Buckets:
| # Visual-OPSD Baseline: Text-only SFT. | |
| # SFT on ThinkMorph text traces, no KL distillation. | |
| # Expected: ~6 hours on 4× H800. | |
| SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" | |
| ROOT_DIR="$(cd "${SCRIPT_DIR}/../.." && pwd)" | |
| cd "${ROOT_DIR}" | |
| export PYTHONPATH="${ROOT_DIR}:${PYTHONPATH:-}" | |
| source .venv/bin/activate | |
| MODEL_PATH=${1:-models/ThinkMorph-7B} | |
| TOTAL_STEPS=${2:-8000} | |
| OUTPUT_NAME=${3:-sft-baseline} | |
| MASTER_ADDR=${ARNOLD_WORKER_0_HOST} | |
| MASTER_PORT=(${ARNOLD_WORKER_0_PORT//,/ }) | |
| NPROC_PER_NODE=${ARNOLD_WORKER_GPU} | |
| NNODES=${ARNOLD_WORKER_NUM} | |
| NODE_RANK=${ARNOLD_ID} | |
| torchrun \ | |
| --nnodes=${NNODES} \ | |
| --node_rank=${NODE_RANK} \ | |
| --nproc_per_node=${NPROC_PER_NODE} \ | |
| --master_addr=${MASTER_ADDR} \ | |
| --master_port=${MASTER_PORT} \ | |
| scripts/visual_opsd/train_visual_opsd_offline.py \ | |
| --mode sft \ | |
| --model_path ${MODEL_PATH} \ | |
| --dataset_config_file ./data/configs/text_reasoning.yaml \ | |
| --finetune_from_hf True \ | |
| --resume_model_only True \ | |
| --finetune_from_ema True \ | |
| --resume_from ${MODEL_PATH} \ | |
| --results_dir results/${OUTPUT_NAME} \ | |
| --checkpoint_dir results/${OUTPUT_NAME}/checkpoints \ | |
| --lr 1e-5 \ | |
| --num_workers 4 \ | |
| --max_num_tokens 10240 \ | |
| --visual_gen False \ | |
| --total_steps ${TOTAL_STEPS} \ | |
| --save_every 1000 \ | |
| --num_shard 4 \ | |
| --wandb_project Visual-OPSD \ | |
| --wandb_name ${OUTPUT_NAME} \ | |
| --wandb_runid 1 \ | |
| --kd_weight 0.0 | |
Xet Storage Details
- Size:
- 1.4 kB
- Xet hash:
- 09fb3c96b65073a2d0b1805dfb897ce96e502afd92fd2f2c75841e2c8ac06332
·
Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.