| import os | |
| os.system("pip install gradio==2.8.0b2") | |
| import gradio as gr | |
| import numpy as np | |
| title = "fairseq S2S" | |
| description = "Gradio Demo for fairseq S2S: speech-to-speech translation models. To use it, simply add your audio, or click one of the examples to load them. Read more at the links below." | |
| article = "<p style='text-align: center'><a href='https://arxiv.org/abs/2010.05171' target='_blank'>fairseq S2T: Fast Speech-to-Text Modeling with fairseq</a> | <a href='https://github.com/pytorch/fairseq/tree/main/examples/speech_to_text' target='_blank'>Github Repo</a></p>" | |
| examples = [ | |
| ["common_voice_es_en.flac","xm_transformer_600m-es_en-multi_domain"], | |
| ["common_voice_ru_18945535.flac","xm_transformer_600m-ru_en-multi_domain"], | |
| ["common_voice_fr_19731305.mp3","xm_transformer_600m-fr_en-multi_domain"], | |
| ["common_voice_en_ru.mp3","xm_transformer_600m-en_ru-multi_domain"], | |
| ["common_voice_en_es.mp3","xm_transformer_600m-en_es-multi_domain"], | |
| ["common_voice_en_ar.mp3","xm_transformer_600m-en_ar-multi_domain"], | |
| ["common_voice_en_tr.mp3","xm_transformer_600m-en_tr-multi_domain"], | |
| ["common_voice_en_fr.mp3","xm_transformer_600m-en_fr-multi_domain"], | |
| ["common_voice_en_vi.mp3","xm_transformer_600m-en_vi-multi_domain"], | |
| ] | |
| io1 = gr.Interface.load("huggingface/facebook/xm_transformer_600m-es_en-multi_domain") | |
| io2 = gr.Interface.load("huggingface/facebook/xm_transformer_600m-ru_en-multi_domain") | |
| io3 = gr.Interface.load("huggingface/facebook/xm_transformer_600m-en_ru-multi_domain") | |
| io4 = gr.Interface.load("huggingface/facebook/xm_transformer_600m-en_es-multi_domain") | |
| io5 = gr.Interface.load("huggingface/facebook/xm_transformer_600m-en_zh-multi_domain") | |
| io6 = gr.Interface.load("huggingface/facebook/xm_transformer_600m-fr_en-multi_domain") | |
| io7 = gr.Interface.load("huggingface/facebook/xm_transformer_600m-en_ar-multi_domain") | |
| io8 = gr.Interface.load("huggingface/facebook/xm_transformer_600m-en_tr-multi_domain") | |
| io9 = gr.Interface.load("huggingface/facebook/xm_transformer_600m-en_fr-multi_domain") | |
| io10 = gr.Interface.load("huggingface/facebook/xm_transformer_600m-en_vi-multi_domain") | |
| def inference(text,model): | |
| if model == "xm_transformer_600m-es_en-multi_domain": | |
| outtext = io1(text) | |
| elif model == "xm_transformer_600m-ru_en-multi_domain": | |
| outtext = io2(text) | |
| elif model == "xm_transformer_600m-en_ru-multi_domain": | |
| outtext = io3(text) | |
| elif model == "xm_transformer_600m-en_es-multi_domain": | |
| outtext = io4(text) | |
| elif model == "xm_transformer_600m-en_zh-multi_domain": | |
| outtext = io5(text) | |
| elif model == "xm_transformer_600m-fr_en-multi_domain": | |
| outtext = io6(text) | |
| elif model == "xm_transformer_600m-en_ar-multi_domain": | |
| outtext = io7(text) | |
| elif model == "xm_transformer_600m-en_tr-multi_domain": | |
| outtext = io8(text) | |
| elif model == "xm_transformer_600m-en_fr-multi_domain": | |
| outtext = io9(text) | |
| else: | |
| outtext = io10(text) | |
| return outtext | |
| gr.Interface( | |
| inference, | |
| [gr.inputs.Audio(label="Input",type="filepath"),gr.inputs.Dropdown(choices=["xm_transformer_600m-es_en-multi_domain","xm_transformer_600m-ru_en-multi_domain","xm_transformer_600m-en_ru-multi_domain","xm_transformer_600m-en_es-multi_domain","xm_transformer_600m-fr_en-multi_domain","xm_transformer_600m-en_ar-multi_domain","xm_transformer_600m-en_tr-multi_domain","xm_transformer_600m-en_fr-multi_domain","xm_transformer_600m-en_vi-multi_domain"], type="value", default="xm_transformer_600m-es_en-multi_domain", label="model") | |
| ], | |
| gr.outputs.Audio(label="Output"), | |
| article=article, | |
| title=title, | |
| examples=examples, | |
| description=description).launch(enable_queue=True,cache_examples=False) |