{
    "training": {
        "wandb_name": "8n_25hz_v45_infill_shared_flow_ns_sfx_ft",
        "wandb_project": "sfx_diff_train",
        "tot_num_steps": 10000000,
        "batch_size": 4,
        "ckpt_every" : 10000,
        "learning_rate": 5e-05,
        "lr_warmup": 10000,
        "lr_scheduler": "constant",
        "model_type": "prefix",
        "weight_decay": 0.001,
        "clip_grad_norm": 0.5,
        "diffusion_objective": "rectified_flow",
        "noise_schedule": "log_snr",
        "betas": [
            0.9,
            0.999
        ],
        "seed_offset": 0,
        "compute_metrics": false,
        "use_ema": true,
        "activation_checkpointing": false,
        "compile": true,
        "preload_checkpoint": "/app2/suno/checkpoints/2025-06-23_12-14-56_s5712/last_ckpt_infer.pt",
        "preload_optimizer": false,
        "check_model_args": false,
        "num_workers": 4
    },
    "model": {
        "io_hz": 25,
        "io_channels": 128,
        "embed_dim": 2048,
        "depth": 32,
        "n_heads": 32,
        "qk_norm": true,
        "block_size": 375,
        "cond_semantic_n_vocab": 4001,
        "cond_semantic_len": 375,
        "cond_text_n_vocab": 60001,
        "cond_text_len": 1536,
        "ctx_len": 375,
        "infill_ctx_len": 0,
        "shared_ctx": true
    },
    "data": {
        "shard_data": true,
        "shard_data_dir": "/mnt/localdisk/tmp",
        "allow_shard_reuse": false,
        "dataset_type": "general_memmap",
        "dataset_dir": "/app2/suno/data/diffusion/sfx/v0",
        "train_vae_memmap_filename": "data_vae_tr.bin",
        "train_semantic_memmap_filename": null,
        "train_metas_filename": "metas_tr.jsonl",
        "val_vae_memmap_filename": "data_vae_val.bin",
        "val_semantic_memmap_filename": null,
        "val_metas_filename": "metas_val.jsonl",
        "vae_dim": 128,
        "semantic_rate_hz": 25,
        "vae_scale_factor": 0.4,
        "foreign_weight" : 1.0,
        "scale_vae_ctx" : true,
        "noise_ctx" : 0.0,
        "semantic_skip_factors" : [],
        "prev_vae_ctx": true,
        "infill_vae_ctx": true,
        "always_pad_semantic": true,
        "aligned_text_prob": 0.0
    }
}