Neura Market
Bloom logo

Bloom

Free

BLOOM by Hugging Face is a model similar to GPT-3 that has been trained on 46 different languages and 13 programming languages. #opensource

FreeFree tier
Inputs: textOutputs: text
Type
Open Source
Company
BigScience Workshop
Visit Website GitHub (40 stars)

About Bloom

BLOOM is an open-source, autoregressive large language model (LLM) developed through the BigScience Workshop. Its architecture is similar to GPT-3, designed for next token prediction. BLOOM has been trained on a diverse dataset covering 46 natural languages and 13 programming languages, making it one of the most multilingual models available. It comes in multiple sizes ranging from 560 million to 176 billion parameters, allowing deployment at various scales. The model is integrated into Hugging Face Transformers, providing easy access through the library's APIs for tasks like text generation, classification, and question answering. BLOOM represents a collaborative effort in open science, with extensive community resources, optimization guides, and inference acceleration techniques.

Key Features

Autoregressive language model for next token prediction (GPT-3-like architecture)
Trained on 46 natural languages and 13 programming languages
Available in multiple model sizes: 560M, 1.1B, 1.7B, 3B, 7.1B, and 176B parameters
Open-source and free to use, hosted on Hugging Face Transformers
Supports causal language modeling, text generation, text classification, token classification, and question answering
Backed by extensive community resources, optimization guides, and inference acceleration tools

Pros & Cons

Pros
  • Open-source and freely available
  • Multilingual and multi-programming language support
  • Multiple size options for different computational budgets
  • Seamless integration with Hugging Face Transformers ecosystem
  • Backed by collaborative BigScience community and active development
Cons
  • Largest model (176B parameters) requires substantial computational resources for inference
  • Not specifically optimized for any single language or domain
  • Comparable to GPT-3 but may lag behind in certain English-only benchmarks
  • Limited fine-tuning documentation compared to some other models

Best For

Multilingual text generation and completionCode generation across 13 programming languagesCausal language modeling tasksText classification and token classificationQuestion answering systemsCross-lingual natural language understanding

FAQ

What is BLOOM?
BLOOM is an open-source autoregressive large language model developed by the BigScience Workshop, with architecture similar to GPT-3. It was trained on 46 natural languages and 13 programming languages.
What sizes of BLOOM are available?
BLOOM is available in several sizes: bloom-560m (560M parameters), bloom-1b1 (1.1B), bloom-1b7 (1.7B), bloom-3b (3B), bloom-7b1 (7.1B), and bloom (176B parameters).
Is BLOOM free to use?
Yes, BLOOM is open-source and available for free under permissive licensing. It can be accessed through Hugging Face Transformers.
What tasks can BLOOM be used for?
BLOOM supports causal language modeling, text generation, text classification, token classification, and question answering. It is particularly suited for multilingual and code generation tasks.