MammoLLM2

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

• Loss: 1.6666

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: 64
• eval_batch_size: 64
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 256
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 1000
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss
1.7888	1.24	500	1.0984
1.1179	2.49	1000	1.0209
1.0556	3.73	1500	0.9913
1.01	4.98	2000	0.9653
0.9723	6.22	2500	0.9608
0.9425	7.47	3000	0.9455
0.9149	8.71	3500	0.9391
0.8877	9.96	4000	0.9253
0.8478	11.2	4500	0.9317
0.8142	12.45	5000	0.9313
0.7814	13.69	5500	0.9299
0.7494	14.93	6000	0.9330
0.7071	16.18	6500	0.9588
0.6774	17.42	7000	0.9704
0.6511	18.67	7500	0.9828
0.6275	19.91	8000	1.0007
0.595	21.16	8500	1.0432
0.5698	22.4	9000	1.0641
0.546	23.65	9500	1.0879
0.523	24.89	10000	1.0982
0.4913	26.14	10500	1.1579
0.4622	27.38	11000	1.1923
0.4378	28.62	11500	1.2152
0.4131	29.87	12000	1.2440
0.3846	31.11	12500	1.3181
0.3592	32.36	13000	1.3497
0.3411	33.6	13500	1.3847
0.324	34.85	14000	1.4070
0.3061	36.09	14500	1.4755
0.2903	37.34	15000	1.5078
0.2795	38.58	15500	1.5351
0.2701	39.83	16000	1.5639
0.2605	41.07	16500	1.5972
0.2521	42.31	17000	1.6191
0.2467	43.56	17500	1.6300
0.2425	44.8	18000	1.6453
0.2386	46.05	18500	1.6554
0.2356	47.29	19000	1.6628
0.2344	48.54	19500	1.6663
0.2333	49.78	20000	1.6666

Framework versions

• Transformers 4.31.0
• Pytorch 2.0.1+cu117
• Datasets 2.14.3
• Tokenizers 0.13.3