Upload folder using huggingface_hub

README.md

@@ -0,0 +1,78 @@
+# PRISM-TP2M-1.4B
+
+PRISM is a diffusion-based 3D human motion generation model for **text-to-motion (T2M)** and **pose-conditioned generation (TP2M)**. This checkpoint (~1.4B parameters) supports single-segment and multi-segment autoregressive generation.
+
+## Model Details
+
+- **Architecture**: Flow-matching DiT transformer with causal spatio-temporal Motion VAE
+- **Text encoder**: UMT5 (T5-style)
+- **Output**: SMPL/SMPL-X body parameters (22 joints, rotation_6d, 30 fps)
+
+## Usage
+
+### Load from Hugging Face
+
+```python
+from mmotion.pipelines.prism_from_pretrained import load_prism_pipeline_from_pretrained
+
+pipe = load_prism_pipeline_from_pretrained("ZeyuLing/PRISM-TP2M-1.4B")
+```
+
+### Text-to-Motion (single segment)
+
+```python
+smplx_dict = pipe(
+    prompts="A person walks forward and waves.",
+    negative_prompt="",
+    num_frames_per_segment=129,
+    num_joints=23,
+    guidance_scale=5.0,
+)
+# smplx_dict: transl, global_orient, body_pose, etc.
+```
+
+### Sequential Multi-Segment
+
+```python
+smplx_dict = pipe(
+    prompts=["A person waves.", "A person walks.", "A person bows."],
+    num_frames_per_segment=[97, 129, 97],
+    guidance_scale=5.0,
+)
+```
+
+### Pose-Conditioned (TP2M)
+
+Provide a first-frame `.npz` as condition:
+
+```python
+smplx_dict = pipe(
+    prompts="The person stands up and walks.",
+    first_frame_motion_path="/path/to/first_frame.npz",
+    num_frames_per_segment=129,
+    guidance_scale=5.0,
+)
+```
+
+## Requirements
+
+- Python ≥3.9
+- PyTorch (CUDA recommended)
+- transformers, diffusers, einops, mmengine
+- SMPL/SMPL-X body model (for full mesh rendering)
+
+## Citation
+
+If you use PRISM in your research, please cite:
+
+```bibtex
+@inproceedings{prism2026,
+  title={PRISM: Text-to-Motion and Sequential Motion Generation},
+  booktitle={ECCV},
+  year={2026},
+}
+```
+
+## License
+
+See the [versatilemotion](https://github.com/...) repository for license terms.

config.json

@@ -0,0 +1,24 @@
+{
+  "model_type": "prism",
+  "transformer": {
+    "patch_size": [
+      1,
+      1
+    ],
+    "attention_head_dim": 128,
+    "cross_attn_norm": true,
+    "added_kv_proj_dim": null,
+    "eps": 1e-06,
+    "ffn_dim": 8960,
+    "freq_dim": 256,
+    "in_channels": 16,
+    "num_attention_heads": 12,
+    "num_layers": 30,
+    "out_channels": 16,
+    "qk_norm": "rms_norm_across_heads",
+    "rope_max_seq_len": 1024,
+    "text_dim": 4096
+  },
+  "vae_scale_factor_temporal": 4,
+  "max_text_length": 256
+}

text_encoder/config.json

@@ -0,0 +1,3 @@
+{
+  "model_type": "umt5_encoder"
+}

text_encoder/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bb0c519d265962df648adcc424a0bff12d0bd23c070ab7867166fc03006e84f6
+size 26924267472

transformer/config.json

@@ -0,0 +1,20 @@
+{
+  "patch_size": [
+    1,
+    1
+  ],
+  "attention_head_dim": 128,
+  "cross_attn_norm": true,
+  "added_kv_proj_dim": null,
+  "eps": 1e-06,
+  "ffn_dim": 8960,
+  "freq_dim": 256,
+  "in_channels": 16,
+  "num_attention_heads": 12,
+  "num_layers": 30,
+  "out_channels": 16,
+  "qk_norm": "rms_norm_across_heads",
+  "rope_max_seq_len": 1024,
+  "text_dim": 4096,
+  "model_type": "prism_transformer"
+}

transformer/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ded6e0d97e9edc5f591969accbcca5315c82cd1f9aeb7080682ad5582e07771f
+size 5675480768

vae/config.json

@@ -0,0 +1,62 @@
+{
+  "model_type": "autoencoder_kl_prism_2d",
+  "base_dim": 96,
+  "decoder_base_dim": null,
+  "z_dim": 16,
+  "dim_mult": [
+    1,
+    2,
+    4,
+    4
+  ],
+  "num_res_blocks": 2,
+  "attn_scales": [],
+  "temporal_downsample": [
+    false,
+    true,
+    true
+  ],
+  "dropout": 0.0,
+  "is_residual": false,
+  "in_channels": 6,
+  "out_channels": 6,
+  "scale_factor_temporal": 4,
+  "latents_mean": [
+    0.0092,
+    -0.0013,
+    -0.0052,
+    0.0025,
+    -0.0012,
+    0.0024,
+    0.0021,
+    -0.0004,
+    -0.0015,
+    0.0013,
+    -0.0002,
+    0.0014,
+    0.0018,
+    -0.0001,
+    -0.0008,
+    0.0009
+  ],
+  "latents_std": [
+    0.993707,
+    1.020968,
+    0.996201,
+    1.025335,
+    0.997547,
+    1.035847,
+    1.008814,
+    0.999811,
+    0.980396,
+    1.000318,
+    1.033794,
+    0.993485,
+    0.998681,
+    1.038657,
+    1.001396,
+    0.997597
+  ],
+  "use_static": false,
+  "use_rollout_trans": true
+}

vae/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aae4b33f4a939c07026ced40ebddf49fa06d5e1d070babd6d050e10d2b3ca51f
+size 69661320