Instructions to use Duke-CEI-SVD/traj-mc with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use Duke-CEI-SVD/traj-mc with Transformers:
# pip install -U transformers accelerate # Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("Duke-CEI-SVD/traj-mc", device_map="auto") - Notebooks
- Google Colab
- Kaggle
Download code/eval/_launch.sh from Duke-CEI-SVD/traj-mc: direct link, hf CLI and curl.
- Browser
- Download file 906 Bytes
-
https://huggingface.co/Duke-CEI-SVD/traj-mc/resolve/main/code/eval/_launch.sh
- Command line
-
hf download hf://Duke-CEI-SVD/traj-mc/code/eval/_launch.sh
-
curl -L -o _launch.sh https://huggingface.co/Duke-CEI-SVD/traj-mc/resolve/main/code/eval/_launch.sh
906 Bytes
| # Clean single-run launcher for ssh/nohup use. | |
| # bash eval/_launch.sh <GPU> <task> <arm> [weights_or_-] [limit_or_-] [nproc] | |
| set -euo pipefail | |
| GPU=$1; TASK=$2; ARM=$3; WEIGHTS=${4:-'-'}; LIMIT=${5:-'-'}; NPROC=${6:-1} | |
| source /home/tl356/miniconda3/etc/profile.d/conda.sh | |
| conda activate llada | |
| cd /home/tl356/LLaDA/trajmc_main | |
| export CUDA_VISIBLE_DEVICES=$GPU | |
| export HF_DATASETS_OFFLINE=1 HF_HUB_OFFLINE=1 HF_ALLOW_CODE_EVAL=1 | |
| export HF_DATASETS_TRUST_REMOTE_CODE=true TOKENIZERS_PARALLELISM=false | |
| export PYTORCH_ALLOC_CONF=expandable_segments:True | |
| ARGS=(--task "$TASK" --arm "$ARM" --num_processes "$NPROC") | |
| [ "$WEIGHTS" != "-" ] && ARGS+=(--weights "$WEIGHTS") | |
| [ "$LIMIT" != "-" ] && ARGS+=(--limit "$LIMIT") | |
| echo "launch host=$(hostname) gpu=$GPU task=$TASK arm=$ARM weights=$WEIGHTS limit=$LIMIT start=$(date)" | |
| python eval/run_lmeval.py "${ARGS[@]}" | |
| echo "done task=$TASK arm=$ARM $(date)" | |