Upload 9 files

README.md

@@ -1,3 +1,99 @@
----
-license: mit
----
+
+---
+license: cc-by-nc-4.0
+tags:
+- mms
+- vits
+pipeline_tag: text-to-speech
+---
+
+# Massively Multilingual Speech (MMS): Swahili Text-to-Speech
+
+This repository contains the **Swahili (swh)** language text-to-speech (TTS) model checkpoint.
+
+This model is part of Facebook's [Massively Multilingual Speech](https://arxiv.org/abs/2305.13516) project, aiming to
+provide speech technology across a diverse range of languages. You can find more details about the supported languages
+and their ISO 639-3 codes in the [MMS Language Coverage Overview](https://dl.fbaipublicfiles.com/mms/misc/language_coverage_mms.html),
+and see all MMS-TTS checkpoints on the Hugging Face Hub: [facebook/mms-tts](https://huggingface.co/models?sort=trending&search=facebook%2Fmms-tts).
+
+MMS-TTS is available in the 🤗 Transformers library from version 4.33 onwards.
+
+## Model Details
+
+VITS (**V**ariational **I**nference with adversarial learning for end-to-end **T**ext-to-**S**peech) is an end-to-end 
+speech synthesis model that predicts a speech waveform conditional on an input text sequence. It is a conditional variational 
+autoencoder (VAE) comprised of a posterior encoder, decoder, and conditional prior.
+
+A set of spectrogram-based acoustic features are predicted by the flow-based module, which is formed of a Transformer-based
+text encoder and multiple coupling layers. The spectrogram is decoded using a stack of transposed convolutional layers,
+much in the same style as the HiFi-GAN vocoder. Motivated by the one-to-many nature of the TTS problem, where the same text 
+input can be spoken in multiple ways, the model also includes a stochastic duration predictor, which allows the model to 
+synthesise speech with different rhythms from the same input text. 
+
+The model is trained end-to-end with a combination of losses derived from variational lower bound and adversarial training. 
+To improve the expressiveness of the model, normalizing flows are applied to the conditional prior distribution. During 
+inference, the text encodings are up-sampled based on the duration prediction module, and then mapped into the 
+waveform using a cascade of the flow module and HiFi-GAN decoder. Due to the stochastic nature of the duration predictor,
+the model is non-deterministic, and thus requires a fixed seed to generate the same speech waveform.
+
+For the MMS project, a separate VITS checkpoint is trained on each langauge.
+
+## Usage
+
+MMS-TTS is available in the 🤗 Transformers library from version 4.33 onwards. To use this checkpoint, 
+first install the latest version of the library:
+
+```
+pip install --upgrade transformers accelerate
+```
+
+Then, run inference with the following code-snippet:
+
+```python
+from transformers import VitsModel, AutoTokenizer
+import torch
+
+model = VitsModel.from_pretrained("facebook/mms-tts-swh")
+tokenizer = AutoTokenizer.from_pretrained("facebook/mms-tts-swh")
+
+text = "some example text in the Swahili language"
+inputs = tokenizer(text, return_tensors="pt")
+
+with torch.no_grad():
+    output = model(**inputs).waveform
+```
+
+The resulting waveform can be saved as a `.wav` file:
+
+```python
+import scipy
+
+scipy.io.wavfile.write("techno.wav", rate=model.config.sampling_rate, data=output)
+```
+
+Or displayed in a Jupyter Notebook / Google Colab:
+
+```python
+from IPython.display import Audio
+
+Audio(output, rate=model.config.sampling_rate)
+```
+
+
+
+## BibTex citation
+
+This model was developed by Vineel Pratap et al. from Meta AI. If you use the model, consider citing the MMS paper:
+
+```
+@article{pratap2023mms,
+    title={Scaling Speech Technology to 1,000+ Languages},
+    author={Vineel Pratap and Andros Tjandra and Bowen Shi and Paden Tomasello and Arun Babu and Sayani Kundu and Ali Elkahky and Zhaoheng Ni and Apoorv Vyas and Maryam Fazel-Zarandi and Alexei Baevski and Yossi Adi and Xiaohui Zhang and Wei-Ning Hsu and Alexis Conneau and Michael Auli},
+    journal={arXiv},
+    year={2023}
+}
+```
+
+## License
+
+The model is licensed as **CC-BY-NC 4.0**.

config.json

@@ -0,0 +1,82 @@
+{
+  "activation_dropout": 0.1,
+  "architectures": [
+    "VitsModel"
+  ],
+  "attention_dropout": 0.1,
+  "depth_separable_channels": 2,
+  "depth_separable_num_layers": 3,
+  "duration_predictor_dropout": 0.5,
+  "duration_predictor_filter_channels": 256,
+  "duration_predictor_flow_bins": 10,
+  "duration_predictor_kernel_size": 3,
+  "duration_predictor_num_flows": 4,
+  "duration_predictor_tail_bound": 5.0,
+  "ffn_dim": 768,
+  "ffn_kernel_size": 3,
+  "flow_size": 192,
+  "hidden_act": "relu",
+  "hidden_dropout": 0.1,
+  "hidden_size": 192,
+  "initializer_range": 0.02,
+  "layer_norm_eps": 1e-05,
+  "layerdrop": 0.1,
+  "leaky_relu_slope": 0.1,
+  "model_type": "vits",
+  "noise_scale": 0.667,
+  "noise_scale_duration": 0.8,
+  "num_attention_heads": 2,
+  "num_hidden_layers": 6,
+  "num_speakers": 1,
+  "posterior_encoder_num_wavenet_layers": 16,
+  "prior_encoder_num_flows": 4,
+  "prior_encoder_num_wavenet_layers": 4,
+  "resblock_dilation_sizes": [
+    [
+      1,
+      3,
+      5
+    ],
+    [
+      1,
+      3,
+      5
+    ],
+    [
+      1,
+      3,
+      5
+    ]
+  ],
+  "resblock_kernel_sizes": [
+    3,
+    7,
+    11
+  ],
+  "sampling_rate": 16000,
+  "speaker_embedding_size": 0,
+  "speaking_rate": 1.0,
+  "spectrogram_bins": 513,
+  "torch_dtype": "float32",
+  "transformers_version": "4.33.0.dev0",
+  "upsample_initial_channel": 512,
+  "upsample_kernel_sizes": [
+    16,
+    16,
+    4,
+    4
+  ],
+  "upsample_rates": [
+    8,
+    8,
+    2,
+    2
+  ],
+  "use_bias": true,
+  "use_stochastic_duration_prediction": true,
+  "vocab_size": 39,
+  "wavenet_dilation_rate": 1,
+  "wavenet_dropout": 0.0,
+  "wavenet_kernel_size": 5,
+  "window_size": 4
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c830aa67ab9199c036e92274ce6b3a31ddf1fa9434b66b70dd485f81b87dc2f8
+size 145228280

preprocessor_config.json

@@ -0,0 +1,11 @@
+{
+  "feature_extractor_type": "VitsFeatureExtractor",
+  "feature_size": 80,
+  "hop_length": 256,
+  "max_wav_value": 32768.0,
+  "n_fft": 1024,
+  "padding_side": "right",
+  "padding_value": 0.0,
+  "return_attention_mask": false,
+  "sampling_rate": 16000
+}

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1cbdf4e40ad12e6801391272b043f30907026891302e5dca4638da1edd14b36e
+size 145389490

special_tokens_map.json

@@ -0,0 +1,4 @@
+{
+  "pad_token": "2",
+  "unk_token": "<unk>"
+}

tokenizer_config.json

@@ -0,0 +1,12 @@
+{
+  "add_blank": true,
+  "clean_up_tokenization_spaces": true,
+  "is_uroman": false,
+  "language": "swh",
+  "model_max_length": 1000000000000000019884624838656,
+  "normalize": true,
+  "pad_token": "2",
+  "phonemize": false,
+  "tokenizer_class": "VitsTokenizer",
+  "unk_token": "<unk>"
+}

vocab.json

@@ -0,0 +1,41 @@
+{
+  " ": 34,
+  "'": 36,
+  "-": 9,
+  "0": 13,
+  "1": 23,
+  "2": 0,
+  "3": 24,
+  "4": 18,
+  "5": 31,
+  "6": 27,
+  "7": 17,
+  "8": 7,
+  "_": 38,
+  "a": 29,
+  "b": 19,
+  "c": 8,
+  "d": 2,
+  "e": 11,
+  "f": 20,
+  "g": 14,
+  "h": 37,
+  "i": 22,
+  "j": 4,
+  "k": 10,
+  "l": 5,
+  "m": 1,
+  "n": 32,
+  "o": 25,
+  "p": 12,
+  "q": 33,
+  "r": 3,
+  "s": 6,
+  "t": 26,
+  "u": 15,
+  "v": 21,
+  "w": 16,
+  "y": 30,
+  "z": 28,
+  "ʼ": 35
+}