Jump to content

Connect SuperML | Leeroopedia MCP: Equip your AI agents with best practices, code verification, and debugging knowledge. Powered by Leeroo — building Organizational Superintelligence. Contact us at founders@leeroo.com.

Implementation:Bentoml BentoML Models Delete

From Leeroopedia
Revision as of 12:07, 16 February 2026 by Admin (talk | contribs) (Auto-imported from implementations/Bentoml_BentoML_Models_Delete.md)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Implementation Metadata
Implementation Name Models Delete
API bentoml.models.delete()
Source src/bentoml/models.py:L53-58 (public), src/bentoml/_internal/store.py:L183-207 (Store.delete)
Workflow Model_Store_Management
Domain ML_Serving, Model_Management
Implements Principle:Bentoml_BentoML_Model_Cleanup
Last Updated 2026-02-13 15:00 GMT

Overview

The bentoml.models.delete() function removes a specific model version from the local BentoML model store. It deletes the model's directory and all contained artifact files, and updates the "latest" symlink if the deleted version was the most recent.

Import

import bentoml

Signature

def delete(tag: Tag | str) -> None

Parameters

Parameter Type Default Description
tag str required The model tag to delete from the local store. Can be "name:version" for a specific version or "name:latest" to delete the latest version.

Inputs and Outputs

Inputs:

  • Model tag to remove from the local store

Outputs:

  • None — the model directory is removed from <bentoml_home>/models/ as a side effect
  • The "latest" symlink is updated if the deleted version was the latest
  • Raises NotFound if the specified model does not exist

Usage Examples

import bentoml

# Delete a specific model version
bentoml.models.delete("text_classifier:abc123")

# Delete the latest version of a model
bentoml.models.delete("text_classifier:latest")

# Bulk cleanup: keep only the 3 most recent versions
models = bentoml.models.list("text_classifier")
for model in models[3:]:
    bentoml.models.delete(model.tag)

# Delete all versions of a model
for model in bentoml.models.list("text_classifier"):
    bentoml.models.delete(model.tag)

Behavior Details

Deletion Process

The delete operation (implemented in Store.delete() at src/bentoml/_internal/store.py:L183-207) performs the following steps:

  1. Resolve the tag: If the tag uses "latest", resolve it to the actual version.
  2. Validate existence: Verify the model exists in the store. Raise NotFound if not.
  3. Remove directory: Delete the model version directory at <bentoml_home>/models/<name>/<version>/.
  4. Update "latest" symlink: If the deleted version was the one pointed to by the "latest" symlink:
    • If other versions remain, update the symlink to point to the next most recent version.
    • If no versions remain, remove the symlink and the model name directory.

Error Handling

Scenario Behavior
Model tag does not exist Raises NotFound error
Delete the latest version "latest" symlink updated to previous version
Delete the only remaining version Model name directory and symlink removed entirely
Delete a non-latest version Only that version's directory removed; "latest" unchanged

Filesystem Effects

Deletion removes the entire version directory, including:

  • Serialized model artifact files (e.g., .pkl, .pt, .onnx)
  • The model.yaml descriptor file
  • Any custom objects stored alongside the model
  • Any framework-specific auxiliary files

Important Notes

  • Irreversible: Deletion is permanent. There is no trash or undo mechanism.
  • Local Only: This operation only affects the local store. Models pushed to BentoCloud are not affected.
  • No Cascade: Deleting a model does not affect any bentos or services that reference it. Those references will fail at resolution time if the model is needed.

Source Reference

  • Public API: src/bentoml/models.py, lines 53-58
  • Store implementation: src/bentoml/_internal/store.py, lines 183-207

Knowledge Sources

Page Connections

Double-click a node to navigate. Hold to expand connections.
Principle
Implementation
Heuristic
Environment