README.md

metadata

base_model: facebook/mbart-large-50-many-to-many-mmt
tags:
  - generated_from_trainer
metrics:
  - bleu
model-index:
  - name: mbart-large-50-many-to-many-mmt-finetuned-hi-to-en
    results: []

mbart-large-50-many-to-many-mmt-finetuned-hi-to-en

This model is a fine-tuned version of facebook/mbart-large-50-many-to-many-mmt on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.5182
• Bleu: 0.0
• Gen Len: 5.5567

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: 4
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 3
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu	Gen Len
2.4071	0.07	500	2.4856	0.0	5.49
2.6727	0.13	1000	2.3521	0.0	5.4733
2.6438	0.2	1500	2.2782	0.0	5.38
2.5672	0.27	2000	2.2432	0.0	5.6367
2.5561	0.34	2500	2.0945	0.0	5.4733
2.5502	0.4	3000	2.0506	0.0	5.4667
2.5326	0.47	3500	2.0163	0.0	5.4367
2.4979	0.54	4000	1.9795	0.0	5.4267
2.4771	0.61	4500	1.9002	0.0	5.28
2.5121	0.67	5000	1.8497	0.0	5.5433
2.4568	0.74	5500	1.7447	0.0	5.66
2.4211	0.81	6000	1.6948	0.0	5.5633
2.4702	0.88	6500	1.5632	0.0	5.4267
2.4161	0.94	7000	1.5122	0.0	5.5633
2.2322	1.01	7500	1.3694	0.0	5.45
1.5244	1.08	8000	1.3296	0.0	5.55
1.5234	1.14	8500	1.2637	0.0	5.5267
1.5902	1.21	9000	1.2621	0.0	5.6133
1.4645	1.28	9500	1.1856	0.0	5.55
1.5426	1.35	10000	1.1528	0.0	5.6267
1.5414	1.41	10500	1.1533	0.0	5.6067
1.4708	1.48	11000	1.1136	0.0	5.5533
1.4625	1.55	11500	1.0736	0.0	5.54
1.5805	1.62	12000	1.0700	0.0	5.5733
1.5295	1.68	12500	1.0171	0.0	5.4867
1.5585	1.75	13000	0.9770	0.0	5.6133
1.5602	1.82	13500	0.9106	0.0	5.5867
1.5361	1.89	14000	0.8823	0.0	5.5333
1.5217	1.95	14500	0.8353	0.0	5.5167
1.3229	2.02	15000	0.7296	0.0	5.4967
0.916	2.09	15500	0.6979	0.0	5.56
0.9075	2.15	16000	0.6797	0.0	5.5333
0.8575	2.22	16500	0.6372	0.0	5.52
0.8702	2.29	17000	0.6288	0.0	5.4967
0.8741	2.36	17500	0.6297	0.0	5.5033
0.8764	2.42	18000	0.6038	0.0	5.5233
0.9152	2.49	18500	0.5853	0.0	5.4933
0.8598	2.56	19000	0.5731	0.0	5.54
0.8586	2.63	19500	0.5558	0.0	5.5533
0.9005	2.69	20000	0.5388	0.0	5.5233
0.889	2.76	20500	0.5269	0.0	5.5067
0.8772	2.83	21000	0.5301	0.0	5.5333
0.8713	2.9	21500	0.5208	0.0	5.5567
0.8962	2.96	22000	0.5182	0.0	5.5567

Framework versions

• Transformers 4.36.2
• Pytorch 2.1.2+cu121
• Datasets 2.16.1
• Tokenizers 0.15.0