longt5_xl_summ_screen_bp_only_30

This model is a fine-tuned version of /exports/eddie/scratch/s1970716/models/summarization/longt5_xl_summ_screen_bp_only/checkpoint-210 on the learn3r/summ_screen_fd_bp dataset. It achieves the following results on the evaluation set:

• Loss: 2.2376
• Rouge1: 40.4388
• Rouge2: 16.4662
• Rougel: 28.0771
• Rougelsum: 38.3405
• Gen Len: 246.7396

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: 0.0005
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 32
• total_train_batch_size: 256
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: constant
• num_epochs: 15.0

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum	Gen Len
0.324	0.97	14	2.2376	40.4388	16.4662	28.0771	38.3405	246.7396
0.2707	1.95	28	2.3204	40.2873	16.7641	27.3895	38.2689	307.3787
0.2217	2.99	43	2.5281	31.9916	13.8136	22.1895	30.623	501.9320
0.1776	3.97	57	2.7530	31.7535	13.8852	22.8653	30.3796	489.6183
0.1424	4.94	71	2.6578	32.117	14.2141	22.3733	30.8328	502.1124
0.1449	5.98	86	2.5508	35.3448	13.8478	24.9044	33.6108	357.3136
0.1191	6.96	100	3.1622	37.2189	16.0076	25.7011	35.294	408.8669
0.0879	8.0	115	2.8510	39.8825	16.8073	27.2428	37.9568	318.2278
0.0899	8.97	129	2.9138	31.7139	13.7066	21.8844	30.5075	500.4053
0.0656	9.95	143	3.1616	33.055	14.5841	22.5883	31.7565	488.1686
0.0542	10.99	158	3.3630	43.7514	18.9011	29.9017	41.6887	198.8077
0.0557	11.97	172	3.3826	42.3089	18.2735	29.0356	40.4154	270.9675
0.0542	12.94	186	3.4408	40.7691	16.529	28.3999	38.9723	186.7308
0.0596	13.98	201	3.5253	37.0037	15.9098	25.2808	35.3868	398.4704
0.0385	14.61	210	3.4990	32.5815	14.2951	22.4501	31.2928	499.3107

Framework versions

• Transformers 4.34.0.dev0
• Pytorch 2.0.1+cu117
• Datasets 2.14.5
• Tokenizers 0.13.3