Drop the per-model READMEs' mentions of the removed round 11 folders and lab paths

260920_fast_efficientvit_b1_352_hexaview_locator/README.md

@@ -1,6 +1,6 @@
 # `fast_efficientvit_b1_352_hexaview_locator` -- the hexaview locator
 
-A deeperfly 2 locator (family `fast`, EfficientViT-B1): it reads a whole hexaview source frame, 2048x1408 scaled by exactly 1/4 to 512x352, and returns where the fly is in each of the three views that frame carries, one channel per view, so a per-frame run can center each view's window on the fly. A run names it as `[pose2d] locator`, beside a hexaview detector such as `260927_mvt_dinov3b_512_hexaview_fly38`.
+A deeperfly 2 locator (family `fast`, EfficientViT-B1): it reads a whole hexaview source frame, 2048x1408 scaled by exactly 1/4 to 512x352, and returns where the fly is in each of the three views that frame carries, one channel per view, so a per-frame run can center each view's window on the fly. A run names it as `[pose2d] locator`, beside the single-fly hexaview detector `261001_mvt_dinov3b_512_hexaview_fly38`. The two-fly detector has a locator of its own, `261007_fast_efficientvit_b1_352_hexaview_pair_locator`.
 
 Exported 2026-09-28 from `hexaview/runs/locf6_all`, trained 2026-09-20 on every labeled moment of the hexaview project then but two mirrored ones.
 
@@ -15,11 +15,22 @@ Exported 2026-09-28 from `hexaview/runs/locf6_all`, trained 2026-09-20 on every
 
 ## Using it
 
+Download it into `models/` at the root of a deeperfly clone, with its detector:
+
+```bash
+uv run hf download tkclam/deeperfly-models --local-dir models \
+  --include "261001_mvt_dinov3b_512_hexaview_fly38/*" \
+  --include "260920_fast_efficientvit_b1_352_hexaview_locator/*"
+```
+
+and name both in the run config, with paths relative to the config file:
+
 ```toml
 [pose2d]
-weights = ["/mnt/upramdya/data/TL/deeperfly-models/260927_mvt_dinov3b_512_hexaview_fly38/mvt_dinov3b_512_hexaview_fly38.pth"]
+weights = ["../models/261001_mvt_dinov3b_512_hexaview_fly38/mvt_dinov3b_512_hexaview_fly38.pth"]
+locator = "../models/260920_fast_efficientvit_b1_352_hexaview_locator/fast_efficientvit_b1_352_hexaview_locator.pth"
 window = "per_frame"
-locator = "/mnt/upramdya/data/TL/deeperfly-models/260920_fast_efficientvit_b1_352_hexaview_locator/fast_efficientvit_b1_352_hexaview_locator.pth"
+tightness = 1.2
 ```
 
-Copying this directory to the machine that runs it and naming the local path is the better habit. Verify a copy with `sha256sum -c SHA256SUMS`.
+Verify a download with `sha256sum -c SHA256SUMS` in this folder.

261001_mvt_dinov3b_512_hexaview_fly38/README.md

@@ -1,6 +1,6 @@
 # `mvt_dinov3b_512_hexaview_fly38` -- the hexaview detector, round 13
 
-A deeperfly 2 detector (`deeperfly-detector-1`, family `mvt`): a multiview transformer on a DINOv3 ViT-B/16 that predicts all 38 points of the `fly38` skeleton in the six views of the hexaview rig, for one fly -- two cameras, each seeing the arena directly and through two prisms. It windows each view on the fly frame by frame, and a video holds three views of one fly, so it runs with its locator, `260920_fast_efficientvit_b1_352_hexaview_locator`, beside it. It supersedes `260927_mvt_dinov3b_512_hexaview_fly38` (round 11): the same architecture and input, with the same recipe retrained on the corrected labels of 2026-10-01.
+A deeperfly 2 detector (`deeperfly-detector-1`, family `mvt`): a multiview transformer on a DINOv3 ViT-B/16 that predicts all 38 points of the `fly38` skeleton in the six views of the hexaview rig, for one fly -- two cameras, each seeing the arena directly and through two prisms. It windows each view on the fly frame by frame, and a video holds three views of one fly, so it runs with its locator, `260920_fast_efficientvit_b1_352_hexaview_locator`, beside it. It supersedes round 11's detector, which is no longer distributed: the same architecture and input, with the same recipe retrained on the corrected labels of 2026-10-01.
 
 Exported 2026-10-09 from `hexaview/runs/r13ship_all`, trained 2026-10-01: warm-started from the octacam ViT-B detector, lr 1e-4, 5,000 steps, with a transductive distillation stream on unlabeled frames, on 541 labeled frames in 17 recordings.
 

261008_mvt_dinov3b_384_octacam_fly38/README.md

@@ -1,6 +1,6 @@
 # `mvt_dinov3b_384_octacam_fly38` -- the octacam detector, round 14
 
-A deeperfly 2 detector (`deeperfly-detector-1`, family `mvt`): a multiview transformer on a DINOv3 ViT-B/16 that predicts all 38 points of the `fly38` skeleton in every view of the octacam rig, computing the views of a frame together. It supersedes `260927_mvt_dinov3b_384_octacam_fly38` (round 11): the same architecture and input, trained on more labels with round 13's recipe.
+A deeperfly 2 detector (`deeperfly-detector-1`, family `mvt`): a multiview transformer on a DINOv3 ViT-B/16 that predicts all 38 points of the `fly38` skeleton in every view of the octacam rig, computing the views of a frame together. It supersedes round 11's detector, which is no longer distributed: the same architecture and input, trained on more labels with round 13's recipe.
 
 Exported 2026-10-09 from `octacam/runs/mvt_dinov3/round14/r14_ship_g10_s0`, trained 2026-10-08: warm-started from a pseudo-labeled ViT-B, lr 1e-4, 5,000 steps, and a distillation stream on unlabeled frames with round 13's detector as the teacher, on the labels of 2026-10-08.