Image-Text-to-Text
Transformers
Safetensors
qwen3_5
vllm
video
multimodal
reinforcement-learning
temporal-grounding
object-tracking
video-segmentation
visual-question-answering
spatial-reasoning
qwen3.5
conversational
Instructions to use OraRL/Video-ORA-9B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use OraRL/Video-ORA-9B with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("image-text-to-text", model="OraRL/Video-ORA-9B") messages = [ { "role": "user", "content": [ {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"}, {"type": "text", "text": "What animal is on the candy?"} ] }, ] pipe(text=messages)# Load model directly from transformers import AutoProcessor, AutoModelForMultimodalLM processor = AutoProcessor.from_pretrained("OraRL/Video-ORA-9B") model = AutoModelForMultimodalLM.from_pretrained("OraRL/Video-ORA-9B", device_map="auto") messages = [ { "role": "user", "content": [ {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"}, {"type": "text", "text": "What animal is on the candy?"} ] }, ] inputs = processor.apply_chat_template( messages, add_generation_prompt=True, tokenize=True, return_dict=True, return_tensors="pt", ).to(model.device) outputs = model.generate(**inputs, max_new_tokens=40) print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:])) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use OraRL/Video-ORA-9B with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "OraRL/Video-ORA-9B" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "OraRL/Video-ORA-9B", "messages": [ { "role": "user", "content": [ { "type": "text", "text": "Describe this image in one sentence." }, { "type": "image_url", "image_url": { "url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg" } } ] } ] }'Use Docker
docker model run hf.co/OraRL/Video-ORA-9B
- SGLang
How to use OraRL/Video-ORA-9B with SGLang:
Install from pip and serve model
# Install SGLang from pip: pip install sglang # Start the SGLang server: python3 -m sglang.launch_server \ --model-path "OraRL/Video-ORA-9B" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "OraRL/Video-ORA-9B", "messages": [ { "role": "user", "content": [ { "type": "text", "text": "Describe this image in one sentence." }, { "type": "image_url", "image_url": { "url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg" } } ] } ] }'Use Docker images
docker run --gpus all \ --shm-size 32g \ -p 30000:30000 \ -v ~/.cache/huggingface:/root/.cache/huggingface \ --env "HF_TOKEN=<secret>" \ --ipc=host \ lmsysorg/sglang:latest \ python3 -m sglang.launch_server \ --model-path "OraRL/Video-ORA-9B" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "OraRL/Video-ORA-9B", "messages": [ { "role": "user", "content": [ { "type": "text", "text": "Describe this image in one sentence." }, { "type": "image_url", "image_url": { "url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg" } } ] } ] }' - Docker Model Runner
How to use OraRL/Video-ORA-9B with Docker Model Runner:
docker model run hf.co/OraRL/Video-ORA-9B
File size: 8,041 Bytes
53c10a4 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 | #!/usr/bin/env bash
set -euo pipefail
SCRIPT_DIR="$(cd -- "$(dirname -- "${BASH_SOURCE[0]}")" && pwd)"
RELEASE_DIR="$(cd -- "${SCRIPT_DIR}/.." && pwd)"
usage() {
cat <<'EOF'
Usage:
launch_multinode.sh [launcher options] -- <orarl-train options>
Launcher options:
--hosts LIST Comma-separated SSH hosts. May also use HOSTS.
--hostfile FILE One SSH host per line. May also use HOSTFILE.
--gpus-per-node N GPUs per host (default: 8).
--ray-port N Ray head port (default: 6379).
--strict-host-key-checking MODE
yes (default), accept-new, ask, or no.
--run Start Ray and training. Default is a safe dry run.
--dry-run Print the launch plan without connecting.
--help Show this message.
The arguments after -- must include --config, --model, --train-data,
--val-data, and --output. The launcher supplies topology and --run.
EOF
}
HOSTS_VALUE="${HOSTS:-}"
HOSTFILE_VALUE="${HOSTFILE:-}"
GPUS_PER_NODE="${GPUS_PER_NODE:-8}"
RAY_PORT="${RAY_PORT:-6379}"
STRICT_MODE="${STRICT_HOST_KEY_CHECKING:-yes}"
EXECUTE=0
while (($# > 0)); do
case "$1" in
--hosts)
[[ $# -ge 2 ]] || { echo "ERROR: --hosts requires a value" >&2; exit 2; }
HOSTS_VALUE="$2"
shift 2
;;
--hostfile)
[[ $# -ge 2 ]] || { echo "ERROR: --hostfile requires a value" >&2; exit 2; }
HOSTFILE_VALUE="$2"
shift 2
;;
--gpus-per-node)
[[ $# -ge 2 ]] || { echo "ERROR: --gpus-per-node requires a value" >&2; exit 2; }
GPUS_PER_NODE="$2"
shift 2
;;
--ray-port)
[[ $# -ge 2 ]] || { echo "ERROR: --ray-port requires a value" >&2; exit 2; }
RAY_PORT="$2"
shift 2
;;
--strict-host-key-checking)
[[ $# -ge 2 ]] || {
echo "ERROR: --strict-host-key-checking requires a value" >&2
exit 2
}
STRICT_MODE="$2"
shift 2
;;
--run)
EXECUTE=1
shift
;;
--dry-run)
EXECUTE=0
shift
;;
--help|-h)
usage
exit 0
;;
--)
shift
break
;;
*)
echo "ERROR: unknown launcher option: $1" >&2
usage >&2
exit 2
;;
esac
done
TRAIN_ARGS=("$@")
if ((${#TRAIN_ARGS[@]} == 0)); then
echo "ERROR: pass orarl-train options after --" >&2
exit 2
fi
if [[ -n "${HOSTS_VALUE}" && -n "${HOSTFILE_VALUE}" ]]; then
echo "ERROR: use only one of HOSTS/--hosts and HOSTFILE/--hostfile" >&2
exit 2
fi
if [[ ! "${GPUS_PER_NODE}" =~ ^[1-9][0-9]*$ ]]; then
echo "ERROR: GPUs per node must be a positive integer" >&2
exit 2
fi
if [[ ! "${RAY_PORT}" =~ ^[0-9]+$ ]] || ((RAY_PORT < 1 || RAY_PORT > 65535)); then
echo "ERROR: Ray port must be between 1 and 65535" >&2
exit 2
fi
case "${STRICT_MODE}" in
yes|accept-new|ask|no) ;;
*)
echo "ERROR: strict host key checking must be yes, accept-new, ask, or no" >&2
exit 2
;;
esac
for argument in "${TRAIN_ARGS[@]}"; do
lowered="${argument,,}"
if [[ "${lowered}" == "--run" || "${lowered}" == "--dry-run" ]]; then
echo "ERROR: training execution mode is controlled by the launcher" >&2
exit 2
fi
if [[ "${lowered}" =~ (^|[._-])(api[_-]?key|auth[_-]?token|access[_-]?token|secret|password|credential|private[_-]?key)(=|$) ]]; then
echo "ERROR: credential-shaped arguments are not forwarded" >&2
exit 2
fi
if [[ "${argument}" =~ ://[^/@[:space:]]+:[^@/[:space:]]+@ ]]; then
echo "ERROR: URLs with embedded authentication are not forwarded" >&2
exit 2
fi
done
declare -a RAW_HOSTS=()
if [[ -n "${HOSTFILE_VALUE}" ]]; then
[[ -f "${HOSTFILE_VALUE}" ]] || {
echo "ERROR: host file does not exist: ${HOSTFILE_VALUE}" >&2
exit 2
}
while IFS= read -r line || [[ -n "${line}" ]]; do
line="${line%%#*}"
line="${line#"${line%%[![:space:]]*}"}"
line="${line%"${line##*[![:space:]]}"}"
[[ -n "${line}" ]] && RAW_HOSTS+=("${line}")
done < "${HOSTFILE_VALUE}"
elif [[ -n "${HOSTS_VALUE}" ]]; then
IFS=',' read -r -a RAW_HOSTS <<< "${HOSTS_VALUE}"
else
echo "ERROR: set HOSTS, set HOSTFILE, or pass the matching option" >&2
exit 2
fi
declare -a HOST_LIST=()
declare -A SEEN_HOSTS=()
for host in "${RAW_HOSTS[@]}"; do
host="${host#"${host%%[![:space:]]*}"}"
host="${host%"${host##*[![:space:]]}"}"
if [[ ! "${host}" =~ ^([A-Za-z0-9._-]+@)?[A-Za-z0-9._-]+$ ]]; then
echo "ERROR: invalid SSH host entry" >&2
exit 2
fi
if [[ -z "${SEEN_HOSTS[${host}]:-}" ]]; then
HOST_LIST+=("${host}")
SEEN_HOSTS["${host}"]=1
fi
done
if ((${#HOST_LIST[@]} == 0)); then
echo "ERROR: no hosts were provided" >&2
exit 2
fi
PYTHON_BIN="${PYTHON_BIN:-$(command -v python || true)}"
if [[ -z "${PYTHON_BIN}" ]]; then
echo "ERROR: active Python executable was not found" >&2
exit 2
fi
if ! "${PYTHON_BIN}" -c 'import sys; sys.exit(sys.version_info < (3, 10))' 2>/dev/null; then
echo "ERROR: OraRL multi-node launch requires Python 3.10 or newer." >&2
echo "Set PYTHON_BIN to the compatible verl environment." >&2
exit 2
fi
export PYTHONPATH="${RELEASE_DIR}${PYTHONPATH:+:${PYTHONPATH}}"
FORWARD_VARIABLES=(
PATH
PYTHONPATH
LD_LIBRARY_PATH
VIRTUAL_ENV
CONDA_PREFIX
CONDA_DEFAULT_ENV
CUDA_VISIBLE_DEVICES
NCCL_DEBUG
NCCL_SOCKET_IFNAME
NCCL_IB_HCA
NCCL_IB_DISABLE
NCCL_P2P_DISABLE
TOKENIZERS_PARALLELISM
HF_HOME
TRANSFORMERS_CACHE
XDG_CACHE_HOME
ORARL_MEDIA_ROOT
ORARL_PREPROCESSED_DIR
ORARL_DATALOADER_WORKERS
ORARL_VIDEO_MAX_PIXELS
ORARL_VIDEO_MIN_PIXELS
ORARL_VIDEO_MAX_FRAMES
ORARL_VIDEO_FPS
ORARL_VIDEO_TOTAL_PIXELS
ORARL_IMAGE_MIN_PIXELS
ORARL_IMAGE_MAX_PIXELS
ORARL_VIDEO_SOURCE_MODE
ORARL_INLINE_VIDEO_TENSORS
ORARL_MAX_PROMPT_LENGTH
ORARL_MAX_RESPONSE_LENGTH
ORARL_ROLLOUT_BATCH_SIZE
ORARL_VAL_BATCH_SIZE
ORARL_ENABLE_THINKING
ORARL_SEED
ORARL_GLOBAL_BATCH_SIZE
ORARL_UPDATE_MICRO_BATCH
ORARL_EXPERIENCE_MICRO_BATCH
ORARL_MAX_TOKEN_LEN_PER_GPU
ORARL_LEARNING_RATE
ORARL_ROLLOUT_BACKEND
)
SSH_OPTIONS=(
-o "BatchMode=yes"
-o "StrictHostKeyChecking=${STRICT_MODE}"
)
build_remote_command() {
local -a parts=("env")
local name
for name in "${FORWARD_VARIABLES[@]}"; do
if [[ -v "${name}" ]]; then
parts+=("${name}=${!name}")
fi
done
parts+=("$@")
local rendered
printf -v rendered '%q ' "${parts[@]}"
printf '%s' "${rendered% }"
}
display_command() {
local host="$1"
local label="$2"
shift 2
local rendered
printf -v rendered '%q ' "$@"
printf '[dry-run] %s on %s: %s\n' "${label}" "${host}" "${rendered% }"
printf ' forwards %d allowlisted environment names when set\n' \
"${#FORWARD_VARIABLES[@]}"
}
run_remote() {
local host="$1"
local label="$2"
shift 2
if ((EXECUTE == 0)); then
display_command "${host}" "${label}" "$@"
return
fi
printf '[run] %s on %s\n' "${label}" "${host}"
remote_command="$(build_remote_command "$@")"
ssh "${SSH_OPTIONS[@]}" "${host}" "${remote_command}"
}
HEAD_HOST="${HOST_LIST[0]}"
HEAD_ADDRESS="${HEAD_HOST##*@}"
RAY_ADDRESS="${HEAD_ADDRESS}:${RAY_PORT}"
NODE_COUNT="${#HOST_LIST[@]}"
run_remote "${HEAD_HOST}" "Ray head" \
"${PYTHON_BIN}" -m ray start \
--head \
"--port=${RAY_PORT}" \
"--num-gpus=${GPUS_PER_NODE}" \
--disable-usage-stats
for ((index = 1; index < NODE_COUNT; index++)); do
run_remote "${HOST_LIST[index]}" "Ray worker" \
"${PYTHON_BIN}" -m ray start \
"--address=${RAY_ADDRESS}" \
"--num-gpus=${GPUS_PER_NODE}" \
--disable-usage-stats
done
TRAIN_COMMAND=(
"${PYTHON_BIN}"
-m
orarl.cli.train
"${TRAIN_ARGS[@]}"
--nodes
"${NODE_COUNT}"
--gpus-per-node
"${GPUS_PER_NODE}"
--run
)
if ((EXECUTE == 0)); then
display_command "${HEAD_HOST}" "trainer (RAY_ADDRESS set at execution)" \
"${TRAIN_COMMAND[@]}"
else
printf '[run] trainer on %s\n' "${HEAD_HOST}"
remote_command="$(build_remote_command "RAY_ADDRESS=${RAY_ADDRESS}" "${TRAIN_COMMAND[@]}")"
ssh "${SSH_OPTIONS[@]}" "${HEAD_HOST}" "${remote_command}"
fi
|