def embed_tformer()

in inference.py [0:0]

• 6 lines of code
• 1 McCabe index (conditional complexity)

def embed_tformer(model, tokenizer, sentences):
    encoded_input = tokenizer(sentences, padding=True, truncation=True, max_length=256, return_tensors='pt')

    #Compute token embeddings
    with torch.no_grad():
        model_output = model(**encoded_input)

    sentence_embeddings = mean_pooling(model_output, encoded_input['attention_mask'])
    return sentence_embeddings