| license: mit | |
| tags: | |
| - text-to-video | |
| - pose-estimation | |
| - human-motion | |
| - lora | |
| # RecCAR (EchoMotion LoRA Checkpoint) | |
| This repository contains the trained **RecCAR** LoRA weights (`RecCAR.pth`) for 3D human motion generation and text-to-video pose alignment. | |
| ## Model Details | |
| - **File Name:** `RecCAR.pth` | |
| - **Architecture:** EchoMotion Transformer + LoRA | |
| - **Training Epochs:** 12 | |
| ## Usage | |
| ```python | |
| import torch | |
| from huggingface_hub import hf_hub_download | |
| # Download the model checkpoint | |
| ckpt_path = hf_hub_download(repo_id="ohad204/RecCAR", filename="RecCAR.pth") | |
| # Load weights into model | |
| state_dict = torch.load(ckpt_path, map_location="cuda") | |
| ``` | |