README.md

metadata

license: mit
tags:
  - generated_from_trainer
model-index:
  - name: rubert-tiny2-srl
    results: []

rubert-tiny2-srl

This model is a fine-tuned version of cointegrated/rubert-tiny2 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.3357
• Benefactive Precision: 1.0
• Benefactive Recall: 0.5
• Benefactive F1: 0.6667
• Benefactive Number: 2
• Causator Precision: 1.0
• Causator Recall: 1.0
• Causator F1: 1.0
• Causator Number: 12
• Cause Precision: 0.4
• Cause Recall: 0.4
• Cause F1: 0.4000
• Cause Number: 5
• Contrsubject Precision: 0.75
• Contrsubject Recall: 0.6667
• Contrsubject F1: 0.7059
• Contrsubject Number: 9
• Deliberative Precision: 1.0
• Deliberative Recall: 1.0
• Deliberative F1: 1.0
• Deliberative Number: 4
• Experiencer Precision: 0.7442
• Experiencer Recall: 0.8101
• Experiencer F1: 0.7758
• Experiencer Number: 79
• Object Precision: 0.7551
• Object Recall: 0.7450
• Object F1: 0.7500
• Object Number: 149
• Predicate Precision: 0.9809
• Predicate Recall: 0.9885
• Predicate F1: 0.9847
• Predicate Number: 260
• Overall Precision: 0.8705
• Overall Recall: 0.8788
• Overall F1: 0.8746
• Overall Accuracy: 0.9411

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.00010372880304918982
• train_batch_size: 1
• eval_batch_size: 1
• seed: 923789
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.29
• num_epochs: 5
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Benefactive Precision	Benefactive Recall	Benefactive F1	Benefactive Number	Causator Precision	Causator Recall	Causator F1	Causator Number	Cause Precision	Cause Recall	Cause F1	Cause Number	Contrsubject Precision	Contrsubject Recall	Contrsubject F1	Contrsubject Number	Deliberative Precision	Deliberative Recall	Deliberative F1	Deliberative Number	Experiencer Precision	Experiencer Recall	Experiencer F1	Experiencer Number	Object Precision	Object Recall	Object F1	Object Number	Predicate Precision	Predicate Recall	Predicate F1	Predicate Number	Overall Precision	Overall Recall	Overall F1	Overall Accuracy
0.3041	1.0	4864	0.3394	0.0	0.0	0.0	2	1.0	0.4167	0.5882	12	0.0	0.0	0.0	5	1.0	0.1111	0.2000	9	0.0	0.0	0.0	4	0.8372	0.4557	0.5902	79	0.8730	0.3691	0.5189	149	0.9884	0.9846	0.9865	260	0.9515	0.6788	0.7924	0.9141
0.3178	2.0	9728	0.2692	0.0	0.0	0.0	2	1.0	1.0	1.0	12	1.0	0.2	0.3333	5	1.0	0.2222	0.3636	9	1.0	0.5	0.6667	4	0.7403	0.7215	0.7308	79	0.8523	0.5034	0.6329	149	0.9808	0.9846	0.9827	260	0.9142	0.7788	0.8411	0.9321
0.124	3.0	14592	0.2990	1.0	0.5	0.6667	2	1.0	1.0	1.0	12	0.0	0.0	0.0	5	0.75	0.6667	0.7059	9	1.0	0.5	0.6667	4	0.7386	0.8228	0.7784	79	0.8	0.6980	0.7455	149	0.9885	0.9885	0.9885	260	0.8904	0.8596	0.8748	0.9435
0.104	4.0	19456	0.2852	1.0	0.5	0.6667	2	0.9231	1.0	0.9600	12	0.4286	0.6	0.5	5	0.6	0.6667	0.6316	9	1.0	0.75	0.8571	4	0.7253	0.8354	0.7765	79	0.7044	0.7517	0.7273	149	0.9847	0.9885	0.9866	260	0.8440	0.8846	0.8638	0.9359
0.0918	5.0	24320	0.3357	1.0	0.5	0.6667	2	1.0	1.0	1.0	12	0.4	0.4	0.4000	5	0.75	0.6667	0.7059	9	1.0	1.0	1.0	4	0.7442	0.8101	0.7758	79	0.7551	0.7450	0.7500	149	0.9809	0.9885	0.9847	260	0.8705	0.8788	0.8746	0.9411

Framework versions

• Transformers 4.28.1
• Pytorch 2.0.0+cu117
• Datasets 2.11.0
• Tokenizers 0.13.3