docs(README): add architecture figure to Model details (#4)

- docs(README): add architecture figure to Model details (cadd264caeeec38c384e7296eef5f442a307b86d)
- assets: express temporal 2x compression (T -> T/2) in the architecture figure (88183663d187907d1ec4c90ed31e3f19ab0028aa)
- assets: temporal-merge cue inline on the video path (2 frames -> 1 token) (68184b4a37a45235e1a686d904db54a127537f57)
- assets: resolve caption collisions and polish figure geometry (d7ffd6b0908451bde7b8d8d2512b0626db179404)
- assets: temporal cue caption T -> T/2 (314a4caa45ac36a6de031eb9b2a1cec246e1246d)

.gitattributes

@@ -34,3 +34,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
 *.gif filter=lfs diff=lfs merge=lfs -text
+assets/architecture.png filter=lfs diff=lfs merge=lfs -text

README.md

@@ -78,6 +78,10 @@ with torch.inference_mode():
 
 ## Model details
 
+<p align="center">
+  <img src="assets/architecture.png" width="820" alt="Motif Vision Encoder architecture: image and video inputs, patch embedding, 40-block transformer stack, and transformer block internals"/>
+</p>
+
 | | |
 |---|---|
 | Backbone | ViT-7B, patch 16, embed 4096, depth 40, heads 32, SwiGLU |