FLOR-6.3B / README.md
mvillegas's picture
Update README.md
9159414
|
raw
history blame
10.9 kB
metadata
language:
  - en
  - es
  - ca
licence:
  - apache-2.0
tags:
  - FLOR
  - bloom
  - spanish
  - catalan
  - english
pipeline_tag: text-generation
widget:
  - text: |-
      Respon a la pregunta següent.
      Pregunta: "Quina és la capital de Suècia?"
      Resposta: "La capital de Suècia és Estocolm."
      ----
      Respon a la pregunta següent.
      Pregunta: "Quina beguda es consumeix als matins per despertar-se?"
      Resposta: "La majoria de gent consumeix cafè per despertar-se."
      ----
      Respon a la pregunta següent.
      Pregunta: "Explica com funciona un motor de combustió"
      Resposta:
    example_title: Pregunta-Resposta
  - text: >-
      Extrae las entidades nombradas del siguiente texto:

      Texto: "Me llamo Wolfgang y vivo en Berlin"

      Entidades: Wolfgang:PER, Berlin:LOC

      ----

      Extrae las entidades nombradas del siguiente texto:

      Texto: "Hoy voy a visitar el parc güell tras salir del barcelona
      supercomputing center"

      Entidades: parc güell:LOC, barcelona supercomputing center:LOC

      ----

      Extrae las entidades nombradas del siguiente texto:

      Texto: "Maria y Miguel no tienen ningún problema contigo"

      Entidades: Maria:PER, Miguel:PER

      ----

      Extrae las entidades nombradas del siguiente texto:

      Texto: "Damián se cortó el pelo"

      Entidades: Damián:PER

      ----

      Extrae las entidades nombradas del siguiente texto:

      Texto: "Lo mejor de Barcelona és el bar de mi amigo Pablo"

      Entidades: Pablo:PER, Barcelona:LOC

      ----

      Extrae las entidades nombradas del siguiente texto:

      Texto: "Carlos comparte piso con Marc"

      Entidades:
    example_title: Entidades-Nombradas

FLOR-6.3B

Table of Contents

Click to expand

Model description

FLOR-6.3B is a 6.3B-parameter transformer-based causal language model for Catalan, Spanish, and English. It is the result of a language adaptation technique performed on BLOOM-7.1B, which involves modifying the model's vocabulary and embedding layer, and continuously pre-training the model with 140B tokens in our target languages.

Intended uses and limitations

The FLOR-6.3B model is ready-to-use only for causal language modeling. It can perform text-generation tasks and be fine-tuned for specific scenarios.

How to use

import torch
from transformers import pipeline, AutoTokenizer, AutoModelForCausalLM

input_text = "Sovint em trobo pensant en tot allò que"

model_id  = "projecte-aina/FLOR-6.3B"
tokenizer = AutoTokenizer.from_pretrained(model_id)
generator = pipeline(
    "text-generation",
    model=model_id,
    tokenizer=tokenizer,
    torch_dtype=torch.bfloat16,
    trust_remote_code=True,
    device_map="auto",
)
generation = generator(
    input_text,
    do_sample=True,
    top_k=10,
    eos_token_id=tokenizer.eos_token_id,
)

print(f"Result: {generation[0]['generated_text']}")

Limitations and bias

At the time of submission, no measures have been taken to estimate the bias and toxicity embedded in the model. However, we are well aware that our models may be biased since the corpora have been collected using crawling techniques on multiple web sources. We intend to conduct research in these areas in the future, and if completed, this model card will be updated.

Training

Language adaptation and training

The language adaptation technique used to create FLOR-6.3B requires the vocabulary of the source model to be adapted before continuing its pre-training with data in the target languages. Specifically, we proceeded as follows:

  1. We trained our own BPE tokenizer for Catalan, Spanish, and English, and replaced the original BLOOM tokenizer and vocabulary with it. This procedure implied a downsizing of the original BLOOM's embedding layer and, therefore, a model compression from 7.1B parameters to 6.3B.
  2. The embeddings corresponding to tokens that are present in both the original and the target vocabulary (matching tokens) were used for initialization.
  3. The embeddings from tokens not present in BLOOM's original vocabulary were initialized as the average of all embeddings.
  4. The model was initialized with the weights from BOOM-7.1B, and with our adapted tokenizer (step 1) and embeddings (steps 2-3).
  5. The model was then trained on a corpus that contains a mixture of Catalan, Spanish, and English data.

Training data

The training corpus is composed of 140B tokens gathered from web crawlings and public domain data.

Dataset Language Words (per-epoch) Epochs Total Tokens
mc4 ca 5,861.79M 1.5 13,452.81M
MaCoCu ca 1,658.89M 2 5,076.21M
CaWac ca 1,286.83M 2.5 4,922.14M
oscar-2301 ca 1,784.57M 1.75 4,778.17M
RacoCatala Articles ca 358.57M 4 2,194.42M
RacoCatala Forums ca 1,301.12M 1 1,990.71M
Tesis (TDX) ca 323.60M 4 1,980.46M
oscar-2201 ca 1,155.35M 1 1,767.69M
Wikipedia ca 266.69M 4 1,632.17M
Nació Digital ca 216.27M 4 1,323.59M
colossal-oscar-05-06-23 ca 207.59M 4 1,270.43M
colossal-oscar-03-04-23 ca 195.43M 4 1,196.01M
colossal-oscar-2022-27 ca 195.03M 4 1,193.59M
Crawling populars ca 683.25M 1 1,045.38M
El Món ca 85.27M 4 521.85M
ACN ca 81.25M 4 497.22M
DOGV ca 76.48M 4 468.05M
DOGC ca 70.51M 4 431.51M
Vilaweb ca 46.90M 4 287.04M
hplt ca 160.27M 1 245.21M
Les Corts Valencianes ca 26.88M 4 164.53M
IB3 ca 15.82M 4 96.82M
BOUA ca 13.42M 4 82.13M
Parlament ca 10.09M 4 61.77M
Aquí Berguedà ca 8.23M 4 50.34M
Wikimedia ca 3.90M 4 23.88M
Gutenberg ca 1.29M 4 7.87M
OSCAR 23.01 es 53,244.56M 0.303 23,070.34M
colossal_oscar_05-06-23 es 5,548.27M 1 7,934.02M
colossal_oscar_03-04-23 es 5,090.46M 1 7,279.36M
All_bio_corpora es 954.85M 2 2,730.88M
Wikipedia es 777.49M 2 2,223.63M
BOE es 1,031.28M 1 1,474.73M
Tesis (TDX) es 268.66M 2 768.37M
Eurlex es 459.19M 1 656.64M
CSIC es 156.76M 2 448.33M
BORME es 63.23M 1 90.42M
colossal_oscar_05-06-23 en 51,615.35M 0.25 21,162.30M
colossal_oscar_03-04-23 en 49,454.01M 0.14 11,354.64M
Wikipedia en 2,116.53M 2 6,942.23M
Gutenberg en 3,513.82M 1 5,762.66M
Eurlex en 438.92M 1 719.83M
legal-mc4 en 417.97M 1 685.47M

Languages

The training data has the same amount of Catalan, Spanish, and English texts. The table below shows the final language distribution:

Language Percentage
Catalan (CA) 33.39%
Spanish (ES) 33.32%
English (EN) 33.29%

Training hyperparameters

TBC

Framework

The training was conducted in 16 Cerebras' CS-2 systems using the cs-2.0.2 release of their software.

Evaluation

FLOR-6.3B has been evaluated in a 5-shot setting, using EleutherAI's LM Evaluation Harness. The evaluation benchmark includes tasks in Catalan, Spanish, and English, with particular emphasis on Catalan datasets.

The tasks were chosen to cover several evaluation areas in order to provide a comprehensive overview of the model's capabilities. The baselines used to compare our results are multilingual and English open-source 7B models and smaller models of the FLOR family of models: TBC.

Our implementation of EleutherAI's LM Evaluation Harness can be found here.

The following is a list of evaluation areas and their respective datasets:

Reading Comprehension and Questions Answering

TBC

Natural Language Inference and Paraphrase Identification

TBC

Commonsense Reasoning and Translation

TBC

Additional information

Author

The Language Technologies Unit from Barcelona Supercomputing Center.

Contact

For further information, please send an email to langtech@bsc.es.

Copyright

Copyright(c) 2023 by Language Technologies Unit, Barcelona Supercomputing Center.

License

Apache License, Version 2.0

Funding

This work was funded by Departament de la Vicepresidència i de Polítiques Digitals i Territori de la Generalitat de Catalunya within the framework of Projecte AINA.

Disclaimer

Click to expand

The model published in this repository is intended for a generalist purpose and is available to third parties under a permissive Apache License, Version 2.0.

Be aware that the model may have biases and/or any other undesirable distortions.

When third parties deploy or provide systems and/or services to other parties using this model (or any system based on it) or become users of the model, they should note that it is their responsibility to mitigate the risks arising from its use and, in any event, to comply with applicable regulations, including regulations regarding the use of Artificial Intelligence.

In no event shall the owner and creator of the model (Barcelona Supercomputing Center) be liable for any results arising from the use made by third parties.