Request for Training Code or Config for Interclip.ckpt Model

[bring-nirachornkul]

Hi, I’m working with the Interclip.ckpt checkpoint for evaluation but running into some issues aligning the setup. Could you please share the training script (e.g., train_interclip.py) or the config used to generate Interclip.ckpt?

The eval config (eval_model.yaml) specifies INPUT_DIM=258, yet the dataset outputs 262 features per motion, and the checkpoint suggests a dual-motion input (516=258*2), conflicting with the single motion provided in eval.

This would help ensure I’m using the correct input dimensions and preprocessing steps. Thanks so much for your assistance!