cstr/

phi3-mini-4k-llamafied-sft-v1

this is only a very quick sft (few minutes in google colab) on a very tiny (60 steps) subset of cstr/capybara_de_sharegpt with unsloth, just as a proof of concept, that you can do this, make the model output in (not very good) german, and prompt with chatml template

Added 2026-08-02 during an account-wide provenance review.

This is a fine-tune, not a format conversion. Most cstr/* repositories are GGUF conversions where the upstream research team remains the provider of the model and conversion changes only the numeric representation of the weights. Training changes the model itself, so under Regulation (EU) 2024/1689 the maintainer of this repository is plausibly the provider of the model it produced, and the obligations that survive the Art. 53(2) free-and-open-source exemption — Art. 53(1)(c) and 53(1)(d) — attach here.

Art. 53(1)(c) — copyright policy. The supervised fine-tuning used cstr/capybara_de_sharegpt, a German ShareGPT-format set published by the same maintainer. Its own provenance is documented at that dataset repository. No text or data mining was performed to assemble a corpus for this repository, so no rights reservation under Art. 4(3) of Directive (EU) 2019/790 was engaged by this step. Any credible claim that this model was trained on material without the necessary rights will be acted on — contact via the Community tab.

Art. 53(1)(d) — training content. Supervised fine-tuning on cstr/capybara_de_sharegpt (German, ShareGPT format), applied to a LLaMAfied Phi-3-mini-4k. No other corpus was introduced by this repository; the remainder of the training content is Microsoft's Phi-3-mini, documented by its provider.

Base model. vonjack/Phi-3-mini-4k-instruct-LLaMAfied — the training built on those weights and inherits their terms.