Update README.md

README.md

@@ -9,77 +9,8 @@ tags:
 pipeline_tag: audio-classification
 ---
 
-import torch
-import torch.nn as nn
-import torch.nn.functional as F
-import torchaudio
-from transformers import AutoConfig, Wav2Vec2Processor
-import librosa
-import IPython.display as ipd
-import numpy as np
-import pandas as pd
-
 model_name_or_path = "quaja/hubert-base-amharic-speech-emotion-recognition"
 config = AutoConfig.from_pretrained(model_name_or_path)
 feature_extractor = Wav2Vec2FeatureExtractor.from_pretrained(model_name_or_path)
 sampling_rate = feature_extractor.sampling_rate
 model = HubertForSpeechClassification.from_pretrained(model_name_or_path)
-
-def speech_file_to_array_fn(path, sampling_rate):
-    speech_array, _sampling_rate = torchaudio.load(path)
-    resampler = torchaudio.transforms.Resample(_sampling_rate)
-    speech = resampler(speech_array).squeeze().numpy()
-    return speech
-
-
-def predict(path, sampling_rate):
-    speech = speech_file_to_array_fn(path, sampling_rate)
-    features = feature_extractor(speech, sampling_rate=sampling_rate, return_tensors="pt", padding=True)
-
-    input_values = features.input_values.to(device)
-
-    with torch.no_grad():
-        logits = model(input_values).logits
-
-    scores = F.softmax(logits, dim=1).detach().cpu().numpy()[0]
-    outputs = [{"Label": config.id2label[i], "Score": f"{round(score * 100, 3):.1f}%"} for i, score in enumerate(scores)]
-    return outputs
-
-
-STYLES = """
-<style>
-div.display_data {
-    margin: 0 auto;
-    max-width: 500px;
-}
-table.xxx {
-    margin: 50px !important;
-    float: right !important;
-    clear: both !important;
-}
-table.xxx td {
-    min-width: 300px !important;
-    text-align: center !important;
-}
-</style>
-""".strip()
-
-def prediction(df_row):
-    path, label = df_row["path"], df_row["emotion"]
-    df = pd.DataFrame([{"Emotion": label, "Sentence": "    "}])
-    setup = {
-        'border': 2,
-        'show_dimensions': True,
-        'justify': 'center',
-        'classes': 'xxx',
-        'escape': False,
-    }
-    ipd.display(ipd.HTML(STYLES + df.to_html(**setup) + "<br />"))
-    speech, sr = torchaudio.load(path)
-    resampler = torchaudio.transforms.Resample(sr)
-    speech = resampler(speech[0]).squeeze().numpy()
-    ipd.display(ipd.Audio(data=np.asarray(speech), autoplay=True, rate=sampling_rate))
-
-    outputs = predict(path, sampling_rate)
-    r = pd.DataFrame(outputs)
-    ipd.display(ipd.HTML(STYLES + r.to_html(**setup) + "<br />"))