Ethan Tang

Results 15 comments of Ethan Tang

PR is failing due to an issue with Torchmetrics 1.4.1 which added a method under `collections.py` called `metrics_state`, see here: `https://github.com/Lightning-AI/torchmetrics/compare/v1.4.0...v1.4.1#diff-2170499c0e9105ceb0b5ccbb04a53dbeb6015a703ca28914d52608427dad98f1`. This method seems to be not Jit-able at least...

blocked on https://github.com/Lightning-AI/torchmetrics/issues/2812

Going to close this PR to await for the next release of torchmetrics which solves the problem.

I don't think this might be related to rocm itself, as streaming I believe is hardware agnostic... can you confirm that said file does not indeed exist in your environment?

Yes please if you are allowed to, as well as the full launch args

just to confirm, are you only wanting to repeat `shard_path: 's3://object/data/shards/LLaVA_Stage2/scienceqa/' ` twice?

@XiaohanZhangCMU this might be related to your PR https://github.com/mosaicml/streaming/pull/813?

AFAIK, if you are using a remote path for a dataset, it is better to just specify the `remote` and don't specify `local` as we can then do custom logic...

@wizyoung if you do manage to reproduce it, can you open up a new issue so we can track it better? This sounds pretty different than what the author of...

@naston wonder if you got a chance to reproduce this issue over the weekend?