Add gradio files

README.md

@@ -1,8 +1,8 @@
 ---
-title: Luganda Asr
-emoji: 🏃
+title: Luganda ASR
+emoji: 🌍
 colorFrom: blue
-colorTo: indigo
+colorTo: yellow
 sdk: gradio
 app_file: app.py
 pinned: false
@@ -10,28 +10,28 @@ pinned: false
 
 # Configuration
 
-`title`: _string_
+`title`: _string_  
 Display title for the Space
 
-`emoji`: _string_
+`emoji`: _string_  
 Space emoji (emoji-only character allowed)
 
-`colorFrom`: _string_
+`colorFrom`: _string_  
 Color for Thumbnail gradient (red, yellow, green, blue, indigo, purple, pink, gray)
 
-`colorTo`: _string_
+`colorTo`: _string_  
 Color for Thumbnail gradient (red, yellow, green, blue, indigo, purple, pink, gray)
 
-`sdk`: _string_
-Can be either `gradio`, `streamlit`, or `static`
+`sdk`: _string_  
+Can be either `gradio` or `streamlit`
 
-`sdk_version` : _string_
+`sdk_version` : _string_  
 Only applicable for `streamlit` SDK.  
 See [doc](https://hf.co/docs/hub/spaces) for more info on supported versions.
 
-`app_file`: _string_
-Path to your main application file (which contains either `gradio` or `streamlit` Python code, or `static` html code).
+`app_file`: _string_  
+Path to your main application file (which contains either `gradio` or `streamlit` Python code).  
 Path is relative to the root of the repository.
 
-`pinned`: _boolean_
+`pinned`: _boolean_  
 Whether the Space stays on top of your list.

app.py

@@ -0,0 +1,49 @@
+import soundfile as sf
+import torch
+from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
+import gradio as gr
+import sox
+import os
+
+
+def convert(inputfile, outfile):
+    sox_tfm = sox.Transformer()
+    sox_tfm.set_output_format(
+        file_type="wav", channels=1, encoding="signed-integer", rate=16000, bits=16
+    )
+    sox_tfm.build(inputfile, outfile)
+
+
+api_token = os.getenv("API_TOKEN")
+model_name = "indonesian-nlp/wav2vec2-luganda"
+processor = Wav2Vec2Processor.from_pretrained(model_name, use_auth_token=api_token)
+model = Wav2Vec2ForCTC.from_pretrained(model_name, use_auth_token=api_token)
+
+
+def parse_transcription(wav_file):
+    filename = wav_file.name.split('.')[0]
+    convert(wav_file.name, filename + "16k.wav")
+    speech, _ = sf.read(filename + "16k.wav")
+    input_values = processor(speech, sampling_rate=16_000, return_tensors="pt").input_values
+    logits = model(input_values).logits
+    predicted_ids = torch.argmax(logits, dim=-1)
+    transcription = processor.decode(predicted_ids[0], skip_special_tokens=True)
+    return transcription
+
+
+output = gr.outputs.Textbox(label="The transcript")
+
+input_ = gr.inputs.Audio(source="microphone", type="file")
+
+gr.Interface(parse_transcription, inputs=input_,  outputs=[output],
+             analytics_enabled=False,
+             show_tips=False,
+             theme='huggingface',
+             layout='vertical',
+             title="Automatic Speech Recognition for Luganda",
+             description="Speech Recognition Live Demo for Luganda",
+             article="This demo was built for the "
+                     "<a href='https://zindi.africa/competitions/mozilla-luganda-automatic-speech-recognition' target='_blank'>Mozilla Luganda Automatic Speech Recognition Competition</a>. "
+                     "It uses the <a href='https://huggingface.co/indonesian-nlp/wav2vec2-luganda' target='_blank'>indonesian-nlp/wav2vec2-luganda</a> model "
+                     "which was fine-tuned on Luganda Common Voice speech datasets.",
+             enable_queue=True).launch( inline=False)

packages.txt

@@ -0,0 +1,2 @@
+libsndfile1
+sox

requirements.txt

@@ -0,0 +1,6 @@
+gradio
+soundfile
+torch
+transformers
+sox
+sentencepiece