texdata/Qwen3.6-27B-slo-med-mt-LoRA-v3

Qwen3.6-27B Slovenian medical — experimental LoRA v3

Trained by [MediaAtlas](https://mediaatlas.si/ai-training.html) — LLM fine-tuning on your own data, trained in the EU, weights delivered. Pricing · All our models

Na kratko: Poskusni adapter LoRA (27B): kratek SFT na sintetičnem slovenskem medicinskem naboru (v3) z naborom za ohranjanje pogovora. Raziskovalni artefakt, ni evalviran, ni za klinično rabo. Za prevajanje uporabite texdata/Qwen3.6-27B-slo-med-mt.

This model is a fine-tuned version of llmfan46/Qwen3.6-27B-uncensored-heretic-v2-Native-MTP-Preserved on the sftsyntheticmedv3 and the sftchat_ret datasets. It achieves the following results on the evaluation set:

• —Loss: 0.6921

Model description

An experimental LoRA adapter from a short supervised fine-tuning run (27 optimizer steps over 3 epochs, effective batch 32) on the internal sft_synthetic_med_v3 set (synthetic Slovenian medical examples) plus sft_chat_ret (Slovenian chat retention). Published for transparency and reproducibility of our training runs.

Intended uses & limitations

Research only. Not evaluated beyond validation loss, not a medical device, not for clinical use; outputs may be wrong. For production-quality models see the main cards linked above.

Training and evaluation data

Internal sets sft_synthetic_med_v3 and sft_chat_ret; held-out validation split of the same data (loss 0.6921). No external benchmark was run.

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• —learning_rate: 2e-05
• —trainbatchsize: 2
• —evalbatchsize: 2
• —seed: 42
• —distributed_type: multi-GPU
• —num_devices: 2
• —gradientaccumulationsteps: 8
• —totaltrainbatch_size: 32
• —totalevalbatch_size: 4
• —optimizer: Use OptimizerNames.ADAMWTORCHFUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• —lrschedulertype: cosine
• —lrschedulerwarmup_steps: 0.05
• —num_epochs: 3.0

Training results

Framework versions

• —PEFT 0.18.1
• —Transformers 5.6.0
• —Pytorch 2.12.1+cu130
• —Datasets 4.0.0
• —Tokenizers 0.22.2