Achal Dave

Zip file shows up as empty in Windows

1

For some reason, the zip file shows up as empty in Windows Explorer; if I open it in 7zip, I can see the `.` directory, but I can't in windows...

Projection formulas are (almost definitely) incorrect

2

I've been trying to use these projection formulas: https://github.com/jysung100/activity_detection/blob/master/FeatureExtractor/featuresRGBD.cpp#L26 ``` cpp // Given (x,y,z) coordinates, converts that point into its x pixel number in the 2D image. int xPixelFromCoords(double x,...

I was trying to parse a large file, and was very surprised to find that underscore was taking so long to return results on the file. It turns out underscore...

Diverged from tomorrow-theme repo

3

It seems this repo has diverged from the vim folder in the tomorrow-theme repo: https://github.com/chriskempson/tomorrow-theme...

Compatibility with :share()'d layers

3

nnlr is super useful, thanks for the package! I was able to use it with a few networks, but I'm having trouble with networks where I'm sharing weights across layers....

For medium-to-large models, if the user doesn't have enough disk space (or, more commonly, has accidentally specified a path on a volume with not enough disk space), we train for...

Figure out why AdamW + gradient accumulation leads to different results for test case

6

In https://github.com/mlfoundations/open_lm/pull/125, we had to switch our gradient accumulation tests from SGD to AdamW to make gradient accumulation tests pass. It's unclear why this is the case; anecdotally, when training...

Minimize how often we load args.resume

1

Currently, we load args.resume potentially up to 3 times. This can be pretty slow for big models, and we should avoid re-loading it in these spots: https://github.com/mlfoundations/open_lm/blob/97d0a4ac592e300c7a1688c4a48584b027b45af8/open_lm/main.py#L110-L156

We should use SHARDED_STATE_DICT when loading/saving checkpoints to avoid loading the entire model in CPU memory, similar to https://github.com/pytorch/pytorch/blob/main/torch/distributed/checkpoint/examples/fsdp_checkpoint_example.py.