metterian/kullm-polyglot-12.8b-v1

KULLM-Polyglot-12.8B-v2

This model is a fine-tuned version of EleutherAI/polyglot-ko-12.8b on a KULLM v2

Detail Codes are available at KULLM Github Repository

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• —learning_rate: 5e-05
• —trainbatchsize: 64
• —seed: 42
• —distributed_type: multi-GPU (A100 80G)
• —num_devices: 4
• —gradientaccumulationsteps: 16
• —totaltrainbatch_size: 256
• —totalevalbatch_size: 32
• —optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• —lrschedulertype: cosine
• —num_epochs: 10.0

Framework versions

• —Transformers 4.28.1
• —Pytorch 2.0.0+cu117
• —Datasets 2.11.0
• —Tokenizers 0.13.3