sandernotenbaert/OKAI-midi-gen-v-001

<!-- 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. -->

OKAI-midi-gen-v-001

This model is a fine-tuned version of [](https://huggingface.co/) on an unknown dataset. It achieves the following results on the evaluation set:

• —Loss: 10.1912
• —Accuracy: 0.0008

Model description

First test with small subset on M1Pro. Generates valid files, notes very clustered with long gaps

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

trainingconfig: vocabsize: 30000 hiddensize: 256 intermediatesize: 512 numhiddenlayers: 4 numattentionheads: 4 numkeyvalueheads: 4 slidingwindow: 4 maxpositionembeddings: 1024 padtokenid: 0 bostokenid: 1 eostokenid: 2

Training hyperparameters

The following hyperparameters were used during training:

• —learning_rate: 0.0001
• —trainbatchsize: 8
• —evalbatchsize: 16
• —seed: 444
• —gradientaccumulationsteps: 3
• —totaltrainbatch_size: 24
• —optimizer: Use adamwtorch with betas=(0.9,0.999) and epsilon=1e-08 and optimizerargs=No additional optimizer arguments
• —lrschedulertype: cosinewithrestarts
• —lrschedulerwarmup_ratio: 0.3
• —training_steps: 2000

Training results

Framework versions

• —Transformers 4.52.3
• —Pytorch 2.6.0
• —Datasets 3.6.0
• —Tokenizers 0.21.1