CoCode-World-Model / trainer_state.json
iapple1's picture
Release CoCode-World step-4000 LoRA and EMA checkpoint
39be9df verified
Raw History Blame Contribute Delete
3.62 kB
{
"checkpoint_version": 1,
"metadata": {
"activation_cpu_offload": true,
"adam_betas": [
0.9,
0.99
],
"attention": "full_bidirectional",
"audio": "disabled",
"context_parallel_degree": 8,
"data_parallel_degree": 4,
"data_schema": "h3-ref-image-video-splitscreen-cache-v2",
"dataset_cursor": 7448,
"dataset_epoch": 1,
"dataset_manifest": "cache/train/manifest.jsonl",
"dataset_manifest_sha256": "8e581827ad4840c721a7771c259ee0fe6b4db7770c9924b87e3deb68be56e8fe",
"dataset_text_encoder_hash": "37b3c4f46ff752e9f3ee1a27e146a101ebcf4afda9251fc83eea4f4a23bf5c8a",
"dataset_transformer_ref_config_hash": "1d1486b61d9c11b92af41840a4a40b83de8a776db8cc0f4c71b40d4e4371c116",
"dataset_vae_config_hash": "76fcd9d9b23b522cd574d1e3adbf138b69f3f6e260e2944e3dbebb4d091cf3ec",
"decfg_adapter_conversion_revision": "official-h3-head-role-dim-qkv-split-swiglu-swap-v1",
"decfg_adapter_fusion_revision": "bf16-matmul-add-like-diffsynth-preset-v1",
"decfg_adapter_model_id": "DiffSynth-Studio/MiniMax-H3-TrainingAdapter",
"decfg_adapter_path": "models/DiffSynth-Studio/MiniMax-H3-TrainingAdapter/model_ref2va.safetensors",
"decfg_adapter_release_revision": "bdddb151d0dfb62b1dfa9e2f8ae6a6e1b179e098",
"decfg_adapter_scale": 1.0,
"decfg_adapter_sha256": "9361683349c1bf78f206c382d91b1e01b91fcddc2408e85ab39c9faa51f0d3d1",
"decfg_adapter_usage": "frozen_training_preset_excluded_from_task_lora_and_inference",
"depth_control": "removed",
"ema_decay": 0.999,
"epoch_tail_policy": "continuous_global_sample_stream",
"global_samples_seen": 16000,
"global_world_size": 32,
"gradient_accumulation": 1,
"learning_rate": 1e-05,
"lora_alpha": 64.0,
"lora_rank": 64,
"lora_targets": "qkvo-ffn",
"lr_schedule": "linear_warmup_then_constant",
"max_gradient_norm": 1.0,
"model_root": "models/MiniMax-H3",
"objective": "target_rows_h3_flow_velocity_mse",
"reference_audio": "disabled",
"reference_image_frozen": true,
"reference_kinds": [
"image",
"video"
],
"reference_noise_aug": 0.999,
"reference_order": "target_first_frame_then_proxy_video",
"sample_order_revision": "epoch-buffer-shuffle-seed-plus-epoch-v1",
"seed": 42,
"shuffle_buffer_size": 32768,
"split_screen": "agent_000_left_agent_001_right",
"target_canvas": [
736,
2560
],
"text_encoder_execution_revision": "qwen3vl-hidden50-hook-layer49-inference-retain-0-through-49-v2",
"text_encoder_layer": 50,
"time_distribution": "uniform_noise_ratio_then_shift",
"transformer_partition": "transformer_ref",
"transformer_ref_config_hash": "1d1486b61d9c11b92af41840a4a40b83de8a776db8cc0f4c71b40d4e4371c116",
"video_shift": 12.0,
"warmup_steps": 200,
"weight_decay": 0.0
},
"parameter_names_sha256": "ba3f0ddd1e25c79475e81da156f0b15f9caf6ecbc7e4675b526c99c925526231",
"payloads": {
"ema.safetensors": {
"bytes": 1383687680,
"sha256": "9d4e0524ee2b9203c2711f2c0b2a133e17d9ea6cb1280aa9974150f0ba7a038c"
},
"lora.safetensors": {
"bytes": 691889848,
"sha256": "e0d4a3cb1c8cb8cb8ef36cefeb45ecc164556404f4003ec39b1af8f435ca602c"
},
"optimizer.pt": {
"bytes": 1384124083,
"sha256": "cfae773db2796b4973eb28d37221eb1ff89f2c3dccdcc09e1a55a2c2b25ff4b1"
},
"rng_state.pt": {
"bytes": 6901,
"sha256": "e4fe9530c60967ab99cab3175013b526f5ec56bbd7e358fe69d9e8f2209f5d71"
}
},
"step": 4000,
"trainable_parameters": 345899008,
"trainable_tensors": 624
}