fix(ema): default EMAConfig.enable to False (opt-in)

EMA was on by default, so every training run on the branch (incl. VLA-JEPA
and other non-flow-matching policies) created a full fp32 shadow copy. EMA
only benefits flow-matching/diffusion policies (pi0/pi05/pi052). Make it
opt-in via --ema.enable=true; the pi05/pi052 recipes already pass that flag.

Co-authored-by: Cursor <cursoragent@cursor.com>

src/lerobot/configs/default.py

@@ -98,14 +98,14 @@ class EMAConfig:
     Cost: 1× model params in fp32 shadow (~13 GB for pi052's 3.3B
     params) + one elementwise update per training step (~1% step time).
 
-    On by default — matches openpi (JAX) which ships EMA on for every
+    Off by default (opt-in): EMA is only beneficial for flow-matching /
-    config, and closes the gap with the openpi PyTorch port which
+    diffusion policies (pi0/pi05/pi052), and the fp32 shadow copy is pure
-    explicitly lists EMA as unsupported. Set ``--ema.enable=false`` to
+    overhead for other policies (e.g. VLA-JEPA). Set ``--ema.enable=true``
-    disable for short runs / memory-constrained training where the
+    to turn it on (the pi05/pi052 training recipes do this). openpi (JAX)
-    extra fp32 shadow copy is the bottleneck.
+    ships EMA on for every config; enable it explicitly to match that.
     """
 
-    enable: bool = True
+    enable: bool = False
     # Target EMA decay β in θ_ema ← β·θ_ema + (1-β)·θ_live (passed to
     # ema-pytorch as ``beta``).
     #   0.999  — last ~1000 steps; pi05_libero default in openpi