metadata
tags:
- coconut
- gpt2
- reasoning
base_model: openai-community/gpt2
gpt2-coconut-checkpoint13
CoCoNut (Chain of Continuous Thought) trained checkpoint 13 based on GPT-2.
Training
- Base model:
openai-community/gpt2 - Training method: CoCoNut (continuous latent reasoning)
- Dataset: GSM8K
Usage
import torch
checkpoint = torch.load("pytorch_model.bin")