cantonese-chinese-translation

This model is a fine-tuned version of fnlp/bart-base-chinese on raptorkwok/cantonese-traditional-chinese-parallel-corpus dataset. It achieves the following results on the evaluation set:

• Loss: 0.2258
• Bleu: 62.1085
• Chrf: 60.1854
• Gen Len: 12.8755

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: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu	Chrf	Gen Len
0.3606	0.48	1000	0.2592	60.9844	58.8851	12.8446
0.3059	0.96	2000	0.2291	61.9606	60.1201	12.8621
0.2296	1.44	3000	0.2254	61.9458	60.0434	12.8578
0.2231	1.92	4000	0.2176	61.9617	59.9299	12.8827
0.174	2.39	5000	0.2290	61.9661	59.8844	12.9068
0.171	2.87	6000	0.2258	62.1085	60.1854	12.8755
0.1346	3.35	7000	0.2334	61.4554	59.5055	12.8175
0.1285	3.83	8000	0.2408	61.3332	59.3276	12.8412
0.1061	4.31	9000	0.2530	61.6505	59.614	12.8566

Framework versions

• Transformers 4.28.1
• Pytorch 2.1.0+cu121
• Datasets 2.14.6
• Tokenizers 0.13.3