dqs-gemma4-e2b-dpo / eval /final /eval_summary.json
alwaysgood's picture
Upload DQS DPO final evaluation results
b9f8827 verified
Raw
History Blame Contribute Delete
1.64 kB
{
"dataset_path": "data/test.jsonl",
"dry_run": false,
"eval_profile": "final",
"filter_fail_ratio": 0.0,
"filter_fail_rows": 0,
"filter_label_counts": {
"clean": 500
},
"filter_pass_rows": 500,
"generation_ok_rows": 500,
"lora_adapter_path": null,
"metrics": {
"bleu": {
"backend": "sacrebleu",
"higher_is_better": true,
"metric": "bleu",
"row_score": "sentence_bleu",
"score": 29.36231018505856
},
"chrf": {
"backend": "sacrebleu",
"higher_is_better": true,
"metric": "chrf",
"row_score": "sentence_chrf",
"score": 49.60279834708574
},
"comet_wmt22_da": {
"backend": "comet",
"higher_is_better": true,
"mean": 0.8950758044719695,
"model": "Unbabel/wmt22-comet-da"
},
"cometkiwi_wmt22_da": {
"backend": "comet",
"higher_is_better": true,
"mean": 0.8539881974458694,
"model": "Unbabel/wmt22-cometkiwi-da"
},
"metricx_24_hybrid_xxl": {
"backend": "metricx",
"higher_is_better": false,
"mean": 3.5188203125,
"model": "google/metricx-24-hybrid-xxl-v2p6-bfloat16"
},
"xcomet_xxl": {
"backend": "comet",
"higher_is_better": true,
"mean": 0.8844950605034828,
"model": "Unbabel/XCOMET-XXL"
}
},
"model_path": "post_training/dqs_preference_training_hf/outputs/gemma4_e2b_dqs_dpo_full_response_seed42/vllm_final",
"output_dir": "post_training/dqs_preference_training_hf/outputs/gemma4_e2b_dqs_dpo_full_response_seed42/eval/final",
"rows": 500,
"run_id": "gemma4_e2b_dqs_dpo_full_response_seed42",
"usage": {}
}