<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->

<!-- in this model i use transfer learning for translate english to Moroccain dialect (darija). -->

<!-- about dataset used for training model : I used about 18,000 pairs of English and Moroccain Dialect. -->

<!-- my model is trained three times, the last being one epoch. -->

Helsinki-NLPopus-mt-tc-big-en-moroccain_dialect

This model was trained from scratch on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.6930
• Bleu: 50.0607
• Gen Len: 14.7048

Model description

MarianConfig { "_name_or_path": "/content/drive/MyDrive/Colab Notebooks/big_helsinki_eng_dar", "activation_dropout": 0.0, "activation_function": "relu", "architectures": [ "MarianMTModel" ], "attention_dropout": 0.0, "bad_words_ids": [ [ 61246 ] ], "bos_token_id": 0, "classifier_dropout": 0.0, "d_model": 1024, "decoder_attention_heads": 16, "decoder_ffn_dim": 4096, "decoder_layerdrop": 0.0, "decoder_layers": 6, "decoder_start_token_id": 61246, "decoder_vocab_size": 61247, "dropout": 0.1, "encoder_attention_heads": 16, "encoder_ffn_dim": 4096, "encoder_layerdrop": 0.0, "encoder_layers": 6, "eos_token_id": 25897, "forced_eos_token_id": 25897, "init_std": 0.02, "is_encoder_decoder": true, "max_length": 512, "max_position_embeddings": 1024, "model_type": "marian", "normalize_embedding": false, "num_beams": 4, "num_hidden_layers": 6, "pad_token_id": 61246, "scale_embedding": true, "share_encoder_decoder_embeddings": true, "static_position_embeddings": true, "torch_dtype": "float32", "transformers_version": "4.28.0", "use_cache": true, "vocab_size": 61247 }

Intended uses & limitations

More information needed

Training and evaluation data

DatasetDict({ train: Dataset({ features: ['input_ids', 'attention_mask', 'labels'], num_rows: 15443 }) test: Dataset({ features: ['input_ids', 'attention_mask', 'labels'], num_rows: 813 }) })

Training procedure

Using transfer learning due to limited data in the Moroccan dialect.

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2e-07
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 4000
• num_epochs: 1
• mixed_precision_training: Native AMP

Training results

Framework versions

• Transformers 4.28.0
• Pytorch 2.0.0+cu118
• Datasets 2.12.0
• Tokenizers 0.13.3