LFM2.5-350M Mental Health (fine-tuned)

Fine-tuned version of LiquidAI/LFM2.5-350M on the MentalChat16K dataset using QLoRA. Exported to ONNX (fp16 + q4) for in-browser inference via Transformers.js.

Note: training data is English-only. The base model is multilingual, but quality in other languages for this fine-tune is unverified.