cortexgrid 0.2.93__tar.gz → 0.2.95__tar.gz

{cortexgrid-0.2.93 → cortexgrid-0.2.95}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: cortexgrid
-Version: 0.2.93
+Version: 0.2.95
 Summary: Connect your ML code to the RoboLab compute cluster — Ray, MLflow, and S3
 Project-URL: Homepage, https://github.com/robodatalab/cortexgrid
 Project-URL: Repository, https://github.com/robodatalab/cortexgrid
@@ -178,6 +178,8 @@ deployed = cortexgrid.deploy_model("qwen", "instruct", saved.run_name, wait=True
 print(deployed.url)
 ```
 
+`save_model` saves a new copy under every run - meant for weights the run produced (e.g. a fine-tune). For a model produced elsewhere (e.g. a pretrained base model), `cortexgrid.import_model(source, MyServeApp, family, suffix)` uploads it once under `run_name=cortexgrid.IMPORTED` and is a no-op on later runs; deploy it with `deploy_model(family, suffix, cortexgrid.IMPORTED)`.
+
 `save_model` is synchronous (registry lifecycle: `uploading` -> `ready`); `deploy_model` schedules the serving lifecycle (`deploying` -> `running`). With `wait=True` a failed deploy raises `cortexgrid.ModelDeployFailed`; `cortexgrid.wait_for_model_serving(family, suffix, run_name, timeout=...)` waits on a deploy started elsewhere, and re-deploying a failed model retries it from scratch. See [model-serving.md](https://github.com/robodatalab/cortexgrid/blob/main/docs/cortexgrid/model-serving.md) for both lifecycles end to end - upload/deploy/undeploy/delete, status queries (`model_registry_status`, `model_serving_status`), and error handling.
 
 ### API reference

{cortexgrid-0.2.93 → cortexgrid-0.2.95}/cortexgrid/__init__.py

@@ -74,12 +74,14 @@ from cortexgrid.ray_util import (
 )
 from cortexgrid.s3_util import delete_prefix, download, get_s3_client, upload, upload_dir
 from cortexgrid.model_storage import (
+    IMPORTED,
     SavedModel,
     delete_model,
     list_models,
     load_model,
     model_registry_status,
 )
+from cortexgrid.model_storage import import_model as _import_model_storage
 from cortexgrid.model_storage import save_model as _save_model_storage
 from cortexgrid.model_serving import (
     Deployment,
@@ -119,7 +121,11 @@ def save_model(
     weights_dir: str | Path, serve_app: type, family: str, suffix: str
 ) -> SavedModel:
     """Persist a weights directory under the current Experiment's run, paired
-    with the serve-app class that will front it at deploy time."""
+    with the serve-app class that will front it at deploy time.
+
+    Every run saves a new copy under its own run_name - meant for weights the
+    run produced (e.g. a fine-tune). For a model produced elsewhere that should
+    be uploaded once and reused across runs, use `import_model`."""
     experiment = Experiment.get_instance()
     return _save_model_storage(
         weights_dir,
@@ -131,6 +137,27 @@ def save_model(
     )
 
 
+def import_model(
+    source: str | Path | Callable[[], str | Path],
+    serve_app: type,
+    family: str,
+    suffix: str,
+) -> SavedModel:
+    """Register a model produced elsewhere once, reuse it on every later call,
+    and record on the current Experiment's run which imported model it used.
+
+    The model belongs to no run (see `cortexgrid.model_storage.import_model`),
+    so the run keeps the link instead: the tag
+    `imported_model/<family>/<suffix>` holds the version's `created_at`, set
+    whether this call uploaded the model or reused it."""
+    experiment = Experiment.get_instance()
+    model = _import_model_storage(source, serve_app, family, suffix)
+    get_mlflow_client().set_tag(
+        experiment.run_id, f"imported_model/{family}/{suffix}", model.created_at
+    )
+    return model
+
+
 __all__ = [
     "Experiment",
     "delete_experiment",
@@ -179,8 +206,10 @@ __all__ = [
     "list_secrets",
     "delete_secret",
     # Model registry
+    "IMPORTED",
     "SavedModel",
     "save_model",
+    "import_model",
     "load_model",
     "list_models",
     "model_registry_status",

{cortexgrid-0.2.93 → cortexgrid-0.2.95}/cortexgrid/model_storage.py

@@ -6,7 +6,14 @@ Mapping cortexgrid taxonomy <-> MLflow Registry:
     family, suffix      -> ModelVersion.tags["family"], ["suffix"]   (denormalized)
     weights blob path   -> ModelVersion.source =
                            "s3://<bucket>/models/<run_name>/<family>/<suffix>/weights/"
-    run linkage         -> ModelVersion.run_id  (built-in MLflow field)
+    run linkage         -> ModelVersion.run_id  (built-in MLflow field; unset
+                           for imported models)
+
+Two ways in: `save_model` registers a fresh copy under the calling run's
+run_name every time it runs (fine-tuned output); `import_model` registers a
+model produced elsewhere once, under the fixed run_name IMPORTED, and is a
+no-op after that. Both write the same layout, so every
+(family, suffix, run_name) consumer - load_model, deploy_model - handles both.
 
 storage.py is pure: it takes run_id/run_name as explicit args and never reads
 the active Experiment singleton. The facade that fills those in lives in
@@ -20,7 +27,7 @@ from dataclasses import dataclass
 from datetime import datetime, timedelta, timezone
 from pathlib import Path
 import tempfile
-from typing import Any
+from typing import Any, Callable
 
 from mlflow.exceptions import MlflowException
 from mlflow.tracking import MlflowClient
@@ -43,12 +50,18 @@ _PHASE_UPLOAD_FAILED = "upload_failed"
 _PHASE_BROKEN = "broken"
 
 # An upload still marked "uploading" this long after the version was created is
-# treated as broken: save_model creates the version immediately before the
-# upload begins, so creation_timestamp is the upload start, and a process that
-# dies mid-upload never flips the tag to "ready"/"upload_failed". Expiry is
+# treated as broken: the version is created before the weights are resolved
+# (for `import_model`, before its download), so creation_timestamp is the
+# start of the whole upload, and a process that dies mid-way never flips the
+# tag to "ready"/"upload_failed". Expiry is
 # derived lazily on read (see `_phase_for`); nothing is written back.
 _UPLOAD_DEADLINE = timedelta(hours=3)
 
+# run_name under which `import_model` registers a model: imported weights belong
+# to no run, so they share one fixed key and outlive the run that imported
+# them. Run names are haikunator "word-word-NN", so no run can take this name.
+IMPORTED = "imported"
+
 
 @dataclass
 class SavedModel:
@@ -137,6 +150,10 @@ def save_model(
     """Upload a weights directory to S3 and register a new MLflow ModelVersion
     paired with the serve-app that fronts it.
 
+    Meant for weights the calling run produced (e.g. a fine-tune): every run
+    saves its own copy under its own run_name, so running the same code twice
+    yields two models. For weights produced elsewhere, use `import_model`.
+
     cortexgrid stores the weights as an opaque directory: it never inspects,
     serializes, or reconstructs their contents, so the on-disk format
     (HuggingFace `save_pretrained`, `torch.save`, ONNX, anything) is entirely
@@ -148,16 +165,69 @@ def save_model(
     Its code is bundled and its import path, bundle URL, and pip list are
     stored as tags on the ModelVersion so `deploy_model` can bind it later
     without the caller holding the class object."""
+    return _upload_model(weights_dir, serve_app, suffix, family, run_id, run_name)
+
+
+def import_model(
+    source: str | Path | Callable[[], str | Path],
+    serve_app: type,
+    family: str,
+    suffix: str,
+) -> SavedModel:
+    """Register a model produced elsewhere (e.g. a pretrained base model) under
+    the fixed key (family, suffix, IMPORTED), once.
+
+    `source` is the weights directory, or a callable returning it; the callable
+    runs only when the upload actually happens, so an expensive download can be
+    skipped on every run after the first. It runs after the version is
+    registered as "uploading", so a concurrent import sees this one in flight
+    while it downloads.
+
+    If a version is already registered under the key:
+      - "ready": no-op, returns it. `source` and `serve_app` are ignored; to
+        replace the weights or the serve-app, `delete_model` it first.
+      - "uploading": raises RuntimeError - another process is importing it.
+      - "upload_failed" / "broken": deleted and imported again.
+
+    The version is linked to no MLflow run, so deleting a run leaves it in
+    place. Deploy it like any saved model:
+    `deploy_model(family, suffix, IMPORTED)`."""
+    existing = model_registry_status(family, suffix, IMPORTED)
+    if existing is not None:
+        if existing.phase == _PHASE_READY:
+            return existing
+        if existing.phase == _PHASE_UPLOADING:
+            raise RuntimeError(
+                f"Model {family}/{suffix}/{IMPORTED} is being imported by "
+                "another process"
+            )
+        delete_model(family, suffix, IMPORTED)
+    return _upload_model(source, serve_app, suffix, family, None, IMPORTED)
+
+
+def _upload_model(
+    weights: str | Path | Callable[[], str | Path],
+    serve_app: type,
+    suffix: str,
+    family: str,
+    run_id: str | None,
+    run_name: str,
+) -> SavedModel:
+    """Register a ModelVersion in "uploading", resolve `weights` to a directory
+    (calling it when it is a callable), upload the weights and the serve-app
+    bundle, and flip it to "ready" (or "upload_failed")."""
     bucket = get_s3_bucket()
     prefix = f"models/{run_name}/{family}/{suffix}"
-    size_bytes = _dir_size_bytes(weights_dir)
     source = f"s3://{bucket}/{prefix}/weights/"
     name = f"{family}__{suffix}"
     client = MlflowClient(tracking_uri=get_mlflow_tracking_uri())
     _ensure_registered_model(client, name)
-    # Register the version up front in the "uploading" phase so the dashboard
-    # can surface a model while its weights are still streaming to storage. The
-    # bundle tags and the flip to "ready" happen only after the upload lands.
+    # Register the version up front in the "uploading" phase, before `weights`
+    # is resolved: the dashboard surfaces the model while it is still being
+    # fetched and uploaded, and a concurrent `import_model` sees the import in
+    # flight for the whole download instead of starting one of its own. The
+    # size is stamped once the directory exists; the bundle tags and the flip
+    # to "ready" happen only after the upload lands.
     version = client.create_model_version(
         name=name,
         source=source,
@@ -166,11 +236,14 @@ def save_model(
             "family": family,
             "suffix": suffix,
             "run_name": run_name,
-            "size_bytes": str(size_bytes),
             _LIFECYCLE_TAG: _PHASE_UPLOADING,
         },
     )
     try:
+        weights_dir = weights() if callable(weights) else weights
+        client.set_model_version_tag(
+            name, version.version, "size_bytes", str(_dir_size_bytes(weights_dir))
+        )
         s3_util.upload_dir(str(weights_dir), dest_path=f"{prefix}/weights")
         bundle_meta = bundle_class(serve_app, family, suffix, run_name)
         for key, value in metadata_to_tags(bundle_meta).items():

{cortexgrid-0.2.93 → cortexgrid-0.2.95}/docs/cortexgrid/README.md

@@ -150,6 +150,8 @@ deployed = cortexgrid.deploy_model("qwen", "instruct", saved.run_name, wait=True
 print(deployed.url)
 ```
 
+`save_model` saves a new copy under every run - meant for weights the run produced (e.g. a fine-tune). For a model produced elsewhere (e.g. a pretrained base model), `cortexgrid.import_model(source, MyServeApp, family, suffix)` uploads it once under `run_name=cortexgrid.IMPORTED` and is a no-op on later runs; deploy it with `deploy_model(family, suffix, cortexgrid.IMPORTED)`.
+
 `save_model` is synchronous (registry lifecycle: `uploading` -> `ready`); `deploy_model` schedules the serving lifecycle (`deploying` -> `running`). With `wait=True` a failed deploy raises `cortexgrid.ModelDeployFailed`; `cortexgrid.wait_for_model_serving(family, suffix, run_name, timeout=...)` waits on a deploy started elsewhere, and re-deploying a failed model retries it from scratch. See [model-serving.md](https://github.com/robodatalab/cortexgrid/blob/main/docs/cortexgrid/model-serving.md) for both lifecycles end to end - upload/deploy/undeploy/delete, status queries (`model_registry_status`, `model_serving_status`), and error handling.
 
 ### API reference

{cortexgrid-0.2.93 → cortexgrid-0.2.95}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "cortexgrid"
-version = "0.2.93"
+version = "0.2.95"
 description = "Connect your ML code to the RoboLab compute cluster — Ray, MLflow, and S3"
 readme = "docs/cortexgrid/README.md"
 license = "Apache-2.0"