beit-base-finetuned-ade-640-640_alpha0.7_temp5.0_t2

This model is a fine-tuned version of c14kevincardenas/ClimBEiT-t2 on the c14kevincardenas/beta_caller_284_person_crop_seq_withlimb dataset. It achieves the following results on the evaluation set:

• Loss: 0.5175
• Accuracy: 0.8340

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: 32
• eval_batch_size: 32
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
1.1235	1.0	180	1.3239	0.3933
0.6303	2.0	360	0.8419	0.6877
0.471	3.0	540	0.7387	0.7322
0.3553	4.0	720	0.6575	0.7737
0.3071	5.0	900	0.6260	0.7826
0.2206	6.0	1080	0.6078	0.7964
0.2027	7.0	1260	0.5722	0.8053
0.1922	8.0	1440	0.5999	0.8093
0.1801	9.0	1620	0.6051	0.7984
0.1758	10.0	1800	0.5721	0.8063
0.154	11.0	1980	0.5764	0.8103
0.1495	12.0	2160	0.5539	0.8172
0.1334	13.0	2340	0.5272	0.8389
0.132	14.0	2520	0.5427	0.8271
0.1284	15.0	2700	0.5345	0.8251
0.1205	16.0	2880	0.5285	0.8350
0.1178	17.0	3060	0.5358	0.8300
0.1199	18.0	3240	0.5190	0.8350
0.112	19.0	3420	0.5260	0.8310
0.1087	20.0	3600	0.5175	0.8340

Framework versions

• Transformers 4.45.2
• Pytorch 2.5.0+cu124
• Datasets 3.0.1
• Tokenizers 0.20.1