docs(finetuning): remove deprecated torchtune references in QLoRA and float8 sections

[Anai-Guo]

Follow-up to #4312, addressing @andrewor14's review request:

Looks like there are still references to torchtune in the QLoRA and quantized finetuning sections. When you get a chance can you fix these as well (in a separate PR)?

torchtune development wound down in 2025 and the library is no longer actively maintained, so this removes the remaining torchtune-specific recipes from finetuning.rst:

QLoRA section

• Removed the deprecated "Option 1: TorchTune Integration" recipe (tune run lora_finetune_single_device).
• Promoted "HuggingFace PEFT Integration" to Option 1 (heading underline length is unchanged, since Option 1:/Option 2: are equal length).
• I intentionally did not swap this to Unsloth (unlike the QAT section in docs(finetuning): replace TorchTune QAT section with Unsloth #4312): Unsloth's QLoRA uses bitsandbytes 4-bit rather than torchao's NF4, so it wouldn't illustrate torchao NF4 in a downstream framework. torchao's NF4 usage is still demonstrated directly via the FrozenNF4Linear / to_nf4 example just above the options.

Float8 Quantized Fine-tuning section

• Removed the deprecated tune run full_finetune_distributed ... enable_fp8_training=true recipe.
• Reworded to point at the shared float8 training APIs (TorchTitan). The measured throughput/accuracy results table and the pre-training tutorial pointer are kept.

One remaining reference, left out of scope: Option 3: TorchAO QAT API still imports from torchtune.models.llama3 import llama3 purely to build a toy model for the runnable example. That's in the QAT section (not the QLoRA/float8 sections you flagged), and replacing the model constructor risks breaking the example, so I left it for now — happy to follow up if you'd like it swapped for a plain nn.Module/transformers model.

🤖 Generated with Claude Code

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/ao/4514

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

This comment was automatically generated by Dr. CI and updates every 15 minutes.