merve/write-with-transformer

1import transformers2import streamlit as st3 4from transformers import AutoTokenizer, AutoModelWithLMHead5  6tokenizer = AutoTokenizer.from_pretrained("gpt2-large")7@st.cache8def load_model(model_name):9    model = AutoModelWithLMHead.from_pretrained("gpt2-large")10    return model11 12model = load_model("gpt2-large")13 14def infer(input_ids, max_length, temperature, top_k, top_p):15 16    output_sequences = model.generate(17        input_ids=input_ids,18        max_length=max_length,19        temperature=temperature,20        top_k=top_k,21        top_p=top_p,22        do_sample=True,23        num_return_sequences=124    )25 26    return output_sequences27default_value = "See how a modern neural network auto-completes your text 🤗 This site, built by the Hugging Face team, lets you write a whole document directly from your browser, and you can trigger the Transformer anywhere using the Tab key. Its like having a smart machine that completes your thoughts 😀 Get started by typing a custom snippet, check out the repository, or try one of the examples. Have fun!"28 29#prompts30st.title("Write with Transformers 🦄")31st.write("The almighty king of text generation, GPT-2 comes in four available sizes, only three of which have been publicly made available. Feared for its fake news generation capabilities, it currently stands as the most syntactically coherent model. A direct successor to the original GPT, it reinforces the already established pre-training/fine-tuning killer duo. From the paper: Language Models are Unsupervised Multitask Learners by Alec Radford, Jeffrey Wu, Rewon Child, David Luan, Dario Amodei and Ilya Sutskever.")32 33sent = st.text_area("Text", default_value, height = 275)34max_length = st.sidebar.slider("Max Length", min_value = 10, max_value=30)35temperature = st.sidebar.slider("Temperature", value = 1.0, min_value = 0.0, max_value=1.0, step=0.05)36top_k = st.sidebar.slider("Top-k", min_value = 0, max_value=5, value = 0)37top_p = st.sidebar.slider("Top-p", min_value = 0.0, max_value=1.0, step = 0.05, value = 0.9)38 39encoded_prompt = tokenizer.encode(sent, add_special_tokens=False, return_tensors="pt")40if encoded_prompt.size()[-1] == 0:41    input_ids = None42else:43    input_ids = encoded_prompt44 45 46output_sequences = infer(input_ids, max_length, temperature, top_k, top_p)47 48 49 50for generated_sequence_idx, generated_sequence in enumerate(output_sequences):51    print(f"=== GENERATED SEQUENCE {generated_sequence_idx + 1} ===")52    generated_sequences = generated_sequence.tolist()53 54    # Decode text55    text = tokenizer.decode(generated_sequence, clean_up_tokenization_spaces=True)56 57    # Remove all text after the stop token58    #text = text[: text.find(args.stop_token) if args.stop_token else None]59 60    # Add the prompt at the beginning of the sequence. Remove the excess text that was used for pre-processing61    total_sequence = (62        sent + text[len(tokenizer.decode(encoded_prompt[0], clean_up_tokenization_spaces=True)) :]63    )64 65    generated_sequences.append(total_sequence)66    print(total_sequence)67 68 69st.write(generated_sequences[-1])70 71 72