Joe Cummings

Update hitting a blocker - can't get this running on M1 Mac with 32GB RAM, PyTorch 2.4.0: I hit the following error: ``` /AppleInternal/Library/BuildRoots/91a344b1-f985-11ee-b563-fe8bc7981bff/Library/Caches/com.apple.xbs/Sources/MetalPerformanceShaders/MPSCore/Types/MPSNDArray.mm:869: failed assertion `[MPSNDArray, initWithBuffer:descriptor:] Error: buffer...

With PyTorch nightlies, I manage to run 4 steps before loss NaNs and buffer issue comes back: ``` 1|4|Loss: nan: 0%| | 4/52002 [00:26

What's your implementation plan @felipemello1 ? I'd prefer not to add a tag "good first issue" without really clear instructions.

I agree this needs to be better thought out how we structure our imports here. However another stopgap we could do is the following: ``` try: import torchvision except ImportError:...

> wouldnt you have to add it in all files that import torchvision? If you want a quick solution, I think that moving these functions from transforms/init --> transforms/vision_transforms/init will...

Hi @dpalmasan! Looks like you might be inheriting from the wrong base class for your template: you should try `InstructTemplate`, not `InstructDataset` like so: ```python class TruthfulQATemplate(InstructTemplate): template = "Instruction:\n{instruction}\n\nInput:\n{input}\n\nResponse:...

Yep, we've definitely got to make our docs more clear - your input is much appreciated! cc @RdoubleA

Closing this in favor of tracking at a high-level which models we want to add, not just specifically coding models.

cc @msaroufim

@felipemello1 Would you mind just also double checking our version guards for AO? We'll need to be extra careful around this now that we're relaxing this pin. CI will only...