Komal-patra/EU_AI_ACT_T5_Model

results

This model is a fine-tuned version of google/flan-t5-base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 2.1507

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: 0.001
• train_batch_size: 4
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 3
• total_train_batch_size: 12
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 16
• training_steps: 1698

Training results

Training Loss	Epoch	Step	Validation Loss
3.0084	0.59	50	2.5425
2.7308	1.18	100	2.4483
2.6435	1.76	150	2.3925
2.5873	2.35	200	2.3558
2.5247	2.94	250	2.3276
2.5323	3.53	300	2.3003
2.4288	4.12	350	2.2771
2.4247	4.71	400	2.2659
2.4014	5.29	450	2.2439
2.3761	5.88	500	2.2336
2.3056	6.47	550	2.2236
2.3443	7.06	600	2.2182
2.2877	7.65	650	2.2066
2.3028	8.24	700	2.1953
2.2589	8.82	750	2.1958
2.2306	9.41	800	2.1834
2.2571	10.0	850	2.1826
2.2109	10.59	900	2.1782
2.2216	11.18	950	2.1802
2.1881	11.76	1000	2.1734
2.1794	12.35	1050	2.1691
2.1933	12.94	1100	2.1654
2.134	13.53	1150	2.1682
2.1698	14.12	1200	2.1564
2.1477	14.71	1250	2.1599
2.1353	15.29	1300	2.1573
2.1206	15.88	1350	2.1525
2.1175	16.47	1400	2.1520
2.1142	17.06	1450	2.1531
2.1152	17.65	1500	2.1529
2.1073	18.24	1550	2.1529
2.099	18.82	1600	2.1520
2.1061	19.41	1650	2.1507

Framework versions

• PEFT 0.8.2
• Transformers 4.38.1
• Pytorch 2.3.0+cu121
• Datasets 2.17.0
• Tokenizers 0.15.2