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 84, 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.
tokenmaxxing training tasks
Harbor task directories used to train Qwen3.5-9B agents with the tvgpo/ trainer
(training-tokenmaxxing repo). One tarball per split; each task is a directory with
instruction.md, task.toml, environment/ (the in-trial check_solution.py and its
tests) and tests/ (the full graded suite).
| file | tasks | contents |
|---|---|---|
leetcode/train1k_python.tar.gz |
1000 | LeetCode, Python, stdin/stdout. 248 hand-reviewed invalid hidden tests dropped (adapters/leetcodedataset/invalid_tests.json). The checker runs half the tests. |
codeforces/train_python_half_1k_exact2.tar.gz |
1000 | Codeforces, Python. The first 1000 tasks of the seed-17 shuffle that exact-match grading can judge (no open-r1 generated checker, no "any valid answer" problems). Graded suite: official tests plus at most 5 validated generated tests. The checker runs half the tests plus the task's largest generated test, and truncates mismatch output and stderr. |
Grading always uses the full suite in tests/config.json. The checker's prompt says so.
Use
hf download nuprl-staging/tokenmaxxing-training --repo-type dataset --local-dir hf_tasks
mkdir -p datasets/leetcodebench datasets/codeforces
tar -xzf hf_tasks/leetcode/train1k_python.tar.gz -C datasets/leetcodebench
tar -xzf hf_tasks/codeforces/train_python_half_1k_exact2.tar.gz -C datasets/codeforces
Every task.toml names its container image by absolute path:
docker_image = "/scratch/wu.zixua/sif_cache/lcb-multi.sif". Point it at your own
build of lcb-multi.sif (environments/lcb_multi/build.sbatch):
sed -i "s|/scratch/wu.zixua/sif_cache/lcb-multi.sif|/scratch/$USER/sif_cache/lcb-multi.sif|" \
datasets/leetcodebench/train1k_python/*/task.toml datasets/codeforces/train_python_half_1k_exact2/*/task.toml
Sources
Problems and tests come from newfacade/LeetCodeDataset and open-r1/codeforces
(plus its generated tests), under those datasets' terms.
- Downloads last month
- -