This model is mbart-large-50-many-to-many-mmt model fine-tuned on English to Zamboanga Chavacano sentences from our own corpus. Chavacano is a Spanish creole, and Zamboanga Chavacano is the most widely spoken Chavacano variety in the Philippines. This translation model aims to leverage the relatedness of Chavacano to Spanish, Tagalog, and other Romance languages through transfer learning to create an English-Zamboanga Chavacano translation model.

The training set contains 5203 sentence, validation set contains 650 sentences, while test set contains 651 sentences. Each sentence is composed of 2-20 words.

The BLEU score of this model on the validation set is 36.37 and 37.74 on the test set.

batch_size = 16
num_train_epochs = 4
learning_rate=2e-4
gradient_accumulation_steps=4
weight_decay=0.01
warmup_steps=500
adam_epsilon=1e-8
max_grad_norm=1.0
fp16=True

Additional Info

  • Source Language: English
  • Target Language: Zamboanga Chavacano
  • BLEU Score: 37.74
  • Brevity Penalty: 0.976
  • Reference Length: 5323
  • Hypothesis Length: 5198
Downloads last month
11
Safetensors
Model size
611M params
Tensor type
F32
ยท
Inference Providers NEW
This model is not currently available via any of the supported Inference Providers.

Model tree for meowyboi/mbart-large-50-en-cbk-finetuned

Finetuned
(130)
this model

Space using meowyboi/mbart-large-50-en-cbk-finetuned 1