rule_learning_margin_3mm_many_negatives_spanpred_attention

This model is a fine-tuned version of enoriega/rule_softmatching on the enoriega/odinsynth_dataset dataset. It achieves the following results on the evaluation set:

• Loss: 0.2196
• Margin Accuracy: 0.8969

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: 5e-05
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 2000
• total_train_batch_size: 8000
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 3.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Margin Accuracy
0.3149	0.16	60	0.3098	0.8608
0.2754	0.32	120	0.2725	0.8733
0.2619	0.48	180	0.2512	0.8872
0.2378	0.64	240	0.2391	0.8925
0.2451	0.8	300	0.2305	0.8943
0.2357	0.96	360	0.2292	0.8949
0.2335	1.12	420	0.2269	0.8952
0.2403	1.28	480	0.2213	0.8957
0.2302	1.44	540	0.2227	0.8963
0.2353	1.6	600	0.2222	0.8961
0.2271	1.76	660	0.2207	0.8964
0.228	1.92	720	0.2218	0.8967
0.2231	2.08	780	0.2201	0.8967
0.2128	2.24	840	0.2219	0.8967
0.2186	2.4	900	0.2202	0.8967
0.2245	2.56	960	0.2205	0.8969
0.2158	2.72	1020	0.2196	0.8969
0.2106	2.88	1080	0.2192	0.8968

Framework versions

• Transformers 4.19.2
• Pytorch 1.11.0
• Datasets 2.2.1
• Tokenizers 0.12.1