README.md

---
tags:
- espnet
- audio
- universa
language: multilingual
datasets:
- universa_unite
license: cc-by-4.0
---

## ESPnet2 universa model

### `espnet/arecho_scale_v0.1-large-decoder`

This model was trained by ftshijt using universa_unite recipe in [espnet](https://github.com/espnet/espnet/).

### Demo: How to use in ESPnet2

Follow the [ESPnet installation instructions](https://espnet.github.io/espnet/installation.html)
if you haven't done that already.

```bash
cd espnet
git checkout 0b68ffd26362f4b50e7c73942c5bbdbc0a220bd4
pip install -e .
cd egs2/universa_unite/uni_versa1
./run.sh --skip_data_prep false --skip_train true --download_model espnet/arecho_scale_v0.1-large-decoder
```



## universa config

<details><summary>expand</summary>

```
config: conf/train_aruniversa_wavlm_large.yaml
print_config: false
log_level: INFO
drop_last_iter: false
dry_run: false
iterator_type: sequence
valid_iterator_type: null
output_dir: exp/universa_universa_ar_overall_scale_token_wavlm_large
ngpu: 1
seed: 777
num_workers: 1
num_att_plot: 0
dist_backend: nccl
dist_init_method: env://
dist_world_size: null
dist_rank: null
local_rank: 0
dist_master_addr: null
dist_master_port: null
dist_launcher: null
multiprocessing_distributed: false
unused_parameters: false
sharded_ddp: false
use_deepspeed: false
deepspeed_config: null
gradient_as_bucket_view: true
ddp_comm_hook: null
cudnn_enabled: true
cudnn_benchmark: false
cudnn_deterministic: false
use_tf32: false
collect_stats: false
write_collected_feats: false
max_epoch: 100
patience: null
val_scheduler_criterion:
- valid
- loss
early_stopping_criterion:
- valid
- loss
- min
best_model_criterion:
-   - train
    - loss
    - min
-   - valid
    - loss
    - min
-   - train
    - acc
    - max
-   - valid
    - acc
    - max
keep_nbest_models: 1
nbest_averaging_interval: 0
grad_clip: -1
grad_clip_type: 2.0
grad_noise: false
accum_grad: 2
no_forward_run: false
resume: true
train_dtype: float32
use_amp: false
log_interval: 50
use_matplotlib: true
use_tensorboard: true
create_graph_in_tensorboard: false
use_wandb: false
wandb_project: null
wandb_id: null
wandb_entity: null
wandb_name: null
wandb_model_log_interval: -1
detect_anomaly: false
use_adapter: false
adapter: lora
save_strategy: all
adapter_conf: {}
pretrain_path: null
init_param: []
ignore_init_mismatch: false
freeze_param:
- frontend.upstream
num_iters_per_epoch: null
batch_size: 16
valid_batch_size: null
batch_bins: 1000000
valid_batch_bins: null
category_sample_size: 10
train_shape_file:
- exp/universa_stats_overall_scale/train/audio_shape
- exp/universa_stats_overall_scale/train/ref_audio_shape
valid_shape_file:
- exp/universa_stats_overall_scale/valid/audio_shape
- exp/universa_stats_overall_scale/valid/ref_audio_shape
batch_type: sorted
valid_batch_type: null
fold_length:
- 256000
sort_in_batch: descending
shuffle_within_batch: false
sort_batch: descending
multiple_iterator: false
chunk_length: 500
chunk_shift_ratio: 0.5
num_cache_chunks: 1024
chunk_excluded_key_prefixes: []
chunk_default_fs: null
chunk_max_abs_length: null
chunk_discard_short_samples: true
train_data_path_and_name_and_type:
-   - dump/raw/overall_scale/wav.scp
    - audio
    - kaldi_ark
-   - dump/raw/overall_scale/metric.scp
    - metrics
    - metric
-   - dump/raw/overall_scale/ref_wav.scp
    - ref_audio
    - kaldi_ark
valid_data_path_and_name_and_type:
-   - dump/raw/overall_dev/wav.scp
    - audio
    - kaldi_ark
-   - dump/raw/overall_dev/metric.scp
    - metrics
    - metric
-   - dump/raw/overall_dev/ref_wav.scp
    - ref_audio
    - kaldi_ark
multi_task_dataset: false
allow_variable_data_keys: false
max_cache_size: 0.0
max_cache_fd: 32
allow_multi_rates: false
valid_max_cache_size: null
exclude_weight_decay: false
exclude_weight_decay_conf: {}
optim: adamw
optim_conf:
    lr: 0.001
scheduler: warmuplr
scheduler_conf:
    warmup_steps: 25000
metric2id: dump/raw/overall_scale/metric2id
metric2type: dump/raw/overall_scale/metric2type
metric_pad_value: -100
token_list: null
metric_token_info: data/token_list/metric_500_percentile_overall_scale_w-numerical/tokens.json
metric_token_pad_value: 0
tokenize_numerical_metric: true
init: null
model_conf: {}
use_ref_audio: true
use_ref_text: false
use_preprocessor: true
token_type: bpe
bpemodel: null
non_linguistic_symbols: null
cleaner: null
g2p: null
sequential_metric: true
randomize_sequential_metric: true
frontend: s3prl
frontend_conf:
    frontend_conf:
        upstream: wavlm_large
    download_dir: ./hub
    multilayer_feature: true
universa: ar_universa
universa_conf:
    embedding_dim: 512
    audio_encoder_type: transformer
    audio_encoder_params:
        num_blocks: 4
        attention_heads: 4
        linear_units: 1024
        dropout_rate: 0.1
        positional_dropout_rate: 0.1
        attention_dropout_rate: 0.1
        input_layer: conv2d
        normalize_before: true
        concat_after: false
        positionwise_layer_type: linear
        positionwise_conv_kernel_size: 1
        layer_drop_rate: 0.1
        qk_norm: false
        use_flash_attn: false
    cross_attention_type: multihead
    cross_attention_params:
        n_head: 2
        dropout_rate: 0.1
    metric_decoder_params:
        num_blocks: 12
        attention_heads: 8
        linear_units: 2048
        dropout_rate: 0.1
        positional_dropout_rate: 0.1
        src_attention_dropout_rate: 0.1
        self_attention_dropout_rate: 0.1
        input_layer: embed
        normalize_before: true
        concat_after: false
        layer_drop_rate: 0.1
        qk_norm: false
        use_flash_attn: false
    use_rope: true
    lsm_weight: 0.1
    sym_sos: <sos>
    sym_eos: <eos>
required:
- output_dir
- metric2id
version: '202503'
distributed: false
```

</details>



### Citing ESPnet

```BibTex
@inproceedings{watanabe2018espnet,
  author={Shinji Watanabe and Takaaki Hori and Shigeki Karita and Tomoki Hayashi and Jiro Nishitoba and Yuya Unno and Nelson Yalta and Jahn Heymann and Matthew Wiesner and Nanxin Chen and Adithya Renduchintala and Tsubasa Ochiai},
  title={{ESPnet}: End-to-End Speech Processing Toolkit},
  year={2018},
  booktitle={Proceedings of Interspeech},
  pages={2207--2211},
  doi={10.21437/Interspeech.2018-1456},
  url={http://dx.doi.org/10.21437/Interspeech.2018-1456}
}






```

or arXiv:

```bibtex
@misc{watanabe2018espnet,
  title={ESPnet: End-to-End Speech Processing Toolkit},
  author={Shinji Watanabe and Takaaki Hori and Shigeki Karita and Tomoki Hayashi and Jiro Nishitoba and Yuya Unno and Nelson Yalta and Jahn Heymann and Matthew Wiesner and Nanxin Chen and Adithya Renduchintala and Tsubasa Ochiai},
  year={2018},
  eprint={1804.00015},
  archivePrefix={arXiv},
  primaryClass={cs.CL}
}
```