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

README.md

@@ -11,9 +11,9 @@ tags:
   - custom_code
 ---
 
-# MOTIF Vision Encoder
+# Motif Vision Encoder
 
-MOTIF Vision Encoder is a **unified image + video** self-supervised vision encoder on a ViT
+Motif Vision Encoder is a **unified image + video** self-supervised vision encoder on a ViT
 backbone. A single **3D-convolutional tokenizer** ingests both modalities — an image is a
 1-frame clip (`T=1`), a video is `T>1` — so the same weights produce **dense patch-level
 features** and a **language-aligned global (CLS) representation**.
@@ -116,7 +116,7 @@ respective dataset licenses; downstream users are responsible for compliance wit
 
 ```bibtex
 @misc{motif_vision_encoder,
-  title  = {MOTIF Vision Encoder: a unified image/video self-supervised vision encoder},
+  title  = {Motif Vision Encoder: a unified image/video self-supervised vision encoder},
   author = {Motif Technologies},
   year   = {2026},
   howpublished = {\url{https://huggingface.co/Motif-Technologies/Motif-Vision-Encoder-7B}}

modeling_motif_vision_encoder.py

@@ -1,7 +1,7 @@
 # Copyright (c) Motif Technologies.
-# Self-contained inference model for MOTIF Vision Encoder (image + video).
+# Self-contained inference model for Motif Vision Encoder (image + video).
 # Auto-assembled from the training repo's inference path; NO training code.
-"""MOTIF Vision Encoder — unified image/video ViT backbone (inference-only).
+"""Motif Vision Encoder — unified image/video ViT backbone (inference-only).
 
 Usage:
     from transformers import AutoModel
@@ -194,7 +194,7 @@ class PatchEmbed(nn.Module):
 class RopePositionEmbedding3D(nn.Module):
     """Full 3D axial RoPE with independent T/H/W frequency bands.
 
-    Unlike the original MOTIF implementation which simply repeats 2D spatial angles
+    Unlike the original Motif implementation which simply repeats 2D spatial angles
     across temporal frames (making temporal positions indistinguishable), this
     implementation partitions the head dimension into three axis groups:
 
@@ -354,7 +354,7 @@ class RopePositionEmbedding3D(nn.Module):
         token_angles_w = angles_w[w_idx]  # [T*H*W, D_W//2]
         angles_half = torch.cat([token_angles_t, token_angles_h, token_angles_w], dim=-1)  # [T*H*W, D_head//2]
 
-        # tile(2) on full concat — matches MOTIF 2D RoPE pattern
+        # tile(2) on full concat — matches Motif 2D RoPE pattern
         # This ensures rotate_half pairs (dim i ↔ dim i+D//2) have identical angles,
         # making the rotation orthogonal (preserves dot products in attention).
         angles = angles_half.tile(2)  # [T*H*W, D_head]
@@ -1347,7 +1347,7 @@ class MotifVisionTransformer(nn.Module):
 # HuggingFace transformers wrapper (inference-only)
 # ============================================================================
 class MotifVisionConfig(PretrainedConfig):
-    """Config for the MOTIF Vision Encoder backbone (image + video)."""
+    """Config for the Motif Vision Encoder backbone (image + video)."""
 
     model_type = "motif_vision"
 
@@ -1407,7 +1407,7 @@ class MotifVisionConfig(PretrainedConfig):
 
 
 class MotifVisionModel(PreTrainedModel):
-    """MOTIF Vision Encoder for HF `AutoModel` (inference). Returns dense + CLS features."""
+    """Motif Vision Encoder for HF `AutoModel` (inference). Returns dense + CLS features."""
 
     config_class = MotifVisionConfig
     base_model_prefix = "motif"