Minilm L6 Mnli

🚀 MiniLM-L6-mnli

This model is designed for text classification and zero-shot classification tasks, offering a fast yet relatively less accurate alternative for natural language inference.

🚀 Quick Start

This model was trained on the MultiNLI dataset. The base model is MiniLM-L6 from Microsoft, which is very fast, but a bit less accurate than other models.

✨ Features

• Fast Inference: Leveraging the MiniLM-L6 base model from Microsoft, it provides rapid processing capabilities.
• Trained on MultiNLI: Ensures effectiveness in natural language inference tasks.

📦 Installation

No specific installation steps are provided in the original README.

💻 Usage Examples

Basic Usage

from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch

model_name = "MoritzLaurer/MiniLM-L6-mnli"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSequenceClassification.from_pretrained(model_name)

premise = "I liked the movie"
hypothesis = "The movie was good."

input = tokenizer(premise, hypothesis, truncation=True, return_tensors="pt")
output = model(input["input_ids"].to(device))  # device = "cuda:0" or "cpu"
prediction = torch.softmax(output["logits"][0], -1).tolist()
label_names = ["entailment", "neutral", "contradiction"]
prediction = {name: round(float(pred) * 100, 1) for pred, name in zip(prediction, label_names)}
print(prediction)

📚 Documentation

Training data

MultiNLI.

Training procedure

MiniLM-L6-mnli-binary was trained using the Hugging Face trainer with the following hyperparameters.

training_args = TrainingArguments(
    num_train_epochs=5,              # total number of training epochs
    learning_rate=2e-05,
    per_device_train_batch_size=32,   # batch size per device during training
    per_device_eval_batch_size=32,    # batch size for evaluation
    warmup_ratio=0.1,                # number of warmup steps for learning rate scheduler
    weight_decay=0.06,               # strength of weight decay
    fp16=True                        # mixed precision training
)

Eval results

The model was evaluated using the (matched) test set from MultiNLI. Accuracy: 0.814

🔧 Technical Details

The model is based on the MiniLM-L6 architecture from Microsoft and is trained on the MultiNLI dataset. It uses the Hugging Face trainer with specific hyperparameters for training.

📄 License

No license information is provided in the original README.

Limitations and bias

Please consult the original MiniLM paper and literature on different NLI datasets for potential biases.

BibTeX entry and citation info

If you want to cite this model, please cite the original MiniLM paper, the respective NLI datasets and include a link to this model on the Hugging Face hub.