refactor(recipes): fold memory into action_execution, drop interjection, fuse smolvla2 forward

Recipe changes:
* action_execution now bundles the memory update as a second
  assistant target gated on a new ``new_memory`` binding (fires
  only at subtask-boundary frames). No "Completed subtask: X"
  filler — the model emits the new subtask AND the updated
  memory back-to-back in one prefix.
* user_interjection_response sub-recipe removed (current
  datasets don't have interjection / say() annotations).
* Standalone memory_update sub-recipe removed (folded above).
* Weights rebalanced: action_execution 0.85, ask_vqa_top/wrist
  0.075 each (sums to 1.0).

Runtime ``_msgs_for_memory`` updated to match the new
boundary-frame prompt layout.

Modeling:
* SmolVLA2Policy now fuses the flow + text losses into a SINGLE
  backbone forward via ``_compute_fused_loss`` (one
  vlm_with_expert pass with [prefix, suffix] embeds, then both
  lm_head CE on lang slice + action_out_proj MSE on suffix).
  Mirrors pi052's existing ``_compute_all_losses_fused`` —
  saves one backbone pass per training step.

Examples:
* Removed the two training SLURM scaffolds; they were
  out-of-date with the recipe refactor.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

examples/training/pi052_hirobot.slurm

@@ -1,75 +0,0 @@
-#!/bin/bash
-#SBATCH --job-name=pi052-hirobot
-#SBATCH --partition=hopper-prod
-#SBATCH --qos=high
-#SBATCH --time=48:00:00
-#SBATCH --ntasks=1
-#SBATCH --gpus-per-task=8
-
-# π0.5 v2 training — reproduces the π0.5 paper's hierarchical recipe.
-#
-# Same recipe blend as the SmolVLA2 stack (recipes/pi052_hirobot.yaml),
-# just on the PaliGemma 2B + Gemma-300m action-expert backbone the
-# paper uses. The text head learns subtask prediction via cross-
-# entropy on supervised spans; the action expert learns the flow
-# field. Paper §IV.D mixes the two losses with α=10, which we encode
-# as flow_loss_weight=10 / text_loss_weight=1.
-
-set -euo pipefail
-
-cd "${LEROBOT_ROOT:-$HOME/lerobot}"
-
-export PATH="$HOME/miniconda3/bin:$HOME/.local/bin:$PATH"
-export LD_LIBRARY_PATH="$HOME/miniconda3/lib:${LD_LIBRARY_PATH:-}"
-export NCCL_TIMEOUT="${NCCL_TIMEOUT:-1800}"
-export HF_HUB_DOWNLOAD_TIMEOUT="${HF_HUB_DOWNLOAD_TIMEOUT:-120}"
-export WANDB_INIT_TIMEOUT="${WANDB_INIT_TIMEOUT:-300}"
-
-DATASET="${DATASET:-pepijn223/super_poulain_full_tool3}"
-POLICY_REPO_ID="${POLICY_REPO_ID:-pepijn223/pi052_hirobot_super_poulain}"
-JOB_NAME="${JOB_NAME:-pi052-hirobot-super-poulain}"
-NUM_PROCESSES="${NUM_PROCESSES:-8}"
-BATCH_SIZE="${BATCH_SIZE:-32}"
-STEPS="${STEPS:-15000}"
-RUN_ID="${SLURM_JOB_ID:-$(date +%Y%m%d_%H%M%S)}"
-OUTPUT_DIR="${OUTPUT_DIR:-/fsx/pepijn/outputs/train/pi052_hirobot_${STEPS}_${RUN_ID}}"
-
-echo "Training pi052 on $DATASET"
-echo "  GPUs:         $NUM_PROCESSES"
-echo "  batch:        $BATCH_SIZE / GPU (global=$((NUM_PROCESSES * BATCH_SIZE)))"
-echo "  steps:        $STEPS"
-echo "  output:       $OUTPUT_DIR"
-echo "  loss mix:     flow_loss_weight=10 (paper α), text_loss_weight=1"
-echo "  augmentation: image_transforms ON, prompt dropout {plan:0.30 memory:0.30 subtask:0.20}"
-
-accelerate launch --multi_gpu --num_processes="$NUM_PROCESSES" \
-    -m lerobot.scripts.lerobot_train \
-    --policy.type=pi052 \
-    --policy.recipe_path=recipes/pi052_hirobot.yaml \
-    --dataset.repo_id="$DATASET" \
-    --dataset.revision=main \
-    --dataset.video_backend=pyav \
-    --output_dir="$OUTPUT_DIR" \
-    --job_name="$JOB_NAME" \
-    --policy.repo_id="$POLICY_REPO_ID" \
-    --policy.compile_model=false \
-    --policy.device=cuda \
-    --policy.tokenizer_max_length=512 \
-    --policy.text_loss_weight=1.0 \
-    --policy.flow_loss_weight=10.0 \
-    --policy.unfreeze_lm_head=true \
-    --steps="$STEPS" \
-    --policy.scheduler_decay_steps="$STEPS" \
-    --batch_size="$BATCH_SIZE" \
-    --wandb.enable=true \
-    --wandb.disable_artifact=true \
-    --wandb.project=hirobot \
-    --log_freq=100 \
-    --save_freq="$STEPS" \
-    --num_workers=0 \
-    --dataset.image_transforms.enable=true \
-    --dataset.image_transforms.max_num_transforms=3 \
-    --dataset.image_transforms.random_order=true \
-    --policy.plan_dropout_prob=0.30 \
-    --policy.memory_dropout_prob=0.30 \
-    --policy.subtask_dropout_prob=0.20

examples/training/smolvla2_hirobot.slurm

@@ -1,82 +0,0 @@
-#!/bin/bash
-#SBATCH --job-name=smolvla2-hirobot
-#SBATCH --partition=hopper-prod
-#SBATCH --qos=high
-#SBATCH --time=48:00:00
-#SBATCH --ntasks=1
-#SBATCH --gpus-per-task=8
-
-# SmolVLA2 training on an annotated dataset.
-#
-# The high_level_subtask recipe (recipes/smolvla2_hirobot.yaml) was
-# fixed in PR3 to supervise the LM head with the *current* active
-# subtask span at every frame, not the next-span target which is
-# empty on stable phases. With the old recipe the head learned to
-# emit ``\n`` on every chunk boundary; the new one supervises a
-# real, scene-grounded string at every frame.
-#
-# Two regularisers are still on:
-#
-#   * --dataset.image_transforms.enable=true: torchvision-v2
-#     ColorJitter + SharpnessJitter + RandomAffine per frame; default
-#     envelope (brightness ±20% etc).
-#   * --policy.{plan,memory,subtask}_dropout_prob: randomly drop the
-#     context messages carrying the named recipe binding so the model
-#     handles missing/stale context. Mirrors Pi0.7 §V.E.
-
-set -euo pipefail
-
-cd "${LEROBOT_ROOT:-$HOME/lerobot}"
-
-export PATH="$HOME/miniconda3/bin:$HOME/.local/bin:$PATH"
-export LD_LIBRARY_PATH="$HOME/miniconda3/lib:${LD_LIBRARY_PATH:-}"
-export NCCL_TIMEOUT="${NCCL_TIMEOUT:-1800}"
-export HF_HUB_DOWNLOAD_TIMEOUT="${HF_HUB_DOWNLOAD_TIMEOUT:-120}"
-export WANDB_INIT_TIMEOUT="${WANDB_INIT_TIMEOUT:-300}"
-
-DATASET="${DATASET:-pepijn223/super_poulain_full_tool3}"
-POLICY_REPO_ID="${POLICY_REPO_ID:-pepijn223/smolvla2_hirobot_super_poulain_tool6}"
-JOB_NAME="${JOB_NAME:-smolvla2-hirobot-super-poulain-tool6}"
-NUM_PROCESSES="${NUM_PROCESSES:-8}"
-BATCH_SIZE="${BATCH_SIZE:-32}"
-STEPS="${STEPS:-15000}"
-RUN_ID="${SLURM_JOB_ID:-$(date +%Y%m%d_%H%M%S)}"
-OUTPUT_DIR="${OUTPUT_DIR:-/fsx/pepijn/outputs/train/smolvla2_hirobot_super_poulain_tool3_${STEPS}_${RUN_ID}}"
-
-echo "Training smolvla2 on $DATASET"
-echo "  GPUs:         $NUM_PROCESSES"
-echo "  batch:        $BATCH_SIZE / GPU (global=$((NUM_PROCESSES * BATCH_SIZE)))"
-echo "  steps:        $STEPS"
-echo "  output:       $OUTPUT_DIR"
-echo "  augmentation: image_transforms ON, prompt dropout {plan:0.30 memory:0.30 subtask:0.20}"
-
-accelerate launch --multi_gpu --num_processes="$NUM_PROCESSES" \
-    -m lerobot.scripts.lerobot_train \
-    --policy.type=smolvla2 \
-    --policy.recipe_path=recipes/smolvla2_hirobot.yaml \
-    --dataset.repo_id="$DATASET" \
-    --dataset.revision=main \
-    --dataset.video_backend=pyav \
-    --output_dir="$OUTPUT_DIR" \
-    --job_name="$JOB_NAME" \
-    --policy.repo_id="$POLICY_REPO_ID" \
-    --policy.compile_model=false \
-    --policy.device=cuda \
-    --policy.tokenizer_max_length=512 \
-    --policy.text_loss_weight=1.0 \
-    --policy.flow_loss_weight=10.0 \
-    --steps="$STEPS" \
-    --policy.scheduler_decay_steps="$STEPS" \
-    --batch_size="$BATCH_SIZE" \
-    --wandb.enable=true \
-    --wandb.disable_artifact=true \
-    --wandb.project=hirobot \
-    --log_freq=100 \
-    --save_freq="$STEPS" \
-    --num_workers=0 \
-    --dataset.image_transforms.enable=true \
-    --dataset.image_transforms.max_num_transforms=3 \
-    --dataset.image_transforms.random_order=true \
-    --policy.plan_dropout_prob=0.30 \
-    --policy.memory_dropout_prob=0.30 \
-    --policy.subtask_dropout_prob=0.20