Polifemus/process-fm
Tabular Classification • Updated
Dataset Viewer
The dataset viewer is not available for this subset.
Cannot get the split names for the config 'default' of the dataset.
Exception: SplitsNotFoundError
Message: The split names could not be parsed from the dataset config.
Traceback: Traceback (most recent call last):
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 291, in _generate_tables
pa_table = paj.read_json(
io.BytesIO(batch), read_options=paj.ReadOptions(block_size=block_size)
)
File "pyarrow/_json.pyx", line 364, in pyarrow._json.read_json
File "pyarrow/error.pxi", line 155, in pyarrow.lib.pyarrow_internal_check_status
File "pyarrow/error.pxi", line 92, in pyarrow.lib.check_status
raise convert_status(status)
pyarrow.lib.ArrowInvalid: JSON parse error: Column() changed from object to number in row 0
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 286, in get_dataset_config_info
for split_generator in builder._split_generators(
~~~~~~~~~~~~~~~~~~~~~~~~~^
StreamingDownloadManager(base_path=builder.base_path, download_config=download_config)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
)
^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 101, in _split_generators
pa_table = next(iter(self._generate_tables(**splits[0].gen_kwargs, allow_full_read=False)))[1]
~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 306, in _generate_tables
batch = json_encode_fields_in_json_lines(original_batch, json_field_paths)
File "/usr/local/lib/python3.14/site-packages/datasets/utils/json.py", line 111, in json_encode_fields_in_json_lines
examples = [ujson_loads(line) for line in original_batch.splitlines()]
~~~~~~~~~~~^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/utils/json.py", line 20, in ujson_loads
return pd.io.json.ujson_loads(*args, **kwargs)
~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^
ValueError: Expected object or value
The above exception was the direct cause of the following exception:
Traceback (most recent call last):
File "/src/services/worker/src/worker/job_runners/config/split_names.py", line 68, in compute_split_names_from_streaming_response
for split in get_dataset_split_names(
~~~~~~~~~~~~~~~~~~~~~~~^
path=dataset,
^^^^^^^^^^^^^
config_name=config,
^^^^^^^^^^^^^^^^^^^
token=hf_token,
^^^^^^^^^^^^^^^
)
^
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 340, in get_dataset_split_names
info = get_dataset_config_info(
path,
...<6 lines>...
**config_kwargs,
)
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 291, in get_dataset_config_info
raise SplitsNotFoundError("The split names could not be parsed from the dataset config.") from err
datasets.inspect.SplitsNotFoundError: The split names could not be parsed from the dataset config.Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
Event Log Mixer — synthetic process-mining corpus
300 synthetic event logs generated by sampling case traces from first-order directly-follows graphs (DFGs) of real process-mining logs. The corpus was built as the synthetic half of the pretraining data for the ProcessGraph-FM project (sequence and graph foundation models for predictive process monitoring).
The 18 real source logs are not redistributed here. Only the generated synthetic traces, the generator, the recipes and the global vocabulary are published.
Composition
| family | logs | cases | events | max activities/log | max resources/log | source pool |
|---|---|---|---|---|---|---|
gmx2 |
30 | 76,800 | 914,327 | 408 | 30 | 17 real logs (pool17) |
gmx2_14 |
10 | 19,000 | 666,712 | 425 | 29 | BPIC 2015 municipalities 2–5 (pool15) |
gmx3 |
260 | 839,410 | 10,730,872 | 468 | 30 | 17 real logs (pool17) |
| total | 300 | 935,210 | 12,311,911 | — | — | — |
Case-level temporal split (inherited from the generator/prep pipeline): 654,517 train / 140,158 val / 140,535 test cases.
Files
| path | content |
|---|---|
npz/*.npz |
300 packed logs (one NPZ per synthetic log) |
global_vocab.json |
global activity/resource vocabulary and log-id mapping |
recipes.json |
exact per-log generation parameters (source pool, k, seed, n_cases) |
stats.json |
per-log and aggregate statistics |
synth_mixer.py |
generator (samples from first-order DFGs of real logs) |
prep.py |
parquet/XES → project NPZ/prefix preparation |
checksums.sha256 |
SHA-256 of every file |
LICENSE |
MIT |
NPZ schema
Each npz/<name>.npz contains the packed event stream of one log:
| key | dtype / shape | meaning |
|---|---|---|
gact |
int32 [n_events] |
global activity id; real activities are 4..776, UNK=1, PAD/BOS/EOS=0/2/3 |
gres |
int32 [n_events] |
global resource id; real resources are 2..2739, UNK=1 |
num |
float32 [n_events, 6] |
log1p(dt_prev)/20, log1p(dt_start)/20, sin(2π·hour/24), cos(2π·hour/24), weekday/6, log1p(amount)/14 |
log_id |
uint16 [n_events] |
global log id (see global_vocab.json → log_ids) |
trace_off |
int64 [n_cases+1] |
case offsets; case i is gact[trace_off[i]:trace_off[i+1]] |
split_of_trace |
int8 [n_cases] |
case split: 0=train, 1=val, 2=test |
Activity and resource names are in global_vocab.json. The mapping used by the generator is the same as in the project code: activities start at id 4, resources at id 2, with id 1 as UNK.
Recipes
recipes.json contains the exact generator arguments. The defaults are max_len=300, dt_range=[30, 21600] seconds, n_res_range=[2, 30], span_days=365, prep --lifecycle none. Examples:
# a gmx2 log: k sources sampled from pool17
python synth_mixer.py --name gmx2_01 --source-pool bpi2013_incidents bpi2013_closed \
bpi2013_open bpi2020_intl bpi2020_dom bpi2020_prepaid bpi2020_rfp sepsis helpdesk \
bpi2019 bpi2018 bpi2014 bpi2015_2 bpi2015_3 bpi2015_4 bpi2015_5 bpi2017 \
--k 1 --seed 7001 --n-cases 820
python prep.py --name gmx2_01 --input data/gmx2_01.parquet --lifecycle none
The generator reads the real logs as out/<log>_prep.npz (created by prep.py from the original XES/parquet files). Those real logs are not included and must be obtained from their respective publishers (4TU.ResearchData, see the model card references).
Intended use
- Pretraining and ablation studies for event-log foundation models (next-activity prediction, next-resource prediction, synthetic-to-real transfer).
- Generating additional logs with different pools/seeds via
synth_mixer.py.
Limitations: the traces are synthetic and inherit only first-order DFG statistics plus a few calibrated distributions; they are not a substitute for real logs when measuring deployment accuracy. Activity/resource names come from the public source logs; case ids and timestamps are synthetic. No real case data is included.
Licensing
The synthetic logs, recipes, statistics and generator are released under MIT. The source logs keep their original licenses (mostly 4TU General Terms of Use; BPIC 2018 is CC0). If you use this corpus, cite it together with the source datasets listed in the model repository Polifemus/prefix-next-activity.
Citation
@misc{event-log-mixer,
title = {Event Log Mixer: A Synthetic Process-Mining Corpus from Directly-Follows Graphs},
author = {Polidori, Filippo},
year = {2026},
howpublished = {\url{https://huggingface.co/datasets/Polifemus/event-log-mixer}}
}
- Downloads last month
- 130