metadata
license: apache-2.0
base_model: Qwen/Qwen3-0.6B-Base
tags:
- fine-tuned
- multiple-choice-qa
- mcqa
- question-answering
- quantized
- 8bit
datasets:
- custom-mcqa-dataset
language:
- en
pipeline_tag: text-generation
MNLP_M2_quantized_model
This model is a quantized fine-tuned version of Qwen/Qwen3-0.6B-Base for Multiple Choice Question Answering (MCQA) tasks.
Model Details
- Base Model: Qwen/Qwen3-0.6B-Base
- Task: Multiple Choice Question Answering
- Model Type: Quantized
- Training Context: Without context
- Evaluation Context: Without context
- Fine-tuning Method: Causal Language Modeling
- Quantization: 8bit
Training Details
- Epochs: 5
- Learning Rate: 5e-05
- Batch Size: 2
- Training Framework: Transformers + PyTorch
- Quantization Method: 8bit
- Quantization Only: No
Performance
| Metric | Baseline | Fine-tuned | Improvement |
|---|---|---|---|
| Accuracy | 69.66% | 70.68% | +1.02% |
Training Data
The model was fine-tuned on a custom MCQA dataset with the following characteristics:
- Format: Multiple choice questions with 4 options (A, B, C, D)
- Context: Not included during training
- Evaluation: Without context
- Quantization: Applied during training and evaluation
Usage
from transformers import AutoTokenizer, AutoModelForCausalLM
tokenizer = AutoTokenizer.from_pretrained("MNLP_M2_quantized_model", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("MNLP_M2_quantized_model", trust_remote_code=True)
# For MCQA tasks, provide the question and options, then generate the answer
prompt = "Question: What is the capital of France?\nA) London\nB) Berlin\nC) Paris\nD) Madrid\nAnswer:"
inputs = tokenizer(prompt, return_tensors="pt")
outputs = model.generate(**inputs, max_new_tokens=5)
answer = tokenizer.decode(outputs[0], skip_special_tokens=True)