m2m100_lr_2e5_gradd_accum_2

This model is a fine-tuned version of facebook/m2m100_418M on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 3.2890
• Bleu: 10.7134
• Gen Len: 45.3067
• Meteor: 0.3064
• Chrf: 33.9112

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: 12
• eval_batch_size: 12
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 24
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 32.0

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu	Gen Len	Meteor	Chrf
3.3178	7.94	2406	2.7937	9.2429	47.6707	0.2869	31.5031
1.6761	15.88	4812	2.8455	10.5222	46.2192	0.3047	33.4306
1.0433	23.82	7218	3.0691	10.4711	46.6063	0.3042	33.6469
0.6522	31.76	9624	3.2890	10.7134	45.3067	0.3064	33.9112

Framework versions

• Transformers 4.30.2
• Pytorch 1.11.0+cu113
• Datasets 2.10.0
• Tokenizers 0.12.1