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/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/harbor/harbor.py", line 171, in _split_generators
                  raise DataFilesNotFoundError("No task.toml or instruction.md files found")
              datasets.exceptions.DataFilesNotFoundError: No task.toml or instruction.md files found
              
              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.

APSP Lite B200 cu128

Research/runtime distribution. Two companion transport files, 22.992 GB each, plus a separate4.572GB dependency runtime. The two parts are one losslessly compressed archive, not two independently extractable model bundles. Weights, model/scorer/tokenizer configs, common encoder, prepared prompt/task manifests, raw prompt-source metadata, source and provenance are included. No full generated training bank is present.

Use the exact commit returned by publication, never main, in download URLs. Company-node transport follows DSH PROXY_DOWNLOAD_GUIDE.md section6: curl, explicit proxy,12 simultaneous byte ranges per part, correct last-range size, bounded temporary downloads and byte resume. No Modal SDK is required on the target nodes.

On instanceA download part1, instanceB download part2. Each needs bash/curl/python3 and its existing proxy environment. Put outputs on shared persistent storage or transfer both verified parts to the same join node. Download and verify download_part_hf.sh and COMPANION_PARTS.json using their pinned hashes in APSP_RELEASE.json before executing. Then invoke bash download_part_hf.sh COMPANION_PARTS.json 1 DEST codemaivanngu/apsp-lite-b200-cu128 PINNED_REV onA, and the same command with2 onB. APSP_DOWNLOAD_THREADS defaults12. No GPU is used by download/join.

Check COMPANION_PARTS.SHA256SUMS, then explicitly concatenate part01 followed by part02 and pipe through gzip -dc to recover apsp-lite-companion-v1.tar. The recovered tar SHA256 must be d11d1ecf591ae59e6e63d80d767e74c3659ded51ec93792139aa34eb667b3b96. Use restore_companion.py restore --archive ... --target ... --sha256 ... and source the generated activate_companion.sh. Restore runtime.tar.gz separately with ensure_runtime.py and runtime-manifest.json. Reserve adequate persistent disk space for ranges, parts, tar and restored payload (about250GB if retained together).

Engineering checks: companion full-file SHA readback PASS; both-part lossless roundtrip PASS; separate companion restore and offline CPU loader/encoder forward PASS. Original three-profile one-prompt×8seed B200 functional qualification and Ubuntu22.04 restore PASS. These do not qualify a company node or complete the dataset. SDXL50 cross-container PNG drift remains unresolved; Lightning/FLUX matched8/8 in that bounded check. Full110400-candidate collection,50-prompt drift study, independent backup and manual prompt QA remain pending. No controller result is claimed.

See NOTICE, UPSTREAM_COMPONENTS.json and LICENSES/ before use. FLUX.1-dev remains under its upstream non-commercial license; public hosting does not remove component restrictions.

Downloads last month
79