metadata

license: mit
base_model: ilos-vigil/bigbird-small-indonesian
tags:
  - generated_from_trainer
metrics:
  - accuracy
model-index:
  - name: test_trainer
    results: []

test_trainer

This model is a fine-tuned version of ilos-vigil/bigbird-small-indonesian on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 0.2500
Accuracy: 0.5

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: 4
eval_batch_size: 4
seed: 42
distributed_type: multi-GPU
num_devices: 4
total_train_batch_size: 16
total_eval_batch_size: 16
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: cosine
num_epochs: 1
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.2611	0.04	500	0.2617	0.5
0.2547	0.07	1000	0.2525	0.5
0.2527	0.11	1500	0.2500	0.5
0.2523	0.14	2000	0.2522	0.5
0.253	0.18	2500	0.2504	0.5
0.252	0.21	3000	0.2553	0.5
0.2662	0.25	3500	0.2501	0.5
0.2869	0.29	4000	0.2705	0.5
0.2534	0.32	4500	0.2505	0.5
0.252	0.36	5000	0.2504	0.5
0.2518	0.39	5500	0.2526	0.5
0.2508	0.43	6000	0.2501	0.5
0.2512	0.46	6500	0.2501	0.5
0.251	0.5	7000	0.2500	0.5
0.2502	0.54	7500	0.2502	0.5
0.2507	0.57	8000	0.2545	0.5
0.2511	0.61	8500	0.2509	0.5
0.2503	0.64	9000	0.2500	0.5
0.2506	0.68	9500	0.2501	0.5
0.2505	0.71	10000	0.2504	0.5
0.2505	0.75	10500	0.2504	0.5
0.2504	0.79	11000	0.2501	0.5
0.2502	0.82	11500	0.2500	0.5
0.2503	0.86	12000	0.2500	0.5
0.2501	0.89	12500	0.2500	0.5
0.2503	0.93	13000	0.2500	0.5
0.2502	0.96	13500	0.2500	0.5
0.2502	1.0	14000	0.2500	0.5

Framework versions

Transformers 4.35.2
Pytorch 2.0.1
Datasets 2.15.0
Tokenizers 0.15.0