Upload README.md with huggingface_hub

README.md

@@ -17,7 +17,7 @@ tags:
 
 openpi `pi05_bi` checkpoint for a bimanual dish-and-sponge task with tactile
 inputs. Step 8000 (1.24 epochs), held-out validation loss
-**0.0535** — the lowest of the run through step 10000.
+**0.0535** (best through step 12000 is step 12000 at 0.0534).
 
 ## Task
 
@@ -87,24 +87,32 @@ each time so successive points are comparable.
 | 6000 | 0.0448 | 0.0565 | 0.0566 | +0.0001 |
 | **8000** | 0.0430 | 0.0513 | **0.0535** | +0.0022 |
 | 10000 | 0.0420 | 0.0514 | 0.0552 | +0.0038 |
+| 12000 | 0.0421 | 0.0493 | 0.0534 | +0.0041 |
 
-`val_unseen` bottomed at step 8000 (0.0535) and rose to 0.0552 at step 10000.
-Notably `val_seen` flattened at the same time (0.0513 -> 0.0514) while the training
-loss kept falling (0.0430 -> 0.0420) -- both validation curves stalling together
-suggests the model has extracted what it can from this data, not merely that
-generalization has saturated.
+`val_unseen` fell steeply through step 8000 (0.0535), read 0.0552 at step 10000,
+then returned to 0.0534 at step 12000 -- the step-10000 reading was noise, not a
+turn. Improvement has nonetheless nearly stopped: the 4,000 steps from 8000 to
+12000 bought 0.0001. Over that same span `val_seen` improved 4% (0.0513 ->
+0.0493) and the gap widened from 0.0022 to 0.0041.
+
+The gap is still small in absolute terms. The sibling `two_tubes_0102` run had a
+gap of 0.0135 at its step 12000 -- more than 3x larger -- and its `val_unseen` had
+been flat for 6,000 steps by then. task2 is overfitting far less at the same step
+count, which is why training continued past this checkpoint.
 
 The early-step `gap` is slightly **negative** (-0.0009 at step 2000). That is not
 evidence of good generalization: each validation pass covers only ~2,560 frames,
 which at ~966 frames per episode is about 2-3 episodes per split, so the sign of
-the gap is dominated by which particular episodes landed in each split. Only the
-`val_unseen` absolute trend is reliable, since the same episodes and the same rng
-are used at every evaluation.
-
-`val_seen` sits below the training loss from step 8000 onward. This is expected:
-the training loss is measured on augmented images (random crop to 95%, +-5 deg
-rotation, colour jitter, applied to all six streams including the four tactile
-ones) while validation runs with `train=False`, which skips augmentation.
+the gap early on is dominated by which particular episodes landed in each split.
+Only the `val_unseen` absolute trend is reliable, since the same episodes and the
+same rng are used at every evaluation.
+
+Note that the training loss is measured on **augmented** images (random crop to
+95%, +-5 deg rotation, colour jitter, applied to all six streams including the
+four tactile ones) while validation runs with `train=False`, which skips
+augmentation. The two numbers are therefore not directly comparable; on this run
+`val_seen` still sits above the training loss at every step, unlike the
+`two_tubes_0102` run where it dropped below after ~2.5 epochs of memorization.
 
 ## Contents