scripts/train_with_caption_balanced.sh

#!/bin/bash
# Atlas online training: detection + planning + caption.
# Default: visual_token_mode=online, live frozen StreamPETR + TopoMLP.
# Data: 28k det + 24k plan + 169k caption = ~220k samples/epoch
# WARNING: caption data is 6x larger than other tasks — consider downsampling.
#
# Usage:
#   bash scripts/train_with_caption_balanced.sh
#   RESUME_CKPT=work_dirs/atlas_with_caption/epoch-4/checkpoint.pt bash scripts/train_with_caption_balanced.sh
#   NUM_GPUS=8 bash scripts/train_with_caption_balanced.sh
set -e

PROJECT_ROOT="$(cd "$(dirname "$0")/.." && pwd)"
cd "$PROJECT_ROOT"
ENV_ROOT="${PROJECT_ROOT}/envs/streampetr"
DEEPSPEED_BIN="${ENV_ROOT}/bin/deepspeed"

if [ ! -x "$DEEPSPEED_BIN" ]; then
    echo "ERROR: deepspeed not found in ${DEEPSPEED_BIN}" >&2
    exit 1
fi

export PATH="${ENV_ROOT}/bin:${PATH}"
export LD_LIBRARY_PATH="${ENV_ROOT}/lib:${LD_LIBRARY_PATH}"

NUM_GPUS=${NUM_GPUS:-4}
PLANNING_TABLE3_MODE=${PLANNING_TABLE3_MODE:-atlas_high_level_ego}
TASK_LOSS_WEIGHTS=${TASK_LOSS_WEIGHTS:-detection=0.35,planning=1.0,caption=0.05}
OUTPUT_DIR=${OUTPUT_DIR:-work_dirs/atlas_with_caption_online}

EXTRA_ARGS=""
if [ -n "$RESUME_CKPT" ]; then
    EXTRA_ARGS="--resume $RESUME_CKPT"
fi

"$DEEPSPEED_BIN" --num_gpus "$NUM_GPUS" train_atlas.py \
    --llm_model pretrained/vicuna-7b-v1.5 \
    --data_json data/atlas_nuscenes_train.json,data/atlas_planning_train_uniad_command.json,data/atlas_caption_train_canonical.json \
    --data_root /home/guoyuanbo/autodl-tmp/data/nuscenes \
    --visual_token_mode online \
    --task_balance_mode scene_unit_111 \
    --task_loss_weights "$TASK_LOSS_WEIGHTS" \
    --planning_table3_mode "$PLANNING_TABLE3_MODE" \
    --streampetr_config configs/streampetr_atlas_aligned.py \
    --streampetr_ckpt pretrained/streampetr/streampetr_eva02_ep24.pth \
    --topomlp_config configs/topomlp_atlas_aligned.py \
    --topomlp_ckpt work_dirs/topomlp_atlas_aligned/epoch_24.pth \
    --deepspeed configs/ds_zero2.json \
    --output_dir "$OUTPUT_DIR" \
    --epochs 8 \
    --lr 2e-5 \
    --weight_decay 1e-4 \
    --batch_size 1 \
    --gradient_accumulation_steps 2 \
    --warmup_ratio 0.03 \
    --max_grad_norm 1.0 \
    --log_steps 100 \
    --save_epochs 1 \
    --keep_last_n_ckpts 3 \
    --seed 42 \
    --num_workers 4 \
    $EXTRA_ARGS