asfahanjaved126/sentiment-classifier-v1

Sentiment Classifier v1

A LoRA fine-tuned adapter for Qwen2.5-3B-Instruct that classifies text as positive or negative sentiment. Trained on Amazon product review data.

Model Details

• Base model: Qwen/Qwen2.5-3B-Instruct
• Fine-tuning method: LoRA (Low-Rank Adaptation), rank 16, alpha 32
• Target modules: q_proj, k_proj, v_proj, o_proj
• Task: Binary sentiment classification (positive / negative)
• Training data: 2,400 examples from Amazon product reviews
• Trainable parameters: 7.4M (0.24% of total model parameters)

Performance

Evaluated on a held-out test set of 300 examples:

Metric	Score
Accuracy	97%
Weighted F1	0.97
Positive precision / recall	0.98 / 0.96
Negative precision / recall	0.97 / 0.98

How to Use

from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel
import torch

base_model = AutoModelForCausalLM.from_pretrained(
    "Qwen/Qwen2.5-3B-Instruct",
    torch_dtype=torch.float16,
    device_map="auto"
)
model = PeftModel.from_pretrained(base_model, "asfahanjaved126/sentiment-classifier-v1")
tokenizer = AutoTokenizer.from_pretrained("asfahanjaved126/sentiment-classifier-v1")

messages = [
    {"role": "system", "content": "Classify as positive or negative. One word only."},
    {"role": "user", "content": "This product completely changed how I work, love it!"}
]
prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
output = model.generate(**inputs, max_new_tokens=5, temperature=0.0, do_sample=False)
result = tokenizer.decode(output[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True)
print(result)

Intended Use

This model is designed for classifying customer reviews, feedback, and similar short-form text into positive/negative sentiment categories. Suitable for e-commerce review analysis, customer feedback triage, and similar use cases.

Limitations

• Trained only on binary sentiment (positive/negative); does not currently classify neutral sentiment.
• Trained on Amazon product review data; may perform differently on other domains without further fine-tuning.
• English language only.

Training Procedure

Fine-tuned using QLoRA (4-bit quantization) on a single T4 GPU via Google Colab. Training used the TRL SFTTrainer with completion-only loss masking, cosine learning rate schedule, and 3 epochs over the training set.