Beta
Explore
Marketplace
Neural Labs
Playground
Wallet
Docs
za-zh-theme – Dataset by FiveC | AlphaNeural AI
You can deploy this model and start earning money today!
FiveC
/
za-zh-theme
like
0
10K<n<100K
parquet
text
datasets
pandas
polars
mlcroissant
us
Views
No views yet
Model card
Files and Versions
Community
API
za-zh-theme
This model is a fine-tuned version of
facebook/mbart-large-50-many-to-many-mmt
on an unknown dataset. It achieves the following results on the evaluation set:
Loss: 2.7864
Sacrebleu: 6.6442
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: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: linear
num_epochs: 3
mixed_precision_training: Native AMP
Training results
Training Loss
Epoch
Step
Validation Loss
Sacrebleu
0.4156
1.0
1678
2.7864
6.6442
0.1474
2.0
3356
2.8686
5.0403
0.0730
3.0
5034
2.9130
6.2466
Framework versions
Transformers 5.0.0
Pytorch 2.10.0+cu128
Datasets 4.0.0
Tokenizers 0.22.2
