On latest splatfacto viewer not coming up

[fasteinke]

Just tried latest nerfstudio in the Google Colab demo notebook; a nerfacto run worked fine, with viewer updating. Trying splatfacto, the training went through to completion, but the viser share URL viewer never showed anything; the loading indicator of the browser kept cycling. A splatfacto run done some days ago, using the previous version of gsplat, worked as expected.

[sumanttyagi]

what was your command while training or using viewer ?

[sumanttyagi]

Also what i have observed if we click on the gui to share again it fail with the same error mentioned here - #3225 that's why it keep on cycling the same traceback is generated

[fasteinke]

I'm using the demo.ipynb as per the repository, the isolation preset scene, and the training command is pasted into the Xterm window as:

ns-train splatfacto --viewer.websocket-port 7007 --viewer.make-share-url True nerfstudio-data --data data/nerfstudio/{scene} --downscale-factor 4

Xterm output:

───────────
[05:24:19] Saving config to: outputs/unnamed/splatfacto/2024-06-16_052419/config.yml experiment_config.py:136
Saving checkpoints to: outputs/unnamed/splatfacto/2024-06-16_052419/nerfstudio_models trainer.py:137
load_3D_points is true, but the dataset was processed with an outdated ns-process-data that didn't convert
colmap points to .ply! Update the colmap dataset automatically? [y/n]: y
Downloading: "https://download.pytorch.org/models/alexnet-owt-7be5be79.pth" to /root/.cache/torch/hub/checkpoints/alexnet-owt-7be5be79.pth
100%|██████████████████████████████████████████████████████████████████████| 233M/233M [00:02<00:00, 113MB/s]
╭─────────────── viser ───────────────╮
│ ╷ │
│ HTTP │ http://0.0.0.0:7007 │
│ Websocket │ ws://0.0.0.0:7007 │
│ ╵ │
╰─────────────────────────────────────╯
(viser) Share URL requested!
/usr/lib/python3.10/multiprocessing/popen_fork.py:66: RuntimeWarning: os.fork() was called. os.fork() is incompatible with multithreaded code, and JAX is multithreaded, so this will likely lead to a deadlock.
self.pid = os.fork()
(viser) Generated share URL (expires in 24 hours, max 32 clients):
https://fresh-unevaluated.share.viser.studio
[05:24:39] Caching / undistorting eval images full_images_datamanager.py:230
[NOTE] Not running eval iterations since only viewer is enabled.
Use --vis {wandb, tensorboard, viewer+wandb, viewer+tensorboard} to run with eval.
No Nerfstudio checkpoint to load, so training from scratch.
Disabled comet/tensorboard/wandb event writers
[05:24:41] Caching / undistorting train images full_images_datamanager.py:230
/usr/lib/python3.10/multiprocessing/popen_fork.py:66: RuntimeWarning: os.fork() was called. os.fork() is incompatible with multithreaded code, and JAX is multithreaded, so this will likely lead to a deadlock.
self.pid = os.fork()
[05:29:53] Printing max of 10 lines. Set flag --logging.local-writer.max-log-size=0 to disable line writer.py:448
wrapping.

The URL fresh-unevaluated.share.viser.studio gives me a loading status, with no output.

Note again, the previous splatfacto version worked fine, with correct viewer interaction

[efyphil]

Same