- Apr 23, 2012
- Reaction score
How many days would it take to train GPT-J-6B model using 1080Ti GPU?
I'm planning to learn more about it. I thought it would cost more to train a model.
GPT-J-6B is a pre-trained model. It can be further fine-tuned.
In order for it to run on a 1080Ti - @Cognitive mentions the 8-bit variation. It is a quantized 8-bit version - https://huggingface.co/hivemind/gpt-j-6B-8bit - that can run on a single GPU
How many days it would take would depend on your dataset.