hf jobs uv run --flavor t4-small -- script.py --early-stopping-patience 3
               └─ Jobs option ┘    └─ script + arguments ─────────────┘

>>> hf jobs uv run train.py

>>> hf jobs uv run python -c 'print("Hello from the cloud!")'

>>> hf jobs uv run --with trl train.py
>>> hf jobs uv run --python 3.12 train.py

>>> hf jobs uv run --with trl-jobs -- trl-jobs sft --model_name Qwen/Qwen3-0.6B --dataset_name trl-lib/Capybara

# /// script
# requires-python = ">=3.11"
# dependencies = ["vllm", "datasets"]
#
# [tool.hf-jobs]
# image   = "vllm/vllm-openai:unlimited-ocr"
# flavor  = "l4x1"
# python  = "/usr/bin/python3"
# secrets = ["HF_TOKEN"]
# ///

>>> hf jobs run ubuntu echo "Hello from the cloud!"

>>> hf jobs run --flavor cpu-basic python:3.12 -- python --help

# Access job environment information
>>> hf jobs run python:3.12 python -c "import os; print(f'Job: {os.environ.get(\"JOB_ID\")}, CPU: {os.environ.get(\"CPU_CORES\")}, Mem: {os.environ.get(\"MEMORY\")}')"

# Pass environment variables
>>> hf jobs uv run -e FOO=foo -e BAR=bar python -c 'import os; print(os.environ["FOO"], os.environ["BAR"])'

# Pass an environment from a local .env file
>>> hf jobs uv run --env-file .env python -c 'import os; print(os.environ["FOO"], os.environ["BAR"])'

# Pass secrets - they will be encrypted server side
>>> hf jobs uv run -s MY_SECRET=psswrd python -c 'import os; print(os.environ["MY_SECRET"])'

# Pass secrets from a local .env.secrets file - they will be encrypted server side
>>> hf jobs uv run --secrets-file .env.secrets python -c 'import os; print(os.environ["MY_SECRET"])'

# Mount a dataset and query it with DuckDB
>>> hf jobs run -v hf://datasets/stanfordnlp/imdb:/dataset \
...     duckdb/duckdb duckdb -c "SELECT * FROM '/dataset/**/*.parquet' LIMIT 5"

# Mount a bucket to save training checkpoints
>>> hf jobs uv run -v hf://buckets/username/my-bucket:/training-outputs \
...     sft.py --output-dir /training-outputs/training-v3-final

>>> hf jobs run -v hf://datasets/username/my-dataset:/data -v hf://buckets/username/my-bucket:/output \
...     python:3.12 python script.py

>>> hf jobs run -v hf://buckets/username/my-bucket:/mnt:ro python:3.12 ls /mnt

>>> hf jobs uv run -v ./pdfs:/input -v ./md-out:/output:rw ocr.py

from huggingface_hub import Volume, run_job

job = run_job(
    image="python:3.12",
    command=["python", "-c", "import os; print(os.listdir('/data'))"],
    volumes=[
        Volume(type="dataset", source="username/my-dataset", mount_path="/data"),
        Volume(type="bucket", source="username/my-bucket", mount_path="/output"),
    ],
)

>>> hf jobs uv run --with torch --flavor a10g-small python -c "import torch; print(f'This code ran with the following GPU: {torch.cuda.get_device_name()}')"

This code ran with the following GPU: NVIDIA A10G

>>> hf jobs uv run --with trl --flavor a10g-small -s HF_TOKEN -- sft.py --model_name_or_path Qwen/Qwen2-0.5B ...

>>> hf jobs hardware
name             pretty name             cpu       ram      storage   accelerator               cost/min  cost/hour
cpu-basic        CPU Basic               2 vCPU    16 GB    50 GB                               $0.0002   $0.01
cpu-upgrade      CPU Upgrade             8 vCPU    32 GB    50 GB                               $0.0005   $0.03
cpu-performance  CPU Performance         32 vCPU   256 GB   1024 GB                             $0.0317   $1.90
cpu-xl           CPU XL                  16 vCPU   124 GB   1000 GB                             $0.0167   $1.00
t4-small         Nvidia T4 - small       4 vCPU    15 GB    50 GB     1x T4 (16 GB)             $0.0067   $0.40
t4-medium        Nvidia T4 - medium      8 vCPU    30 GB    100 GB    1x T4 (16 GB)             $0.0100   $0.60
a10g-small       Nvidia A10G - small     4 vCPU    15 GB    110 GB    1x A10G (24 GB)           $0.0167   $1.00
a10g-large       Nvidia A10G - large     12 vCPU   46 GB    200 GB    1x A10G (24 GB)           $0.0250   $1.50
a10g-largex2     2x Nvidia A10G - large  24 vCPU   92 GB    1000 GB   2x A10G (48 GB)           $0.0500   $3.00
a10g-largex4     4x Nvidia A10G - large  48 vCPU   184 GB   2000 GB   4x A10G (96 GB)           $0.0833   $5.00
a100-large       Nvidia A100 - large     12 vCPU   142 GB   1000 GB   1x A100 (80 GB)           $0.0417   $2.50
a100x4           4x Nvidia A100          48 vCPU   568 GB   4000 GB   4x A100 (320 GB)          $0.1667   $10.00
a100x8           8x Nvidia A100          96 vCPU   1136 GB  8000 GB   8x A100 (640 GB)          $0.3333   $20.00
h200             Nvidia H200             23 vCPU   256 GB   3000 GB   1x H200 (141 GB)          $0.0833   $5.00
h200x2           Nvidia H200             46 vCPU   512 GB   6000 GB   2x H200 (282 GB)          $0.1667   $10.00
h200x4           Nvidia H200             92 vCPU   1024 GB  12000 GB  4x H200 (564 GB)          $0.3333   $20.00
h200x8           Nvidia H200             184 vCPU  2048 GB  24000 GB  8x H200 (1128 GB)         $0.6667   $40.00
rtx-pro-6000     Nvidia RTX PRO 6000     23 vCPU   256 GB   475 GB    1x RTX PRO 6000 (96 GB)   $0.0458   $2.75
rtx-pro-6000x2   Nvidia RTX PRO 6000     46 vCPU   512 GB   950 GB    2x RTX PRO 6000 (192 GB)  $0.0917   $5.50
rtx-pro-6000x4   Nvidia RTX PRO 6000     92 vCPU   1024 GB  1900 GB   4x RTX PRO 6000 (384 GB)  $0.1833   $11.00
rtx-pro-6000x8   Nvidia RTX PRO 6000     184 vCPU  2048 GB  3800 GB   8x RTX PRO 6000 (768 GB)  $0.3667   $22.00
l4x1             1x Nvidia L4            8 vCPU    30 GB    400 GB    1x L4 (24 GB)             $0.0133   $0.80
l4x4             4x Nvidia L4            48 vCPU   186 GB   3200 GB   4x L4 (96 GB)             $0.0633   $3.80
l40sx1           1x Nvidia L40S          8 vCPU    62 GB    380 GB    1x L40S (48 GB)           $0.0300   $1.80
l40sx4           4x Nvidia L40S          48 vCPU   382 GB   3200 GB   4x L40S (192 GB)          $0.1383   $8.30
l40sx8           8x Nvidia L40S          192 vCPU  1534 GB  6500 GB   8x L40S (384 GB)          $0.3917   $23.50

curl -H "Authorization: Bearer $HF_TOKEN" https://<job_id>--<port>.hf.jobs/

# Expose a web server running on port 8000
>>> hf jobs run --expose 8000 python:3.12 python -m http.server 8000
✓ Job started
  id: 6a2aa7cec4f53f9fc5aa4cff
  url: https://huggingface.co/jobs/Wauplin/6a2aa7cec4f53f9fc5aa4cff
Hint: Exposed ports are reachable at (requires an HF token with read access to the job):
  https://6a2aa7cec4f53f9fc5aa4cff--8000.hf.jobs
Serving HTTP on 0.0.0.0 port 8000 (http://0.0.0.0:8000/) ...

>>> from huggingface_hub import run_job
>>> job = run_job(image="python:3.12", command=["python", "-m", "http.server", "8000"], expose=[8000])
>>> job.status.expose_urls
['https://6a2ab384c4f53f9fc5aa4d4f--8000.hf.jobs']

# Expose port 8000 (requires an HF token)
>>> hf jobs expose <job_id> 8000

# Expose port 8000 (requires an HF token) and port 9000 publicly
>>> hf jobs expose <job_id> 8000 --public 9000

# Close all exposed ports
>>> hf jobs expose <job_id> --clear

# Start a job with SSH enabled
>>> hf jobs run --ssh --detach --timeout 10m python:3.12 sleep infinity
✓ Job started
  id: 6a2bd1f1871c005b5352ad31
  url: https://huggingface.co/jobs/Wauplin/6a2bd1f1871c005b5352ad31
Hint: Use `hf jobs ssh 6a2bd1f1871c005b5352ad31` to open an SSH session into the job.

# Open an SSH session into the job
>>> hf jobs ssh 6a2bd1f1871c005b5352ad31

>>> hf jobs ssh 6a2bd1f1871c005b5352ad31 --dry-run
ssh 6a2bd1f1871c005b5352ad31@ssh.hf.jobs

>>> hf jobs ssh 6a2bd1f1871c005b5352ad31 -i ~/.ssh/id_ed25519

>>> from huggingface_hub import run_job
>>> job = run_job(image="python:3.12", command=["sleep", "infinity"], ssh=True)
>>> job.status.ssh_url
'ssh://6a2bd1f1871c005b5352ad31@ssh.hf.jobs'

# Forward local port 6006 to the Job's TensorBoard on port 6006
>>> ssh -L 6006:localhost:6006 6a2bd1f1871c005b5352ad31@ssh.hf.jobs

# Make your local port 8080 reachable from inside the Job on port 8080
>>> ssh -R 8080:localhost:8080 6a2bd1f1871c005b5352ad31@ssh.hf.jobs

# Start a server, reachable by the other members of the group as "master"
>>> hf jobs run --detach --network-group train --network-alias master python:3.12 python -m http.server 8000

# Start a client in the same group
>>> hf jobs run --detach --network-group train python:3.12 sh -c 'curl --retry 10 --retry-connrefused "http://${HF_NETWORK_GROUP_PREFIX}master:8000/"'

>>> from huggingface_hub import run_job
>>> server = run_job(
...     image="python:3.12",
...     command=["python", "-m", "http.server", "8000"],
...     network_group="train",
...     network_aliases=["master"],
... )
>>> client = run_job(
...     image="python:3.12",
...     command=["sh", "-c", 'curl --retry 10 --retry-connrefused "http://${HF_NETWORK_GROUP_PREFIX}master:8000/"'],
...     network_group="train",
... )

>>> hf jobs uv run --timeout 7200 --with torch --flavor a10g-large train.py

>>> hf jobs uv run --timeout 2h --with torch --flavor a10g-large train.py

--timeout 30m    # 30 minutes
--timeout 1.5h   # 1.5 hours
--timeout 1d     # 1 day
--timeout 3600s  # 3600 seconds

>>> hf jobs uv run --attempts 3 --flavor a10g-large train.py

>>> hf jobs uv run --namespace my-org-name python -c "print('Running in an org account')"

>>> hf jobs uv run --namespace my-org-name --token hf_xxx python -c "print('Running in an org account')"

hf jobs uv run --label fine-tuning --label model=Qwen3-06B --label dataset=Capybara ...

hf jobs run --name daily-report python:3.12 python report.py

# Replace the labels on a Job
hf jobs labels <job_id> --label env=prod --label team=ml

# Name an existing Job (keeps its other labels)
hf jobs labels <job_id> --name daily-report

# Remove all labels from a Job
hf jobs labels <job_id> --clear