Collin McCarthy

I stumbled upon the root of the problem here. I still stand by my solution above but the issue is stemming from `Visualizer.get_image()` ```python @master_only def get_image(self) -> np.ndarray: """Get...

I just run `unset LD_LIBRARY_PATH` in my bash script which launches the job, or comment out where I set this env var in my ~/.bashrc. But depending on how your...

Wow, I'm so sorry I thought I responded to this, but clearly I did not. @hiyyg I am using a registry even with the python-style config, so I'm just looking...

I ended up implementing this in a similar way to the "${key}" evaluation above. I can use something like the following the config file: ```python num_classes = 80 num_det_layers =...

Thank you for following up. I understand the string syntax could be too complex for most users. The other issue is that it's easy to make a mistake, and debugging...

I forgot this is how I did it before. 3. When saving a resume checkpoint before a validation iteration, save metadata like `needs_val=True`, then after validation save a new checkpoint...

I was able to get this working with approach 3 above but it required a few different changes: - New PreTrainValHook to check if we need to run validation before...

The issue here for me was that the rank 0 GPU was not returning from the forward pass, and for whatever reason the broadcast wasn't blocking. So the returned `tmpdir`...

Hey @luisbergua, sorry for the delay. Setting `WANDB_CONFIG_DIR` to a different directory did not work. However if I noticed that when my init args use `resume=False` instead of `resume='auto'` the...

Also interested in understanding this better, as I'm trying to combine a few things in VILA and LLaVA-NeXt and this makes me concerned that something might break in an unexpected...