metadata

license: apache-2.0
base_model: facebook/bart-base
tags:
  - generated_from_trainer
metrics:
  - rouge
model-index:
  - name: bart-base-ckb
    results: []

bart-base-ckb

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

Loss: 0.6353
Rouge1: 0.0
Rouge2: 0.0
Rougel: 0.0
Rougelsum: 0.0
Cer: 4.7349
Gen Len: 13.2035

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: 320
eval_batch_size: 256
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
num_epochs: 5.0

Training results

Training Loss	Epoch	Step	Cer	Gen Len	Validation Loss	Rougel	Rougelsum
3.5765	0.11	500	10.3648	16.358	3.4338	0.0	0.0
3.1749	0.22	1000	10.955	15.828	3.0110	0.0	0.0
2.8696	0.33	1500	11.2579	15.358	2.6943	0.0	0.0
2.6279	0.43	2000	11.6734	14.8565	2.4300	0.0	0.0
2.4166	0.54	2500	11.028	14.523	2.2248	0.0	0.0
2.2444	0.65	3000	10.4379	14.4185	2.0490	0.0	0.0
2.0902	0.76	3500	10.0905	14.242	1.8756	0.0	0.0
1.9565	0.87	4000	9.7629	14.042	1.7377	0.0	0.0
1.8319	0.98	4500	9.4737	13.877	1.6244	0.0	0.0
1.7104	1.09	5000	9.153	13.825	1.5353	0.0	0.0
1.6148	1.2	5500	8.8125	13.726	1.4422	0.0	0.0
1.5459	1.3	6000	8.5589	13.681	1.3755	0.0	0.0
1.4837	1.41	6500	8.2717	13.6225	1.3035	0.0	0.0
1.4219	1.52	7000	8.0684	13.549	1.2407	0.0	0.0
1.3743	1.63	7500	7.7684	13.502	1.1865	0.0	0.0
1.3318	1.74	8000	7.5247	13.509	1.1503	0.0	0.0
1.2893	1.85	8500	7.3826	13.456	1.1085	0.0	0.0
1.2228	2.0	9198	1.0506	0.0	0.0	7.0411	13.3935
0.9343	3.0	13797	0.7769	0.0	0.0	5.5303	13.2935
0.7915	4.0	18396	0.6663	0.0	0.0	4.8928	13.209
0.7436	5.0	22995	0.6353	0.0	0.0	4.7349	13.2035

Framework versions

Transformers 4.34.0.dev0
Pytorch 1.13.1
Datasets 2.14.5
Tokenizers 0.13.3