datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
shaer-sft-test-generations-k5
Shaer SFT Test Generations K=5
This dataset contains generated outputs from the final Shaer SFT adapter on the held-out test split.
Source dataset: Shaer-AI/ashaar-with-enhanced-descriptions-baseform-final-sft-lte20-min500-splits
Source split: test
Source rows: 3,481
Samples per source row: 5
Total generated rows: 17,405
Base model: Navid-AI/Yehia-7B-preview
Adapter repo: Shaer-AI/Shaer-adapters
Adapter subfolder: adapters/fresh_sft/train/best
Generation backend: vLLM with LoRA… See the full description on the dataset page: https://huggingface.co/datasets/Shaer-AI-2/shaer-sft-test-generations-k5.shaer-sft-test-generations-k5-meter-count
Shaer SFT Test Generations K=5 Meter/Count Scores
This dataset enriches Shaer-AI/shaer-sft-test-generations-k5 with automatic meter and requested-line-count evaluation.
Rows: 17,405
Metric pass: 4BiLSTM meter classifier plus line-count adherence
Core metric columns: meter, count_adherence, parsed_num_lines, requested_num_lines, meter_eval_status, count_eval_status
Overall meter mean: 0.6520384872952881
Overall count-adherence mean: 0.9783539094873928
The automatic meter score… See the full description on the dataset page: https://huggingface.co/datasets/Shaer-AI/shaer-sft-test-generations-k5-meter-count.
