#!/bin/bash
# CoT Generation - Nebula Submission
# Usage: sh cot_gen/run_nebula.sh

CLUSTER_CONFIG='./cluster.json'
cat <<EOT > ${CLUSTER_CONFIG}
{
    "worker": {
        "gpu": 100,
        "cpu": 1200,
        "memory": 50000
    }
}
EOT

WORKER_COUNT=8

args="--resume"
nebulactl run mdl --engine=xdl \
                  --refresh \
                  --ignore=data/*,.git/* \
                  --queue=pailitao_algo_high_2 \
                  --worker_count=$WORKER_COUNT \
                  --odps_project=pailitao_algo_high_2 \
                  --entry="./cot_gen/generate_cot.py" \
                  --oss_access_id=LTAI5tMZb5HHH6HHgAExAyiL \
                  --oss_access_key=6MVP8JnTE01t4TLUHhPoFV0swZNTrb \
                  --oss_bucket=pailitao-ai \
                  --oss_endpoint=oss-cn-zhangjiakou-internal.aliyuncs.com \
                  --algo_name=pytorch2100 \
                  --user_params="$args" \
                  --job_name=shaixuan-cot-gen-qwen36-27b \
                  --file.cluster_file=$CLUSTER_CONFIG \
                  --force \
                  --max_failover_times=2 \
                  --requirements_file_name="./requirements.txt"

rm ${CLUSTER_CONFIG}