Filter
Python package: neptune-fetcher
Filter used to specify criteria when fetching experiments or runs.
Methods for attributes
The following functions create a criterion based on the value of an attribute:
| Method | Description | Example |
|---|---|---|
name_eq() | Run or experiment name equals the provided string. | name_eq("flying-123") |
name_in() | Run or experiment name equals any of the provided strings. | name_in("flying-123", "swimming-77") |
eq() | Attribute value equals an int, float, str, or datetime value. | eq("lr", 0.001) |
ne() | Attribute value doesn't equal an int, float, str, or datetime value. | ne("sys/owner", "bot@my-workspace") |
gt() | Attribute value is greater than an int, float, str, or datetime value. | gt("acc", 0.9) |
ge() | Attribute value is greater than or equal to an int, float, str, or datetime value. | ge("acc", 0.91) |
lt() | Attribute value is less than an int, float, str, or datetime value. | lt("loss", 0.1) |
le() | Attribute value is less than or equal to an int, float, str, or datetime value. | le("loss", 0.11) |
matches_all() | String attribute value matches a regex pattern or all in a list of regexes. | matches_all("optimizer", r"^Ada.*") |
matches_none() | String attribute value doesn't match a regex pattern or any in a of list of regexes. | matches_none("optimizer", [r"momentum", r"RMS"]) |
contains_all() | contains_all("sys/tags", ["best", "v2.1"]) | |
contains_none() | contains_none("tokenizer", "bpe") | |
exists() | Attribute exists in the run or experiment. | exists("metric7") |
Methods for filters
The following methods take already defined Filter objects as arguments:
| Method | Description | Example |
|---|---|---|
negate() | Negate a filter. Equivalent to prepending ~ to the filter. | negate(Filter) |
all() | Concatenation (AND). Equivalent to joining filters with &. | all(Filter1, Filter2) |
any() | Alternation (OR). Equivalent to joining filters with |. | any(Filter1, Filter2) |
Examples
import neptune_fetcher.alpha as npt
from neptune_fetcher.alpha.filters import Filter
Fetch experiments with a validation accuracy no higher than 0.6
low_val_acc = Filter.le("metrics/val/acc", 0.6)
npt.fetch_experiments_table(experiments=low_val_acc)
Various criteria
owned_by_me = Filter.eq("sys/owner", "vidar")
loss_filter = Filter.lt("validation/loss", 0.1)
tag_filter = Filter.contains_none("sys/tags", ["test", "buggy"])
dataset_check = Filter.exists("dataset_version")
my_interesting_experiments = owned_by_me & loss_filter & tag_filter & dataset_check
npt.fetch_experiments_table(experiments=my_interesting_experiments)
tip
You can also pass an exact list directly to the experiments, runs, or attributes argument:
npt.fetch_experiments_table(
experiments=["daring-kittiwake_week-41", "discreet-kittiwake_week-42"]
)
For details, see Construct filters with Fetcher API.