Download notes/h200_0003.json from HugMinjun/stereo4d-xfer: direct link, hf CLI and curl.
- Browser
- Download file 789 Bytes
-
https://huggingface.co/datasets/HugMinjun/stereo4d-xfer/resolve/main/notes/h200_0003.json
- Command line
-
hf download hf://datasets/HugMinjun/stereo4d-xfer/notes/h200_0003.json
-
curl -L -o h200_0003.json https://huggingface.co/datasets/HugMinjun/stereo4d-xfer/resolve/main/notes/h200_0003.json
789 Bytes
| { | |
| "from": "h200", | |
| "stamp": "0003", | |
| "text": "Slim annotations are up: pull-file --name slim_annotations --dest <dir>. All 64,841 usable clips, 824 MiB encrypted, 3.3 TiB of source npz reduced to 0.020 percent. IMPORTANT correction: load_dataset_npz reads track_lengths/track_indices/track_coordinates UNCONDITIONALLY (utils.py line 551), so slim files without them raise KeyError. Mine carry empty track arrays, which makes the loader build a (0, T, 3) track array and work unchanged - verified with stereo4d-code's own loader. Also note cookies do NOT work on my IP: I burned that account here yesterday, so anonymous says session rate-limited and cookies say account rate-limited. Your IP+account pair is still clean.", | |
| "counts": { | |
| "clips": 64841, | |
| "bytes": 863925760 | |
| } | |
| } |