Spaces:
Runtime error
Runtime error
| import gradio | |
| import numpy as np | |
| import pandas as pd | |
| import re | |
| import fasttext | |
| from huggingface_hub import hf_hub_download | |
| model_path = hf_hub_download(repo_id="facebook/fasttext-id-vectors", filename="model.bin") | |
| model = fasttext.load_model(model_path) | |
| new_catalog = pd.read_csv('new_catalog.csv') | |
| new_prodcatalog_vec = np.load('product_catalog_vec.npy') | |
| def preprocess(text): | |
| # remove units | |
| regex_f = r'(\d+[.]{0,1}\d*\s*(kg|KG|Kg|ml|ML|Ml|mL|mil|ltr|LTR|liter|lt|Lt|LT|L|l|kilograms|%|gram|Gram|grm|GR|Gr|gr|g|G|wp|Wp|WP|METER|sp|Amper|Ahm|A|Btr|btr|ons|kapsul|cm|inc|cc|SL|sl))' | |
| text = re.sub(regex_f, "", text) | |
| # replace extra characters | |
| text = text.replace('@', ' ') | |
| text = text.replace('.', ' ') | |
| text = text.replace('-', ' ') | |
| text = text.replace('(', ' ') | |
| text = text.replace(')', ' ') | |
| text = text.strip() | |
| # lowercase | |
| text = text.lower() | |
| return text | |
| def extract_features(text): | |
| return model.get_sentence_vector(text) | |
| def cosine_similarity(v1, v2): | |
| return np.dot(v1,v2) / (np.linalg.norm(v1)*np.linalg.norm(v2)) | |
| def get_most_similar(target, reference, n=5): | |
| score = [] | |
| for r in reference: | |
| if np.count_nonzero(r) == 0: # untuk menghindari score nan | |
| score.append(0.) | |
| else: | |
| score.append(cosine_similarity(target, r)) | |
| # ranking descending | |
| idx = np.array(score).argsort()[::-1][:n] | |
| top_score = np.take(score, idx) | |
| return idx, top_score | |
| def predict_catalog(text): | |
| text = preprocess(text) | |
| X = extract_features(text) | |
| idx, top_score = get_most_similar(X, new_prodcatalog_vec) | |
| catalog = np.take(new_catalog['Product SKU'].values, idx)[0] | |
| return catalog | |
| demo = gradio.Interface( | |
| fn=predict_catalog, | |
| inputs=gradio.Textbox(placeholder="Enter a product name here..."), | |
| outputs="label", | |
| examples=[["NPK Kujang 30.6.8 @ 25 Kg"]]) | |
| demo.launch(share=True) |