🎸 🥁 Rockbot 🎤 🎧
A GPT-2 based lyrics generator fine-tuned on the writing styles of 16000 songs by 270 artists across MANY genres (not just rock).
Instructions: Type in a fake song title, pick an artist, click "Generate".
Note: Due to the nature of language models, lyrics bleed across artists and you may see NSFW lyrics unexpectedly (e.g., from The Beatles), especially if you change the configuration to allow more entropy. I have made no attempt to censor lyrics whatsoever.
Finally, these lyrics are computer generated. Not all of these will be non-repetitive and/or coherent. Just have fun.
GPT-2 124M version Model page on Hugging Face
DistilGPT2 version Model page on Hugging Face This is leaner but the lyrics are much more simplistic.
🎹 🪘 🎷 🎺 🪗 🪕 🎻
Background
With the shutdown of Google Play Music I used Google's takeout function to gather the metadata from artists I've listened to over the past several years. I wanted to take advantage of this bounty to build something fun. I scraped the top 50 lyrics for artists I'd listened to at least once from Genius, then fine tuned GPT-2's 124M token model using the AITextGen framework after considerable post-processing. For more on generation, see here.
Full Tech Stack
Google Play Music (R.I.P.). Python. Streamlit. GPT-2. AITextGen. Pandas. LyricsGenius. Google Colab (GPU based Training). Knime (data cleaning).
How to Use The Model
Please refer to AITextGen for much better documentation.
Training Parameters Used
ai.train("lyrics.txt",
line_by_line=False,
from_cache=False,
num_steps=10000,
generate_every=2000,
save_every=2000,
save_gdrive=False,
learning_rate=1e-3,
batch_size=3,
eos_token="<|endoftext|>",
#fp16=True
)
To Use
Generate With Prompt (Use Title Case):
Song Name
BY
Artist Name