You can deploy this model and start earning money today!

RichardErkhov/
cosimoiaia_-_Loquace-410m-gguf

Name: cosimoiaia_-_Loquace-410m-gguf
Availability: InStock

endpoints_compatible

gguf

template

Views

No views yet

Quantization made by Richard Erkhov.

Github

Discord

Request more models

Loquace-410m - GGUF

Model creator: https://huggingface.co/cosimoiaia/
Original model: https://huggingface.co/cosimoiaia/Loquace-410m/

Name	Quant method	Size
Loquace-410m.Q2_K.gguf	Q2_K	0.16GB
Loquace-410m.IQ3_XS.gguf	IQ3_XS	0.18GB
Loquace-410m.IQ3_S.gguf	IQ3_S	0.18GB
Loquace-410m.Q3_K_S.gguf	Q3_K_S	0.18GB
Loquace-410m.IQ3_M.gguf	IQ3_M	0.2GB
Loquace-410m.Q3_K.gguf	Q3_K	0.21GB
Loquace-410m.Q3_K_M.gguf	Q3_K_M	0.21GB
Loquace-410m.Q3_K_L.gguf	Q3_K_L	0.22GB
Loquace-410m.IQ4_XS.gguf	IQ4_XS	0.22GB
Loquace-410m.Q4_0.gguf	Q4_0	0.23GB
Loquace-410m.IQ4_NL.gguf	IQ4_NL	0.23GB
Loquace-410m.Q4_K_S.gguf	Q4_K_S	0.23GB
Loquace-410m.Q4_K.gguf	Q4_K	0.25GB
Loquace-410m.Q4_K_M.gguf	Q4_K_M	0.25GB
Loquace-410m.Q4_1.gguf	Q4_1	0.25GB
Loquace-410m.Q5_0.gguf	Q5_0	0.27GB
Loquace-410m.Q5_K_S.gguf	Q5_K_S	0.27GB
Loquace-410m.Q5_K.gguf	Q5_K	0.28GB
Loquace-410m.Q5_K_M.gguf	Q5_K_M	0.28GB
Loquace-410m.Q5_1.gguf	Q5_1	0.29GB
Loquace-410m.Q6_K.gguf	Q6_K	0.31GB
Loquace-410m.Q8_0.gguf	Q8_0	0.4GB

Original model description:

license: cc-by-nc-2.0 datasets:

cosimoiaia/Loquace-102k language:
it pipeline_tag: conversational tags:
alpaca
llama
llm
finetune
Italian
qlora

Model Card for Loquace-410m

🇮🇹 Loquace-410m 🇮🇹

An exclusively Italian speaking, instruction finetuned, Large Language model. 🇮🇹

The Loquace Italian LLM models are created as a proof-of-concept to evaluate on how language tuning can be achieved using QLoRa by instruct-tunings foundational LLMs using dataset of a specific language.

The QLoRa (https://github.com/artidoro/qlora) method of fine-tuning significantly lower the resources requirements compared to any other methods available, this allow to easily execute the process on significanly larger dataset while still using consumers GPUs and still achieve high accuracy.

Model Description

Loquace-410m is the second smallest model of the Loquace family. It was trained using QLoRa on a large dataset of 102k question/answer pairs exclusively in Italian using pythia-410m as base.

The related code can be found at: https://github.com/cosimoiaia/Loquace

Loquace-410m is part of the big Loquace family:

https://huggingface.co/cosimoiaia/Loquace-70m - Based on pythia-70m https://huggingface.co/cosimoiaia/Loquace-410m - Based on pythia-410m https://huggingface.co/cosimoiaia/Loquace-7B - Based on Falcon-7B. https://huggingface.co/cosimoiaia/Loquace-12B - Based on pythia-12B https://huggingface.co/cosimoiaia/Loquace-20B - Based on gpt-neox-20B

Usage

python

from transformers import (
    AutoTokenizer,
    AutoModelForCausalLM,
    BitsAndBytesConfig
)

tokenizer = AutoTokenizer.from_pretrained("cosimoiaia/Loquace-410m", padding_side="right", use_fast=True)
model = AutoModelForCausalLM.from_pretrained(
    "cosimoiaia/Loquace-410m",
    load_in_8bit=True,
    device_map="auto",
    quantization_config=BitsAndBytesConfig(
      load_in_4bit=True,
      llm_int8_has_fp16_weight=False
    )
)

Training

Loquace-410m was trained on a conversational dataset comprising 102k question/answer pairs in Italian language. The training data was constructed by putting together translations from the original alpaca Dataset and other sources like the OpenAssistant dataset. The model was trained for only 10000 iterations and took 9 hours on a single RTX 3090, kindly provided by Genesis Cloud. (https://gnsiscld.co/26qhlf)

Limitations

Loquace-410m may not handle complex or nuanced queries well and may struggle with ambiguous or poorly formatted inputs.
The model may generate responses that are factually incorrect or nonsensical. It should be used with caution, and outputs should be carefully verified.
The training data primarily consists of conversational examples and may not generalize well to other types of tasks or domains.

Dependencies

PyTorch
Transformers library by Hugging Face
Bitsandbites
QLoRa

RichardErkhov/cosimoiaia_-_Loquace-410m-gguf