[DRAFT] update evaluate to be concurrent #685
isahers1Annotations
1 warning and 2 notices
|
benchmark
ubuntu-latest pipelines will use ubuntu-24.04 soon. For more details, see https://github.com/actions/runner-images/issues/10636
|
|
Benchmark results:
python/langsmith/evaluation/_arunner.py#L1
...........
WARNING: the benchmark result may be unstable
* the standard deviation (73.7 ms) is 11% of the mean (669 ms)
Try to rerun the benchmark with more runs, values and/or loops.
Run 'python -m pyperf system tune' command to reduce the system jitter.
Use pyperf stats, pyperf dump and pyperf hist to analyze results.
Use --quiet option to hide these warnings.
create_5_000_run_trees: Mean +- std dev: 669 ms +- 74 ms
...........
WARNING: the benchmark result may be unstable
* the standard deviation (143 ms) is 10% of the mean (1.38 sec)
Try to rerun the benchmark with more runs, values and/or loops.
Run 'python -m pyperf system tune' command to reduce the system jitter.
Use pyperf stats, pyperf dump and pyperf hist to analyze results.
Use --quiet option to hide these warnings.
create_10_000_run_trees: Mean +- std dev: 1.38 sec +- 0.14 sec
...........
WARNING: the benchmark result may be unstable
* the standard deviation (232 ms) is 17% of the mean (1.40 sec)
Try to rerun the benchmark with more runs, values and/or loops.
Run 'python -m pyperf system tune' command to reduce the system jitter.
Use pyperf stats, pyperf dump and pyperf hist to analyze results.
Use --quiet option to hide these warnings.
create_20_000_run_trees: Mean +- std dev: 1.40 sec +- 0.23 sec
...........
dumps_class_nested_py_branch_and_leaf_200x400: Mean +- std dev: 690 us +- 23 us
...........
dumps_class_nested_py_leaf_50x100: Mean +- std dev: 25.0 ms +- 0.5 ms
...........
dumps_class_nested_py_leaf_100x200: Mean +- std dev: 104 ms +- 2 ms
...........
dumps_dataclass_nested_50x100: Mean +- std dev: 25.4 ms +- 0.2 ms
...........
WARNING: the benchmark result may be unstable
* the standard deviation (15.1 ms) is 21% of the mean (70.6 ms)
Try to rerun the benchmark with more runs, values and/or loops.
Run 'python -m pyperf system tune' command to reduce the system jitter.
Use pyperf stats, pyperf dump and pyperf hist to analyze results.
Use --quiet option to hide these warnings.
dumps_pydantic_nested_50x100: Mean +- std dev: 70.6 ms +- 15.1 ms
...........
dumps_pydanticv1_nested_50x100: Mean +- std dev: 196 ms +- 2 ms
|
|
Comparison against main:
python/langsmith/evaluation/_arunner.py#L1
+-----------------------------------------------+----------+------------------------+
| Benchmark | main | changes |
+===============================================+==========+========================+
| dumps_pydanticv1_nested_50x100 | 217 ms | 196 ms: 1.10x faster |
+-----------------------------------------------+----------+------------------------+
| create_5_000_run_trees | 721 ms | 669 ms: 1.08x faster |
+-----------------------------------------------+----------+------------------------+
| dumps_class_nested_py_branch_and_leaf_200x400 | 695 us | 690 us: 1.01x faster |
+-----------------------------------------------+----------+------------------------+
| dumps_class_nested_py_leaf_100x200 | 103 ms | 104 ms: 1.00x slower |
+-----------------------------------------------+----------+------------------------+
| create_10_000_run_trees | 1.37 sec | 1.38 sec: 1.01x slower |
+-----------------------------------------------+----------+------------------------+
| dumps_class_nested_py_leaf_50x100 | 24.8 ms | 25.0 ms: 1.01x slower |
+-----------------------------------------------+----------+------------------------+
| dumps_dataclass_nested_50x100 | 25.2 ms | 25.4 ms: 1.01x slower |
+-----------------------------------------------+----------+------------------------+
| create_20_000_run_trees | 1.36 sec | 1.40 sec: 1.02x slower |
+-----------------------------------------------+----------+------------------------+
| dumps_pydantic_nested_50x100 | 64.9 ms | 70.6 ms: 1.09x slower |
+-----------------------------------------------+----------+------------------------+
| Geometric mean | (ref) | 1.01x faster |
+-----------------------------------------------+----------+------------------------+
|