idefics2_ft_augmented_dataset

This model is a fine-tuned version of HuggingFaceM4/idefics2-8b on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.1871

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: 2.5e-05
• train_batch_size: 4
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 25
• training_steps: 5000

Training results

Training Loss	Epoch	Step	Validation Loss
0.6338	0.0234	100	0.6474
0.5748	0.0468	200	0.5899
0.4678	0.0702	300	0.5603
0.428	0.0936	400	0.5249
0.3798	0.1170	500	0.4984
0.3665	0.1404	600	0.4733
0.4406	0.1637	700	0.4510
0.4723	0.1871	800	0.4245
0.4807	0.2105	900	0.4158
0.4196	0.2339	1000	0.3971
0.3443	0.2573	1100	0.3738
0.4133	0.2807	1200	0.3631
0.2838	0.3041	1300	0.3334
0.4134	0.3275	1400	0.3264
0.2838	0.3509	1500	0.3125
0.275	0.3743	1600	0.2944
0.4141	0.3977	1700	0.2839
0.2498	0.4211	1800	0.2749
0.2817	0.4444	1900	0.2606
0.2899	0.4678	2000	0.2526
0.2695	0.4912	2100	0.2521
0.2619	0.5146	2200	0.2424
0.2238	0.5380	2300	0.2373
0.3049	0.5614	2400	0.2301
0.1308	0.5848	2500	0.2292
0.1936	0.6082	2600	0.2190
0.2479	0.6316	2700	0.2191
0.1575	0.6550	2800	0.2165
0.193	0.6784	2900	0.2107
0.2526	0.7018	3000	0.2114
0.1574	0.7251	3100	0.2087
0.1989	0.7485	3200	0.2051
0.1761	0.7719	3300	0.2013
0.2223	0.7953	3400	0.1996
0.2127	0.8187	3500	0.1966
0.2477	0.8421	3600	0.1923
0.1931	0.8655	3700	0.1908
0.182	0.8889	3800	0.1888
0.1693	0.9123	3900	0.1878
0.1346	0.9357	4000	0.1853
0.1484	0.9591	4100	0.1849
0.1217	0.9825	4200	0.1838
0.0669	1.0058	4300	0.1844
0.1292	1.0292	4400	0.1877
0.1106	1.0526	4500	0.1876
0.0828	1.0760	4600	0.1875
0.0485	1.0994	4700	0.1871
0.0624	1.1228	4800	0.1874
0.0895	1.1462	4900	0.1871
0.1	1.1696	5000	0.1871

Framework versions

• Transformers 4.41.2
• Pytorch 2.3.1+cu121
• Datasets 2.19.2
• Tokenizers 0.19.1