Hugging Face's logo

beomi
/
KoAlpaca-Polyglot-12.8B

Text Generation
Transformers
PyTorch
Safetensors
Korean
gpt_neox
Generated from Trainer
polyglot-ko
gpt-neox
KoAlpaca
text-generation-inference
Inference Endpoints
Model card Files Community
1
KoAlpaca-Polyglot-12.8B / README.md
beomi's picture
beomi
Update README.md
c53f71b
preview code
|
raw
history blame
1.12 kB
metadata
license: apache-2.0
tags:
  - generated_from_trainer
  - polyglot-ko
  - gpt-neox
  - KoAlpaca
model-index:
  - name: KoAlpaca-Polyglot-12.8B
    results: []
language:
  - ko
datasets:
  - KoAlpaca-v1.1b
pipeline_tag: text-generation

Update @ 2023.06.01

  • Add Safetensor sharded model weight (max shard = 1GB)

KoAlpaca-Polyglot-12.8B (v1.1b)

This model is a fine-tuned version of EleutherAI/polyglot-ko-12.8b on a KoAlpaca Dataset v1.1b

Detail Codes are available at KoAlpaca Github Repository

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 5e-05
  • train_batch_size: 1
  • seed: 42
  • distributed_type: multi-GPU (A100 80G)
  • num_devices: 4
  • gradient_accumulation_steps: 64
  • total_train_batch_size: 256
  • total_eval_batch_size: 32
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 2.0

Framework versions

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