{
    "data": {
        "dataset_type": "speech_translation_multitask",
        "audio_encoder_name": "UsefulSensors/moonshine-streaming-medium",
        "data_path": "NhutP/VietSpeech",
        "src_lang": "",
        "tgt_lang": "",
        "max_length": 128,
        "packing": false,
        "src_column": "en",
        "tgt_column": "vi",
        "dedupe": false,
        "remove_wiki": false,
        "fix_ftfy": false,
        "normalize_punct": false,
        "detokenize": false,
        "remove_bracketed": false,
        "dereify": false,
        "task_tokens": [
            "<vi_en>",
            "<vi_zh>",
            "<vi_ko>"
        ],
        "precomputed_data_dir": "precomputed_data"
    },
    "generator": {
        "max_iterations": 10,
        "mbr": 1,
        "length_beam": 1,
        "oracle_length": true,
        "strategy": "reparam-uncond-deterministic-cosine",
        "argmax_decoding": true,
        "bpe": "sentencepiece",
        "bleu_tokenize": "13a",
        "return_history": false,
        "temperature": 1.0
    },
    "model": {
        "num_diffusion_timesteps": 50,
        "diffusion_type": "absorbing",
        "pretrained": "FacebookAI/xlm-roberta-base",
        "cache_dir": "cache",
        "prefix_lm": false,
        "attention_strategy": "full",
        "vocab_pad_to_multiple": 1,
        "lora": false,
        "lora_target_modules": [
            "query",
            "value"
        ],
        "lora_alpha": 16,
        "lora_rank": 16,
        "lora_bias": "none",
        "lora_dropout": 0
    },
    "train": {
        "output_dir": "outputs/vi_multitask",
        "per_device_train_batch_size": 128,
        "num_train_epochs": 3.0,
        "max_steps": 100000,
        "learning_rate": 0.0001,
        "lr_scheduler_type": "cosine",
        "warmup_steps": 0.1,
        "optim": "adamw_torch_fused",
        "weight_decay": 0.01,
        "adam_beta1": 0.9,
        "adam_beta2": 0.98,
        "adam_epsilon": 1e-08,
        "gradient_accumulation_steps": 1,
        "average_tokens_across_devices": true,
        "max_grad_norm": 1.0,
        "label_smoothing_factor": 0,
        "bf16": true,
        "fp16": false,
        "bf16_full_eval": true,
        "fp16_full_eval": false,
        "gradient_checkpointing": false,
        "torch_compile": false,
        "use_liger_kernel": false,
        "use_cache": false,
        "auto_find_batch_size": false,
        "logging_strategy": "steps",
        "logging_steps": 200,
        "logging_first_step": false,
        "log_on_each_node": true,
        "logging_nan_inf_filter": true,
        "include_num_input_tokens_seen": "no",
        "log_level": "info",
        "log_level_replica": "warning",
        "disable_tqdm": false,
        "report_to": [
            "wandb"
        ],
        "run_name": "vi_multitask_en_zh_ko",
        "project": "huggingface",
        "eval_strategy": "steps",
        "eval_steps": 5000,
        "eval_delay": 0,
        "per_device_eval_batch_size": 16,
        "prediction_loss_only": false,
        "eval_on_start": false,
        "eval_do_concat_batches": true,
        "eval_use_gather_object": false,
        "include_for_metrics": [],
        "batch_eval_metrics": false,
        "save_only_model": false,
        "save_strategy": "steps",
        "save_steps": 10000,
        "save_on_each_node": false,
        "save_total_limit": 3,
        "enable_jit_checkpoint": false,
        "push_to_hub": false,
        "hub_strategy": "every_save",
        "hub_always_push": false,
        "load_best_model_at_end": true,
        "metric_for_best_model": "wer",
        "greater_is_better": true,
        "ignore_data_skip": false,
        "restore_callback_states_from_checkpoint": false,
        "full_determinism": false,
        "seed": 42,
        "use_cpu": false,
        "dataloader_drop_last": false,
        "dataloader_num_workers": 16,
        "dataloader_pin_memory": true,
        "dataloader_persistent_workers": false,
        "remove_unused_columns": false,
        "train_sampling_strategy": "random",
        "length_column_name": "length",
        "ddp_timeout": 1800,
        "debug": [],
        "skip_memory_metrics": true,
        "do_train": false,
        "do_eval": true,
        "do_predict": false,
        "local_rank": -1,
        "batch_by_tokens": false,
        "eval_metric": "wer",
        "eval_metrics": [
            "bleu",
            "wer"
        ],
        "weighting": "linear",
        "mask_on_source": false,
        "mask_on_paddings": false,
        "mask_ratio_sampler": "diffusion",
        "train_length": false,
        "wandb_project": "diffusion-speech-recognition",
        "mixed_precision": "bf16",
        "_n_gpu": 1
    }
}