whr778/gliner2-multi-v1-cmnee

gliner2-multi-v1_cmnee

A GLiNER2 multi-task information-extraction model (entities, relations, events, and classification) fine-tuned from fastino/gliner2-multi-v1.

⚠️ License at a glance

• Effective license: Unverified — review required
• Commercial use: Unverified
• All dataset licenses verified: No

See License for the full determination and per-dataset terms.

Model details

• Base model: fastino/gliner2-multi-v1
• Library: gliner2
• Tasks: entity, relation, event, and classification extraction
• Experiment: gliner2-multi-v1_cmnee

Training data

1 dataset used for this run. 9,284 training records (val: 1,606, test: 2,727).

Dataset	Task(s)	Train	Val	Test	Language	License	Source
CMNEE	Event extraction (Chinese military)	9,284	1,606	2,727	zh	see source	link

Dataset notes

• CMNEE — Chinese military news event extraction; 8 event types, 11 argument roles.

Training procedure

Setting	Value
Trained on	2026-07-19
Duration	1h 17m
Throughput	17.9 samples/s
Epochs	15
Batch size	16 (× 2 grad-accum)
Encoder LR	1e-05
Task-head LR	0.0003
Weight decay	0.01
Scheduler	cosine_restarts (warmup 0.05)
Precision	bf16
Max grad norm	1.0
Best-checkpoint metric	eval_event_strict_micro_f1
Seed	42

Evaluation

Decision threshold: 0.5 (config default).

Blind test (held-out test splits)

Micro precision / recall / F1, strict → relaxed.

Category	Precision	Recall	F1	Support
event_type	1.000 → 1.000	0.972 → 0.972	0.986 → 0.986	3819
event_trigger	0.921 → 0.924	0.832 → 0.835	0.874 → 0.877	4848
event_argument	0.268 → 0.838	0.188 → 0.614	0.221 → 0.709	18432
event	0.533 → 0.885	0.414 → 0.709	0.466 → 0.787	27099

Best checkpoint (validation)

Micro precision / recall / F1, strict → relaxed.

Category	Precision	Recall	F1	Support
event_type	1.000 → 1.000	0.981 → 0.981	0.990 → 0.990	2248
event_trigger	0.894 → 0.914	0.837 → 0.856	0.864 → 0.884	2643
event_argument	0.308 → 0.816	0.227 → 0.623	0.261 → 0.706	9366
event	0.565 → 0.872	0.459 → 0.726	0.506 → 0.792	14257

License

Effective license: Unverified — review required. This model is a derivative of its base model and every training dataset, so the most restrictive term across all of them governs the whole model.

• Commercial use: Unverified
• Share-alike obligation: No
• All licenses verified: No
• Base model: gliner2-multi-v1 — see model card

Unverified — verify the upstream terms before redistribution

• CMNEE (see source)
• gliner2-multi-v1 (see model card)

License strings are copied verbatim from each dataset's card/source and from tools/train/dataset_registry.yaml. "see card"/"see source"/"other" mean the upstream declares no clear license — treat as unverified. This summary is informational, not legal advice; confirm terms before redistribution or commercial use.

Citation

If you use this model, please cite GLiNER2 and the underlying datasets (linked in Training data).

---

Model card generated automatically at the end of training (2026-07-19).