Download Latest Version
v0.22.0 source code.tar.gz (28.4 MB)
Get an email when there's a new version of LLM Foundry
| Name | Modified | Size | Downloads / Week |
|---|---|---|---|
| Parent folder | |||
| README.md | 2025-07-29 | 1.1 kB | |
| v0.22.0 source code.tar.gz | 2025-07-29 | 28.4 MB | |
| v0.22.0 source code.zip | 2025-07-29 | 28.6 MB | |
| Totals: 3 Items | 57.1 MB | 0 | |
What's Changed
- Change test to use foundry images by @bowenyang008 in https://github.com/mosaicml/llm-foundry/pull/1839
- Bias in QKV input projections for supporting Qwen2 model series by @gupta-abhay in https://github.com/mosaicml/llm-foundry/pull/1837
- Fix backwards compatibility with attention_bias by @dakinggg in https://github.com/mosaicml/llm-foundry/pull/1840
- Add Head Dim as a configurable parameter for Attention and Rope by @jdchang1 in https://github.com/mosaicml/llm-foundry/pull/1842
- Support for reusing the input to W_k and W_v by @ShashankMosaicML in https://github.com/mosaicml/llm-foundry/pull/1710
- Display data type in error message by @emmanuel-ferdman in https://github.com/mosaicml/llm-foundry/pull/1834
- updated composer and streaming versions by @ethantang-db in https://github.com/mosaicml/llm-foundry/pull/1857
- updated and restricted our composer and mlflow libraries by @ethantang-db in https://github.com/mosaicml/llm-foundry/pull/1859
Full Changelog: https://github.com/mosaicml/llm-foundry/compare/v0.21.0...v0.22.0