YAML Metadata
Warning:
The pipeline tag "text2text-generation" is not in the official list: text-classification, token-classification, table-question-answering, question-answering, zero-shot-classification, translation, summarization, feature-extraction, text-generation, fill-mask, sentence-similarity, text-to-speech, text-to-audio, automatic-speech-recognition, audio-to-audio, audio-classification, audio-text-to-text, voice-activity-detection, depth-estimation, image-classification, object-detection, image-segmentation, text-to-image, image-to-text, image-to-image, image-to-video, unconditional-image-generation, video-classification, reinforcement-learning, robotics, tabular-classification, tabular-regression, tabular-to-text, table-to-text, multiple-choice, text-ranking, text-retrieval, time-series-forecasting, text-to-video, image-text-to-text, visual-question-answering, document-question-answering, zero-shot-image-classification, graph-ml, mask-generation, zero-shot-object-detection, text-to-3d, image-to-3d, image-feature-extraction, video-text-to-text, keypoint-detection, visual-document-retrieval, any-to-any, video-to-video, other
🧠 FLAN-T5 QLoRA (Prompt Tuned) - Country Capital QA
This model is a fine-tuned version of google/flan-t5-base using QLoRA and Prompt Tuning on a hybrid QA dataset.
📌 Highlights
- 🔍 Correct & incorrect (hallucinated) QA pairs
- ⚙️ Trained using 4-bit QLoRA with PEFT
- 🔧 Prompt tuning enables parameter-efficient adaptation
🏗️ Training
- Base Model:
google/flan-t5-base - Method: QLoRA + Prompt Tuning with PEFT
- Quantization: 4-bit NF4
- Frameworks: 🤗 Transformers, PEFT, Accelerate
- Evaluation: BLEU = 92.5, ROUGE = 87.3
📚 Dataset
Mixture of 20 correct and 3 incorrect QA samples from Pravesh390/country-capital-mixed.
📦 Usage
from transformers import pipeline
pipe = pipeline("text2text-generation", model="Pravesh390/flan-t5-qlora-countryqa-v1")
pipe("What is the capital of Brazil?")
📈 Intended Use
- Evaluate hallucinations in QA systems
- Robust model development for real-world QA
- Academic research or education
🏷️ License
Apache 2.0 — Free for research and commercial use.
Dataset used to train Pravesh390/flan-t5-qlora-countryqa-v3
Evaluation results
- bleu on Country-Capital Mixed QAself-reported92.500
- rouge on Country-Capital Mixed QAself-reported87.300