Download scripts/test/resume_deplot_multigpu.sh from Jack04810/agentic-rl-main: direct link, hf CLI and curl.
- Browser
- Download file 948 Bytes
-
https://huggingface.co/Jack04810/agentic-rl-main/resolve/main/scripts/test/resume_deplot_multigpu.sh
- Command line
-
hf download hf://Jack04810/agentic-rl-main/scripts/test/resume_deplot_multigpu.sh
-
curl -L -o resume_deplot_multigpu.sh https://huggingface.co/Jack04810/agentic-rl-main/resolve/main/scripts/test/resume_deplot_multigpu.sh
948 Bytes
| set -euo pipefail | |
| ROOT_DIR="${ROOT_DIR:-$(cd "$(dirname "${BASH_SOURCE[0]}")/../.." && pwd)}" | |
| cd "${ROOT_DIR}" | |
| DEPLOT_MODEL="${DYME_DEPLOT_MODEL:-models/deplot}" | |
| LOG_PATH=/tmp/train_medium_vf_full_deplot.nohup.log | |
| { | |
| echo "[start] $(date -Is) resume multi-GPU DePlot" | |
| echo "[config] devices=cuda:3,cuda:7 batch_size=4 worker_chunk_size=32 cache=/tmp/train_medium_vf_full_deplot_cache.json" | |
| } > "${LOG_PATH}" | |
| PYTHONUNBUFFERED=1 python \ | |
| scripts/build_visual_facts_chartqa_deplot.py \ | |
| --input data/chartqa/train_medium_vf_full.json \ | |
| --output /tmp/train_medium_vf_full_deplot.json \ | |
| --cache /tmp/train_medium_vf_full_deplot_cache.json \ | |
| --batch-size 4 \ | |
| --worker-chunk-size 32 \ | |
| --max-new-tokens 384 \ | |
| --model-id "${DEPLOT_MODEL}" \ | |
| --devices cuda:3,cuda:7 \ | |
| --dtype float32 \ | |
| --no-progress >> "${LOG_PATH}" 2>&1 | |
| status=$? | |
| echo "[exit] $(date -Is) status=${status}" >> "${LOG_PATH}" | |
| exit "${status}" | |