docs(README): fix leftover img alt text to match tracking caption

docs(README): note Motif > V-JEPA 2.1 in point-tracking demo

docs(README): correct demo caption to point-tracking via feature similarity

docs(README): condense results prose to DAVIS-SOTA + competitive summary

docs(README): underline second-best value per column

docs(README): V-JEPA 2.1 KITTI = 2.5

docs(README): Motif K400 = 87.4 (eval complete)

weights: publish ckpt-199 (phase3 366999base_w_gram_6n_syn, iter 199)

docs(README): blank OpenCLIP zero-shot cell; VideoMAEv2 data as 1.35M

docs(README): swap DINOv2 for PEcore/OpenCLIP/VideoMAEv2, drop caption block ref

docs(README): add attention demo GIF + data-efficiency headline

docs(README): unify Training data column to B units

docs(README): Training-data column + corrected ImageNet numbers

docs(README): DAVIS S/M/L table, Publisher column, updated metrics

docs: add training-data scale (~0.47B samples, 96% image / 4% video) + data-efficiency vs DINOv3 LVD-1689M

docs: brand name MOTIF -> Motif (title, intro, citation, docstrings)

docs: update repo id to Motif-Vision-Encoder-7B in usage + citation

rename modeling_motif.py -> modeling_motif_vision_encoder.py (update auto_map + README)

refactor: consolidate all imports to module top (dedupe per-section imports)

License->MIT; strip training-only code (init_weights, torch._dynamo config, stochastic depth). Inference forward bit-identical (verified).

remove linear-probe eval ckpt

add linear-probe classifier ckpt (ImageNet-1K, best top1=87.266) for eval-only reproduction

Initial release: MOTIF vision encoder 7B (bf16 safetensors + modeling_motif.py)

initial commit