salman

We're working on better surfacing documentation for the finetuning techniques we support. This issue tracks the different recipes we still need to add documentation for. If you're interesting in helping...

Currently, if we specify multiple tasks for the eval recipe, and one of the tasks is a generation task which uses KV-cacheing, then the cache is still enabled for non-generation...

Following the instructions in `/root/torchtune/recipes/configs/llama3/8B_qlora_single_device.yaml` ```bash $ tune download meta-llama/Meta-Llama-3-8B-Instruct --output-dir /tmp/Meta-Llama-3-8B-Instruct --hf-token $ tune run lora_finetune_single_device --config llama3/8B_qlora_single_device checkpointer.checkpoint_dir=/tmp/Meta-Llama-3-8B-Instruct File "/usr/local/lib/python3.11/dist-packages/torchtune/training/checkpointing/_utils.py", line 161, in get_path raise ValueError(f"No file with...

cc @RdoubleA we'll have to re-benchmark here

Fixing docstring linter

1

#### Context What is the purpose of this PR? Is it to - [ ] add a new feature - [ ] fix a bug - [x] update tests and/or...

This should be straightforward. The main issue I see coming up is with compile - similar to how we attempt to compile the reference and policy model in our single...

PPO Performance Improvements

3

Closes #1425 This PR provides various performance improvements to our PPO single device recipe. | Branch | Total training time (hours)* | Peak memory allocated (GB) | | --------------- |...

```[tasklist] ### Tasks - [ ] https://github.com/pytorch/torchtune/issues/2082 - [ ] Full-finetune distributed DPO recipe #1966 - [ ] #1262 - [ ] PPO tutorial/deep dive - [ ] DPO tutorial/deep...