enoriega/rule_learning_1mm_many_negatives_spanpred_mse_attention

rule_learning_1mm_many_negatives_spanpred_avf

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: 0.0731

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
0.1215	0.16	20	0.1191
0.1091	0.32	40	0.1079
0.0993	0.48	60	0.0993
0.0938	0.64	80	0.0952
0.085	0.8	100	0.0858
0.0837	0.96	120	0.0842
0.0811	1.12	140	0.0827
0.0799	1.28	160	0.0809
0.078	1.44	180	0.0786
0.0792	1.6	200	0.0781
0.0797	1.76	220	0.0765
0.0775	1.92	240	0.0758
0.0735	2.08	260	0.0748
0.0704	2.24	280	0.0744
0.0744	2.4	300	0.0737
0.0752	2.56	320	0.0733
0.075	2.72	340	0.0738
0.0701	2.88	360	0.0732

Framework versions

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