Jennny/llama-3.1-coherence-reg-adapter

<!-- 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. -->

llama-3.1-coherence-reg-adapter

This model is a fine-tuned version of meta-llama/Llama-3.1-8B-Instruct on an unknown dataset. It achieves the following results on the evaluation set:

• —Loss: 0.3659
• —Mse: 0.3659
• —Rmse: 0.6049
• —Mae: 0.4649
• —R2: 0.1011
• —Rounded Accuracy: 0.6873
• —Mae Class 0: 2.9010
• —Mse Class 0: 8.5636
• —Mae Class 1: 2.2816
• —Mse Class 1: 5.3301
• —Mae Class 2: 1.4010
• —Mse Class 2: 2.0718
• —Mae Class 3: 0.6129
• —Mse Class 3: 0.4212
• —Mae Class 4: 0.3256
• —Mse Class 4: 0.1442
• —Pred Count 0: 0
• —Pred Percent 0: 0.0
• —Pred Count 1: 1
• —Pred Percent 1: 0.0246
• —Pred Count 2: 14
• —Pred Percent 2: 0.3444
• —Pred Count 3: 671
• —Pred Percent 3: 16.5068
• —Pred Count 4: 3379
• —Pred Percent 4: 83.1242

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: 1
• —evalbatchsize: 1
• —seed: 42
• —gradientaccumulationsteps: 8
• —totaltrainbatch_size: 8
• —optimizer: Use OptimizerNames.ADAMWTORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizerargs=No additional optimizer arguments
• —lrschedulertype: cosine
• —lrschedulerwarmup_ratio: 0.1
• —num_epochs: 1
• —mixedprecisiontraining: Native AMP

Training results

Framework versions

• —PEFT 0.13.2
• —Transformers 4.49.0
• —Pytorch 2.5.1+cu124
• —Datasets 3.3.2
• —Tokenizers 0.21.1