m2m100_418M-finetuned-en-ru

This model is a fine-tuned version of facebook/m2m100_418M on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.6798
• Bleu: 51.7753
• Gen Len: 61.5271

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: 4
• eval_batch_size: 3
• seed: 42
• gradient_accumulation_steps: 50
• total_train_batch_size: 200
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu	Gen Len
0.8981	1.0	9218	0.8277	48.5181	61.4701
0.8167	2.0	18437	0.7696	49.6956	61.4641
0.7718	3.0	27656	0.7396	50.4045	61.5042
0.7373	4.0	36875	0.7203	50.8134	61.5313
0.713	5.0	46094	0.7069	51.136	61.5931
0.6937	6.0	55312	0.6967	51.4274	61.5911
0.6787	7.0	64531	0.6889	51.5404	61.5012
0.6659	8.0	73750	0.6844	51.7187	61.5229
0.6584	9.0	82969	0.6809	51.8218	61.5392
0.649	10.0	92180	0.6798	51.7753	61.5271

Framework versions

• Transformers 4.33.2
• Pytorch 2.2.0.dev20230920+cu121
• Datasets 2.14.4
• Tokenizers 0.13.3