Upload README.md with huggingface_hub

README.md

@@ -0,0 +1,39 @@
+---
+license: apache-2.0
+library_name: hftrainer
+pipeline_tag: text-to-motion
+tags:
+- motion-generation
+- text-to-motion
+- autoregressive
+- diffusion
+- humanml3d
+- motionstreamer
+---
+
+# MotionStreamer (hftrainer reproduction)
+
+Self-contained, `ref_repo`-independent reproduction of **MotionStreamer** (Xiao
+et al., ICCV 2025), packaged as an [hftrainer](https://github.com/) `ModelBundle`
+artifact. Causal TAE + LLaMA autoregressive transformer + per-token diffusion
+head are vendored in-tree; the `save_pretrained`/`from_pretrained` round-trip is
+**bit-identical**.
+
+- **Task:** Text-to-Motion · **Representation:** MotionStreamer-272 (30 fps)
+- **Text encoder:** SentenceT5-XXL (`sentence-transformers/sentence-t5-xxl`, frozen, reloaded by name)
+- **Paper:** [arXiv:2503.15451](https://arxiv.org/abs/2503.15451) · **Code:** https://github.com/zju3dv/MotionStreamer
+
+## Usage
+
+```python
+from hftrainer.models.motionstreamer import MotionStreamerBundle
+from hftrainer.pipelines.motionstreamer import MotionStreamerPipeline
+
+bundle = MotionStreamerBundle.from_pretrained("ZeyuLing/hftrainer-motionstreamer-humanml272", device="cuda")
+pipe   = MotionStreamerPipeline(bundle, device="cuda")
+motions = pipe.infer_t2m(["a person walks forward then turns around"], [120])  # list of (T, 272)
+```
+
+Files: `tae.safetensors` · `ar.safetensors` (LLaMA AR + diffusion head) ·
+`ms_config.json` · `Mean.npy` / `Std.npy`. The SentenceT5-XXL text encoder is
+reloaded by name and not duplicated here.