# mosaicml/llm-foundry

LLM training code for Databricks foundation models

Repository: https://github.com/mosaicml/llm-foundry
Canonical: https://ross.abutalabs.com/products/llm-foundry
Homepage: https://www.databricks.com/blog/introducing-dbrx-new-state-art-open-llm
Language: Python
License: Apache-2.0
License Family: permissive
Topics: deep-learning, llm, neural-networks, nlp, pytorch
Last push: 2026-03-25T17:22:01+00:00

## Health v2 (maintenance only)
Score: 65/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 74, release rhythm 40, longevity 87
- inputs: {"age_days": 1223, "days_push": 161, "days_rel": 401, "gap_med": 12.5, "n_releases_24m": 19}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 4441, forks 590 (observed 2026-08-28T04:08:49.398525+00:00)

## What it is
LLM Foundry is a PyTorch-based codebase for training, finetuning, evaluating, and deploying large language models from 125M to 70B+ parameters, built on Composer and the MosaicML/Databricks platform. It includes data preparation scripts, training and inference benchmarking tools, and model conversion to HuggingFace or ONNX formats.

## Use cases
- finetune an LLM on my own data
- pretrain a large language model on a GPU cluster
- convert a trained LLM to ONNX for inference
- evaluate an LLM on academic in-context learning tasks
- benchmark LLM training throughput and MFU
- prepare text data for LLM pretraining
- profile LLM inference latency and throughput

## When to choose
- you want to pretrain or finetune LLMs at scale on GPUs with an efficient, battle-tested codebase
- you are on the Databricks/MosaicML platform and want integrated workload launching
- you need end-to-end LLM workflows: data prep, training, eval, and export

## When to avoid
- you only need to run inference with an existing model without training
- you want a lightweight single-GPU hobbyist finetuning tool with minimal setup
- you are not working in the PyTorch/Composer ecosystem

## Facets
- artifact type: library
- maturity: active
- function: llm-training, machine-learning, deep-learning, nlp, benchmarking, etl
- domain: large-language-models, deep-learning, machine-learning, gpu-computing
- platform: python, cloud
- tags: finetuning, pytorch, mosaicml, databricks, onnx-export, in-context-learning-eval, streaming-dataset, natural-language-processing, gpu, docker

## Member repositories
- mosaicml/llm-foundry (main) score 65

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:08:49.398525+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T18:20:50.086458+00:00, confidence not recorded.
  - readme: https://github.com/mosaicml/llm-foundry (fetched 2026-08-28T04:08:49.398525+00:00, sha 63dd3c9088b6)
  - homepage: https://www.databricks.com/blog/introducing-dbrx-new-state-art-open-llm (fetched 2026-08-29T09:07:56.740200+00:00, sha 44136fa355b3)
  - registry_pypi: https://pypi.org/pypi/llm-foundry/json (fetched 2026-08-29T09:07:56.749123+00:00, sha fa17a20a80cc)
- Data as of 2026-08-30T08:39:29.467469+00:00.