add files

README.md

@@ -0,0 +1,120 @@
+---
+language:
+- en
+- fa
+datasets:
+- covost2
+tags:
+- audio
+- speech-translation
+- automatic-speech-recognition
+license: MIT
+---
+
+
+# S2T-SMALL-COVOST2-EN-FA-ST
+
+`s2t-small-covost2-en-fa-st` is a Speech to Text Transformer (S2T) model trained for end-to-end Speech Translation (ST).
+The S2T model was proposed in [this paper](https://arxiv.org/abs/2010.05171) and released in
+[this repository](https://github.com/pytorch/fairseq/tree/master/examples/speech_to_text)
+
+
+## Model description
+
+S2T is a transformer-based seq2seq (encoder-decoder) model designed for end-to-end Automatic Speech Recognition (ASR) and Speech
+Translation (ST). It uses a convolutional downsampler to reduce the length of speech inputs by 3/4th before they are
+fed into the encoder. The model is trained with standard autoregressive cross-entropy loss and generates the
+transcripts/translations autoregressively.
+
+## Intended uses & limitations
+
+This model can be used for end-to-end English speech to Farsi text translation.
+See the [model hub](https://huggingface.co/models?filter=speech_to_text) to look for other S2T checkpoints.
+
+
+### How to use
+
+As this a standard sequence to sequence transformer model, you can use the `generate` method to generate the
+transcripts by passing the speech features to the model.
+
+*Note: The `Speech2TextProcessor` object uses [torchaudio](https://github.com/pytorch/audio)  to extract the
+filter bank features. Make sure to install the `torchaudio` package before running this example.*
+
+You could either install those as extra speech dependancies with
+`pip install transformers"[speech, sentencepiece]"` or install the packages seperatly 
+with `pip install torchaudio sentencepiece`.
+
+
+```python
+import torch
+from transformers import Speech2TextProcessor, Speech2TextForConditionalGeneration
+from datasets import load_dataset
+import soundfile as sf
+
+model = Speech2TextForConditionalGeneration.from_pretrained("facebook/s2t-small-covost2-en-fa-st")
+processor = Speech2TextProcessor.from_pretrained("facebook/s2t-small-covost2-en-fa-st")
+
+def map_to_array(batch):
+    speech, _ = sf.read(batch["file"])
+    batch["speech"] = speech
+    return batch
+
+ds = load_dataset(
+    "patrickvonplaten/librispeech_asr_dummy",
+    "clean",
+    split="validation"
+)
+ds = ds.map(map_to_array)
+
+inputs = processor(
+    ds["speech"][0],
+    sampling_rate=48_000,
+    return_tensors="pt"
+)
+generated_ids = model.generate(input_ids=inputs["input_features"], attention_mask=inputs["attention_mask"])
+
+translation = processor.batch_decode(generated_ids, skip_special_tokens=True)
+```
+
+
+## Training data
+
+The s2t-small-covost2-en-fa-st is trained on English-Farsi subset of [CoVoST2](https://github.com/facebookresearch/covost).
+CoVoST is a large-scale multilingual ST corpus based on [Common Voice](https://arxiv.org/abs/1912.06670), created to to foster 
+ST research with the largest ever open dataset
+
+
+## Training procedure
+
+### Preprocessing
+
+The speech data is pre-processed by extracting Kaldi-compliant 80-channel log mel-filter bank features automatically from
+WAV/FLAC audio files via PyKaldi or torchaudio. Further utterance-level CMVN (cepstral mean and variance normalization)
+is applied to each example.
+
+The texts are lowercased and tokenized using character based SentencePiece vocab.
+
+
+### Training
+
+The model is trained with standard autoregressive cross-entropy loss and using [SpecAugment](https://arxiv.org/abs/1904.08779).
+The encoder receives speech features, and the decoder generates the transcripts autoregressively. To accelerate
+model training and for better performance the encoder is pre-trained for English ASR.
+
+## Evaluation results
+
+CoVOST2 test results for en-fa (BLEU score): 11.43
+
+
+
+### BibTeX entry and citation info
+
+```bibtex
+@inproceedings{wang2020fairseqs2t,
+  title = {fairseq S2T: Fast Speech-to-Text Modeling with fairseq},
+  author = {Changhan Wang and Yun Tang and Xutai Ma and Anne Wu and Dmytro Okhonko and Juan Pino},
+  booktitle = {Proceedings of the 2020 Conference of the Asian Chapter of the Association for Computational Linguistics (AACL): System Demonstrations},
+  year = {2020},
+}
+
+```

config.json

@@ -0,0 +1,46 @@
+{
+  "activation_dropout": 0.1,
+  "activation_function": "relu",
+  "architectures": [
+    "Speech2TextForConditionalGeneration"
+  ],
+  "attention_dropout": 0.1,
+  "bos_token_id": 0,
+  "classifier_dropout": 0.0,
+  "conv_channels": 1024,
+  "conv_kernel_sizes": [
+    5,
+    5
+  ],
+  "d_model": 256,
+  "decoder_attention_heads": 4,
+  "decoder_ffn_dim": 2048,
+  "decoder_layerdrop": 0.0,
+  "decoder_layers": 6,
+  "decoder_start_token_id": 2,
+  "dropout": 0.1,
+  "early_stopping": true,
+  "encoder_attention_heads": 4,
+  "encoder_ffn_dim": 2048,
+  "encoder_layerdrop": 0.0,
+  "encoder_layers": 12,
+  "eos_token_id": 2,
+  "gradient_checkpointing": false,
+  "init_std": 0.02,
+  "input_channels": 1,
+  "input_feat_per_channel": 80,
+  "is_encoder_decoder": true,
+  "max_length": 200,
+  "max_source_positions": 6000,
+  "max_target_positions": 1024,
+  "model_type": "speech_to_text",
+  "num_beams": 5,
+  "num_conv_layers": 2,
+  "num_hidden_layers": 12,
+  "pad_token_id": 1,
+  "scale_embedding": true,
+  "tie_word_embeddings": false,
+  "transformers_version": "4.4.0.dev0",
+  "use_cache": true,
+  "vocab_size": 226
+}

preprocessor_config.json

@@ -0,0 +1,11 @@
+{
+  "do_ceptral_normalize": true,
+  "feature_size": 80,
+  "normalize_means": true,
+  "normalize_vars": true,
+  "num_mel_bins": 80,
+  "padding_side": "right",
+  "padding_value": 0.0,
+  "return_attention_mask": true,
+  "sampling_rate": 48000
+}

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cac7668055b6192f23161ad487765e11212897280956aa34aa22500af01a01ba
+size 108490245

sentencepiece.bpe.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:25987ef895641543e2d555789dac1871bf90f9032916e4979aa44d93983dec32
+size 240137

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 @@
+{"bos_token": "<s>", "eos_token": "</s>", "unk_token": "<unk>", "pad_token": "<pad>", "do_upper_case": false, "do_lower_case": false, "tgt_lang": null, "lang_codes": null, "tokenizer_file": null}

vocab.json

@@ -0,0 +1,228 @@
+{
+  "<s>": 0,
+  "<pad>": 1,
+  "</s>": 2,
+  "<unk>": 3,
+  "\u2581": 4,
+  "\u0627": 5,
+  "\u06cc": 6,
+  "\u0631": 7,
+  "\u062f": 8,
+  "\u0646": 9,
+  "\u0648": 10,
+  "\u0647": 11,
+  "\u0645": 12,
+  "\u062a": 13,
+  "\u0628": 14,
+  "\u0633": 15,
+  "\u06a9": 16,
+  "\u0644": 17,
+  "\u0634": 18,
+  ".": 19,
+  "\u0632": 20,
+  "\u06af": 21,
+  "\u0641": 22,
+  "\u062e": 23,
+  "\u0642": 24,
+  "\u062c": 25,
+  "\u0639": 26,
+  "\u067e": 27,
+  "\u0622": 28,
+  "\u060c": 29,
+  "\u062d": 30,
+  "\u0637": 31,
+  "\u0686": 32,
+  "\u0635": 33,
+  "e": 34,
+  "a": 35,
+  "\u063a": 36,
+  "\"": 37,
+  "r": 38,
+  "o": 39,
+  "i": 40,
+  "n": 41,
+  "\u0636": 42,
+  "\u0630": 43,
+  "\u0638": 44,
+  "l": 45,
+  "t": 46,
+  "s": 47,
+  "\u00ab": 48,
+  "\u00bb": 49,
+  "\u062b": 50,
+  "\u0626": 51,
+  "u": 52,
+  "\u0698": 53,
+  "h": 54,
+  "\u061f": 55,
+  "(": 56,
+  ")": 57,
+  "d": 58,
+  "\u064b": 59,
+  "\u06d5": 60,
+  "c": 61,
+  "m": 62,
+  "g": 63,
+  "y": 64,
+  "S": 65,
+  "k": 66,
+  "-": 67,
+  "C": 68,
+  "p": 69,
+  ":": 70,
+  "!": 71,
+  "B": 72,
+  "M": 73,
+  "b": 74,
+  "\u0643": 75,
+  "T": 76,
+  "A": 77,
+  "w": 78,
+  "P": 79,
+  "v": 80,
+  "\u0623": 81,
+  "\u0650": 82,
+  "R": 83,
+  "f": 84,
+  "\u064a": 85,
+  "D": 86,
+  "H": 87,
+  "G": 88,
+  "L": 89,
+  "W": 90,
+  "F": 91,
+  "E": 92,
+  "K": 93,
+  "N": 94,
+  "J": 95,
+  "O": 96,
+  "z": 97,
+  "\u06c0": 98,
+  "I": 99,
+  "\u201d": 100,
+  "\u201c": 101,
+  "\u0621": 102,
+  "V": 103,
+  "x": 104,
+  "<": 105,
+  ">": 106,
+  "\u061b": 107,
+  "\u0624": 108,
+  "\u064f": 109,
+  "'": 110,
+  ",": 111,
+  "\u0649": 112,
+  "U": 113,
+  "Y": 114,
+  "j": 115,
+  "\u064e": 116,
+  "q": 117,
+  "\u2013": 118,
+  "Z": 119,
+  "\u0640": 120,
+  "X": 121,
+  "Q": 122,
+  "\u0651": 123,
+  "[": 124,
+  "]": 125,
+  "/": 126,
+  "\u0654": 127,
+  "1": 128,
+  "_": 129,
+  "\u2019": 130,
+  "0": 131,
+  "5": 132,
+  "\u2014": 133,
+  "\u0629": 134,
+  "\u064c": 135,
+  "2": 136,
+  "9": 137,
+  "3": 138,
+  "6": 139,
+  "?": 140,
+  "\u06ce": 141,
+  "4": 142,
+  "&": 143,
+  "\u06f1": 144,
+  "\u06f2": 145,
+  "8": 146,
+  "7": 147,
+  "*": 148,
+  "\\": 149,
+  "\u00ad": 150,
+  "\u00f1": 151,
+  "\u00ac": 152,
+  "\u00e1": 153,
+  "\u2018": 154,
+  "\u00e9": 155,
+  "\u00f6": 156,
+  "\u0670": 157,
+  ";": 158,
+  "\u00f3": 159,
+  "\u00e4": 160,
+  "\u014d": 161,
+  "\u06f0": 162,
+  "\u06f9": 163,
+  "\u0625": 164,
+  "\u06f5": 165,
+  "\u00fc": 166,
+  "\u0652": 167,
+  "\u00ed": 168,
+  "\u06f4": 169,
+  "=": 170,
+  "\u0101": 171,
+  "\u06b5": 172,
+  "\u06f6": 173,
+  "\u00d6": 174,
+  "\u00fa": 175,
+  "\u06f3": 176,
+  "\u06f8": 177,
+  "\u00c9": 178,
+  "\u00e3": 179,
+  "\u00e8": 180,
+  "\u00f0": 181,
+  "\u00f8": 182,
+  "\u0103": 183,
+  "\u0142": 184,
+  "\u03c0": 185,
+  "\u0438": 186,
+  "\u043a": 187,
+  "\u064d": 188,
+  "\u066c": 189,
+  "\u06a4": 190,
+  "%": 191,
+  "\u00c3": 192,
+  "\u00d3": 193,
+  "\u00e2": 194,
+  "\u00e6": 195,
+  "\u00f4": 196,
+  "\u00fe": 197,
+  "\u0107": 198,
+  "\u010c": 199,
+  "\u010d": 200,
+  "\u012b": 201,
+  "\u0153": 202,
+  "\u0160": 203,
+  "\u0161": 204,
+  "\u016b": 205,
+  "\u02bb": 206,
+  "\u0412": 207,
+  "\u0435": 208,
+  "\u0437": 209,
+  "\u0439": 210,
+  "\u043b": 211,
+  "\u043d": 212,
+  "\u044c": 213,
+  "\u044f": 214,
+  "\u0695": 215,
+  "\u0696": 216,
+  "\u06c6": 217,
+  "\u06f7": 218,
+  "\u1e43": 219,
+  "\u2261": 220,
+  "\u2500": 221,
+  "\u4eac": 222,
+  "\u5927": 223,
+  "\u90fd": 224,
+  "\u962a": 225
+}