Added configuration files

README.md

@@ -8,6 +8,12 @@ sdk_version: 4.42.0
 app_file: app.py
 pinned: false
 license: cc-by-nc-sa-4.0
+models:
+- asr-africa/wav2vec2-xls-r-1b-naijavoices-hausa-500hr-v0
+- asr-africa/wav2vec2-xls-r-1b-naijavoices-igbo-500hr-v0
+- asr-africa/wav2vec2-xls-r-1b-naijavoices-yoruba-500hr-v0
+datasets:
+- mozilla-foundation/common_voice_17_0
+- naijavoices/naijavoices-dataset
+- google/fleurs
 ---
-
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

app.py

@@ -0,0 +1,30 @@
+import gradio as gr
+from transformers import pipeline, Wav2Vec2ProcessorWithLM
+
+def transcribe(audio, model_id):
+    # load processor
+    p = Wav2Vec2ProcessorWithLM.from_pretrained(model_id, revision="lm")
+    # load eval pipeline
+    asr = pipeline("automatic-speech-recognition", model=model_id, tokenizer=p.tokenizer, feature_extractor=p.feature_extractor, decoder=p.decoder)
+
+    text = asr(audio)["text"]
+    return text
+
+asr_app = gr.Interface(
+    fn=transcribe, 
+    inputs=[
+        gr.Audio(sources=["upload", "microphone"], type="filepath"),
+        gr.Dropdown(
+            [
+                "asr-africa/wav2vec2-xls-r-1b-naijavoices-hausa-500hr-v0",
+                "asr-africa/wav2vec2-xls-r-1b-naijavoices-igbo-500hr-v0",
+                "asr-africa/wav2vec2-xls-r-1b-naijavoices-yoruba-500hr-v0"
+            ]
+        )
+    ], 
+    outputs="text",
+    title="NaijaVoices ASR",
+    description="Realtime demo for Hausa, Igbo and Yoruba speech recognition using a fine-tuned Wav2Vec-XLS-R 1B model.",
+)
+
+asr_app.launch()

packages.txt

@@ -0,0 +1,14 @@
+libavcodec-extra
+sox
+libsndfile1
+build-essential
+libboost-system-dev
+libboost-thread-dev
+libboost-program-options-dev
+libboost-test-dev
+libeigen3-dev
+zlib1g-dev
+libbz2-dev
+liblzma-dev
+git-lfs
+tree

requirements.txt

@@ -0,0 +1,8 @@
+transformers
+torch
+torchaudio
+accelerate
+soundfile
+librosa
+pyctcdecode
+https://github.com/kpu/kenlm/archive/master.zip