LLM_Teached_Pegasus_From_Scratch

This model is a fine-tuned version of google/pegasus-large on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.5146
• Rouge1: 0.4863
• Rouge2: 0.2348
• Rougel: 0.4011
• Rougelsum: 0.4012
• Gen Len: 27.5716
• Precision: 0.9118
• Recall: 0.9131
• F1: 0.9122

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: 2e-05
• train_batch_size: 24
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 96
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 16
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	F1	Gen Len	Validation Loss	Precision	Recall	Rouge1	Rouge2	Rougel	Rougelsum
2.0443	1.0	521	0.9049	28.3633	1.7046	0.9041	0.9061	0.4488	0.203	0.3633	0.3633
1.7826	2.0	1042	0.9072	28.1949	1.6347	0.9062	0.9085	0.4616	0.2133	0.3761	0.3758
1.7134	3.0	1563	0.9084	28.5218	1.5991	0.9072	0.91	0.4683	0.2186	0.3824	0.3822
1.6664	4.0	2084	0.9096	28.2498	1.5767	0.9087	0.9109	0.4738	0.2233	0.3878	0.3876
1.6296	5.0	2605	0.9103	28.2396	1.5595	0.9093	0.9117	0.4775	0.2265	0.3911	0.391
1.5984	6.0	3126	0.9109	28.28	1.5468	0.9098	0.9124	0.4805	0.2284	0.3941	0.3938
1.5738	7.0	3647	1.5370	0.4807	0.2296	0.3945	0.3946	27.8378	0.9105	0.9124	0.9113
1.5476	8.0	4168	1.5308	0.4823	0.2315	0.3963	0.3965	27.7364	0.9108	0.9125	0.9114
1.535	9.0	4689	1.5261	0.4829	0.2309	0.3974	0.3974	27.6535	0.911	0.9125	0.9116
1.52	10.0	5210	1.5231	0.4847	0.2332	0.3992	0.3993	27.816	0.911	0.9128	0.9117
1.5145	11.0	5731	1.5200	0.4851	0.2339	0.4004	0.4006	27.3604	0.9119	0.9127	0.9121
1.5028	12.0	6252	1.5178	0.4858	0.2345	0.4001	0.4002	27.4625	0.9118	0.9129	0.9122
1.4946	13.0	6773	1.5164	0.4859	0.2341	0.4004	0.4005	27.6789	0.9115	0.9131	0.9121
1.4877	14.0	7294	1.5151	0.4868	0.235	0.4013	0.4013	27.5804	0.9119	0.9131	0.9123
1.4855	15.0	7815	1.5146	0.4863	0.2349	0.4014	0.4016	27.5844	0.9117	0.9131	0.9122
1.4782	16.0	8336	1.5146	0.4863	0.2348	0.4011	0.4012	27.5716	0.9118	0.9131	0.9122

Framework versions

• Transformers 4.36.0
• Pytorch 2.0.1+cu117
• Datasets 2.14.5
• Tokenizers 0.15.0