# Question answering
```elixir
Mix.install(
[
{:bumblebee, "~> 0.5"},
{:nx, "~> 0.9", override: true},
{:exla, "~> 0.9"},
{:kino, "~> 0.15"}
],
system_env: [
{"XLA_TARGET", "cuda12"},
{"EXLA_TARGET", "cuda"}
],
config: [nx: [default_backend: EXLA.Backend]]
)
```
## 設定
```elixir
cache_dir = "/tmp/bumblebee_cache"
```
## モデルのダウンロード
```elixir
{:ok, roberta} =
Bumblebee.load_model({
:hf,
"deepset/roberta-base-squad2",
cache_dir: cache_dir
})
```
```elixir
{:ok, tokenizer} =
Bumblebee.load_tokenizer({
:hf,
"roberta-base",
cache_dir: cache_dir
})
```
## 文章の準備
```elixir
question_input =
Kino.Input.text("QUESTION",
default: "What industries does Elixir help?"
)
```
```elixir
context_input =
Kino.Input.textarea("CONTEXT",
default:
~s/Elixir is a dynamic, functional language for building scalable and maintainable applications. Elixir runs on the Erlang VM, known for creating low-latency, distributed, and fault-tolerant systems. These capabilities and Elixir tooling allow developers to be productive in several domains, such as web development, embedded software, data pipelines, and multimedia processing, across a wide range of industries./
)
```
```elixir
question = Kino.Input.read(question_input)
context = Kino.Input.read(context_input)
```
```elixir
inputs = Bumblebee.apply_tokenizer(tokenizer, {question, context})
```
```elixir
outputs = Axon.predict(roberta.model, roberta.params, inputs)
```
```elixir
answer_start_index =
outputs.start_logits
|> Nx.argmax()
|> Nx.to_number()
```
```elixir
answer_end_index =
outputs.end_logits
|> Nx.argmax()
|> Nx.to_number()
```
```elixir
answer_tokens =
inputs["input_ids"][[0, answer_start_index..answer_end_index]]
|> Nx.to_flat_list()
```
```elixir
Bumblebee.Tokenizer.decode(tokenizer, answer_tokens)
```
## 時間計測
```elixir
proc = fn question, context ->
inputs = Bumblebee.apply_tokenizer(tokenizer, {question, context})
outputs = Axon.predict(roberta.model, roberta.params, inputs)
answer_start_index =
outputs.start_logits
|> Nx.argmax()
|> Nx.to_number()
answer_end_index =
outputs.end_logits
|> Nx.argmax()
|> Nx.to_number()
answer_tokens =
inputs["input_ids"][[0, answer_start_index..answer_end_index]]
|> Nx.to_flat_list()
Bumblebee.Tokenizer.decode(tokenizer, answer_tokens)
end
```
```elixir
1..10
|> Enum.map(fn _ ->
{time, _} = :timer.tc(proc, [question, context])
time
end)
|> then(&(Enum.sum(&1) / 10))
```
