Add Claude Skill for PEFT/LoRA fine-tuning guidance

[jeeth-kataria]

What does this PR do?

Adds a Claude SKILL for PEFT (Parameter-Efficient Fine-Tuning) and LoRA guidance, complementing the transformers-api skill in #43340.

Why is this needed?

PEFT/LoRA is one of the most popular integrations with Transformers for memory-efficient fine-tuning. This skill helps Claude Code users with:

• LoRA configuration (r, lora_alpha, target_modules)
• QLoRA setup for consumer GPUs
• Adapter operations (save, load, merge, multi-adapter)
• Training integration with Trainer/SFTTrainer
• Common troubleshooting patterns

Structure

.claude/skills/peft-lora/
├── SKILL.md # Main entry point
├── reference/areas/
│ ├── lora-basics.md # LoRA fundamentals
│ ├── qlora-quantization.md # 4-bit/8-bit training
│ ├── adapter-operations.md # Load/save/merge
│ ├── training-integration.md # Trainer & SFTTrainer
│ └── troubleshooting.md # Common errors
└── templates/
└── minimal_repro.md # Bug report template

Related PRs

• Claude code skills for transformers-api #43340 - Claude code skills for transformers-api (this complements that PR)
• Addresses issue Please create a Huggingface Transformers SKILL for Claude #42971