hfdask 0.1.0__tar.gz

hfdask-0.1.0/.gitignore

@@ -0,0 +1,16 @@
+.venv/
+.cache/
+.hfdask/
+__pycache__/
+*.py[cod]
+.pytest_cache/
+.mypy_cache/
+.ruff_cache/
+dist/
+docs/
+*.egg-info/
+.env
+.env.*
+node-*.env
+mesh.json
+!.env.example

hfdask-0.1.0/LICENSE

@@ -0,0 +1,203 @@
+Copyright 2026 Clause Logic Inc.
+
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

hfdask-0.1.0/PKG-INFO

@@ -0,0 +1,278 @@
+Metadata-Version: 2.5
+Name: hfdask
+Version: 0.1.0
+Summary: Run Dask workloads on Hugging Face Jobs
+Project-URL: Homepage, https://github.com/Hanno-Labs/hfdask
+Project-URL: Repository, https://github.com/Hanno-Labs/hfdask
+Project-URL: Issues, https://github.com/Hanno-Labs/hfdask/issues
+Author: Stephen Solka
+License-Expression: Apache-2.0
+License-File: LICENSE
+Requires-Python: >=3.11
+Requires-Dist: dask[distributed]<2027,>=2025.1
+Requires-Dist: huggingface-hub<2,>=1.29
+Requires-Dist: iroh==1.1.0
+Requires-Dist: pydantic<3,>=2.12
+Requires-Dist: pyyaml<7,>=6
+Description-Content-Type: text/markdown
+
+# hfdask
+
+Run ordinary Dask programs across Hugging Face Jobs from one YAML cluster definition.
+hfdask ships the current Git working tree, starts a coordinator Job and worker Jobs,
+connects them through an authenticated encrypted mesh, and cleans up the paid Jobs
+when the program finishes.
+
+> **Status:** pre-release (`0.1.0`). APIs and configuration may change before the
+> first stable release.
+
+## Why hfdask
+
+We like working in the Hugging Face ecosystem and find ourselves using it for more
+and more of our work. Hugging Face Jobs gives us on-demand CPUs and GPUs alongside
+native access to models, datasets, Spaces, and buckets.
+
+Much of that work combines CPU-heavy data transformation with GPU-heavy inference.
+We like expressing these pipelines in Dask: ordinary Python task graphs make it easy
+to parallelize transformations, preserve dependencies between stages, and schedule
+each task on suitable hardware. What was missing was a way to run that same Dask
+programming model across Hugging Face Jobs.
+
+hfdask connects the two. It turns a YAML cluster definition into a temporary Dask
+cluster on Hugging Face Jobs, runs an ordinary Dask script, and cleans up the Jobs
+when the program finishes. This lets us:
+
+- use CPU and GPU Jobs together in one Dask task graph;
+- keep data transformation, inference, and aggregation in one program;
+- use native Hugging Face model, dataset, Space, and bucket mounts;
+- ship a locked project environment and the current Git working tree;
+- connect the cluster over an authenticated, encrypted mesh without exposing a
+  public Dask scheduler;
+- recover and clean up paid Jobs when a run fails or is interrupted.
+
+```text
+submitting machine
+    hfdask CLI
+        │  source archive + Job definitions
+        ▼
+coordinator HF Job                 worker HF Job(s)
+script + Dask scheduler  ◀─Iroh─▶  one Dask worker per CPU core
+optional CPU workers               optional exclusive GPU assignments
+```
+
+## CPU-only quickstart
+
+Create a Git-backed uv project. Add the libraries used by the Dask program as
+normal project dependencies, then add hfdask to the deployment dependency group
+selected by the cluster YAML.
+
+```sh
+mkdir hfdask-quickstart
+cd hfdask-quickstart
+git init
+uv init --bare --python 3.12
+uv add "dask[dataframe,distributed]>=2025.1,<2027" pandas
+uv add --group deploy hfdask
+curl -fL https://raw.githubusercontent.com/Hanno-Labs/hfdask/v0.1.0/examples/cpu.py -o cpu.py
+curl -fL https://raw.githubusercontent.com/Hanno-Labs/hfdask/v0.1.0/examples/cpu.yaml -o cluster.yaml
+```
+
+[`examples/cpu.py`](https://github.com/Hanno-Labs/hfdask/blob/v0.1.0/examples/cpu.py) is a normal, unannotated Dask DataFrame
+program. It creates four partitions per live worker, so the Dask scheduler can use
+the full CPU pool. The script imports Dask and pandas, not hfdask.
+
+1. Authenticate the submitting machine with `uv run --group deploy hf auth login`.
+2. Set your HF namespace in `cluster.yaml`.
+3. Review the two `cpu-basic` Jobs, the 15-minute timeout, and the public relay
+   consent, then run:
+
+```sh
+uv run --group deploy hfdask run --cluster cluster.yaml cpu.py
+```
+
+The coordinator hosts the script and scheduler. Because `coordinator.worker: true`,
+it reserves one CPU core for the scheduler and starts workers on its remaining
+cores. The separate worker Job uses every complete CPU core. Neither the YAML nor
+the DataFrame graph contains Hugging Face-specific task annotations.
+
+For heterogeneous CPU → GPU → CPU execution with mounted Hub data and worker-local
+vLLM engines, see the [AG News inference example](https://github.com/Hanno-Labs/hfdask/blob/v0.1.0/examples/README.md#gpu-inference).
+
+## Cluster configuration
+
+The CLI reads a strict YAML definition before staging source or submitting Jobs:
+
+| Field | Purpose |
+|---|---|
+| `namespace` | HF user or organization that owns the paid Jobs |
+| `coordinator.flavor` | Hardware for the scheduler and script |
+| `coordinator.worker` | Run workers beside the scheduler, reserving one core for it |
+| `workers.flavor`, `workers.count` | Remote worker hardware and machine count |
+| `environment.image` | Bootstrap image containing Python and uv |
+| `environment.groups` | Locked dependency groups installed in every Job |
+| `mounts` | Hub repositories or buckets mounted into every Job |
+| `timeout` | HF Job lifetime such as `15m` or `2h` |
+| `network.public_relays` | Explicit consent to public discovery and relay fallback |
+
+Unknown keys and coerced scalar types are rejected. Counts must be YAML integers,
+booleans must be YAML booleans, and `workers.count` must be between 1 and 63.
+Clusters support at most 64 Jobs including the coordinator.
+
+Mount sources use HF CLI paths:
+
+- `hf://models/namespace/repo`
+- `hf://datasets/namespace/repo`
+- `hf://spaces/namespace/repo`
+- `hf://buckets/namespace/bucket`
+
+Repository mounts are read-only and may pin a branch, tag, or commit with
+`revision`. Bucket mounts do not accept revisions; set `read_only: false` explicitly
+for outputs. Keep permissions narrow and use a fresh output prefix for each run.
+
+## Workers and task placement
+
+Each machine detects its CPU affinity and cgroup quota at startup. Worker-only
+machines start one process for every complete available core. A coordinator with
+`worker: true` reserves exactly one core for the scheduler and uses the rest; a
+one-core coordinator therefore contributes no worker.
+
+Every worker has one Dask thread. Visible NVIDIA GPUs are assigned exclusively to
+the first worker processes, one GPU per process, while remaining processes are
+CPU-only. GPU discovery requires `nvidia-smi`; MIG partitions are not supported.
+Detected worker counts are exchanged before the mesh allocates Dask service ports.
+
+Unannotated Dask tasks are eligible for every worker. Use ordinary Dask resources
+for numeric reservations such as `resources={"GPU": 1}`. For categorical placement,
+`hfdask.routing.workers_with` and `submit_on` select workers by detected hardware or
+custom group tags without consuming artificial resource slots.
+
+## Dependencies and bootstrap
+
+Declare the libraries imported by your Dask program as normal project dependencies,
+independent of where the program will run:
+
+```sh
+uv add "dask[dataframe,distributed]>=2025.1,<2027" pandas
+```
+
+Add hfdask to a deployment dependency group, then list that group under
+`environment.groups` in the cluster YAML. Dependency groups stay local to the
+project instead of becoming published package extras:
+
+```sh
+uv add --group deploy hfdask
+```
+
+```yaml
+environment:
+  groups: [deploy]
+```
+
+Enable the same group when submitting so the hfdask CLI is available locally:
+
+```sh
+uv run --group deploy hfdask run --cluster cluster.yaml cpu.py
+```
+
+hfdask ships the project's `pyproject.toml` and `uv.lock` to every Job and runs
+`uv sync --locked --no-dev --group deploy`. That installs the base workload
+together with hfdask's remote runner in one locked environment.
+
+An isolated `uvx hfdask` invocation cannot replace the dependency group because it
+does not add the remote runner to the submitted project's lockfile.
+
+## Source shipping
+
+Run the CLI from a Git project containing `pyproject.toml`, `uv.lock`, and the
+project-relative script. hfdask snapshots the current working tree, including
+eligible untracked files—not just committed `HEAD`.
+
+Ignored files and recognized secret paths are excluded, and symlinks are rejected.
+Filename filtering cannot identify every secret, so review the working tree before
+submission. Keep data and model weights in mounts. The source limits are 8 MiB
+compressed, 8 MiB uncompressed, and 2,000 files.
+
+The launcher creates or reuses the namespace's private `jobs-artifacts` bucket,
+uploads the archive under a unique prefix, mounts it read-only, and verifies its
+checksum before extraction. The CLI prints the artifact URI. Artifacts remain after
+the run and can incur storage charges until removed.
+
+## Cleanup and recovery
+
+The CLI writes public recovery handles to `.hfdask/run-*.json`, or to an unused path
+selected with `--manifest`. It saves the growing manifest after every acknowledged
+submission, waits for the driver, and verifies termination of known Jobs. Failures
+and interrupts attempt cleanup without hiding an unverified result.
+
+If the submitting process is lost or cleanup cannot be verified, keep the manifest
+and inspect every recorded Job. With HF credentials available:
+
+```python
+import json
+from pathlib import Path
+
+from hfdask.cluster import Cluster
+
+manifest = json.loads(Path(".hfdask/run-<id>.json").read_text())
+Cluster.from_manifest(manifest).close()
+```
+
+`Cluster.close()` cancels known Jobs and polls them to a terminal state. The
+manifest contains no private node keys, is not a workload checkpoint, and does not
+resolve an ambiguous submission that returned no handle; reconcile those Jobs by
+their cluster labels before retrying.
+
+## Security and limitations
+
+Iroh provides authenticated encrypted QUIC, NAT traversal, and relay transport.
+Dask services bind only to loopback, and each node admits identities from its fixed
+cluster roster. `network.public_relays: true` is required explicitly; discovery and
+relay services can observe connection metadata.
+
+Run only trusted workloads and images. Dask tasks execute arbitrary Python and can
+access their mounted data. HF credentials remain on the submitting machine; Jobs
+receive distinct Iroh keys through Job secrets.
+
+Job-local disks and scheduler state are ephemeral. Automatic whole-cluster resume
+is not implemented. Long workloads should write independently recoverable shards
+to durable storage. The example output writes are not an atomic checkpoint protocol.
+
+## API documentation
+
+From a repository checkout, generate the pdoc reference locally:
+
+```sh
+uv sync --locked --group docs
+mise run docs
+open docs/hfdask.html
+```
+
+The generated `docs/` directory is ignored build output. The tracked
+[`pdoc-templates`](https://github.com/Hanno-Labs/hfdask/tree/v0.1.0/pdoc-templates) directory is source configuration. The reference
+documents the public package and implementation modules but intentionally excludes
+the standalone pre-installation `hfdask.bootstrap` script.
+
+Start with `hfdask.jobs` for a single Job, `hfdask.cluster` for multi-Job lifecycle,
+`hfdask.client` for persistent connections, and `hfdask.routing` for placement.
+
+## Development
+
+```sh
+uv sync --locked --group dev
+mise run check-format
+mise run lint
+mise run test
+mise run docs
+mise run build
+```
+
+CI runs formatting, Ruff, ty, the test suite, pdoc generation, and distribution
+builds on pull requests and pushes to `main`. The opt-in encrypted transport test is:
+
+```sh
+HFDASK_TEST_KEYS=1 uv run pytest tests/test_iroh_integration.py
+```
+
+## License
+
+[Apache-2.0](https://github.com/Hanno-Labs/hfdask/blob/v0.1.0/LICENSE).