dest stringlengths 30 44 | size int64 9 19.8M | sha256 stringlengths 64 64 |
|---|---|---|
data/session0001-clip000000.audio.pcm.zst | 1,410,648 | 02e42325cebfd61ea749e734e62f95f7d1c9abcf35ff867711eeee2832ae2ec2 |
data/session0001-clip000000.events.jsonl.zst | 9 | 6fb8543883dafcc2ed0934866b8b3f775231f5b0933add08bf502e9e84944cbd |
data/session0001-clip000000.frames.jsonl.zst | 11,920 | dff16a49dc9b5605f2488957f750fae4e516d769180afbb8f62df38c5fc096e4 |
data/session0001-clip000000.manifest.json | 471 | cdf3bc902905d9499e9b8cfe6fc9a74b59221386ad5c92407e8f9a7033719ea1 |
data/session0001-clip000000.meta.json | 5,270 | cd6caf52ddb559054255bcba3866ad66e3ae0c885a723adf6064fb788b00efe4 |
data/session0001-clip000000.video.mp4 | 19,829,447 | d5e5bc113627940f2ab020cb7dbe243d5b73274c05ec8ce871f5d664d338c392 |
data/session0001-clip000001.audio.pcm.zst | 1,550,111 | a049a306e734e2a81eaba81a00e21d7e33d5c13a05aa641bc05de2f7709989cd |
data/session0001-clip000001.events.jsonl.zst | 9 | 6fb8543883dafcc2ed0934866b8b3f775231f5b0933add08bf502e9e84944cbd |
data/session0001-clip000001.frames.jsonl.zst | 11,792 | 441b750c7cd03079cb0078f768744d2e2c0c679e11bdb6bbbb1c3fc9284eb961 |
data/session0001-clip000001.manifest.json | 471 | 028bb045fad45ddd1f9eeaa6cab877e088fa6b619d6ac7f38b8df78034ac1a91 |
data/session0001-clip000001.meta.json | 5,273 | eaa553f13449e837781800c4fed9efb01159e3bab4328d3e7a1305d63d90a5c1 |
data/session0001-clip000001.video.mp4 | 19,632,996 | 65f8866f0ba4b56769a15320ae55d06026e881a36b7040f7c357f214a24f020f |
data/session0001-clip000002.audio.pcm.zst | 1,517,640 | 9a4f424b0d7c6997f0636aa3c86ebfa79d9f663fc88f39d6782ddb3cd4f45453 |
data/session0001-clip000002.events.jsonl.zst | 9 | 6fb8543883dafcc2ed0934866b8b3f775231f5b0933add08bf502e9e84944cbd |
data/session0001-clip000002.frames.jsonl.zst | 11,469 | 4e042ab7a6db3e67713897305dfb34cc107291fb57cb86f8b7ecd1e8ecc4694d |
data/session0001-clip000002.manifest.json | 471 | 106db302b06679dc3489b08f9918d05f55d66f0b5e2bfd9e3939247ca38657b3 |
data/session0001-clip000002.meta.json | 5,274 | a3b51bb28ba0eec6c692e26c0b285c4789e387234136b8d7e2215911380caea1 |
data/session0001-clip000002.video.mp4 | 16,050,370 | e2e5d18f4d4020951c30168028fce449d158f463c89b8eeecb5c163efea4d196 |
data/session0001-clip000003.audio.pcm.zst | 1,191,998 | 6a01d0f4d53b15f3a56fa47d1b37f1ba0b1ce70317a653c302c14c754c359069 |
data/session0001-clip000003.events.jsonl.zst | 9 | 6fb8543883dafcc2ed0934866b8b3f775231f5b0933add08bf502e9e84944cbd |
data/session0001-clip000003.frames.jsonl.zst | 11,295 | 4c4bd59f7b39b8a2e1fc0748aba1631233a3efefc584e4867b90c89575ad79d9 |
data/session0001-clip000003.manifest.json | 471 | 528a740a5fafa658c97ea70cbc38785fb1aab20e31a33b68aed1f7a5a562ba2c |
data/session0001-clip000003.meta.json | 5,274 | 4de0e35c7a2ceaa9e237f8886fd158ead91f8df18196a057c61952b4294c18d8 |
data/session0001-clip000003.video.mp4 | 14,552,700 | e7827c62eedd8e0b5ca72168228cf7f5e091c2868f4f758c6cb39c7a720c3cc7 |
data/session0001-clip000004.audio.pcm.zst | 1,634,676 | a005a55e79bc2e4acc8c69512b480e63de87cb65094797d7c2a2c51b82b5e1d5 |
data/session0001-clip000004.events.jsonl.zst | 9 | 6fb8543883dafcc2ed0934866b8b3f775231f5b0933add08bf502e9e84944cbd |
data/session0001-clip000004.frames.jsonl.zst | 11,207 | e72da168a0bcd747ab269545948e0e113348348295825c2c05546a66c6d2e369 |
data/session0001-clip000004.manifest.json | 471 | 10f8eb62d674ac9ed61fad2d9d4afc4fbee72ce44059367c9b79293819a8049f |
data/session0001-clip000004.meta.json | 5,274 | 92259423a2582c26efec4751e2e5f99c614c6b6bf2380c62d472a99a962ead4b |
data/session0001-clip000004.video.mp4 | 19,216,504 | 81ac68d47e9fd3fac03fa45344cd4c401d7ed486f754219f4acb384642ed8bf7 |
data/session0001-clip000005.audio.pcm.zst | 1,491,939 | 983c5d3224269cba9b66ee6db3eb2f218231dac259f86749ba7baf10ad34e30f |
data/session0001-clip000005.events.jsonl.zst | 9 | 6fb8543883dafcc2ed0934866b8b3f775231f5b0933add08bf502e9e84944cbd |
data/session0001-clip000005.frames.jsonl.zst | 11,318 | 4fc621189f3d669e67760eb4db0c0e5501be0e2492700835692ab0100bcf010b |
data/session0001-clip000005.manifest.json | 471 | 01c3efc78e612897281e56727b23309a5a30839c2b61add82abed9696da27004 |
data/session0001-clip000005.meta.json | 5,274 | d6d93c6465c607617dd9047df12096229e89c3e6c7f4b32ef860d5583f2aca74 |
data/session0001-clip000005.video.mp4 | 14,849,871 | 9c565e1f7968f7b95818feb5cc9eccaff1248e9f7ae03b5dd39b91e755ef7b8c |
data/session0001-clip000006.audio.pcm.zst | 399,561 | 15d9e9a6a35517ac1291112f8adc9da4b7cceb97eb6692ae6599603dc05954c0 |
data/session0001-clip000006.events.jsonl.zst | 1,416 | 732bb5e77b211ca9807803510b5c7026ccfa21419fa204cea2fa823ced07741f |
data/session0001-clip000006.frames.jsonl.zst | 7,124 | 4f95046f103d3cfcbb9a5d86482e40ddc33c1598b917e8980f1764fe87383e64 |
data/session0001-clip000006.manifest.json | 471 | ee8f5308f48eda3be9d495cdbbc052b2f1675d66b0b031ba8c53adba502b1581 |
data/session0001-clip000006.meta.json | 5,278 | 39207fca6c9c6a3793f7fc297c5f1f3bb31d717701971f6ce7fe40e4c2aa1fcf |
data/session0001-clip000006.video.mp4 | 6,973,879 | 7c857ab5a7a14fe2423d084567422c47b9c961b3278673a4d10b08e1227a3320 |
qc/session0001.acceptance.json | 3,341 | 024ef6eb7992e981256f89f36745fb3864c1c382e40620aa458e5c2e0b9d961d |
Diogenes WUCHANG Gameplay Archive v01
Public project-closeout capture archive. The seven clips preserve one complete recorded session of WUCHANG: Fallen Feathers, totaling 65.2 seconds, 3,912 frames, 1920×1080 at 60 fps. No account or access request is required to download. This is an inspection and reproducibility archive, including the capture's recorded failures. It is not presented as certified training data.
The collection has 43 original source files totaling 120,423,490 bytes. It contains seven video files, seven process-loopback audio files, seven frame-input traces, seven event traces, seven clip metadata files, seven digest manifests, and one session acceptance/QC report. No simulator/game binary, model weights, account records, or microphone track is included.
Layout and fields
Each tar shard groups six modalities by the same WebDataset-style sample key:
data/session0001-clip000000, through clip000006.
| Suffix | Content |
|---|---|
.video.mp4 |
Original H.264 gameplay video. |
.audio.pcm.zst |
Zstandard-compressed, headerless signed 16-bit little-endian PCM, 48 kHz, stereo; process loopback, not a microphone recording. |
.frames.jsonl.zst |
One JSON row per recorded frame: frame/session indexes, timestamps, duplicate/drop markers, keys/buttons/actions and pointer movement. |
.events.jsonl.zst |
Timestamped input events; six of the seven event files are empty. |
.meta.json |
dm-local-dev-recording-2 capture schema, game/catalog identifiers, encoder and audio parameters, health counters, per-modality hashes and historical eligibility flags. |
.manifest.json |
Original per-clip SHA-256 map using the original logical modality names such as video.mp4. These names correspond to sample field suffixes. |
qc/session0001.acceptance.json is the full recorded session/QC report.
MANIFEST.jsonl inside each shard and the repository's MANIFEST.jsonl provide
public member paths, byte sizes and SHA-256 hashes. They are index metadata, not
additional gameplay samples. Filter to members under data/ when iterating clips.
A dependency-free Python example to inspect one sample without extracting paths:
import json
import tarfile
with tarfile.open("downloaded-shard.tar", "r:") as archive:
samples = {}
for member in archive:
if member.isfile() and member.name.startswith("data/"):
key, field = member.name.split(".", 1)
samples.setdefault(key, {})[field] = member.name
first = samples[sorted(samples)[0]]
metadata = json.load(archive.extractfile(first["meta.json"]))
print(metadata["n_frames"], metadata["fps_num"], metadata["fps_den"])
To read a .zst modality, decompress its bytes with Zstandard before parsing
JSON Lines or PCM. frames.jsonl.zst frame indexes align with the clip video;
session_frame preserves the continuous session frame index. Input labels and
QC values are preserved as recorded and must not be assumed correct.
Known limitations and retained provenance
All seven clips retain non_trainable: true, production_certified: false, and
catalog.mode: dev_non_trainable. The historical acceptance report failed QC:
98.1% of frames have no input, duplicate frames are 21.37%, only one distinct key
is recorded, and several expected action channels are missing. Six event traces
are empty. These are retained research/debugging observations, not remediated
labels or claims of training suitability. There is no curated train/test split.
Structured metadata was read in full and scanned for account, email, token, physical-device identifiers and private filesystem paths. Generic capture-device profile names and opaque session/target identifiers remain for reproducibility. Representative video frames were inspected; this is not exhaustive scene-level annotation. Audio provenance is the recorded game-process loopback setting.
All file bytes are preserved. Only public archive member names are normalized; private bucket paths and operator-machine paths are omitted from the public index. The source file SHA-256 values are checked during transfer, and the resulting tar's byte count and SHA-256 are checked against Hugging Face LFS at a pinned commit before any source copy is eligible for deletion.
Capture/schema tooling: https://github.com/CuddlerAI/Diogenes-Archive.
See NOTICE.md for third-party rights and historical usage constraints.
- Downloads last month
- 81