Hey there,
First, thanks for the excellent work! We are inspired by your methodology and would like to build upon them.
To further our research, we are interested in exploring other SFT methods based on the continued pretrained model. So would it be possible to share the model checkpoints for Qwen2.5-Base+CPT? This would be incredibly helpful for our experiments.
We completely understand if sharing is not feasible and are happy to follow any guidelines if needed. Thank you for your time and consideration!
Best regards