Views
No views yet
| Field | Value |
|---|---|
| Training step | 200,000 |
| Input dim | 668 |
| Latent dim | 512 |
| Codebooks | 6 × 512 |
| Window length | 64 frames |
| Downsample factor | 4 |
model.safetensors — model weights only (for inference)checkpoint.pt — full training checkpoint (optimizer + scheduler + args)config.json — architecture hyperparameterstraining_metadata.json — training step and CLI args1import json
2import torch
3from safetensors.torch import load_file
4from huggingface_hub import hf_hub_download
5from rvqvae import MotionResidualRVQVAE
6
7repo_id = "Signvrse/signvrse-momask-rvqvae"
8config_path = hf_hub_download(repo_id, "config.json")
9weights_path = hf_hub_download(repo_id, "model.safetensors")
10
11with open(config_path) as f:
12 cfg = json.load(f)
13
14model = MotionResidualRVQVAE(
15 input_dim=cfg["input_dim"],
16 latent_dim=cfg["latent_dim"],
17 num_codebooks=cfg["num_codebooks"],
18 codebook_size=cfg["codebook_size"],
19 quant_dropout_prob=cfg["quant_dropout_prob"],
20 ema_decay=cfg["ema_decay"],
21)
22model.load_state_dict(load_file(weights_path))
23model.eval()
24
25# x: [batch, 668, seq_len] float32 motion features
26# out = model(x)rvqvae package, or copy rvqvae/momask_rvqvae.py into your project.