Model Card for Minerva-1B-base-v1.0

Minerva is the first family of truly-open (data and model) Italian-English LLMs pretrained from scratch developed by Sapienza NLP in collaboration with Future Artificial Intelligence Research (FAIR) and CINECA. Notably, approximately half of its pretraining data includes Italian text.

This model is in preview: currently, we are limiting the access to this model to the members of the Future Artificial Intelligence Research community. If you are a member of this community and you would like to get access to Minerva, please fill the form and update your account to use your institutional email address. If you are not member, Minerva will be accessible to everyone in the next few weeks!

Description

This is the model card for Minerva-1B-base-v1.0, a 1 billion parameter model trained on 200 billion tokens (100 billion in Italian, 100 billion in English).

This model is part of the Minerva LLM family:

• Minerva-350M-base-v1.0
• Minerva-1B-base-v1.0
• Minerva-3B-base-v1.0 (currently training!)

How to use Minerva with Hugging Face transformers

import transformers
import torch

model_id = "sapienzanlp/Minerva-1B-base-v1.0"

# Initialize the pipeline.
pipeline = transformers.pipeline(
    "text-generation",
    model=model_id,
    model_kwargs={"torch_dtype": torch.bfloat16},
    device_map="auto",
)

# Input text for the model.
input_text = "La capitale dell'Italia è"

# Compute the outputs.
output = pipeline(
  input_text,
  max_new_tokens=128,
)

# Output:
# [{'generated_text': "La capitale dell'Italia è la città di Roma, che si trova a [...]"}]

Model Architecture

Minerva-1B-base-v1.0 is a Transformer model based on the Mistral architecture, where the number of layers, number of heads, and the hidden states dimension are modified to reach 1B parameters. Please, take a look at the configuration file for a detailed breakdown of the hyperparameters we chose for this model.

The Minerva LLM family is composed of:

Model Name	Tokens	Layers	Hidden Size	Attention Heads	KV Heads	Sliding Window	Max Context Length
Minerva-350M-base-v1.0	70B (35B it + 35B en)	16	1152	16	4	2048	16384
Minerva-1B-base-v1.0	200B (100B it + 100B en)	16	2048	16	4	2048	16384
Minerva-3B-base-v1.0	660B (330B it + 330B en)	32	2560	32	8	2048	16384

Model Training

Minerva-1B-base-v1.0 was trained using llm-foundry 0.6.0 from MosaicML. The hyperparameters used are the following:

Model Name	Optimizer	lr	betas	eps	weight decay	Scheduler	Warmup Steps	Batch Size (Tokens)	Total Steps
Minerva-350M-base-v1.0	Decoupled AdamW	2e-4	(0.9, 0.95)	1e-8	0.0	Cosine	2%	4M	16,690
Minerva-1B-base-v1.0	Decoupled AdamW	2e-4	(0.9, 0.95)	1e-8	0.0	Cosine	2%	4M	47,684
Minerva-3B-base-v1.0	Decoupled AdamW	2e-4	(0.9, 0.95)	1e-8	0.0	Cosine	2%	4M	157,357

Model Evaluation

We assessed our model using the LM-Evaluation-Harness library, which serves as a comprehensive framework for testing generative language models across a wide range of evaluation tasks.

All the reported benchmark data was already present in the LM-Evaluation-Harness suite.

Italian Data:

Task	Accuracy
xcopa (0-shot)	0.602
Hellaswag (5-shot)	0.3847
Belebele (5-shot)	0.25
TruthfulQA MC 1 (0-shot)	0.2375
TruthfulQA MC 2 (0-shot)	0.3996
M MMLU (5-shot)	0.2546
arc challenge (5-shot)	0.2549

English Data:

Task	Accuracy
Hellaswag (5-shot)	0.43
Belebele (5-shot)	0.26
TruthfulQA MC 1 (0-shot)	0.21
TruthfulQA MC 2 (0-shot)	0.38
M MMLU (5-shot)	0.26
arc challenge (5-shot)	0.26
arc easy (5-shot)	0.4966
piqa (5-shot)	0.6975
sciq (5-shot)	0.830

Training Data

Minerva-1B-base-v1.0 was trained on 100B Italian tokens and 100B English tokens sampled from CulturaX. The data was selected from the following sources:

• OSCAR-2201
• OSCAR-2301
• mC4

We have extracted some statistics on Italian (115B tokens) and English (210B tokens) documents from CulturaX on the selected sources:

Proportion of number of tokens per domain (Italian)

Proportion of number of tokens per domain (English)

Tokenizer Fertility

The tokenizer fertility measures the average amount of tokens produced per tokenized word. A tokenizer displaying high fertility values in a particular language typically indicates that it segments words in that language extensively. The tokenizer fertility is strictly correlated with the inference speed of the model with respect to a specific language, as higher values mean longer sequences of tokens to generate and thus lower inference speed.

Fertility computed over a sample of Cultura X (CX) data and Wikipedia (Wp):

Model	Voc. Size	Fertility IT (CX)	Fertility EN (CX)	Fertility IT (Wp)	Fertility EN (Wp)
Mistral-7B-v0.1	32000	1.87	1.32	2.05	1.57
gemma-7b	256000	1.42	1.18	1.56	1.34
Minerva-1B-base-v1.0	32768	1.39	1.32	1.66	1.59

Notice

Minerva-350M-base-v1.0 is a pretrained base model and, therefore, has no moderation mechanisms.

The Sapienza NLP Team

• Riccardo Orlando: data preprocessing, model training
• Pere-Lluis Huguet Cabot: data preprocessing, vocabulary, evaluation
• Luca Moroni: data curation, data analysis, downstream tasks, evaluation
• Simone Conia: data curation, evaluation, project supervision
• Edoardo Barba: data preprocessing, downstream tasks, project supervision
• Roberto Navigli: project coordinator

Special thanks for their support

• Giuseppe Fiameni, Nvidia
• Sergio Orlandini, CINECA

Acknowledgments

This work was funded by the PNRR MUR project PE0000013-FAIR. We acknowledge the CINECA award "IscB_medit" under the ISCRA initiative, for the availability of high performance computing resources and support.