Massimo Caccia

RL then Multi-task and task-incremental RL

+1 on all the examples that the user should be able to launch from the command-line. +1 for Tim's reframing. Seems pretty intuitive. Not sure what else to add, sorry....

> Once we fine-tune the model, it would be great if there's a command line option available to convert the pt weights to hf and upload them to Hugging Face....

> https://gist.github.com/monk1337/925a5a44c431ed1f1d3927141f31b6d2 I tried this w/ for LLAMA-3-8b and got the following error: ``` File "/home/toolkit/ui-copilot/finetuning/utils/convert_llama_weights_to_hf2.py", line 447, in main write_model( File "/home/toolkit/ui-copilot/finetuning/utils/convert_llama_weights_to_hf2.py", line 195, in write_model f"model.layers.{layer_i}.input_layernorm.weight": loaded[0][ KeyError:...

> As I commented above, the FullModelCheckpointer does this conversion for you. But seems like you're still running into issues? Yes, it's still unclear to me how to use the...

hi @leo-young To debug, i.e. directly launch python code instead of using the CLI (which won't hit break points) you can use the following code I wrote (wIll need some...

could be that I'm using Vscode's visual debugger!

Hi @webcsm . Sorry for the late reply. I added the dataset [here](https://drive.google.com/drive/folders/13Yf_RYE3hX-y43OZXLhJrjPHXt8BSuDr?usp=sharing). I don't think it's a viable long-term solution, as Google will ask me permission every time someone...