Vit-GPT2-COCO2017Flickr-115k-12

This model is a fine-tuned version of NourFakih/Vit-GPT2-COCO2017Flickr-115k-12 on an unknown dataset. It achieves the following results on the evaluation set:

• Gen Len: 11.9339
• Loss: 0.5262
• Rouge1: 40.547
• Rouge2: 14.9559
• Rougel: 36.6474
• Rougelsum: 36.655

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: 4
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 3.0

Training results

Training Loss	Epoch	Step	Gen Len	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum
0.3804	0.0696	500	11.8970	0.4475	41.2997	15.8517	37.5088	37.5135
0.383	0.1391	1000	11.4465	0.4486	41.0786	15.592	37.2557	37.247
0.3804	0.2087	1500	11.7144	0.4462	41.0521	15.5552	37.3142	37.3142
0.376	0.2783	2000	11.7237	0.4503	41.0712	15.4215	37.1593	37.1483
0.3742	0.3478	2500	12.056	0.4424	41.0197	15.5533	37.0838	37.0828
0.3702	0.4174	3000	11.51	0.4476	41.3443	15.849	37.5657	37.573
0.3682	0.4870	3500	11.9582	0.4470	41.5477	16.1138	37.5701	37.5784
0.367	0.5565	4000	11.3945	0.4440	41.244	15.7945	37.5079	37.5147
0.3609	0.6261	4500	11.8043	0.4479	41.0531	15.81	37.3273	37.3317
0.3638	0.6957	5000	11.6248	0.4414	41.3129	15.94	37.5627	37.583
0.3543	0.7652	5500	11.5961	0.4482	41.1584	15.7318	37.4062	37.4104
0.3521	0.8348	6000	11.7494	0.4445	41.3954	15.8539	37.6111	37.6208
0.3537	0.9043	6500	11.8193	0.4438	41.6655	16.1228	37.7752	37.7812
0.3459	0.9739	7000	11.614	0.4427	41.4156	15.8963	37.5839	37.5837
0.3092	1.0435	7500	11.9294	0.4620	41.2316	15.7114	37.3554	37.3522
0.2924	1.1130	8000	11.8922	0.4626	40.8321	15.5115	37.0416	37.0319
0.2886	1.1826	8500	11.8237	0.4692	40.9025	15.347	37.0047	36.9942
0.2901	1.2522	9000	11.8817	0.4649	41.0739	15.3744	37.1782	37.1775
0.2868	1.3217	9500	11.8390	0.4668	41.2378	15.6205	37.3887	37.3857
0.2825	1.3913	10000	11.6785	0.4680	41.0405	15.5335	37.1249	37.1293
0.2843	1.4609	10500	11.792	0.4753	40.7387	15.0593	36.8401	36.8406
0.2819	1.5305	11000	11.7928	0.4718	41.1479	15.5299	37.2971	37.2975
0.2791	1.6001	11500	11.7897	0.4728	40.8974	15.2068	37.0748	37.0794
0.2756	1.6696	12000	11.7343	0.4776	40.9051	15.3769	37.1201	37.122
0.275	1.7392	12500	11.6749	0.4799	41.0987	15.4789	37.1856	37.1813
0.2703	1.8088	13000	11.6395	0.4787	41.211	15.6686	37.34	37.3342
0.2733	1.8783	13500	11.7356	0.4808	41.1137	15.5692	37.2625	37.2592
0.27	1.9479	14000	11.7816	0.4818	41.4823	15.8925	37.6161	37.6165
0.255	2.0175	14500	11.6703	0.5038	40.9089	15.2776	36.9295	36.9275
0.2292	2.0871	15000	11.8069	0.5076	41.0044	15.3659	37.0441	37.0438
0.2226	2.1567	15500	11.8068	0.5129	40.8428	15.2553	36.9787	36.9642
0.2236	2.2262	16000	11.8470	0.5153	40.6711	15.13	36.8865	36.8805
0.2215	2.2958	16500	11.8765	0.5200	40.7621	15.0155	36.8522	36.85
0.2215	2.3654	17000	11.7364	0.5186	40.6314	15.1297	36.8185	36.8123
0.2202	2.4349	17500	11.9915	0.5208	40.6588	15.0601	36.5692	36.5715
0.2126	2.5045	18000	11.8697	0.5218	40.6073	14.8542	36.6365	36.6408
0.2178	2.5741	18500	11.9752	0.5221	40.2978	14.671	36.4154	36.4278
0.2179	2.6436	19000	11.9277	0.5201	40.5832	14.9498	36.6925	36.7034
0.2165	2.7132	19500	11.9769	0.5252	40.5697	15.0438	36.7188	36.7223
0.2134	2.7827	20000	11.938	0.5269	40.6224	15.0783	36.6861	36.6882
0.2157	2.8523	20500	11.8804	0.5289	40.4897	14.9113	36.6069	36.6136
0.2142	2.9219	21000	11.8809	0.5260	40.4523	14.952	36.6097	36.616
0.2124	2.9914	21500	11.9339	0.5262	40.547	14.9559	36.6474	36.655

Framework versions

• Transformers 4.42.3
• Pytorch 2.1.2
• Datasets 2.20.0
• Tokenizers 0.19.1