updating the repo with the fine-tuned model

.gitattributes

@@ -32,3 +32,4 @@ saved_model/**/* 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 +1,82 @@
+---
+license: apache-2.0
+tags:
+- automatic-speech-recognition
+- experiments/data/atcosim_uwb_atcc/train
+- generated_from_trainer
+metrics:
+- wer
+model-index:
+- name: 0.0ld_0.0ad_0.0attd_0.05fpd_0.075mtp_12mtl_0.0mfp_12mfl_1acc
+  results: []
+---
+
+<!-- This model card has been generated automatically according to the information the Trainer had access to. You
+should probably proofread and complete it, then remove this comment. -->
+
+# 0.0ld_0.0ad_0.0attd_0.05fpd_0.075mtp_12mtl_0.0mfp_12mfl_1acc
+
+This model is a fine-tuned version of [facebook/wav2vec2-xls-r-300m](https://huggingface.co/facebook/wav2vec2-xls-r-300m) on the EXPERIMENTS/DATA/ATCOSIM_UWB_ATCC/TRAIN - NA dataset.
+It achieves the following results on the evaluation set:
+- Loss: 0.5595
+- Wer: 0.1687
+
+## Model description
+
+More information needed
+
+## Intended uses & limitations
+
+More information needed
+
+## Training and evaluation data
+
+More information needed
+
+## 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   | 3.0458          | 1.0    |
+| 2.9181        | 1.27  | 1000  | 1.1503          | 0.4723 |
+| 2.9181        | 1.9   | 1500  | 0.8275          | 0.3500 |
+| 0.7646        | 2.53  | 2000  | 0.6990          | 0.2845 |
+| 0.7646        | 3.17  | 2500  | 0.5828          | 0.2509 |
+| 0.5394        | 3.8   | 3000  | 0.5363          | 0.2487 |
+| 0.5394        | 4.44  | 3500  | 0.5467          | 0.2171 |
+| 0.4558        | 5.07  | 4000  | 0.5290          | 0.2090 |
+| 0.4558        | 5.7   | 4500  | 0.4992          | 0.2046 |
+| 0.3773        | 6.34  | 5000  | 0.4934          | 0.2052 |
+| 0.3773        | 6.97  | 5500  | 0.4700          | 0.1983 |
+| 0.3301        | 7.6   | 6000  | 0.4938          | 0.1874 |
+| 0.3301        | 8.24  | 6500  | 0.5364          | 0.1893 |
+| 0.2938        | 8.87  | 7000  | 0.5170          | 0.1830 |
+| 0.2938        | 9.51  | 7500  | 0.5408          | 0.1815 |
+| 0.2674        | 10.14 | 8000  | 0.5581          | 0.1733 |
+| 0.2674        | 10.77 | 8500  | 0.5389          | 0.1719 |
+| 0.24          | 11.41 | 9000  | 0.5344          | 0.1714 |
+| 0.24          | 12.04 | 9500  | 0.5503          | 0.1686 |
+| 0.211         | 12.67 | 10000 | 0.5595          | 0.1687 |
+
+
+### 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": 0.5595183372497559,
+    "eval_runtime": 161.956,
+    "eval_samples": 4723,
+    "eval_samples_per_second": 29.162,
+    "eval_steps_per_second": 2.433,
+    "eval_wer": 0.16865713938706936,
+    "train_loss": 0.6397389862060547,
+    "train_runtime": 13054.0758,
+    "train_samples": 18925,
+    "train_samples_per_second": 18.385,
+    "train_steps_per_second": 0.766
+}

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 +1,107 @@
+{
+  "_name_or_path": "facebook/wav2vec2-xls-r-300m",
+  "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": 768,
+  "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_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": 768,
+  "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": 0.5595183372497559,
+    "eval_runtime": 161.956,
+    "eval_samples": 4723,
+    "eval_samples_per_second": 29.162,
+    "eval_steps_per_second": 2.433,
+    "eval_wer": 0.16865713938706936
+}

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 vgnf004
+# Started at Wed 30 Nov 07:37:30 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-xls-r-300m/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=18 threads=1
+# Finished at Wed 30 Nov 07:37:48 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,
+  "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 sha256:955e67ecbe9217132afe903b7188c9207bbfd552090feecd9fa0f1076b3dc3d8
+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": 0.6397389862060547,
+    "train_runtime": 13054.0758,
+    "train_samples": 18925,
+    "train_samples_per_second": 18.385,
+    "train_steps_per_second": 0.766
+}

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": 3.0458309650421143,
+      "eval_runtime": 143.841,
+      "eval_samples_per_second": 32.835,
+      "eval_steps_per_second": 2.739,
+      "eval_wer": 1.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 9.98e-05,
+      "loss": 2.9181,
+      "step": 1000
+    },
+    {
+      "epoch": 1.27,
+      "eval_loss": 1.1502786874771118,
+      "eval_runtime": 143.5317,
+      "eval_samples_per_second": 32.906,
+      "eval_steps_per_second": 2.745,
+      "eval_wer": 0.4722885713129023,
+      "step": 1000
+    },
+    {
+      "epoch": 1.9,
+      "eval_loss": 0.8274601101875305,
+      "eval_runtime": 142.4484,
+      "eval_samples_per_second": 33.156,
+      "eval_steps_per_second": 2.766,
+      "eval_wer": 0.3499858305331768,
+      "step": 1500
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 8.89111111111111e-05,
+      "loss": 0.7646,
+      "step": 2000
+    },
+    {
+      "epoch": 2.53,
+      "eval_loss": 0.6990081667900085,
+      "eval_runtime": 145.8955,
+      "eval_samples_per_second": 32.372,
+      "eval_steps_per_second": 2.701,
+      "eval_wer": 0.2845228938099672,
+      "step": 2000
+    },
+    {
+      "epoch": 3.17,
+      "eval_loss": 0.5828294157981873,
+      "eval_runtime": 147.7617,
+      "eval_samples_per_second": 31.964,
+      "eval_steps_per_second": 2.666,
+      "eval_wer": 0.2508602890571232,
+      "step": 2500
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 7.780000000000001e-05,
+      "loss": 0.5394,
+      "step": 3000
+    },
+    {
+      "epoch": 3.8,
+      "eval_loss": 0.5363064408302307,
+      "eval_runtime": 147.8939,
+      "eval_samples_per_second": 31.935,
+      "eval_steps_per_second": 2.664,
+      "eval_wer": 0.24873486903364236,
+      "step": 3000
+    },
+    {
+      "epoch": 4.44,
+      "eval_loss": 0.5467007756233215,
+      "eval_runtime": 151.8112,
+      "eval_samples_per_second": 31.111,
+      "eval_steps_per_second": 2.595,
+      "eval_wer": 0.21711671592243229,
+      "step": 3500
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 6.66888888888889e-05,
+      "loss": 0.4558,
+      "step": 4000
+    },
+    {
+      "epoch": 5.07,
+      "eval_loss": 0.5289521813392639,
+      "eval_runtime": 152.3052,
+      "eval_samples_per_second": 31.01,
+      "eval_steps_per_second": 2.587,
+      "eval_wer": 0.20901987773774341,
+      "step": 4000
+    },
+    {
+      "epoch": 5.7,
+      "eval_loss": 0.4992470443248749,
+      "eval_runtime": 148.0991,
+      "eval_samples_per_second": 31.891,
+      "eval_steps_per_second": 2.66,
+      "eval_wer": 0.20464758511801143,
+      "step": 4500
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 5.558888888888889e-05,
+      "loss": 0.3773,
+      "step": 5000
+    },
+    {
+      "epoch": 6.34,
+      "eval_loss": 0.49340707063674927,
+      "eval_runtime": 155.8909,
+      "eval_samples_per_second": 30.297,
+      "eval_steps_per_second": 2.527,
+      "eval_wer": 0.2051738796000162,
+      "step": 5000
+    },
+    {
+      "epoch": 6.97,
+      "eval_loss": 0.47003018856048584,
+      "eval_runtime": 146.5312,
+      "eval_samples_per_second": 32.232,
+      "eval_steps_per_second": 2.689,
+      "eval_wer": 0.19827132504756892,
+      "step": 5500
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 4.447777777777778e-05,
+      "loss": 0.3301,
+      "step": 6000
+    },
+    {
+      "epoch": 7.6,
+      "eval_loss": 0.4938233494758606,
+      "eval_runtime": 154.4839,
+      "eval_samples_per_second": 30.573,
+      "eval_steps_per_second": 2.55,
+      "eval_wer": 0.18738107768916237,
+      "step": 6000
+    },
+    {
+      "epoch": 8.24,
+      "eval_loss": 0.5363861322402954,
+      "eval_runtime": 163.224,
+      "eval_samples_per_second": 28.936,
+      "eval_steps_per_second": 2.414,
+      "eval_wer": 0.18934456094894944,
+      "step": 6500
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.336666666666667e-05,
+      "loss": 0.2938,
+      "step": 7000
+    },
+    {
+      "epoch": 8.87,
+      "eval_loss": 0.5169841051101685,
+      "eval_runtime": 164.5878,
+      "eval_samples_per_second": 28.696,
+      "eval_steps_per_second": 2.394,
+      "eval_wer": 0.18300878506943039,
+      "step": 7000
+    },
+    {
+      "epoch": 9.51,
+      "eval_loss": 0.5407580733299255,
+      "eval_runtime": 163.5687,
+      "eval_samples_per_second": 28.875,
+      "eval_steps_per_second": 2.409,
+      "eval_wer": 0.18149062790980122,
+      "step": 7500
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 2.225555555555556e-05,
+      "loss": 0.2674,
+      "step": 8000
+    },
+    {
+      "epoch": 10.14,
+      "eval_loss": 0.5580623149871826,
+      "eval_runtime": 162.7994,
+      "eval_samples_per_second": 29.011,
+      "eval_steps_per_second": 2.42,
+      "eval_wer": 0.173272337152342,
+      "step": 8000
+    },
+    {
+      "epoch": 10.77,
+      "eval_loss": 0.538887083530426,
+      "eval_runtime": 163.5851,
+      "eval_samples_per_second": 28.872,
+      "eval_steps_per_second": 2.409,
+      "eval_wer": 0.1719161167564066,
+      "step": 8500
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 1.1144444444444445e-05,
+      "loss": 0.24,
+      "step": 9000
+    },
+    {
+      "epoch": 11.41,
+      "eval_loss": 0.5344488024711609,
+      "eval_runtime": 163.7329,
+      "eval_samples_per_second": 28.846,
+      "eval_steps_per_second": 2.406,
+      "eval_wer": 0.17141006436986356,
+      "step": 9000
+    },
+    {
+      "epoch": 12.04,
+      "eval_loss": 0.5502966046333313,
+      "eval_runtime": 154.8226,
+      "eval_samples_per_second": 30.506,
+      "eval_steps_per_second": 2.545,
+      "eval_wer": 0.16863689729160763,
+      "step": 9500
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 3.3333333333333334e-08,
+      "loss": 0.211,
+      "step": 10000
+    },
+    {
+      "epoch": 12.67,
+      "eval_loss": 0.5595183372497559,
+      "eval_runtime": 157.1471,
+      "eval_samples_per_second": 30.055,
+      "eval_steps_per_second": 2.507,
+      "eval_wer": 0.16865713938706936,
+      "step": 10000
+    },
+    {
+      "epoch": 12.67,
+      "step": 10000,
+      "total_flos": 2.7136568655380324e+19,
+      "train_loss": 0.6397389862060547,
+      "train_runtime": 13054.0758,
+      "train_samples_per_second": 18.385,
+      "train_steps_per_second": 0.766
+    }
+  ],
+  "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:f2cb36695fb66bf763f5c29a45137dfc2f7f8480ce24a65a0341e1f64851ddc3
+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
+}