mlx-community/sapiens2-pointmap-1b-bf16

bf16 MLX conversion of facebook/sapiens2-pointmap-1b (Meta's Sapiens2, ICLR 2026): per-pixel XYZ pointmaps + scale. Converted with mlx-vlm 0.7.0; the original float32 checkpoint is 2x this size.

What is in model.safetensors (3.26 GB):

• every parameter in bfloat16 (the reference runs inference in bf16 mixed precision);
• the q/k/v projections merged into one wqkv tensor per block, as the mlx-vlm Sapiens2 model expects.

Refer to the original model card for the model description, intended use and license.

Use with mlx-vlm

pip install -U mlx-vlm

from mlx_vlm import load
from mlx_vlm.models.sapiens2.generate import Sapiens2Predictor, read_image

model, _ = load("mlx-community/sapiens2-pointmap-1b-bf16")
predictor = Sapiens2Predictor(model)
output = predictor.infer(read_image("image.jpg"))
pointmaps, scale = output["pointmaps"], output["scales"]  # (H, W, 3), (1,)

Outputs are numpy arrays at the input resolution (dense tasks) or in source-image pixel coordinates (pose). See the mlx-vlm Sapiens2 README for preprocessing details and the per-task output keys.