Instructions to use WitneyWW/acts-sensoronly-ablation with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Cosmos
How to use WitneyWW/acts-sensoronly-ablation with Cosmos:
# No code snippets available yet for this library. # To use this model, check the repository files and the library's documentation. # Want to help? PRs adding snippets are welcome at: # https://github.com/huggingface/huggingface.js
- Notebooks
- Google Colab
- Kaggle
Download scripts/train.sh from WitneyWW/acts-sensoronly-ablation: direct link, hf CLI and curl.
- Browser
- Download file 1.57 kB
-
https://huggingface.co/WitneyWW/acts-sensoronly-ablation/resolve/refs%2Fpr%2F2/scripts/train.sh
- Command line
-
hf download hf://WitneyWW/acts-sensoronly-ablation@refs/pr/2/scripts/train.sh
-
curl -L -o train.sh https://huggingface.co/WitneyWW/acts-sensoronly-ablation/resolve/refs%2Fpr%2F2/scripts/train.sh
1.57 kB
| # Train one of the released models (single node, one DDP rank per GPU). | |
| # bash scripts/train.sh configs/nano_action_causal_rope_v1.yaml [key=value ...] | |
| # NGPU=4 bash scripts/train.sh configs/nano_action_causal_pusht_v1.yaml training.epochs=1 | |
| # Under Slurm (sbatch/srun with --ntasks-per-node == #GPUs) it uses srun instead of torchrun: | |
| # sbatch -N1 --ntasks-per-node=4 --gres=gpu:4 scripts/train.sh configs/...yaml | |
| # Paper runs: 4x H100 80GB, batch_size 2/GPU, 20 epochs (~2.3 h (pushT) - 3.3 h (rope) per epoch). | |
| # Re-launching with the same run_name auto-resumes from $ACTS_RUNS_ROOT/<run_name>/last.pt. | |
| set -euo pipefail | |
| HERE="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)" | |
| source "${ACTS_REPO:-$HERE}/env.sh" | |
| cd "$ACTS_REPO" | |
| CONFIG="${1:?usage: train.sh <config.yaml> [key=value ...]}"; shift || true | |
| if [[ -n "${SLURM_JOB_ID:-}" && "${SLURM_NTASKS:-1}" -gt 1 ]]; then | |
| export MASTER_ADDR="$(scontrol show hostnames "$SLURM_JOB_NODELIST" | head -n 1)" | |
| export MASTER_PORT=$((29500 + SLURM_JOB_ID % 35500)) | |
| echo "[train.sh] slurm job=$SLURM_JOB_ID ntasks=$SLURM_NTASKS master=$MASTER_ADDR:$MASTER_PORT config=$CONFIG $*" | |
| exec srun --ntasks="$SLURM_NTASKS" --cpu-bind=none "$PYTHON_BIN" -u -m vtm_cosmos.train.train --config "$CONFIG" "$@" | |
| fi | |
| NGPU="${NGPU:-$("$PYTHON_BIN" -c 'import torch; print(max(1, torch.cuda.device_count()))')}" | |
| echo "[train.sh] torchrun nproc=$NGPU config=$CONFIG $*" | |
| exec "$PYTHON_BIN" -m torch.distributed.run --standalone --nproc_per_node="$NGPU" \ | |
| -m vtm_cosmos.train.train --config "$CONFIG" "$@" | |