metadata
dataset_info:
features:
- name: text
dtype: string
- name: len_tokens
dtype: int64
splits:
- name: train
num_bytes: 44143916
num_examples: 48016
download_size: 8526144
dataset_size: 44143916
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
#tokenizer: google/gemma-2b-it
hist len_tokens
<bos><start_of_turn>system
You are a helpful AI assistant.
solo responde en formato json.
lista de codigos linguisticos disponibles: ["es", "en", "fr", "de"].<end_of_turn>
<start_of_turn>user
{ "input": "es",
"targets": "fr",
"inputs_es": "¿Qué presidente de los Estados Unidos nunca se ha casado?"
}<end_of_turn>
<start_of_turn>model
{ "targets": "fr",
"targets_fr": "James Buchanan est le seul président qui ne s'est jamais marié."
}<end_of_turn><eos>
describe.
percentil.
