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.
Diogenes Cyberpunk 2077 Raw Depth Experiments Archive v01
The project has closed. This owner-authorized archive preserves 82,247 original .d32 frame files and 106 complete JSONL indices / 681,416,337,178 source bytes from historical development attempts. Transfer is in progress. The manifest lists intended members, not proof that all bytes have already arrived.
Format and original limitations
The index schema is dm-cp2077-raw-depth-frame-index/v1, with original present sequence/QPC timing, arm/capture/resource labels, dimensions, row pitch, byte length, relative frame name, readback lag and RGB pairing description. Indexed pixel format is DXGI_FORMAT_D32_FLOAT_LE: little-endian float32, row-major. There are 82,070 stored 1920×1080 frames and 177 stored 1280×720 frames. Each file's exact byte count matches its indexed dimensions × 4. These are native diagnostic depth buffers, not calibrated linear-metre geometry.
Branches named raw-depth, raw-depth-all-presents and raw-depth-selected preserve original selection stages. Their frames can overlap; the count is stored files, not unique observations or independent trajectories. Every indexed frame is present in the reviewed source inventory. An index resolves its file name inside the corresponding branch directory, not directly beside the index. The original rgb_pairing value is same_present_seq_gpu_watermark_submitted; this records capture-side evidence, not a new end-to-end pairing certification.
A historical failed smoke attempt is retained. The projectionfix-short-smoke-20260812 branch contains 177 lower-resolution frames. A sample of 4,096 float slots from its middle frame contains 1,174 finite values outside [0,1] and 132 non-finite values. It must not be treated as validated depth. No values were replaced or repaired. The other 105 sampled branches contained finite values within [0,1], which is only a representative format check and does not establish usable geometry throughout each recording.
Related RGB/audio, camera observations, broader diagnostics, package ZIPs and other index copies are outside this initial raw-frame manifest and should not be assumed present. Historical training/production gates are not upgraded by public access. No absent eligibility field was filled with an invented boolean. See NOTICE.md for rights limits.
Shards and review
TAR shards under archive/ keep at most 100 complete consecutive frames from one branch; its original index is in the first shard of that branch. No individual frame is split or repacked into a new large ZIP. Each tar ends with a per-member SHA-256/size MANIFEST.jsonl. Collect all branch shards to reconstruct the original paths and global index. The complete reviewed collection is published and verified in 877 TAR shards. MANIFEST.jsonl records the actual SHA-256 and byte size of every member; ARCHIVE_INDEX.json links each immutable verification revision and TAR hash. All corresponding R2 source objects were removed only after pinned Hub LFS metadata and anonymous-download checks passed.
All 106 indices were read in full, their structure and strings reviewed, and their paths matched one-to-one against 82,247 stored frames. A 16 KiB sample from the middle frame of every indexed branch was decoded for the format/quality observations above. The review preserves failed experiments and does not imply exhaustive geometric validation or new training authorization.
Read a frame with numpy.frombuffer(bytes, dtype='<f4').reshape(height,width) using its own index dimensions. Interpret projection/calibration and the known failed attempt before use.
Capture source: Diogenes-Archive.
Retired depth and capture packages
Additional original ZIP packages preserve historical experiments and delivery structures. The supplement records original failures, incomplete inputs, and qualification limits separately from the original collection.
- Downloads last month
- 283