Upload checkpoint, sanitized config, and transcripts for ctc-baseline_xlsr_set_4

README.md

@@ -0,0 +1,41 @@
+---
+title: "CTC-DRO XLSR-based ASR model - set 4"
+language: multilingual
+tags:
+  - asr
+  - ctc-dro
+  - XLSR
+license: cc-by-nc-4.0
+---
+
+# CTC-Baseline XLSR-based ASR model - set 4
+
+This repository contains a CTC-Baseline XLSR-based automatic speech recognition (ASR) model trained with ESPnet.  
+The model was trained on balanced training data from set 4.
+
+## Intended Use
+
+This model is intended for ASR. Users can run inference using the provided checkpoint (`valid.loss.best.pth`) and configuration file (`config.yaml`):
+```bash
+import soundfile as sf
+from espnet2.bin.asr_inference import Speech2Text
+
+asr_train_config = "ctc-baseline_xlsr_set_4/config.yaml"
+asr_model_file = "ctc-baseline_xlsr_set_4/valid.loss.best.pth"
+
+model = Speech2Text.from_pretrained(
+    asr_train_config=asr_train_config,
+    asr_model_file=asr_model_file
+)
+
+speech, _ = sf.read("input.wav")
+text, *_ = model(speech)[0]
+
+print("Recognized text:", text)
+```
+
+## How to Use
+
+1. Clone this repository.
+2. Use ESPnet’s inference scripts with the provided `config.yaml` and checkpoint file.
+3. Ensure any external resources referenced in `config.yaml` are available at the indicated relative paths.

config.yaml

@@ -0,0 +1,342 @@
+accum_grad: 16
+adapter: lora
+adapter_conf: {}
+allow_multi_rates: false
+allow_variable_data_keys: false
+aux_ctc_tasks: []
+batch_bins: 1000000
+batch_size: 4
+batch_type: duration_language
+best_model_criterion:
+- - valid
+  - loss
+  - min
+bpemodel: null
+chunk_default_fs: null
+chunk_excluded_key_prefixes: []
+chunk_length: 500
+chunk_shift_ratio: 0.5
+cleaner: null
+collect_stats: false
+create_graph_in_tensorboard: false
+ctc_conf:
+  ctc_type: builtin
+cudnn_benchmark: false
+cudnn_deterministic: true
+cudnn_enabled: true
+decoder: null
+decoder_conf: {}
+detect_anomaly: false
+distributed: false
+drop_last_iter: false
+dry_run: false
+duration_batch_length: -1
+early_stopping_criterion:
+- valid
+- loss
+- min
+encoder: transformer
+encoder_conf:
+  attention_dropout_rate: 0.1
+  attention_heads: 8
+  dropout_rate: 0.1
+  input_layer: conv2d2
+  linear_units: 1024
+  normalize_before: true
+  num_blocks: 2
+  output_size: 256
+  positional_dropout_rate: 0.1
+exclude_weight_decay: false
+exclude_weight_decay_conf: {}
+fold_length:
+- 80000
+- 150
+freeze_param: []
+frontend: s3prl
+frontend_conf:
+  download_dir: ./hub
+  frontend_conf:
+    upstream: xls_r_300m
+  fs: 16k
+  multilayer_feature: true
+g2p: null
+grad_clip: 5.0
+grad_clip_type: 2.0
+grad_noise: false
+ignore_init_mismatch: false
+init: xavier_uniform
+init_param: []
+input_size: null
+iterator_type: sequence
+joint_net_conf: null
+keep_nbest_models: 3
+log_interval: null
+log_level: INFO
+max_cache_fd: 32
+max_cache_size: 0.0
+max_epoch: 40
+model: espnet
+model_conf:
+  ctc_weight: 1.0
+multiple_iterator: false
+multiprocessing_distributed: false
+nbest_averaging_interval: 0
+ngpu: 1
+no_forward_run: false
+noise_apply_prob: 1.0
+noise_db_range: '13_15'
+noise_scp: null
+non_linguistic_symbols: ./nlsyms.txt
+normalize: utterance_mvn
+normalize_conf: {}
+num_att_plot: 3
+num_cache_chunks: 1024
+num_iters_per_epoch: 1200
+num_workers: 4
+optim: adam
+optim_conf:
+  lr: 0.0001
+  weight_decay: 1.0e-06
+output_dir: ./inference_results
+patience: null
+postencoder: null
+postencoder_conf: {}
+preencoder: linear
+preencoder_conf:
+  input_size: 1024
+  output_size: 80
+preprocessor: default
+preprocessor_conf: {}
+pretrain_path: null
+print_config: false
+required:
+- output_dir
+- token_list
+resume: true
+rir_apply_prob: 1.0
+rir_scp: null
+save_strategy: all
+scheduler: null
+scheduler_conf: {}
+seed: 0
+sharded_ddp: false
+short_noise_thres: 0.5
+shuffle_within_batch: false
+sort_batch: descending
+sort_in_batch: descending
+specaug: specaug
+specaug_conf:
+  apply_freq_mask: true
+  apply_time_mask: true
+  apply_time_warp: true
+  freq_mask_width_range:
+  - 0
+  - 27
+  num_freq_mask: 2
+  num_time_mask: 10
+  time_mask_width_ratio_range:
+  - 0.0
+  - 0.05
+  time_warp_mode: bicubic
+  time_warp_window: 5
+speech_volume_normalize: null
+token_list:
+- <blank>
+- <unk>
+- <space>
+- E
+- A
+- O
+- N
+- S
+- I
+- ا
+- L
+- T
+- R
+- و
+- D
+- ن
+- ر
+- ی
+- ي
+- M
+- U
+- H
+- P
+- ک
+- م
+- C
+- А
+- Ӹ
+- Н
+- B
+- ت
+- س
+- ل
+- J
+- K
+- ہ
+- Т
+- ے
+- G
+- Ш
+- К
+- Е
+- Л
+- Ы
+- V
+- М
+- ج
+- Ӓ
+- ه
+- ب
+- د
+- О
+- Y
+- '[slv]'
+- Р
+- ڪ
+- پ
+- Z
+- '[mrj]'
+- F
+- گ
+- И
+- В
+- ئ
+- Д
+- '[sot]'
+- ں
+- '[spa]'
+- W
+- Q
+- П
+- Г
+- ف
+- ق
+- С
+- ع
+- ش
+- Ж
+- ز
+- ھ
+- آ
+- Č
+- Í
+- У
+- ح
+- '[urd]'
+- Š
+- ٹ
+- چ
+- Ь
+- ٽ
+- '[snd]'
+- ڻ
+- Й
+- ط
+- ص
+- ٿ
+- Ц
+- خ
+- Ó
+- Я
+- Á
+- É
+- Ч
+- ۾
+- '0'
+- Ž
+- З
+- '1'
+- ۽
+- –
+- ڏ
+- Э
+- ڊ
+- —
+- ڈ
+- ء
+- Ñ
+- ڙ
+- ِ
+- '2'
+- ٻ
+- Х
+- Ӱ
+- ظ
+- ض
+- ث
+- ڳ
+- ،
+- X
+- ¡
+- غ
+- ڑ
+- Ӧ
+- ذ
+- ¿
+- '5'
+- ڌ
+- '3'
+- ڀ
+- ُ
+- '9'
+- Ú
+- '4'
+- '8'
+- ۔
+- '6'
+- ٺ
+- Ю
+- »
+- Б
+- «
+- ڇ
+- ً
+- ڃ
+- '7'
+- ڄ
+- ؤ
+- ڍ
+- Ф
+- َ
+- ٰ
+- ّ
+- ڱ
+- ”
+- ژ
+- ڦ
+- Ё
+- ؛
+- ٍ
+- Щ
+- ؟
+- ’
+- ‘
+- °
+- ۃ
+- إ
+- Ć
+- <sos/eos>
+token_type: char
+train_dtype: float32
+unused_parameters: true
+use_adapter: false
+use_amp: false
+use_lang_prompt: false
+use_matplotlib: true
+use_nlp_prompt: false
+use_preprocessor: true
+use_tensorboard: true
+val_scheduler_criterion:
+- valid
+- loss
+valid_batch_bins: null
+valid_batch_size: null
+valid_batch_type: null
+valid_iterator_type: null
+valid_max_cache_size: null
+version: '202402'
+write_collected_feats: false

valid.loss.best.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6ea23bca7b78e6588073f319b3b5fe03d7560f607fa29eddd13369f1b032fe13
+size 1288666400