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/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/webdataset/webdataset.py", line 80, in _split_generators
raise ValueError(
...<2 lines>...
)
ValueError: The TAR archives of the dataset should be in WebDataset format, but the files in the archive don't share the same prefix or the same types.
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.
TMAX
Release complete. All 10,040 image archives are available and verified. Pin this dataset's completed commit when downloading assets.
Prepared task assets and Linux amd64 container images for the
filter14-envfixed-4e9c5217 release:
| Content | Count |
|---|---|
| Tasks | 10,109 |
| Required image archives | 10,040 |
| Original images | 9,275 |
| Images with environment dependency repairs | 765 |
The image archives total 1,594,052,809,390 bytes, approximately 1.45 TiB. Only images required by this prepared task set are included.
Source and processing
The upstream dataset is
allenai/tmax-15k-open-instruct
at revision 7b090eca98bf351356bc1c64290c5c4a09f2f98c.
This is a filtered and environment-repaired release derived from that dataset.
It is not the complete upstream 15K task set. The final processing stage
retained 10,109 of its 10,254 input tasks, excluding 145 tasks and updating
references to 765 repaired images. Earlier filtering is already incorporated
in that stage's input.
The task archive, task configuration checksums, processing reports, and repaired image references are preserved from the prepared release. Downloaders do not need to repeat filtering or image repair. Task limits and tests in the prepared archive are retained.
Download tasks and image metadata
Pin the completed release's Hugging Face commit for reproducibility.
hf download daixuancheng/tmax \
release-status.json READY.json \
releases/filter14-envfixed-4e9c5217/tasks.tar.gz \
releases/filter14-envfixed-4e9c5217/tasks-manifest.json \
releases/filter14-envfixed-4e9c5217/images/manifest.json \
--repo-type dataset --local-dir ./tmax
The task manifest maps each task to its image reference. The image manifest records each archive's Hugging Face path, size, SHA-256, image ID, architecture, and archive tags. A runner can download the task set and manifests first, then download and import only the images required by its next tasks. All mappings are supplied in this repository; consumers do not need to construct an index.
For each required image, download its hf_path, verify its SHA-256, and import
the archive with docker load --input <archive> or podman load --input <archive>.
Some archives carry a localhost/ tag prefix. Use the recorded image ID to
assign the exact task image reference with docker tag <image-id> <image> or
podman tag <image-id> <image>. Start containers from the imported image with
registry pulling disabled.
Images are downloaded as files from this Hugging Face dataset and imported
locally. This image transfer path requires no Docker Hub access. Hugging Face
dataset URLs cannot be used directly with docker pull.
Contents and runtime requirements
tasks.tar.gzcontains the final task instructions, task configuration, and tests.tasks-manifest.jsonrecords task-to-image mappings and task configuration hashes.images/manifest.jsonrecords the independently downloadable image archives.provenance/records the final filtering and environment repair stage.- The root release inventory and
SHA256SUMSallow file integrity verification.
A compatible task runner and a Linux amd64 Docker or Podman environment are required. This repository supplies task assets and images; it does not include a complete host training environment or guarantee that every task can run without additional runtime dependencies or network access.
Attribution and licensing
The upstream dataset identifies its data license as Open Data Commons
Attribution License 1.0 (odc-by). See LICENSE_NOTICE.md.
Container images retain their original software licenses and notices.
These tasks include tests and evaluation information. Results on tasks used for training must be reported as training-set results.
- Downloads last month
- 141