Hello all, I want to summarize a text with bart-large-cnn but I cant find any up to date documentation on how to do it. All the infos I find online seems to be outdated. This is my code:
def summarize_text(text) do
# Load the BART model for summarization
{:ok, model_info} = Bumblebee.load_model({:hf, "facebook/bart-large-cnn"})
{:ok, tokenizer} = Bumblebee.load_tokenizer({:hf, "facebook/bart-large-cnn"})
{:ok, generation_config} = Bumblebee.load_generation_config({:hf, "facebook/bart-large-cnn"})
generation_config = Bumblebee.configure(generation_config, max_new_tokens: 100)
serving =
Bumblebee.Text.generation(model_info, tokenizer, generation_config,
compile: [batch_size: 1, sequence_length: 512],
defn_options: [
compiler: EXLA,
cache: Path.join(System.tmp_dir!(), "bumblebee_examples/text_summarization")
]
)
# Generate summary
{:ok, result} = Nx.Serving.run(serving, %{text: text})
summary = Enum.map(result.results, fn r -> r.text end) |> Enum.join(" ")
summary
end
thats the error
Yts.Audio.summarize_text("Elixir leverages the Erlang VM, known for running low-latency, distributed, and fault-tolerant systems. Elixir is successfully used in web development, embedded software, data ingestion, and multimedia processing, across a wide range of industries")
[debug] the following parameters were missing:
* language_modeling_head.logits_bias.bias
W0000 00:00:1743201754.366949 53443 onednn_matmul.cc:319] [Perf]: MatMul reference implementation being executed ** (MatchError) no match of right hand side value: %{results: [%{text: "The first time a person seess when a person is looking for a place to go to a nightclub.", token_summary: %{input: 53, output: 89, padding: 459}}]}
(yts 0.1.0) lib/yts/audio.ex:51: Yts.Audio.summarize_text/1
iex:35: (file)