peft

This is the repo for prompt tuning a language model to improve the given prompt (vague).

This project leverages FLAN-T5 from Hugging Face to perform dialogue summarization, fine-tuning with ROUGE, and detoxifying summaries using PPO and PEFT.

SpeechCueLLM Implementation: Enabling Llama-3 to detect emotions from speech vocal nuances via text description mapping. | Python, PyTorch, HuggingFace, Llama-3, LoRA, Praat | AI, NLP, Multimodal Speech Emotion Recognition

Implementation of Low-Rank Adaptation (LoRA) for parameter-efficient fine-tuning of GPT-2 on the SQuAD dataset for question answering, exploring training efficiency, loss masking, and performance metrics like F1 and Exact Match. Final Course project for Deep Learning at University of Kerman, Spring 2025.

GAWS: Parameter efficient fine-tuning technique that reduces the inference latency and maintains performance comparable to LoRA.

Parameter-Efficient Fine-Tuning of TinyLlama-1.1B for bank support LoRA adapters — training only 0.10% of parameters | Streamlit Chat UI | Cloud-to-Local Pipeline

Clean LoRA fine-tune of Llama-3.2-11B-Vision for astronomy captioning on one T4, plus a documented audit of five defects in the published pipeline it reproduces.

Assistente de codigo bilingue (PT/EN): fine-tune LoRA/QLoRA do Qwen2.5-Coder, com dataset, treino, API FastAPI e notebook do Colab.

What: LoRA/QLoRA lab (adapters + scripts). For: small Spark fine-tunes; weights stay local. How: ./setup.sh

Evaluation and gradient-aligned optimization (P-Trojan) of backdoor persistence in code LLMs under benign post-training.

Safety-gated emotion classifier: fine-tuned DistilBERT (92.2% acc) + retrieval-grounded agent with ordered safety gates

Fine-Tuning Google's Vision Transformer LoRA technique. Two different LoRA adapters are tuned for Separate Classification (Food and Human actions). A simple Gradio interface is implemented to run the Inference.

Modular LLM alignment framework for SFT, DPO, and GRPO with LoRA/QLoRA, configurable training pipelines, and reproducible experimentation.

Demonstration of LLM techniques such as prompt engineering, full finetuning, PEFT (LoRA) etc.

This project is based on fine-tuning LLM models (FLAN-T5) for text summarisation task using PEFT approach. All evaluation metrics being computed on ROUGE scoring and LoRA optimisation techniques being used for fine-tuning.

Official repository for "A Systematic Literature Review of Parameter-Efficient Fine-Tuning for Large Code Models" (arXiv:2504.21569)

This repository contains hands on tutorials on fine tuning LLMs

End-to-end LLM engineering pipeline: LoRA fine-tuning on Qwen3-8B/14B, FP8 quantization, vLLM concurrency benchmarking, and BFCL evaluation — built for finance function-calling on H100

Fine-tuning Mistral-7B and LLaMA using QLoRA & PEFT for efficient LLM customization on a single GPU

UAV doctrine corpus and small language model fine-tuning pipeline targeting onboard