初始化项目，由ModelHub XC社区提供模型

Model: Jzuluaga/wav2vec2-large-960h-lv60-self-en-atc-uwb-atcc-and-atcosim
Source: Original Platform

.gitattributes

@@ -0,0 +1,35 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text
+language_model/atcosim_uwb_atcc_4g.binary filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,244 @@
+---
+language: en
+license: apache-2.0
+tags:
+- audio
+- automatic-speech-recognition
+- en-atc
+- en
+- generated_from_trainer
+datasets:
+- Jzuluaga/atcosim_corpus
+- Jzuluaga/uwb_atcc
+metrics:
+- wer
+base_model: facebook/wav2vec2-large-960h-lv60-self
+model-index:
+- name: wav2vec2-large-960h-lv60-self-en-atc-uwb-atcc-and-atcosim
+  results:
+  - task:
+      type: automatic-speech-recognition
+      name: Speech Recognition
+    dataset:
+      name: UWB-ATCC dataset (Air Traffic Control Communications)
+      type: Jzuluaga/uwb_atcc
+      config: test
+      split: test
+    metrics:
+    - type: wer
+      value: 17.48
+      name: TEST WER
+      verified: false
+    - type: wer
+      value: 14.26
+      name: TEST WER (+LM)
+      verified: false
+  - task:
+      type: automatic-speech-recognition
+      name: Speech Recognition
+    dataset:
+      name: ATCOSIM corpus (Air Traffic Control Communications)
+      type: Jzuluaga/atcosim_corpus
+      config: test
+      split: test
+    metrics:
+    - type: wer
+      value: 1.85
+      name: TEST WER
+      verified: false
+    - type: wer
+      value: 1.13
+      name: TEST WER (+LM)
+      verified: false
+---
+
+
+This model is a fine-tuned version of [facebook/wav2vec2-large-960h-lv60-self](https://huggingface.co/facebook/wav2vec2-large-960h-lv60-self) on the EXPERIMENTS/DATA/ATCOSIM_UWB_ATCC/TRAIN - NA dataset.
+It achieves the following results on the evaluation set:
+
+# wav2vec2-large-960h-lv60-self-en-atc-uwb-atcc-and-atcosim
+
+This model is a fine-tuned version of [facebook/wav2vec2-large-960h-lv60-self](https://huggingface.co/facebook/wav2vec2-large-960h-lv60-self) on two corpus:
+- [UWB-ATCC corpus](https://huggingface.co/datasets/Jzuluaga/uwb_atcc), and
+- [ATCOSIM corpus](https://huggingface.co/datasets/Jzuluaga/atcosim_corpus). 
+
+<a href="https://colab.research.google.com/github/idiap/w2v2-air-traffic/blob/main/src/eval_xlsr_atc_model.ipynb">
+    <img alt="GitHub" src="https://colab.research.google.com/assets/colab-badge.svg\">
+</a>
+<a href="https://github.com/idiap/w2v2-air-traffic">
+    <img alt="GitHub" src="https://img.shields.io/badge/GitHub-Open%20source-green\">
+</a>
+
+
+It achieves the following results on the evaluation set (two tests sets joined together: UWB-ATCC and ATCOSIM):
+- Loss: 0.4042
+- Wer: 0.1049
+
+Paper: [How Does Pre-trained Wav2Vec 2.0 Perform on Domain Shifted ASR? An Extensive Benchmark on Air Traffic Control Communications](https://arxiv.org/abs/2203.16822).
+
+Authors: Juan Zuluaga-Gomez, Amrutha Prasad, Iuliia Nigmatulina, Saeed Sarfjoo, Petr Motlicek, Matthias Kleinert, Hartmut Helmke, Oliver Ohneiser, Qingran Zhan
+
+Abstract: Recent work on self-supervised pre-training focus</b> on leveraging large-scale unlabeled speech data to build robust end-to-end (E2E)acoustic models (AM) that can be later fine-tuned on downstream tasks e.g., automatic speech recognition (ASR). Yet, few works investigated the impact on performance when the data properties substantially differ between the pre-training and fine-tuning phases, termed domain shift. We target this scenario by analyzing the robustness of Wav2Vec 2.0 and XLS-R models on downstream ASR for a completely unseen domain, air traffic control (ATC) communications. We benchmark these two models on several open-source and challenging ATC databases with signal-to-noise ratio between 5 and 20 dB. Relative word error rate (WER) reductions between 20% to 40% are obtained in comparison to hybrid-based ASR baselines by only fine-tuning E2E acoustic models with a smaller fraction of labeled data. We analyze WERs on the low-resource scenario and gender bias carried by one ATC dataset.
+
+Code — GitHub repository: https://github.com/idiap/w2v2-air-traffic
+
+## Usage
+
+You can use our Google Colab notebook to run and evaluate our model: https://github.com/idiap/w2v2-air-traffic/blob/master/src/eval_xlsr_atc_model.ipynb
+
+## Intended uses & limitations
+
+This model was fine-tuned on air traffic control data. We don't expect that it keeps the same performance on some others datasets, e.g., LibriSpeech or CommonVoice.
+
+
+## Training and evaluation data
+
+See Table 1 (page 3) in our paper: [How Does Pre-trained Wav2Vec 2.0 Perform on Domain Shifted ASR? An Extensive Benchmark on Air Traffic Control Communications](https://arxiv.org/abs/2203.16822). We described there the partitions of how to use our model. 
+
+- We use the UWB-ATCC + ATCOSIM corpus to fine-tune this model. You can download the raw data here:
+  - https://lindat.mff.cuni.cz/repository/xmlui/handle/11858/00-097C-0000-0001-CCA1-0 and,
+  - https://www.spsc.tugraz.at/databases-and-tools/atcosim-air-traffic-control-simulation-speech-corpus.html
+- However, do not worry, we have prepared the database in `Datasets format`:
+  -  Here, [UWB-ATCC corpus on HuggingFace](https://huggingface.co/datasets/Jzuluaga/uwb_atcc).
+  -  Here: [ATCOSIM CORPUS on HuggingFace](https://huggingface.co/datasets/Jzuluaga/atcosim_corpus). 
+- If you want to prepare a database in HuggingFace format, you can follow the data loader script in: [data_loader_atc.py](https://huggingface.co/datasets/Jzuluaga/uwb_atcc/blob/main/atc_data_loader.py).
+
+## Writing your own inference script
+
+If you use language model, you need to install the KenLM bindings with:
+
+```bash
+conda activate your_environment
+pip install https://github.com/kpu/kenlm/archive/master.zip
+```
+
+The snippet of code:
+
+```python
+from datasets import load_dataset, load_metric, Audio
+import torch
+from transformers import AutoModelForCTC, Wav2Vec2Processor, Wav2Vec2ProcessorWithLM
+import torchaudio.functional as F
+
+USE_LM = False
+DATASET_ID = "Jzuluaga/uwb_atcc"
+MODEL_ID = "Jzuluaga/wav2vec2-large-960h-lv60-self-en-atc-uwb-atcc-and-atcosim"
+
+# 1. Load the dataset
+# we only load the 'test' partition, however, if you want to load the 'train' partition, you can change it accordingly
+uwb_atcc_corpus_test = load_dataset(DATASET_ID, "test", split="test")
+
+# 2. Load the model
+model = AutoModelForCTC.from_pretrained(MODEL_ID)
+
+# 3. Load the processors, we offer support with LM, which should yield better resutls
+if USE_LM:
+    processor = Wav2Vec2ProcessorWithLM.from_pretrained(MODEL_ID)
+else:
+    processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
+# 4. Format the test sample
+sample = next(iter(uwb_atcc_corpus_test))
+file_sampling_rate = sample['audio']['sampling_rate']
+# resample if neccessary
+if file_sampling_rate != 16000:
+    resampled_audio = F.resample(torch.tensor(sample["audio"]["array"]), file_sampling_rate, 16000).numpy()
+else:
+    resampled_audio = torch.tensor(sample["audio"]["array"]).numpy()
+input_values = processor(resampled_audio, return_tensors="pt").input_values
+
+# 5. Run the forward pass in the model
+with torch.no_grad():
+    logits = model(input_values).logits
+    
+# get the transcription with processor
+if USE_LM:
+    transcription = processor.batch_decode(logits.numpy()).text
+else:
+    pred_ids = torch.argmax(logits, dim=-1)
+    transcription = processor.batch_decode(pred_ids)
+# print the output
+print(transcription)
+```
+
+# Cite us
+
+If you use this code for your research, please cite our paper with:
+
+```
+@article{zuluaga2022how,
+    title={How Does Pre-trained Wav2Vec2. 0 Perform on Domain Shifted ASR? An Extensive Benchmark on Air Traffic Control Communications},
+    author={Zuluaga-Gomez, Juan and Prasad, Amrutha and Nigmatulina, Iuliia and Sarfjoo, Saeed and others},
+    journal={IEEE Spoken Language Technology Workshop (SLT), Doha, Qatar},
+    year={2022}
+  }
+```
+and, 
+
+```
+@article{zuluaga2022bertraffic,
+  title={BERTraffic: BERT-based Joint Speaker Role and Speaker Change Detection for Air Traffic Control Communications},
+  author={Zuluaga-Gomez, Juan and Sarfjoo, Seyyed Saeed and Prasad, Amrutha and others},
+  journal={IEEE Spoken Language Technology Workshop (SLT), Doha, Qatar},
+  year={2022}
+  }
+```
+
+and, 
+
+```
+@article{zuluaga2022atco2,
+  title={ATCO2 corpus: A Large-Scale Dataset for Research on Automatic Speech Recognition and Natural Language Understanding of Air Traffic Control Communications},
+  author={Zuluaga-Gomez, Juan and Vesel{\`y}, Karel and Sz{\"o}ke, Igor and Motlicek, Petr and others},
+  journal={arXiv preprint arXiv:2211.04054},
+  year={2022}
+}
+```
+
+
+## Training procedure
+
+### Training hyperparameters
+
+The following hyperparameters were used during training:
+- learning_rate: 0.0001
+- train_batch_size: 24
+- eval_batch_size: 12
+- seed: 42
+- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
+- lr_scheduler_type: linear
+- lr_scheduler_warmup_steps: 1000
+- training_steps: 10000
+- mixed_precision_training: Native AMP
+
+### Training results
+
+| Training Loss | Epoch | Step  | Validation Loss | Wer    |
+|:-------------:|:-----:|:-----:|:---------------:|:------:|
+| No log        | 0.63  | 500   | 2.2638          | 0.9359 |
+| 2.6089        | 1.27  | 1000  | 0.7277          | 0.2407 |
+| 2.6089        | 1.9   | 1500  | 0.5800          | 0.1745 |
+| 0.6019        | 2.53  | 2000  | 0.4887          | 0.1514 |
+| 0.6019        | 3.17  | 2500  | 0.4666          | 0.1421 |
+| 0.4722        | 3.8   | 3000  | 0.4426          | 0.1451 |
+| 0.4722        | 4.44  | 3500  | 0.4176          | 0.1248 |
+| 0.4278        | 5.07  | 4000  | 0.4365          | 0.1239 |
+| 0.4278        | 5.7   | 4500  | 0.3816          | 0.1177 |
+| 0.369         | 6.34  | 5000  | 0.4113          | 0.1172 |
+| 0.369         | 6.97  | 5500  | 0.3863          | 0.1230 |
+| 0.341         | 7.6   | 6000  | 0.3850          | 0.1116 |
+| 0.341         | 8.24  | 6500  | 0.4014          | 0.1141 |
+| 0.3119        | 8.87  | 7000  | 0.3953          | 0.1078 |
+| 0.3119        | 9.51  | 7500  | 0.4018          | 0.1080 |
+| 0.3008        | 10.14 | 8000  | 0.3964          | 0.1074 |
+| 0.3008        | 10.77 | 8500  | 0.3917          | 0.1078 |
+| 0.2741        | 11.41 | 9000  | 0.3961          | 0.1057 |
+| 0.2741        | 12.04 | 9500  | 0.3974          | 0.1053 |
+| 0.2531        | 12.67 | 10000 | 0.4042          | 0.1049 |
+
+
+### Framework versions
+
+- Transformers 4.24.0
+- Pytorch 1.13.0+cu117
+- Datasets 2.6.1
+- Tokenizers 0.13.2

added_tokens.json

@@ -0,0 +1,4 @@
+{
+  "</s>": 30,
+  "<s>": 29
+}

all_results.json

@@ -0,0 +1,14 @@
+{
+    "epoch": 12.67,
+    "eval_loss": 0.40417608618736267,
+    "eval_runtime": 163.5465,
+    "eval_samples": 4723,
+    "eval_samples_per_second": 28.879,
+    "eval_steps_per_second": 2.409,
+    "eval_wer": 0.10485405449172099,
+    "train_loss": 0.5960800704956055,
+    "train_runtime": 13155.4001,
+    "train_samples": 18925,
+    "train_samples_per_second": 18.243,
+    "train_steps_per_second": 0.76
+}

alphabet.json

@@ -0,0 +1 @@
+{"labels": [" ", "a", "b", "c", "d", "e", "f", "g", "h", "i", "j", "k", "l", "m", "n", "o", "p", "q", "r", "s", "t", "u", "v", "w", "x", "y", "z", "\u2047", "", "<s>", "</s>"], "is_bpe": false}

config.json

@@ -0,0 +1,108 @@
+{
+  "_name_or_path": "facebook/wav2vec2-large-960h-lv60-self",
+  "activation_dropout": 0.0,
+  "adapter_kernel_size": 3,
+  "adapter_stride": 2,
+  "add_adapter": false,
+  "apply_spec_augment": true,
+  "architectures": [
+    "Wav2Vec2ForCTC"
+  ],
+  "attention_dropout": 0.0,
+  "bos_token_id": 1,
+  "classifier_proj_size": 256,
+  "codevector_dim": 256,
+  "contrastive_logits_temperature": 0.1,
+  "conv_bias": true,
+  "conv_dim": [
+    512,
+    512,
+    512,
+    512,
+    512,
+    512,
+    512
+  ],
+  "conv_kernel": [
+    10,
+    3,
+    3,
+    3,
+    3,
+    2,
+    2
+  ],
+  "conv_stride": [
+    5,
+    2,
+    2,
+    2,
+    2,
+    2,
+    2
+  ],
+  "ctc_loss_reduction": "mean",
+  "ctc_zero_infinity": true,
+  "diversity_loss_weight": 0.1,
+  "do_stable_layer_norm": true,
+  "eos_token_id": 2,
+  "feat_extract_activation": "gelu",
+  "feat_extract_dropout": 0.0,
+  "feat_extract_norm": "layer",
+  "feat_proj_dropout": 0.05,
+  "feat_quantizer_dropout": 0.0,
+  "final_dropout": 0.0,
+  "hidden_act": "gelu",
+  "hidden_dropout": 0.0,
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 1024,
+  "initializer_range": 0.02,
+  "intermediate_size": 4096,
+  "layer_norm_eps": 1e-05,
+  "layerdrop": 0.0,
+  "mask_feature_length": 12,
+  "mask_feature_min_masks": 0,
+  "mask_feature_prob": 0.0,
+  "mask_time_length": 12,
+  "mask_time_min_masks": 2,
+  "mask_time_prob": 0.075,
+  "model_type": "wav2vec2",
+  "num_adapter_layers": 3,
+  "num_attention_heads": 16,
+  "num_codevector_groups": 2,
+  "num_codevectors_per_group": 320,
+  "num_conv_pos_embedding_groups": 16,
+  "num_conv_pos_embeddings": 128,
+  "num_feat_extract_layers": 7,
+  "num_hidden_layers": 24,
+  "num_negatives": 100,
+  "output_hidden_size": 1024,
+  "pad_token_id": 28,
+  "proj_codevector_dim": 256,
+  "tdnn_dilation": [
+    1,
+    2,
+    3,
+    1,
+    1
+  ],
+  "tdnn_dim": [
+    512,
+    512,
+    512,
+    512,
+    1500
+  ],
+  "tdnn_kernel": [
+    5,
+    3,
+    3,
+    1,
+    1
+  ],
+  "torch_dtype": "float32",
+  "transformers_version": "4.24.0",
+  "use_weighted_layer_sum": false,
+  "vocab_size": 31,
+  "xvector_output_dim": 512
+}

eval_results.json

@@ -0,0 +1,9 @@
+{
+    "epoch": 12.67,
+    "eval_loss": 0.40417608618736267,
+    "eval_runtime": 163.5465,
+    "eval_samples": 4723,
+    "eval_samples_per_second": 28.879,
+    "eval_steps_per_second": 2.409,
+    "eval_wer": 0.10485405449172099
+}

language_model/atcosim_uwb_atcc_4g.binary

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:974c8f8db889c30ac4060a741144847c7b71cad0a8bd6714d81a265794b235d0
+size 1360621

language_model/attrs.json

@@ -0,0 +1 @@
+{"alpha": 0.5, "beta": 1.5, "unk_score_offset": -10.0, "score_boundary": true}

log/eval_model_train

@@ -0,0 +1,15 @@
+# Running on vgnf007
+# Started at Wed 30 Nov 07:36:30 CET 2022
+# python3 src/eval_model.py --language-model experiments/data/atcosim_uwb_atcc/train/lm/atcosim_uwb_atcc_4g.binary --pretrained-model experiments/results/bs_exp/linear/wav2vec2-large-960h-lv60-self/atcosim_uwb_atcc/0.0ld_0.0ad_0.0attd_0.05fpd_0.075mtp_12mtl_0.0mfp_12mfl_1acc/ --print-output true --test-set experiments/data/atcosim_corpus/train 
+Integrating a LM by shallow fusion, results should be better
+*** Loading the Wav2Vec 2.0 model, loading... ***
+Traceback (most recent call last):
+  File "/remote/idiap.svm/temp.speech01/jzuluaga/experiments/journal/asr/github/w2v2-air-traffic/src/eval_model.py", line 250, in <module>
+    main()
+  File "/remote/idiap.svm/temp.speech01/jzuluaga/experiments/journal/asr/github/w2v2-air-traffic/src/eval_model.py", line 152, in main
+    processor, processor_ctc_kenlm, model = get_kenlm_processor(path_model, path_lm)
+  File "/remote/idiap.svm/temp.speech01/jzuluaga/experiments/journal/asr/github/w2v2-air-traffic/src/eval_model.py", line 65, in get_kenlm_processor
+    path_tokenizer + "/vocab.json",
+TypeError: unsupported operand type(s) for +: 'PosixPath' and 'str'
+# Accounting: time=25 threads=1
+# Finished at Wed 30 Nov 07:36:55 CET 2022 with status 1

preprocessor_config.json

@@ -0,0 +1,10 @@
+{
+  "do_normalize": true,
+  "feature_extractor_type": "Wav2Vec2FeatureExtractor",
+  "feature_size": 1,
+  "padding_side": "right",
+  "padding_value": 0.0,
+  "processor_class": "Wav2Vec2ProcessorWithLM",
+  "return_attention_mask": true,
+  "sampling_rate": 16000
+}

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:854b50dd536a9bf991e47cf4bf6bea5b84ecc0cdad160cf46ba711c69359da2b
+size 1262028973

special_tokens_map.json

@@ -0,0 +1,6 @@
+{
+  "bos_token": "<s>",
+  "eos_token": "</s>",
+  "pad_token": "[PAD]",
+  "unk_token": "[UNK]"
+}

tokenizer_config.json

@@ -0,0 +1,11 @@
+{
+  "bos_token": "<s>",
+  "do_lower_case": false,
+  "eos_token": "</s>",
+  "pad_token": "[PAD]",
+  "processor_class": "Wav2Vec2ProcessorWithLM",
+  "replace_word_delimiter_char": " ",
+  "tokenizer_class": "Wav2Vec2CTCTokenizer",
+  "unk_token": "[UNK]",
+  "word_delimiter_token": "|"
+}

train_results.json

@@ -0,0 +1,8 @@
+{
+    "epoch": 12.67,
+    "train_loss": 0.5960800704956055,
+    "train_runtime": 13155.4001,
+    "train_samples": 18925,
+    "train_samples_per_second": 18.243,
+    "train_steps_per_second": 0.76
+}

trainer_state.json

@@ -0,0 +1,265 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 12.67427122940431,
+  "global_step": 10000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.63,
+      "eval_loss": 2.2638113498687744,
+      "eval_runtime": 149.9992,
+      "eval_samples_per_second": 31.487,
+      "eval_steps_per_second": 2.627,
+      "eval_wer": 0.9359337678636492,
+      "step": 500
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 9.970000000000001e-05,
+      "loss": 2.6089,
+      "step": 1000
+    },
+    {
+      "epoch": 1.27,
+      "eval_loss": 0.7277476787567139,
+      "eval_runtime": 148.226,
+      "eval_samples_per_second": 31.864,
+      "eval_steps_per_second": 2.658,
+      "eval_wer": 0.24067851503987692,
+      "step": 1000
+    },
+    {
+      "epoch": 1.9,
+      "eval_loss": 0.5800275802612305,
+      "eval_runtime": 148.0487,
+      "eval_samples_per_second": 31.902,
+      "eval_steps_per_second": 2.661,
+      "eval_wer": 0.1745475891664305,
+      "step": 1500
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 8.892222222222223e-05,
+      "loss": 0.6019,
+      "step": 2000
+    },
+    {
+      "epoch": 2.53,
+      "eval_loss": 0.48867326974868774,
+      "eval_runtime": 148.4526,
+      "eval_samples_per_second": 31.815,
+      "eval_steps_per_second": 2.654,
+      "eval_wer": 0.15135014776729688,
+      "step": 2000
+    },
+    {
+      "epoch": 3.17,
+      "eval_loss": 0.466572642326355,
+      "eval_runtime": 148.0238,
+      "eval_samples_per_second": 31.907,
+      "eval_steps_per_second": 2.662,
+      "eval_wer": 0.14213999433221328,
+      "step": 2500
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 7.78111111111111e-05,
+      "loss": 0.4722,
+      "step": 3000
+    },
+    {
+      "epoch": 3.8,
+      "eval_loss": 0.44257038831710815,
+      "eval_runtime": 149.7003,
+      "eval_samples_per_second": 31.55,
+      "eval_steps_per_second": 2.632,
+      "eval_wer": 0.14505485607870128,
+      "step": 3000
+    },
+    {
+      "epoch": 4.44,
+      "eval_loss": 0.41759932041168213,
+      "eval_runtime": 149.0588,
+      "eval_samples_per_second": 31.685,
+      "eval_steps_per_second": 2.643,
+      "eval_wer": 0.12481276061697907,
+      "step": 3500
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 6.671111111111111e-05,
+      "loss": 0.4278,
+      "step": 4000
+    },
+    {
+      "epoch": 5.07,
+      "eval_loss": 0.4364745318889618,
+      "eval_runtime": 148.5505,
+      "eval_samples_per_second": 31.794,
+      "eval_steps_per_second": 2.652,
+      "eval_wer": 0.12388162422573985,
+      "step": 4000
+    },
+    {
+      "epoch": 5.7,
+      "eval_loss": 0.3815610408782959,
+      "eval_runtime": 149.3382,
+      "eval_samples_per_second": 31.626,
+      "eval_steps_per_second": 2.638,
+      "eval_wer": 0.1177280272053763,
+      "step": 4500
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 5.560000000000001e-05,
+      "loss": 0.369,
+      "step": 5000
+    },
+    {
+      "epoch": 6.34,
+      "eval_loss": 0.4113306403160095,
+      "eval_runtime": 159.7075,
+      "eval_samples_per_second": 29.573,
+      "eval_steps_per_second": 2.467,
+      "eval_wer": 0.11716124853244808,
+      "step": 5000
+    },
+    {
+      "epoch": 6.97,
+      "eval_loss": 0.3862614035606384,
+      "eval_runtime": 152.4609,
+      "eval_samples_per_second": 30.978,
+      "eval_steps_per_second": 2.584,
+      "eval_wer": 0.1230112141208858,
+      "step": 5500
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 4.448888888888889e-05,
+      "loss": 0.341,
+      "step": 6000
+    },
+    {
+      "epoch": 7.6,
+      "eval_loss": 0.384976863861084,
+      "eval_runtime": 159.7493,
+      "eval_samples_per_second": 29.565,
+      "eval_steps_per_second": 2.466,
+      "eval_wer": 0.1116149143759362,
+      "step": 6000
+    },
+    {
+      "epoch": 8.24,
+      "eval_loss": 0.401400089263916,
+      "eval_runtime": 155.917,
+      "eval_samples_per_second": 30.292,
+      "eval_steps_per_second": 2.527,
+      "eval_wer": 0.11406420792680458,
+      "step": 6500
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.337777777777778e-05,
+      "loss": 0.3119,
+      "step": 7000
+    },
+    {
+      "epoch": 8.87,
+      "eval_loss": 0.39530250430107117,
+      "eval_runtime": 165.7417,
+      "eval_samples_per_second": 28.496,
+      "eval_steps_per_second": 2.377,
+      "eval_wer": 0.10782964252459415,
+      "step": 7000
+    },
+    {
+      "epoch": 9.51,
+      "eval_loss": 0.4018384516239166,
+      "eval_runtime": 163.4263,
+      "eval_samples_per_second": 28.9,
+      "eval_steps_per_second": 2.411,
+      "eval_wer": 0.10801182138374965,
+      "step": 7500
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 2.2277777777777778e-05,
+      "loss": 0.3008,
+      "step": 8000
+    },
+    {
+      "epoch": 10.14,
+      "eval_loss": 0.3963571786880493,
+      "eval_runtime": 172.5171,
+      "eval_samples_per_second": 27.377,
+      "eval_steps_per_second": 2.284,
+      "eval_wer": 0.10744504271082142,
+      "step": 8000
+    },
+    {
+      "epoch": 10.77,
+      "eval_loss": 0.39167362451553345,
+      "eval_runtime": 165.6389,
+      "eval_samples_per_second": 28.514,
+      "eval_steps_per_second": 2.379,
+      "eval_wer": 0.10780940042913242,
+      "step": 8500
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 1.1166666666666668e-05,
+      "loss": 0.2741,
+      "step": 9000
+    },
+    {
+      "epoch": 11.41,
+      "eval_loss": 0.3961273431777954,
+      "eval_runtime": 164.9191,
+      "eval_samples_per_second": 28.638,
+      "eval_steps_per_second": 2.389,
+      "eval_wer": 0.10568398040565159,
+      "step": 9000
+    },
+    {
+      "epoch": 12.04,
+      "eval_loss": 0.39744970202445984,
+      "eval_runtime": 164.8733,
+      "eval_samples_per_second": 28.646,
+      "eval_steps_per_second": 2.39,
+      "eval_wer": 0.10529938059187888,
+      "step": 9500
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 5.555555555555556e-08,
+      "loss": 0.2531,
+      "step": 10000
+    },
+    {
+      "epoch": 12.67,
+      "eval_loss": 0.40417608618736267,
+      "eval_runtime": 164.6908,
+      "eval_samples_per_second": 28.678,
+      "eval_steps_per_second": 2.392,
+      "eval_wer": 0.10485405449172099,
+      "step": 10000
+    },
+    {
+      "epoch": 12.67,
+      "step": 10000,
+      "total_flos": 2.7136568655380324e+19,
+      "train_loss": 0.5960800704956055,
+      "train_runtime": 13155.4001,
+      "train_samples_per_second": 18.243,
+      "train_steps_per_second": 0.76
+    }
+  ],
+  "max_steps": 10000,
+  "num_train_epochs": 13,
+  "total_flos": 2.7136568655380324e+19,
+  "trial_name": null,
+  "trial_params": null
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3c84ee358d1ff2b725c335fdd3efaeb0ad280010e6d18e558b2ffafcfdf3c6f8
+size 3771

vocab.json

@@ -0,0 +1,31 @@
+{
+  "[PAD]": 28,
+  "[UNK]": 27,
+  "a": 1,
+  "b": 2,
+  "c": 3,
+  "d": 4,
+  "e": 5,
+  "f": 6,
+  "g": 7,
+  "h": 8,
+  "i": 9,
+  "j": 10,
+  "k": 11,
+  "l": 12,
+  "m": 13,
+  "n": 14,
+  "o": 15,
+  "p": 16,
+  "q": 17,
+  "r": 18,
+  "s": 19,
+  "t": 20,
+  "u": 21,
+  "v": 22,
+  "w": 23,
+  "x": 24,
+  "y": 25,
+  "z": 26,
+  "|": 0
+}