⚡️ Speed up method MetadataToFloatCollectionInvocation.invoke by 16%

[codeflash-ai]

📄 16% (0.16x) speedup for MetadataToFloatCollectionInvocation.invoke in invokeai/app/invocations/metadata_linked.py

⏱️ Runtime : 22.1 microseconds → 19.1 microseconds (best of 64 runs)

📝 Explanation and details

The optimized code achieves a 15% speedup by eliminating unnecessary object creation and string conversions in the metadata extraction path. Here are the key optimizations:

What was optimized:

1. Eliminated intermediate dictionary allocation: The original code always created a data dictionary (either empty {} or self.metadata.root), which required memory allocation and assignment. The optimized version uses direct conditional logic to avoid this.

2. Avoided redundant string conversion: The original code unconditionally called str() on the label, even when it was already a string. The optimized version includes a type check to only convert non-string keys, with proper exception handling as a fallback.

3. Reduced temporary variable assignments: The optimized code eliminates the intermediate output variable and directly passes the result to FloatCollectionOutput().

Why this leads to speedup:

• Memory allocation reduction: Creating dictionaries and intermediate objects has overhead in Python's memory manager
• String conversion optimization: isinstance(key, str) is faster than unconditional str() conversion when the input is already a string
• Fewer variable assignments: Direct value passing reduces Python's variable lookup and assignment overhead

Performance benefits based on test cases:

• Tests show consistent 8-27% improvements across different scenarios
• The test_large_scale_metadata_none_large_default case (8% improvement) demonstrates the optimization works well even with large default values
• Basic metadata-None cases show 24-27% improvements, indicating the conditional logic optimization is particularly effective for the common case where metadata is absent

The optimization is especially valuable since this appears to be a metadata extraction utility that could be called frequently in AI inference pipelines, where even small per-call improvements compound significantly.

✅ Correctness verification report:

Test	Status
⚙️ Existing Unit Tests	🔘 None Found
🌀 Generated Regression Tests	✅ 6 Passed
⏪ Replay Tests	🔘 None Found
🔎 Concolic Coverage Tests	🔘 None Found
📊 Tests Coverage	100.0%

🌀 Generated Regression Tests and Runtime

from dataclasses import dataclass # function to test # (Paste from the provided code for context, but in real use, this would be imported from the actual module) from typing import Any, Dict, List, Optional # imports import pytest # used for our unit tests from invokeai.app.invocations.metadata_linked import \ MetadataToFloatCollectionInvocation # Simulated minimal stubs for required classes (as we cannot import actual InvokeAI types here) class FloatCollectionOutput: def __init__(self, collection: List[float]): self.collection = collection @dataclass class MetadataStub: root: Dict[str, Any] @dataclass class InvocationContext: # Minimal stub, not used in logic pass CUSTOM_LABEL: str = "* CUSTOM LABEL *" from invokeai.app.invocations.metadata_linked import \ MetadataToFloatCollectionInvocation # ------------------- UNIT TESTS ------------------- # Basic Test Cases def test_basic_no_metadata_returns_default(): """Test: metadata is None, returns default_value.""" node = MetadataToFloatCollectionInvocation(label="cfg_scale", default_value=[3.14], metadata=None) codeflash_output = node.invoke(InvocationContext()); result = codeflash_output # 4.98μs -> 3.92μs (26.9% faster) # Edge Test Cases #------------------------------------------------ from typing import Any, Dict, Optional # imports import pytest # used for our unit tests from invokeai.app.invocations.metadata_linked import \ MetadataToFloatCollectionInvocation # function to test # Minimal stubs for required classes and constants CUSTOM_LABEL: str = "* CUSTOM LABEL *" class FloatCollectionOutput: def __init__(self, collection): self.collection = collection class InvocationContext: pass class MetadataStub: """Stub for metadata object with a .root property.""" def __init__(self, root: Dict[str, Any]): self.root = root from invokeai.app.invocations.metadata_linked import \ MetadataToFloatCollectionInvocation # ------------------ UNIT TESTS ------------------ # Basic Test Cases def test_basic_none_metadata_returns_default(): """Test when metadata is None, should always return default.""" invocation = MetadataToFloatCollectionInvocation(label="cfg_scale", default_value=[2.2], metadata=None) codeflash_output = invocation.invoke(InvocationContext()); result = codeflash_output # 5.21μs -> 4.17μs (24.9% faster) # Edge Test Cases def test_large_scale_metadata_none_large_default(): """Test with None metadata and large default_value.""" large_default = [float(i) for i in range(1000)] invocation = MetadataToFloatCollectionInvocation(label="cfg_scale", default_value=large_default, metadata=None) codeflash_output = invocation.invoke(InvocationContext()); result = codeflash_output # 11.9μs -> 11.0μs (8.07% faster) # codeflash_output is used to check that the output of the original code is the same as that of the optimized code.

To edit these changes git checkout codeflash/optimize-MetadataToFloatCollectionInvocation.invoke-mhllpo9d and push.