Normalize Motius Pipeline.from_pretrained metadata

README.md

@@ -1,5 +1,5 @@
 ---
-library_name: hftrainer
+library_name: motius
 pipeline_tag: other
 tags:
 - motion-generation
@@ -13,11 +13,11 @@ license: other
 
 # MotionStreamer
 
-Streaming/autoregressive text-to-motion baseline integrated into the hftrainer
+Streaming/autoregressive text-to-motion baseline integrated into the motius
 Model Zoo. Our reproduction is **fully self-contained and independent of
 `ref_repo`**: the causal TAE, the LLaMA autoregressive transformer, the
 per-token diffusion head and the OpenAI-style Gaussian-diffusion sampler are all
-vendored into `hftrainer.models.motion.motionstreamer._ms`. The `save_pretrained` /
+vendored into `motius.models.motion.motionstreamer._ms`. The `save_pretrained` /
 `from_pretrained` round-trip is **bit-identical** (`max-abs-diff = 0.0` for both
 the TAE and the AR weights).
 
@@ -25,7 +25,7 @@ the TAE and the AR weights).
 |---|---|
 | **Task** | Text-to-Motion (T2M) |
 | **Bundle / Pipeline** | `MotionStreamerBundle` / `MotionStreamerPipeline` |
-| **Processed HF artifact** | [`ZeyuLing/hftrainer-motionstreamer-humanml272`](https://huggingface.co/ZeyuLing/hftrainer-motionstreamer-humanml272) |
+| **Processed HF artifact** | [`ZeyuLing/Motius-MotionStreamer-HumanML272`](https://huggingface.co/ZeyuLing/Motius-MotionStreamer-HumanML272) |
 | **Motion representation** | **MotionStreamer-272** (272-dim, 30 fps) |
 | **Text encoder** | SentenceT5-XXL (`sentence-transformers/sentence-t5-xxl`, frozen) |
 | **Paper** | *MotionStreamer: Streaming Motion Generation via Diffusion-based Autoregressive Model*, 2025 — [arXiv:2503.15451](https://arxiv.org/abs/2503.15451) |
@@ -35,20 +35,20 @@ the TAE and the AR weights).
 
 ## Weights
 
-Current hftrainer artifact (diffusers-style `from_pretrained`):
+Current motius artifact (diffusers-style `from_pretrained`):
 
 | Artifact | Location | Contents | Status |
 |---|---|---|---|
-| MotionStreamer HumanML3D-272 | [`ZeyuLing/hftrainer-motionstreamer-humanml272`](https://huggingface.co/ZeyuLing/hftrainer-motionstreamer-humanml272) | `tae.safetensors` + `ar.safetensors` + `ms_config.json` + `Mean.npy` / `Std.npy` | public Hub artifact; complete SentenceT5 packaging pending |
+| MotionStreamer HumanML3D-272 | [`ZeyuLing/Motius-MotionStreamer-HumanML272`](https://huggingface.co/ZeyuLing/Motius-MotionStreamer-HumanML272) | `tae.safetensors` + `ar.safetensors` + `ms_config.json` + `Mean.npy` / `Std.npy` | public Hub artifact; complete SentenceT5 packaging pending |
 | local mirror | `checkpoints/motionstreamer/t2m_humanml272` | same layout | optional local cache |
 
 **Use directly from the Hub:**
 
 ```python
-from hftrainer.pipelines.motionstreamer import MotionStreamerPipeline
+from motius.pipelines.motionstreamer import MotionStreamerPipeline
 
 pipe = MotionStreamerPipeline.from_pretrained(
-    "ZeyuLing/hftrainer-motionstreamer-humanml272",
+    "ZeyuLing/Motius-MotionStreamer-HumanML272",
     device="cuda",
 )
 motions = pipe.infer_t2m(["a person walks forward then turns around"], [120])  # list of (T, 272)
@@ -62,7 +62,7 @@ resolved by name rather than stored inside the repo.
 **Or download to disk first:**
 
 ```bash
-huggingface-cli download ZeyuLing/hftrainer-motionstreamer-humanml272 \
+huggingface-cli download ZeyuLing/Motius-MotionStreamer-HumanML272 \
     --local-dir checkpoints/motionstreamer/t2m_humanml272
 ```
 
@@ -79,7 +79,7 @@ text -> SentenceT5-XXL -> LLaMA AR (CFG, per-token diffusion sampling)
      -> latent tokens (dim 16) -> causal TAE decoder (×4 upsample) -> 272-dim motion
 ```
 
-Convert to/from HumanML3D-263 with `hftrainer.motion.representation.convert`
+Convert to/from HumanML3D-263 with `motius.motion.representation.convert`
 (`hml263_to_motion272`, etc.).
 
 ---
@@ -100,14 +100,14 @@ python3 scripts/eval/eval_ms_h3d272.py --pred_dir outputs/evaluation/ms_h3d272/m
 
 ### MotionStreamer-272 evaluator (native space)
 
-The hftrainer `MotionStreamer272Evaluator` is the same TMR-style evaluator used
+The motius `MotionStreamer272Evaluator` is the same TMR-style evaluator used
 in the paper (matching feature scale: MM-Dist ≈ 15, Diversity ≈ 27). Paper
 numbers below are from the ICCV 2025 HumanML3D test-set table.
 
-> _Full-set generation (7412 pairs, 8 GPUs) is in progress; the `hftrainer`
+> _Full-set generation (7412 pairs, 8 GPUs) is in progress; the `motius`
 > column is filled in once scoring completes._
 
-| Metric | hftrainer | MotionStreamer paper (ICCV'25) |
+| Metric | motius | MotionStreamer paper (ICCV'25) |
 |---|---|---|
 | FID ↓ | _pending_ | 11.790 |
 | R-Precision Top-1 / 2 / 3 ↑ | _pending_ | 0.631 / 0.802 / 0.859 |
@@ -122,10 +122,18 @@ confirming the evaluator; the model row follows once generation finishes.
 
 ## Implementation notes
 
-- **Vendored, ref_repo-independent**: `hftrainer/models/motionstreamer/_ms/` holds
+- **Vendored, ref_repo-independent**: `motius/models/motionstreamer/_ms/` holds
   `tae.py` / `causal_cnn.py` / `resnet.py` (causal TAE), `llama_model.py` (LLaMA
   AR), `diffloss.py` + `diffusion/` (per-token diffusion head). Only relative
   imports were changed from the upstream source.
 - **Text encoder reloaded by name**: SentenceT5-XXL is frozen and not duplicated
   into the artifact (like CLIP for MDM).
 - **Guidance**: classifier-free, default scale `4.0`, token unit length `4`.
+
+## Direct Loading
+
+```python
+from motius import Pipeline
+
+pipeline = Pipeline.from_pretrained("ZeyuLing/Motius-MotionStreamer-HumanML272")
+```

model_index.json

@@ -0,0 +1,28 @@
+{
+  "_class_name": "MotionStreamerPipeline",
+  "_library_name": "motius",
+  "format_version": 1,
+  "pipeline_class": "motius.pipelines.motionstreamer.MotionStreamerPipeline",
+  "bundle_class": "motius.models.motionstreamer.MotionStreamerBundle",
+  "tasks": [
+    "text_to_motion",
+    "temporal_motion_completion",
+    "sequential_text_to_motion"
+  ],
+  "required_files": [
+    "Mean.npy",
+    "Std.npy",
+    "ar.safetensors",
+    "ms_config.json",
+    "tae.safetensors"
+  ],
+  "api": {
+    "loader": "motius.Pipeline.from_pretrained",
+    "task_methods": [
+      "infer_text_to_motion",
+      "infer_temporal_motion_completion",
+      "infer_sequential_text_to_motion"
+    ]
+  },
+  "artifact_format": "motius-motionstreamer-v1"
+}