Edit model card


Model description

This model was trained on the MultiNLI dataset. The model was trained for binary NLI, which means that the "neutral" and "contradiction" classes were merged into one class. The model therefore predicts "entailment" or "not_entailment". The base model is MiniLM-L6 from Microsoft, which is very fast, but a bit less accurate than other models.

Intended uses & limitations

How to use the model

from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch

model_name = "MoritzLaurer/MiniLM-L6-mnli-binary"
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", "not_entailment"]
prediction = {name: round(float(pred) * 100, 1) for pred, name in zip(prediction, label_names)}

Training data


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
    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 binary (matched) test set from MultiNLI. Accuracy: 0.886

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.

Downloads last month