Note: this is a very involved task, the generation can take a long time if you run it on a CPU. To achieve a better quality increase the number of steps.
{:ok, model_info} = Bumblebee.load_model({:hf, "gpt2"})
{:ok, tokenizer} = Bumblebee.load_tokenizer({:hf, "gpt2"})
serving =
Bumblebee.Text.generation(model_info, tokenizer,
max_new_tokens: 10,
compile: [batch_size: 1, sequence_length: 100],
defn_options: [compiler: EXLA]
)
text_input = Kino.Input.textarea("Text", default: "Yesterday, I was reading a book and")
form = Kino.Control.form([text: text_input], submit: "Run")
frame = Kino.Frame.new()
form
|> Kino.Control.stream()
|> Kino.listen(fn %{data: %{text: text}} ->
Kino.Frame.render(frame, Kino.Markdown.new("Running..."))
%{results: [%{text: generated_text}]} = Nx.Serving.run(serving, text)
Kino.Frame.render(frame, Kino.Markdown.new(generated_text))
end)
Kino.Layout.grid([form, frame], boxed: true, gap: 16)Livebook runs inside a Space, so you can use their GPUs to train and run models.
Nx.Serving.run(serving, "The Elixir programming language")
