baddu/sotto-cleanup-lfm25-350m-GGUF

Sotto Cleanup LFM2.5 350M — GGUF

Model lineage

LiquidAI/LFM2.5-350M-Base
→ juanquivilla/sotto-cleanup-lfm25-350m
→ GGUF conversion + quantization
→ this repository

Modifications

• —Converted the upstream BF16 Safetensors checkpoint to GGUF.
• —Quantized using llama.cpp.
• —No additional fine-tuning or training was performed for this repository.
• —These files are modified derivative versions of the upstream weights.

Intended use

Local ASR transcript cleanup and post-processing.

Prompt format

Input:

{raw transcript}

Output:

Recommended settings

do_sample=false
repetition_penalty=1.05

Upstream references

• —Sotto Cleanup LFM2.5 350M
• —LiquidAI LFM2.5-350M-Base
• —llama.cpp

License

The upstream Sotto fine-tune currently identifies itself as MIT. However, it is based on the LiquidAI/LFM2.5-350M-Base weights, which are subject to the LFM Open License v1.0. This GGUF distribution contains modified derivative versions of those upstream weights and preserves the LFM Open License v1.0. See LICENSE for the exact license text.

No endorsement / disclaimer

This repository is an independent conversion and distribution. It is not officially endorsed by LiquidAI, the Sotto author, or the llama.cpp project. The upstream names and links are provided for attribution and provenance only. The files are provided under the applicable license and without any additional warranty.