File size: 2,509 Bytes
a463c68 dcf5730 a463c68 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 |
---
language:
- en
tags:
- text-classification
- zero-shot-classification
metrics:
- accuracy
widget:
- text: "I liked the movie. [SEP] The movie was good."
---
# MiniLM-L6-mnli-binary
## Model description
This model was trained on the [MultiNLI](https://huggingface.co/datasets/multi_nli) 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
```python
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)}
print(prediction)
```
### Training data
[MultiNLI](https://huggingface.co/datasets/multi_nli).
### 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 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. |