Qwen3-Omni LoRA Evaluation

Base model: /path/to/ropedia_workspace/modelscope_models/Qwen__Qwen3-Omni-30B-A3B-Instruct
Adapter: checkpoints/xperience10m_qwen3_omni_32ep_lora/adapter_lora
Dataset: results/omni_finetune/xperience10m_qwen3_omni_32ep_dataset/dataset.jsonl
Eval split: train
Samples: 128
Episodes: 1
Accuracy: 0.0000
Macro-F1: 0.0000
Unseen eval labels: 0

Artifacts include metrics.json, predictions.csv, per_class_metrics.csv, and confusion_matrix.csv.