gpt2-NaturalQuestions_4000-ep20

This model is a fine-tuned version of gpt2 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.2717

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: 48
• eval_batch_size: 96
• seed: 1799
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss
1.7832	0.6	50	1.3857
1.4409	1.19	100	1.3070
1.312	1.79	150	1.2716
1.2412	2.38	200	1.2514
1.1902	2.98	250	1.2283
1.1171	3.57	300	1.2307
1.0728	4.17	350	1.2228
1.022	4.76	400	1.2145
1.0001	5.36	450	1.2225
0.9645	5.95	500	1.2183
0.9175	6.55	550	1.2185
0.911	7.14	600	1.2179
0.877	7.74	650	1.2218
0.8365	8.33	700	1.2260
0.8387	8.93	750	1.2223
0.8047	9.52	800	1.2289
0.7822	10.12	850	1.2304
0.7652	10.71	900	1.2353
0.7502	11.31	950	1.2370
0.7275	11.9	1000	1.2411
0.6998	12.5	1050	1.2515
0.7128	13.1	1100	1.2465
0.6865	13.69	1150	1.2553
0.6748	14.29	1200	1.2544
0.6661	14.88	1250	1.2563
0.6636	15.48	1300	1.2592
0.6403	16.07	1350	1.2630
0.6309	16.67	1400	1.2679
0.6281	17.26	1450	1.2667
0.6237	17.86	1500	1.2692
0.621	18.45	1550	1.2708
0.6195	19.05	1600	1.2711
0.6123	19.64	1650	1.2713

Framework versions

• Transformers 4.29.2
• Pytorch 1.10.0+cu111
• Datasets 2.5.1
• Tokenizers 0.13.3