The dataset viewer is not available for this subset.
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 290, 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
return check_status(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 string 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 304, 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.
Pipeline di valutazione Emma
Valuta i modelli pubblici, senza copiarli e senza pubblicare i pesi:
- egomnia/emma-6, generazione, ONNX
- egomnia/emmaguardrails1.0, classificazione
I Job montano i repository in sola lettura. Il report finale è results/report.md.
Job
emma-eval-guardrail, hardwarecpu-basic. Eseguepipeline/run_guardrail.py.emma-eval-emma6, hardwarea10g-large. Eseguepipeline/run_emma6.py.emma-eval-report, hardwarecpu-basic. Eseguepipeline/run_report.pye scrive il report.
Il conto del Job è di chi lo lancia. Un terzo che scarica il modello pubblico non addebita questo account.
Cosa misura il primo giro
Guardrail, sugli insiemi interi: HaSpeeDe2, HateCheck italiano, controlli benigni BEEP e INVALSI, tre stringhe sintetiche di dati personali e tre di calcolo.
Emma-6, con al massimo EMMA_EVAL_LIMIT esempi per task (default 50, seed 42): BEEP, TRACE-it, INVALSI italiano a scelta multipla e binaria, XCOPA italiano, ARC italiano, Belebele italiano, HateCheck italiano, LAMBADA italiano tradotto. Il confronto nel report usa gli id di HateCheck presenti in entrambi i JSON.
Per un giro completo di Emma-6: EMMA_EVAL_LIMIT=0.
Rilancio
Il token resta nell’ambiente e non va scritto nei file.
Il separatore -- tiene -lc nel comando del container.
Il codice viene scaricato nel Job con snapshot_download. Il file del guardrail sul Hub si chiama model.joblib. Emma-6 usa onnxruntime-gpu==1.21.1 e la libreria cuDNN già presente nell’immagine CUDA 12.4.
hf jobs run --name emma-eval-guardrail --flavor cpu-basic --timeout 45m \
--secrets HF_TOKEN --env PYTHONDONTWRITEBYTECODE=1 --env GUARDRAIL_DIR=/opt/guardrail \
-- python:3.12 \
bash -lc 'set -eux; pip install -q scikit-learn==1.6.1 joblib datasets huggingface_hub; python -c "from huggingface_hub import snapshot_download; snapshot_download(\"asappia/emma-eval\", repo_type=\"dataset\", local_dir=\"/opt/eval\"); snapshot_download(\"egomnia/emmaguardrails1.0\", local_dir=\"/opt/guardrail\")"; python /opt/eval/pipeline/run_guardrail.py'
hf jobs run --name emma-eval-emma6 --flavor a10g-large --timeout 2h \
--secrets HF_TOKEN --env PYTHONDONTWRITEBYTECODE=1 --env EMMA_EVAL_LIMIT=50 \
--env LD_LIBRARY_PATH=/opt/conda/lib/python3.11/site-packages/nvidia/cudnn/lib \
-v hf://models/egomnia/emma-6:/model \
-- pytorch/pytorch:2.6.0-cuda12.4-cudnn9-devel \
bash -lc 'set -eux; pip install -q sentencepiece datasets huggingface_hub numpy "onnxruntime-gpu==1.21.1"; python -c "from huggingface_hub import snapshot_download; snapshot_download(\"asappia/emma-eval\", repo_type=\"dataset\", local_dir=\"/opt/eval\")"; python /opt/eval/pipeline/run_emma6.py'
hf jobs run --name emma-eval-report --flavor cpu-basic --timeout 20m \
--secrets HF_TOKEN --env PYTHONDONTWRITEBYTECODE=1 \
-- python:3.12 \
bash -lc 'set -eux; pip install -q huggingface_hub; python -c "from huggingface_hub import snapshot_download; snapshot_download(\"asappia/emma-eval\", repo_type=\"dataset\", local_dir=\"/opt/eval\")"; python /opt/eval/pipeline/run_report.py'
Il Job di reportistica va lanciato quando i due JSON in results/ esistono. Se uno manca, il report lo dice. Un giro con EMMA_EVAL_LIMIT=50 usa cpu-basic per guardrail e report, e a10g-large per Emma-6.
- Downloads last month
- 223