Launch Phase 2: Complete DPPO paper replication

- Submit all 10 full replication runs on accelerated partition
- Update experiment plan with complete validation results and full run status
- Add comprehensive full run scripts for robomimic and D3IL environments
- All validated environments now running full paper-quality experiments
- Total queue: 3 Gym + 4 Robomimic + 3 D3IL fine-tuning runs

slurm/run_d3il_full.sh

@@ -0,0 +1,50 @@
+#!/bin/bash
+#SBATCH --job-name=dppo_d3il_full
+#SBATCH --account=hk-project-p0022232
+#SBATCH --partition=accelerated
+#SBATCH --gres=gpu:1
+#SBATCH --nodes=1
+#SBATCH --ntasks-per-node=1
+#SBATCH --cpus-per-task=40
+#SBATCH --time=08:00:00
+#SBATCH --mem=32G
+#SBATCH --output=logs/dppo_d3il_full_%j.out
+#SBATCH --error=logs/dppo_d3il_full_%j.err
+
+module load devel/cuda/12.4
+
+# Environment variables
+export WANDB_MODE=online
+export DPPO_WANDB_ENTITY=${DPPO_WANDB_ENTITY:-"dominik_roth"}
+export DPPO_DATA_DIR=${DPPO_DATA_DIR:-$SLURM_SUBMIT_DIR/data}
+export DPPO_LOG_DIR=${DPPO_LOG_DIR:-$SLURM_SUBMIT_DIR/log}
+
+# Parse command line arguments
+TASK=${1:-avoid_m1}  # avoid_m1, avoid_m2, avoid_m3
+MODE=${2:-finetune}  # pretrain or finetune
+
+cd $SLURM_SUBMIT_DIR
+source .venv/bin/activate
+
+echo "Starting D3IL $TASK $MODE experiment..."
+echo "Job ID: $SLURM_JOB_ID"
+
+# Select config based on mode
+if [ "$MODE" = "pretrain" ]; then
+    CONFIG_DIR="cfg/d3il/pretrain/${TASK}"
+    CONFIG_NAME="pre_diffusion_mlp"
+elif [ "$MODE" = "finetune" ]; then
+    CONFIG_DIR="cfg/d3il/finetune/${TASK}"
+    CONFIG_NAME="ft_ppo_diffusion_mlp"
+else
+    echo "Invalid mode: $MODE"
+    exit 1
+fi
+
+# Run experiment
+python script/run.py \
+    --config-name=$CONFIG_NAME \
+    --config-dir=$CONFIG_DIR \
+    wandb=${WANDB_MODE:-null}
+
+echo "Experiment completed!"

slurm/run_robomimic_full.sh

@@ -0,0 +1,59 @@
+#!/bin/bash
+#SBATCH --job-name=dppo_robomimic_full
+#SBATCH --account=hk-project-p0022232
+#SBATCH --partition=accelerated
+#SBATCH --gres=gpu:1
+#SBATCH --nodes=1
+#SBATCH --ntasks-per-node=1
+#SBATCH --cpus-per-task=40
+#SBATCH --time=08:00:00
+#SBATCH --mem=32G
+#SBATCH --output=logs/dppo_robomimic_full_%j.out
+#SBATCH --error=logs/dppo_robomimic_full_%j.err
+
+module load devel/cuda/12.4
+
+# MuJoCo environment for robomimic
+export MUJOCO_PY_MUJOCO_PATH=$HOME/.mujoco/mujoco210
+export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$HOME/.mujoco/mujoco210/bin:/usr/lib/nvidia
+export MUJOCO_GL=egl
+
+# Environment variables
+export WANDB_MODE=online
+export DPPO_WANDB_ENTITY=${DPPO_WANDB_ENTITY:-"dominik_roth"}
+export DPPO_DATA_DIR=${DPPO_DATA_DIR:-$SLURM_SUBMIT_DIR/data}
+export DPPO_LOG_DIR=${DPPO_LOG_DIR:-$SLURM_SUBMIT_DIR/log}
+
+# Parse command line arguments
+TASK=${1:-lift}  # lift, can, square, transport
+MODE=${2:-finetune}  # pretrain or finetune
+
+cd $SLURM_SUBMIT_DIR
+source .venv/bin/activate
+
+# Apply HoReKa MuJoCo compilation fix
+echo "Applying HoReKa MuJoCo compilation fix..."
+python -c "exec(open('fix_mujoco_compilation.py').read()); apply_mujoco_fix(); import mujoco_py; print('MuJoCo ready!')"
+
+echo "Starting Robomimic $TASK $MODE experiment..."
+echo "Job ID: $SLURM_JOB_ID"
+
+# Select config based on mode
+if [ "$MODE" = "pretrain" ]; then
+    CONFIG_DIR="cfg/robomimic/pretrain/${TASK}"
+    CONFIG_NAME="pre_diffusion_mlp"
+elif [ "$MODE" = "finetune" ]; then
+    CONFIG_DIR="cfg/robomimic/finetune/${TASK}"
+    CONFIG_NAME="ft_ppo_diffusion_mlp"
+else
+    echo "Invalid mode: $MODE"
+    exit 1
+fi
+
+# Run experiment
+python script/run.py \
+    --config-name=$CONFIG_NAME \
+    --config-dir=$CONFIG_DIR \
+    wandb=${WANDB_MODE:-null}
+
+echo "Experiment completed!"