CoolFace
Modelpublic

thunderboltc/mbart50_sanlish-to-bangla_1934_lr2

sourceHugging Faceupdated 1mo agoView on Hugging Face
0likes2.6kdownloads
Model Card

<!-- 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. -->

mbart50sanlish-to-bangla1934_lr2

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: 3.4931
  • —Bleu: 17.9652
  • —Chrf: 22.9651
  • —Gen Len: 4.6287

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
  • —trainbatchsize: 4
  • —evalbatchsize: 4
  • —seed: 42
  • —optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • —lrschedulertype: linear
  • —num_epochs: 25
  • —mixedprecisiontraining: Native AMP

Training results

Training LossEpochStepValidation LossBleuChrfGen Len
3.83771.03343.228716.233443.76824.4701
2.3872.06682.882012.703329.23394.4611
1.50483.010022.852210.682222.96514.6856
0.93064.013363.036317.286042.75624.7695
0.55895.016703.137017.965224.46384.7275
0.35226.020043.19229.287528.96864.5569
0.2537.023383.24209.820422.38494.7246
0.19298.026723.25558.257922.36814.6287
0.14199.030063.29118.257922.36814.7186
0.104710.033403.32757.809822.14844.6168
0.075711.036743.373411.478756.20264.7186
0.057812.040083.41039.652425.80174.8234
0.045913.043423.388113.042428.69004.7575
0.02914.046763.426617.965271.20644.6257
0.023415.050103.44739.820422.38494.5988
0.017916.053443.43607.809823.53474.6048
0.012717.056783.453717.965222.96514.8234
0.009218.060123.465117.965222.96514.7485
0.007619.063463.504117.965222.96514.5569
0.005720.066803.488417.965222.96514.6377
0.005221.070143.475717.965222.96514.6946
0.004322.073483.490817.965222.96514.6228
0.003123.076823.490917.965222.96514.6976
0.002824.080163.49249.287523.70704.6737
0.002525.083503.493117.965222.96514.6287

Framework versions

  • —Transformers 4.44.2
  • —Pytorch 2.11.0+cu128
  • —Datasets 4.0.0
  • —Tokenizers 0.19.1