SmolLM2 minor issues

Hello,

On a fresh conda env install following the README, I get two minor annoyances when trying to run SFT on SmolLM2 with

```
ACCELERATE_LOG_LEVEL=info accelerate launch --config_file recipes/accelerate_configs/deepspeed_zero3.yaml scripts/run_sft.py recipes/smollm2/sft/config_smol.yaml
```

1. I get an unused parameter exception for `use_flash_attention_2` from `H4ArgumentParser`. Deleting the corresponding line from the config solves the problem (I think).
2. I get a `ValueError: Chat template is already added to the tokenizer. If you want to overwrite it, please set it to None` from the setup_chat_format [[here](https://github.com/huggingface/alignment-handbook/blob/main/scripts/run_sft.py#L127)](url). Adding
```{python}
if hasattr(tokenizer, "chat_template") and tokenizer.chat_template is not None:
tokenizer.chat_template = None # Reset the chat template
```
solves the problem (I think).

I don't know if I am doing something wrong or if the solutions I mentioned are actual solutions.

Thanks