Fine tuning model

[dingtutu01]

How do I fine-tune my own dataset?

[frank-xwang]

Hi,

To fine-tune the model on your own datasets, you'll need to follow our pipeline and resume model training by loading our pretrained model.

Specifically:

1. Prepare your datasets by following the instructions in [#training-data-generation](https://github.com/frank-xwang/InstanceDiffusion#training-data-generation).

2. Fine-tune the model by adapting the process described in [#instancediffusion-model-training](https://github.com/frank-xwang/InstanceDiffusion#instancediffusion-model-training). Note: Make sure to "resume" the model training via loading the pretrained checkpoint and fine-tune the model on your own dataset. For more details, refer to

   InstanceDiffusion/utils/checkpoint.py

   Line 104 in 7306022

   def load_autoresume_ckpt(checkpoint, config, model, ema, opt, scheduler):

   to understand how the model resume function works.

Let me know if you have further questions! 😊