ec5d3678298505eae00490456ebc365df995324cc33703",
"model_provenance.py": "8b5409504258657fdab9665dbd0a51a09e5503e468a9b0e7eb76ad0c2f55ab20",
"probe_student_lora_ab.py": "6ff009a751dc07c54cf8dedc98b4061f36080eec93c5068996c5bcec448f0150",
"train_dmd.py": "a2f2c71ead08514d96898019eee6c39536192eeb5b017bb2c7454d1cfb0e7172",
"train_dmd_longrun.py": "5843b7c57eea8c602bd06cae4580d4efae6fc15be4d370b4a2951cf60baf6ecb",
"train_dmd_longrun_4step.py": "5a751327012031d7a51e5004dc3d47094d7810e52ed72d05b04b513bedcee1bf",
"train_dmd_longrun_4step_autoresume.py": "42abbef8788b0d86ae6d2d6094bec6144d754d355748eefea6a7f189b7bd8918",
"train_dmd_longrun_8step.py": "2f0302f973ba0729040687741eed95edbf321bd4c85745ec1ca977fe91c9e7a0",
"train_dmd_longrun_8step_autoresume.py": "21b033ae059b73587257f6923db3587ce6b16c18aa4e8b6f5a91a615905d68dd",
"train_dmd_longrun_v2.py": "2ef1d89d400df60a86be10b8828ea7c4b87fb96112538bfd25e7baaf5e8bc357",
"train_dmd_longrun_v3.py": "ca1905890c1e3c849ccb7f189166d7d7ddaa22b9d6162c629beab8e0b5c65c81",
"train_dmd_longrun_v4.py": "e27328e359c6ff4fccd38623fb7eadb84c48f131fb547ca82bd7364c3f2dd44c",
"train_dmd_longrun_v5.py": "14778be13523721a38fb539d973c6d0ad7eff8054ab146a76adcb684bdd9c7cf",
"train_utils.py": "fd64c02074680470776fa9e8f58578bde7fdb90603d2cad3eab5e9216c6f8e9c",
"visualize_teacher_x0.py": "962f86b0fb699accecc36eb0b57fdc278835c22731f78f5e4d5fe36c3a33d49f"
},
"initialization": "original_Base_fa9c8ab1_new_student_and_critic_LoRA_A_random_B_zero",
"old_lora_warm_start": false,
"operational_resume_scope": "this_new_experiment_only",
"audio_objective": {
"dmd_weight": 1,
"teacher_endpoint_nfe": 50,
"endpoint_weight": 1,
"spectral_weight": 0,
"label": "新初始化 · 50% 音频描述数据 · DMD + 教师音频目标",
"endpoint_target": "50_frozen_Base_Euler_calls_from_exact_selected_student_noisy_AV_input_to_zero_sigma",
"endpoint_is_not_cached_Base50_output": true,
"endpoint_grid": "official_shared_base_sigma_subgrid_video_shift12_audio_shift3",
"spectral_ffts": [],
"spectral_hop": "n_fft/4",
"spectral_log_floor": 0.0001,
"spectral_decoder": null,
"waveform_gain_normalization": false,
"inference_postprocessing": false
},
"fresh_source_sha256": {
"train_fresh.py": "6b478c00a64c5bb405823960f541fa3d0faf4cc14b17db5cc762043a71ddec10",
"common.py": "60d1bd3d7774dca1ac1c54234ac05fd2fcb7a0ad1a01450d789fe882f7ad2ef7",
"audio_objectives.py": "4e22fd00a67f5e7e7ce03bec2607911079aad7129fc6b50c88cae3385bc1c605"
},
"milestone_evaluation_every": 200,
"pilot_training_budget_is_not_goal_completion": true,
"original_prompt_component": {
"prompt_dataset": {
"bytes": 145875068,
"line_count": 248221,
"path": "prompts/vidprom_filtered_extended.txt",
"repo_id": "Efficient-Large-Model/LongLive-1.3B",
"revision": "cda9138d93872ef109d9a74c270184eb3a5acc6e",
"row_policy": "nonempty_python_str_strip_then_python_str_rstrip",
"sha256": "7896742f468bc8aef9e4547424d1ce0a951acdb2a82233790155401a99bf5aa5"
},
"prompt_dataset_row_count": 248221,
"prompt_sampler": {
"batch_size_per_rank": 1,
"dataset_size": 248221,
"drop_last": true,
"epoch": 0,
"global_steps": 2000,
"name": "torch.utils.data.distributed.DistributedSampler",
"seed": 20260806,
"shuffle": true,
"world_size": 16
},
"prompt_schedule_audit": {
"dataset_size": 248221,
"global_steps": 2000,
"scheduled_rows": 32000,
"unique_source_indices": 32000,
"world_size": 16
},
"prompt_collection_provenance": {
"loader": "minimax_h3.longlive_prompt_distributed.load_prompt_cache_collection",
"path": "/lustre/fs1/portfolios/nvr/projects/nvr_elm_llm/users/shyang/LongLive-LoRA/Minimax-H3/training/LongLive-LoRA/artifacts/minimax_h3/prompt-cache-v5-w16-s2000/collection.json",
"bytes": 4539,
"sha256": "590c072c951049c660ad8cbe3399746d0224029eb1ed4083119a9b50e4eb8e46",
"payload_schema_version": 1,
"artifact_type": "minimax_h3_longlive_distributed_prompt_cache_collection",
"model_revision": "fa9c8ab1eaa21c8ae25e7e40b83b2e6002f340af",
"shards": [
{
"bytes": 2340850895,
"conditions": 2000,
"path": "rank-00000-of-00016.pt",
"rank": 0,
"sha256": "ac25de38c7aa2a1f0f8e0ccd76d662eeba514bc3e807b43877ab4b3f6c408bac"
},
{
"bytes": 2336362127,
"conditions": 2000,
"path": "rank-00001-of-00016.pt",
"rank": 1,
"sha256": "ef800ca248af17d3d05c10354b1e1fed2431f63c7fd7b6bfeb4cb9b40ca36025"
},
{
"bytes": 2339642127,
"conditions": 2000,
"path": "rank-00002-of-00016.pt",
"rank": 2,
"sha256": "e258544f5b2c1eb79912418445242f728d44bf0bc4026f3930e2bfa035e6b0ed"
},
{
"bytes": 2340954703,
"conditions": 2000,
"path": "rank-00003-of-00016.pt",
"rank": 3,
"sha256": "d9ef8adb270eca7923354dc9f6df0cb606604e6beb5aa750fdc463af0a41fd4d"
},
{
"bytes": 2333896015,
"conditions": 2000,
"path": "rank-00004-of-00016.pt",
"rank": 4,
"sha256": "12a94a253a5f9a94441e5c89c3c097cdfe665ea23200557b9caad7462b3952dd"
},
{
"bytes": 2337396047,
"conditions": 2000,
"path": "rank-00005-of-00016.pt",
"rank": 5,
"sha256": "c0f349ca80065522c529f40581b32c928b1e04dc4245a6dd54dbc52c4b184482"
},
{
"bytes": 2352031311,
"conditions": 2000,
"path": "rank-00006-of-00016.pt",
"rank": 6,
"sha256": "3e74eb0061eda77f3e620680fccea76cc1e596a871b27c9436a5cc964a11cfd6"
},
{
"bytes": 2346638735,
"conditions": 2000,
"path": "rank-00007-of-00016.pt",
"rank": 7,
"sha256": "51cccd347d6055dafc27c6734d81d23f93d202681d8d857ac3d40e2de0d052fe"
},
{
"bytes": 2336423759,
"conditions": 2000,
"path": "rank-00008-of-00016.pt",
"rank": 8,
"sha256": "b97823b091e0530b21a925f0b4103a8e022e9dd276422dd182079a2827d6a100"
},
{
"bytes": 2327823759,
"conditions": 2000,
"path": "rank-00009-of-00016.pt",
"rank": 9,
"sha256": "d5307b0250d532ac54484726d18e0f53fee3043164c6f634b26b196dd85be4f8"
},
{
"bytes": 2349843407,
"conditions": 2000,
"path": "rank-00010-of-00016.pt",
"rank": 10,
"sha256": "33e47a61beb520341cce9313ca278946c6d978a626186cc18a2bab4016700417"
},
{
"bytes": 2318440719,
"conditions": 2000,
"path": "rank-00011-of-00016.pt",
"rank": 11,
"sha256": "5810fcddac96e8e8ed91a0a3c60630b97aa03ccf5f4006ff00bf166917c13046"
},
{
"bytes": 2330704527,
"conditions": 2000,
"path": "rank-00012-of-00016.pt",
"rank": 12,
"sha256": "5815f9a772903857ead5d6d70a5c0135e73e475b545746a4a449f5e2dc2b1c4a"
},
{
"bytes": 2326816463,
"conditions": 2000,
"path": "rank-00013-of-00016.pt",
"rank": 13,
"sha256": "894effe10ed17b74d6b5d6dbd20e11da7d49385fbe8b99bb912d0983dc1a5298"
},
{
"bytes": 2327963919,
"conditions": 2000,
"path": "rank-00014-of-00016.pt",
"rank": 14,
"sha256": "3b6c62e0619f1e280720ea776b3b888b4b4882e7f75b032c140d160e805a3724"
},
{
"bytes": 2350203407,
"conditions": 2000,
"path": "rank-00015-of-00016.pt",
"rank": 15,
"sha256": "6fc63df45b01b08172fecb3b9229ea338b6083af62f3fdc5cfef1aacb287af4d"
}
],
"validate_only_full_shard_sha256_readback": true,
"production_manifest_verifies_all_shard_sizes": true,
"production_every_rank_verifies_own_shard_sha256": true
},
"prompt_runtime_selection": {
"policy": "rank_local_absolute_global_step_v1",
"rank_cache_loader": "minimax_h3.longlive_prompt_distributed.load_rank_prompt_cache",
"index_formula": "global_step-1",
"absolute_global_step_resume_guard": true,
"dataset_rows": 248221,
"sampler_seed_shared_across_ranks": 20260806,
"sampler_epoch": 0,
"sampler_shuffle": true,
"sampler_drop_last": true,
"sampler_rank_epoch_rows": 15513,
"sampler_epoch_dropped_rows": 13,
"world_size": 16,
"conditions_per_rank": 2000,
"scheduled_conditions": 32000,
"shared_prompt_across_ranks": false,
"rank_distinct_prompt": true,
"rank_distinct_noise": true,
"critic_conditions_per_rank": 1500,
"student_update_rule": "(global_step-1)%5==0",
"student_conditions_per_rank": 300,
"student_uses_same_rank_condition_as_critic": true
}
},
"prompt_dataset": {
"kind": "deterministic_mixture",
"original_fraction": 0.5,
"authored_audio_fraction": 0.5,
"authored_candidate_manifest": "/lustre/fs1/portfolios/nvr/projects/nvr_elm_llm/users/shyang/LongLive-LoRA/Minimax-H3/experiments/audio_fresh_quality_20260914/audio_data_candidates.json",
"authored_candidate_manifest_sha256": "4142bb4f21ba8ec2513e9839d8d4a793f6b4cca6580cc4e8bf09397351863f0d",
"authored_embedding_collection": "/lustre/fs1/portfolios/nvr/projects/nvr_elm_llm/users/shyang/LongLive-LoRA/Minimax-H3/experiments/audio_fresh_quality_20260914/audio_candidate_cache/COLLECTION.json",
"authored_embedding_collection_sha256": "dd3f32d488f0b7eb40d0f485c994d5c881c20e491c8e258b6c6b83fce55c90a3",
"authored_cache_contract": {
"artifact_type": "h3_authored_audio_candidate_embeddings_v1",
"schema_version": 1,
"model_revision": "fa9c8ab1eaa21c8ae25e7e40b83b2e6002f340af",
"workflow": "positive_text_encoding",
"IR": false,
"candidate_manifest_sha256": "4142bb4f21ba8ec2513e9839d8d4a793f6b4cca6580cc4e8bf09397351863f0d",
"qualified_teacher_receipt_sha256": "39eea24416458a42a59fa9cf38b3e36df1fa361058f713f55332d772fd673db3",
"qualification_scope": "Sampled transcript gate only, not perceptual or whole-pool quality.",
"sources": {
"audio_candidate_cache.py": "234d4a9cfad2952b2f98b76f70c77a46399483e9d6da5030ac80639e1fb4b616",
"encode_audio_candidates.py": "5adcb38b74cb83423ba39b6c690e4c9a9294f9a7cc0ed3796a83f7964352de18",
"audio_data_candidates.py": "28cebc27e0c8d914cd716401f9ebb9c03133c5803b362165ddb2e56aa7744274"
},
"official_encoder_sha256": "dc37019187d287d1346bb0a0f7d080bb9d1443ce1ae4b0f65dba57b40a75d8eb",
"text_dim": 5120
},
"authored_prompt_strings": 256,
"authored_core_scenes": 32,
"evaluation_text_and_reference_phrases_excluded": true,
"semantic_near_duplicate_exclusion_proven": false
},
"prompt_cache_argument_scope": "original_LongLive_component_and_preserved_noise_seed_schedule_only",
"data_trial_arm": "audio_mixed_endpoint_latent",
"matched_loss_recipe": "endpoint_latent",
"data_trial_source_sha256": {
"train_audio_mixed.py": "6193189881a1fef24144c4ddf57aa77965bf69308a1392724c826720e8d6c338",
"audio_mixed_conditions.py": "2b41ff3d1249ccdd7857d7b384476e8b65ec09593a44c810e931c6106d863dfd",
"audio_candidate_cache.py": "234d4a9cfad2952b2f98b76f70c77a46399483e9d6da5030ac80639e1fb4b616",
"audio_data_candidates.py": "28cebc27e0c8d914cd716401f9ebb9c03133c5803b362165ddb2e56aa7744274"
},
"training_budget_extension": {
"same_run_not_foreign_LoRA_warm_start": true,
"pilot_config_sha256": "7114d47ab94108ab0e54a972e96dd536d442ec5ce5dc406b41bf999440e19ab1",
"original_planned_outer_iterations": 400,
"extended_outer_iterations": 1500,
"transition_checkpoint_step": 50,
"transition_checkpoint_success_sha256": "a702baf7e06d8c823bce49399ee85126bc68a517a1e12d2dc486e8edc98db00d",
"optimizer_and_per_rank_rng_restored": true,
"checkpoint_payloads_are_never_rewritten": true,
"evaluation_outer_iterations": [
200,
400,
600,
800,
1000,
1200,
1400,
1500
],
"source_sha256": "309b8877de66a25c3ecdc85ffad46ad65aa007f978b04ed441fb4d855e37c40b",
"registry_source_sha256": "138bd3bc2a4f9fd7a386d55d1d4c6641e880d921b20182e76d8babbf6a68a66f"
},
"config_sha256": "916da7041fe91de966917d8600ac7c1ec64a9b9d4171c457151c687efeb7bf5a"
}