mosaicml/composer

Health v2 · maintenance only

Adoption not part of the score

5499 stars · 470 forks observed · 2026-09-22

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

Composer is an open-source PyTorch-based deep learning training library by MosaicML (now Databricks) for training neural networks faster and at lower cost. It abstracts distributed training, parallelism, data loading, and memory optimization for large-scale multi-node GPU clusters.

Use cases

• train large language models on multi-node GPU clusters
• speed up PyTorch training of transformer models
• fine-tune BERT-style embedding models efficiently
• train diffusion models at scale
• run distributed CNN training experiments faster
• reduce GPU training costs for deep learning workloads

When to choose

• you train neural networks on multi-GPU or multi-node clusters
• you want distributed training best practices without writing parallelism code yourself
• you are training LLMs, diffusion models, or transformers in PyTorch

When to avoid

• you need a lightweight single-GPU training loop with minimal dependencies
• your project uses TensorFlow or JAX instead of PyTorch
• you need inference or serving rather than training

Facets

library · maturity active

deep-learning llm-training machine-learning gpu-computing deep-learning machine-learning large-language-models gpu-computing python cloud pytorch distributed-training training-efficiency mosaicml neural-networks gpu linux docker

Member repositories

For agents

markdown · JSON · MCP: product_card(name="mosaicml/composer")