jonatasgrosman/exp_w2v2r_de_xls-r_age_teens-8_sixties-2_s136

.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
+cv7_transcriptions.json filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,17 @@
+---
+language:
+- de
+license: apache-2.0
+tags:
+- automatic-speech-recognition
+- de
+datasets:
+- mozilla-foundation/common_voice_7_0
+---
+# exp_w2v2r_de_xls-r_age_teens-8_sixties-2_s136
+
+Fine-tuned [facebook/wav2vec2-xls-r-300m](https://huggingface.co/facebook/wav2vec2-xls-r-300m) for speech recognition using the train split of [Common Voice 7.0 (de)](https://huggingface.co/datasets/mozilla-foundation/common_voice_7_0).
+When using this model, make sure that your speech input is sampled at 16kHz.
+
+This model has been fine-tuned by the [HuggingSound](https://github.com/jonatasgrosman/huggingsound) tool.
+    

config.json

@@ -0,0 +1,108 @@
+{
+  "_name_or_path": "facebook/wav2vec2-xls-r-300m",
+  "activation_dropout": 0.05,
+  "adapter_kernel_size": 3,
+  "adapter_stride": 2,
+  "add_adapter": false,
+  "apply_spec_augment": true,
+  "architectures": [
+    "Wav2Vec2ForCTC"
+  ],
+  "attention_dropout": 0.05,
+  "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": false,
+  "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": 0.05,
+  "gradient_checkpointing": false,
+  "hidden_act": "gelu",
+  "hidden_dropout": 0.05,
+  "hidden_size": 1024,
+  "initializer_range": 0.02,
+  "intermediate_size": 4096,
+  "layer_norm_eps": 1e-05,
+  "layerdrop": 0.05,
+  "mask_feature_length": 10,
+  "mask_feature_min_masks": 0,
+  "mask_feature_prob": 0.0,
+  "mask_time_length": 10,
+  "mask_time_min_masks": 2,
+  "mask_time_prob": 0.05,
+  "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": 0,
+  "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": "4.18.0",
+  "use_weighted_layer_sum": false,
+  "vocab_size": 38,
+  "xvector_output_dim": 512
+}

cv7_evaluation.json

@@ -0,0 +1,396 @@
+{
+  "validation": [
+    {
+      "subset_type": "all",
+      "subset": "all",
+      "cer": 0.04492097390914287,
+      "wer": 0.18271583124932042
+    },
+    {
+      "subset_type": "age",
+      "subset": "",
+      "cer": 0.04427924472032959,
+      "wer": 0.18112412684933105
+    },
+    {
+      "subset_type": "age",
+      "subset": "twenties",
+      "cer": 0.047698760163349396,
+      "wer": 0.1942563330825182
+    },
+    {
+      "subset_type": "age",
+      "subset": "thirties",
+      "cer": 0.04144505770195685,
+      "wer": 0.16447368421052633
+    },
+    {
+      "subset_type": "age",
+      "subset": "teens",
+      "cer": 0.04390695647436945,
+      "wer": 0.18682545589877186
+    },
+    {
+      "subset_type": "age",
+      "subset": "seventies",
+      "cer": 0.0432113920942794,
+      "wer": 0.195578231292517
+    },
+    {
+      "subset_type": "age",
+      "subset": "fourties",
+      "cer": 0.06328320802005012,
+      "wer": 0.22857142857142856
+    },
+    {
+      "subset_type": "age",
+      "subset": "eighties",
+      "cer": 0.05057096247960848,
+      "wer": 0.23170731707317074
+    },
+    {
+      "subset_type": "age",
+      "subset": "fifties",
+      "cer": 0.04711268645694875,
+      "wer": 0.18745785569790965
+    },
+    {
+      "subset_type": "age",
+      "subset": "sixties",
+      "cer": 0.051874907366236844,
+      "wer": 0.20020909566126502
+    },
+    {
+      "subset_type": "gender",
+      "subset": "",
+      "cer": 0.04427881875049604,
+      "wer": 0.18112563483473482
+    },
+    {
+      "subset_type": "gender",
+      "subset": "female",
+      "cer": 0.04561535566428303,
+      "wer": 0.1796267496111975
+    },
+    {
+      "subset_type": "gender",
+      "subset": "other",
+      "cer": 0.0765639589169001,
+      "wer": 0.2804878048780488
+    },
+    {
+      "subset_type": "gender",
+      "subset": "male",
+      "cer": 0.04798255873080443,
+      "wer": 0.19083405734144224
+    },
+    {
+      "subset_type": "accent",
+      "subset": "",
+      "cer": 0.04483434875755759,
+      "wer": 0.18282481699036537
+    },
+    {
+      "subset_type": "accent",
+      "subset": "canada",
+      "cer": 0.08455882352941177,
+      "wer": 0.325
+    },
+    {
+      "subset_type": "accent",
+      "subset": "france",
+      "cer": 0.05067567567567568,
+      "wer": 0.22916666666666666
+    },
+    {
+      "subset_type": "accent",
+      "subset": "belgium",
+      "cer": 0.05405405405405406,
+      "wer": 0.2692307692307692
+    },
+    {
+      "subset_type": "accent",
+      "subset": "slovakia",
+      "cer": 0.04722222222222222,
+      "wer": 0.22807017543859648
+    },
+    {
+      "subset_type": "accent",
+      "subset": "netherlands",
+      "cer": 0.05723905723905724,
+      "wer": 0.21739130434782608
+    },
+    {
+      "subset_type": "accent",
+      "subset": "switzerland",
+      "cer": 0.05278263039680559,
+      "wer": 0.2141623488773748
+    },
+    {
+      "subset_type": "accent",
+      "subset": "luxembourg",
+      "cer": 0.08353510895883777,
+      "wer": 0.3140495867768595
+    },
+    {
+      "subset_type": "accent",
+      "subset": "czechia",
+      "cer": 0.038527397260273974,
+      "wer": 0.14545454545454545
+    },
+    {
+      "subset_type": "accent",
+      "subset": "turkey",
+      "cer": 0.13022113022113022,
+      "wer": 0.4915254237288136
+    },
+    {
+      "subset_type": "accent",
+      "subset": "russia",
+      "cer": 0.0674061433447099,
+      "wer": 0.24064171122994651
+    },
+    {
+      "subset_type": "accent",
+      "subset": "austria",
+      "cer": 0.04820165094339623,
+      "wer": 0.19238578680203045
+    },
+    {
+      "subset_type": "accent",
+      "subset": "poland",
+      "cer": 0.04390243902439024,
+      "wer": 0.18181818181818182
+    },
+    {
+      "subset_type": "accent",
+      "subset": "united_kingdom",
+      "cer": 0.08549222797927461,
+      "wer": 0.3162393162393162
+    },
+    {
+      "subset_type": "accent",
+      "subset": "other",
+      "cer": 0.05748502994011976,
+      "wer": 0.216
+    },
+    {
+      "subset_type": "accent",
+      "subset": "germany",
+      "cer": 0.04285577551964522,
+      "wer": 0.1729335672001754
+    },
+    {
+      "subset_type": "accent",
+      "subset": "united_states",
+      "cer": 0.0861797162375197,
+      "wer": 0.303886925795053
+    }
+  ],
+  "test": [
+    {
+      "subset_type": "all",
+      "subset": "all",
+      "cer": 0.05115450300694033,
+      "wer": 0.1985494106980961
+    },
+    {
+      "subset_type": "age",
+      "subset": "",
+      "cer": 0.05031743978013447,
+      "wer": 0.1961445598436362
+    },
+    {
+      "subset_type": "age",
+      "subset": "twenties",
+      "cer": 0.053694690969240066,
+      "wer": 0.21451717433570966
+    },
+    {
+      "subset_type": "age",
+      "subset": "fifties",
+      "cer": 0.046683399067766224,
+      "wer": 0.19602977667493796
+    },
+    {
+      "subset_type": "age",
+      "subset": "teens",
+      "cer": 0.0642579129029889,
+      "wer": 0.23496993987975953
+    },
+    {
+      "subset_type": "age",
+      "subset": "seventies",
+      "cer": 0.06611368626494267,
+      "wer": 0.2571912013536379
+    },
+    {
+      "subset_type": "age",
+      "subset": "fourties",
+      "cer": 0.07876517064429421,
+      "wer": 0.25041551246537397
+    },
+    {
+      "subset_type": "age",
+      "subset": "eighties",
+      "cer": 0.06788990825688074,
+      "wer": 0.2876712328767123
+    },
+    {
+      "subset_type": "age",
+      "subset": "thirties",
+      "cer": 0.05276852147715342,
+      "wer": 0.19738442822384428
+    },
+    {
+      "subset_type": "age",
+      "subset": "sixties",
+      "cer": 0.07248348231636223,
+      "wer": 0.2547683923705722
+    },
+    {
+      "subset_type": "gender",
+      "subset": "",
+      "cer": 0.05039616258610655,
+      "wer": 0.19636973811238642
+    },
+    {
+      "subset_type": "gender",
+      "subset": "female",
+      "cer": 0.06805109886783314,
+      "wer": 0.2554438860971524
+    },
+    {
+      "subset_type": "gender",
+      "subset": "other",
+      "cer": 0.1245819397993311,
+      "wer": 0.2556818181818182
+    },
+    {
+      "subset_type": "gender",
+      "subset": "male",
+      "cer": 0.054841443604774345,
+      "wer": 0.20961814986419555
+    },
+    {
+      "subset_type": "accent",
+      "subset": "",
+      "cer": 0.051013207868536255,
+      "wer": 0.19819005212528304
+    },
+    {
+      "subset_type": "accent",
+      "subset": "austria",
+      "cer": 0.05869682283252558,
+      "wer": 0.22437673130193905
+    },
+    {
+      "subset_type": "accent",
+      "subset": "united_states",
+      "cer": 0.0743651753325272,
+      "wer": 0.24066390041493776
+    },
+    {
+      "subset_type": "accent",
+      "subset": "belgium",
+      "cer": 0.07614213197969544,
+      "wer": 0.38095238095238093
+    },
+    {
+      "subset_type": "accent",
+      "subset": "luxembourg",
+      "cer": 0.045774647887323945,
+      "wer": 0.18421052631578946
+    },
+    {
+      "subset_type": "accent",
+      "subset": "denmark",
+      "cer": 0.07352941176470588,
+      "wer": 0.4444444444444444
+    },
+    {
+      "subset_type": "accent",
+      "subset": "czechia",
+      "cer": 0.15176151761517614,
+      "wer": 0.4716981132075472
+    },
+    {
+      "subset_type": "accent",
+      "subset": "united_kingdom",
+      "cer": 0.11442006269592477,
+      "wer": 0.35294117647058826
+    },
+    {
+      "subset_type": "accent",
+      "subset": "france",
+      "cer": 0.07856450048496605,
+      "wer": 0.28
+    },
+    {
+      "subset_type": "accent",
+      "subset": "slovakia",
+      "cer": 0.047979797979797977,
+      "wer": 0.1896551724137931
+    },
+    {
+      "subset_type": "accent",
+      "subset": "bulgaria",
+      "cer": 0.0,
+      "wer": 0.0
+    },
+    {
+      "subset_type": "accent",
+      "subset": "russia",
+      "cer": 0.16835016835016836,
+      "wer": 0.4666666666666667
+    },
+    {
+      "subset_type": "accent",
+      "subset": "greece",
+      "cer": 0.09615384615384616,
+      "wer": 0.2222222222222222
+    },
+    {
+      "subset_type": "accent",
+      "subset": "netherlands",
+      "cer": 0.06429652042360061,
+      "wer": 0.26424870466321243
+    },
+    {
+      "subset_type": "accent",
+      "subset": "switzerland",
+      "cer": 0.06454231784047088,
+      "wer": 0.2338935574229692
+    },
+    {
+      "subset_type": "accent",
+      "subset": "turkey",
+      "cer": 0.1935483870967742,
+      "wer": 0.4
+    },
+    {
+      "subset_type": "accent",
+      "subset": "poland",
+      "cer": 0.09866220735785954,
+      "wer": 0.32608695652173914
+    },
+    {
+      "subset_type": "accent",
+      "subset": "other",
+      "cer": 0.0,
+      "wer": 0.0
+    },
+    {
+      "subset_type": "accent",
+      "subset": "germany",
+      "cer": 0.04803580176625578,
+      "wer": 0.18911123129619614
+    },
+    {
+      "subset_type": "accent",
+      "subset": "hungary",
+      "cer": 0.06153846153846154,
+      "wer": 0.42857142857142855
+    }
+  ]
+}

cv7_transcriptions.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:03f719b8bfa255a880e980c91acbc8c8d49248ee3ab3f5f3a20c04f78f9f8f82
+size 138771755

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": "Wav2Vec2Processor",
+  "return_attention_mask": true,
+  "sampling_rate": 16000
+}

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eb8476e7705919db4544f6d9aa9caf7c515e0a8ec17e1fe4a7f7f2293dd084e8
+size 1262054513

special_tokens_map.json

@@ -0,0 +1 @@
+{"bos_token": "<s>", "eos_token": "</s>", "unk_token": "<unk>", "pad_token": "<pad>"}

tokenizer_config.json

@@ -0,0 +1 @@
+{"unk_token": "<unk>", "bos_token": "<s>", "eos_token": "</s>", "pad_token": "<pad>", "do_lower_case": false, "word_delimiter_token": "|", "replace_word_delimiter_char": " ", "special_tokens_map_file": null, "name_or_path": "/tmp/tmpxe2f7vu4", "tokenizer_class": "Wav2Vec2CTCTokenizer", "processor_class": "Wav2Vec2Processor"}

vocab.json

@@ -0,0 +1 @@
+{"<pad>": 0, "<s>": 1, "</s>": 2, "<unk>": 3, "|": 4, "'": 5, "-": 6, "a": 7, "b": 8, "c": 9, "d": 10, "e": 11, "f": 12, "g": 13, "h": 14, "i": 15, "j": 16, "k": 17, "l": 18, "m": 19, "n": 20, "o": 21, "p": 22, "q": 23, "r": 24, "s": 25, "t": 26, "u": 27, "v": 28, "w": 29, "x": 30, "y": 31, "z": 32, "ß": 33, "á": 34, "ä": 35, "ö": 36, "ü": 37}