How to beginner · 3 min read

How to use AutoModelForSequenceClassification

Q: How to use AutoModelForSequenceClassification

Use AutoModelForSequenceClassification from Hugging Face Transformers to load a pretrained model for text classification by specifying the model name or path. Combine it with AutoTokenizer to preprocess input text, then pass tokenized inputs to the model to get classification logits.

Quick answer

Use AutoModelForSequenceClassification from Hugging Face Transformers to load a pretrained model for text classification by specifying the model name or path. Combine it with AutoTokenizer to preprocess input text, then pass tokenized inputs to the model to get classification logits.

PREREQUISITES

Python 3.8+
pip install transformers torch
Basic knowledge of PyTorch

Setup

Install the transformers and torch libraries to use Hugging Face models. Set up your Python environment with the required packages.

bash

pip install transformers torch

Step by step

This example shows how to load a pretrained sequence classification model and tokenizer, tokenize input text, and get classification logits.

python

from transformers import AutoModelForSequenceClassification, AutoTokenizer
import torch

# Load pretrained model and tokenizer
model_name = "distilbert-base-uncased-finetuned-sst-2-english"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSequenceClassification.from_pretrained(model_name)

# Prepare input text
texts = ["I love using Hugging Face!", "This is a bad example."]
inputs = tokenizer(texts, padding=True, truncation=True, return_tensors="pt")

# Forward pass
outputs = model(**inputs)
logits = outputs.logits

# Convert logits to probabilities
probs = torch.nn.functional.softmax(logits, dim=-1)

print("Logits:", logits)
print("Probabilities:", probs)

output

Logits: tensor([[ 3.2, -3.1],
        [-2.9,  2.7]])
Probabilities: tensor([[9.98e-01, 2.02e-03],
        [1.90e-03, 9.98e-01]])

Common variations

Use different pretrained models by changing model_name to any sequence classification checkpoint on Hugging Face Hub.
Run inference asynchronously using asyncio with PyTorch's async support.
Use Trainer API for fine-tuning the model on your own dataset.

Troubleshooting

If you get a ModelNotFoundError, verify the model name is correct and available on Hugging Face Hub.
For CUDA errors, ensure PyTorch is installed with GPU support and your GPU drivers are up to date.
If tokenization fails, check that input texts are strings and tokenizer is properly loaded.

✅

Key Takeaways

Use AutoModelForSequenceClassification.from_pretrained() to load any pretrained classification model.
Always pair the model with the matching AutoTokenizer for correct input preprocessing.
Pass tokenized inputs as keyword arguments to the model to get logits for classification.
Convert logits to probabilities with softmax for interpretable outputs.
Check model availability and environment setup to avoid common errors.

Verified 2026-04 · distilbert-base-uncased-finetuned-sst-2-english

Verify ↗