Event JSON
{
"id": "19a4e594c6bc7b6ccaf3f2246f84aaaf24ca1366fb8372e58fdcfac6d967f2fb",
"pubkey": "f87a90442cc0e4352b7e3ba21b2807df9a507086a528ddd11e01b5b96c7081b9",
"created_at": 1794288497,
"kind": 1,
"tags": [
[
"proxy",
"https://fosstodon.org/@TheASF/117590490976843691",
"web"
],
[
"proxy",
"https://fosstodon.org/users/TheASF/statuses/117590490976843691",
"activitypub"
],
[
"L",
"pink.momostr"
],
[
"l",
"pink.momostr.activitypub:https://fosstodon.org/users/TheASF/statuses/117590490976843691",
"pink.momostr"
],
[
"-"
]
],
"content": "How can we know what data was used to train a language model?\n\nApache Sourcelume is developing machine-readable provenance records for datasets and models. A sample training pipeline demonstrates how a model's training process can generate a provenance record, linking to dataset records and documenting token counts, data shares, and loss. \n\nRead the blog to learn more about what the project is working towards: https://news.apache.org/foundation/entry/tiny-model-full-provenance-where-apache-sourcelume-is-and-where-it-is-heading",
"sig": "f16bb4cfb3046f8300a2b5a5fc8922db4aeaae2acea56675a1f94d7a6ddfa12c2e5ab0aa4490812cbce86ce700f4855b9ee9d7ea399175a7f111634af7c8948b"
}