This folder contains actively maintained examples of use of πŸ€— Transformers organized along NLP tasks. If you are looking for an example that used to be in this folder, it may have moved to our research projects subfolder (which contains frozen snapshots of research projects).

Important noteΒΆ


To make sure you can successfully run the latest versions of the example scripts, you have to install the library from source and install some example-specific requirements. To do this, execute the following steps in a new virtual environment:

git clone https://github.com/huggingface/transformers
cd transformers
pip install .

Then cd in the example folder of your choice and run

pip install -r requirements.txt

To browse the examples corresponding to released versions of πŸ€— Transformers, click on the line below and then on your desired version of the library:

Examples for older versions of πŸ€— Transformers

Alternatively, you can find switch your cloned πŸ€— Transformers to a specific version (for instance with v3.5.1) with

git checkout tags/v3.5.1

and run the example command as usual afterward.

The Big Table of TasksΒΆ

Here is the list of all our examples:

  • with information on whether they are built on top of Trainer/TFTrainer (if not, they still work, they might just lack some features),

  • whether or not they leverage the πŸ€— Datasets library.

  • links to Colab notebooks to walk through the scripts and run them easily,

Task Example datasets Trainer support TFTrainer support πŸ€— Datasets Colab
language-modeling Raw text βœ… - βœ… Open In Colab
multiple-choice SWAG, RACE, ARC βœ… βœ… βœ… Open In Colab
question-answering SQuAD βœ… βœ… βœ… Open In Colab
summarization CNN/Daily Mail βœ… - - -
text-classification GLUE, XNLI βœ… βœ… βœ… Open In Colab
text-generation - n/a n/a - Open In Colab
token-classification CoNLL NER βœ… βœ… βœ… Open In Colab
translation WMT βœ… - - -

Distributed training and mixed precisionΒΆ

All the PyTorch scripts mentioned above work out of the box with distributed training and mixed precision, thanks to the Trainer API. To launch one of them on n GPUS, use the following command:

python -m torch.distributed.launch \
    --nproc_per_node number_of_gpu_you_have path_to_script.py \

As an example, here is how you would fine-tune the BERT large model (with whole word masking) on the text classification MNLI task using the run_glue script, with 8 GPUs:

python -m torch.distributed.launch \
    --nproc_per_node 8 text-classification/run_glue.py \
    --model_name_or_path bert-large-uncased-whole-word-masking \
    --task_name mnli \
    --do_train \
    --do_eval \
    --max_seq_length 128 \
    --per_device_train_batch_size 8 \
    --learning_rate 2e-5 \
    --num_train_epochs 3.0 \
    --output_dir /tmp/mnli_output/

If you have a GPU with mixed precision capabilities (architecture Pascal or more recent), you can use mixed precision training with PyTorch 1.6.0 or latest, or by installing the Apex library for previous versions. Just add the flag --fp16 to your command launching one of the scripts mentioned above!

Using mixed precision training usually results in 2x-speedup for training with the same final results (as shown in this table for text classification).

Running on TPUsΒΆ

When using Tensorflow, TPUs are supported out of the box as a tf.distribute.Strategy.

When using PyTorch, we support TPUs thanks to pytorch/xla. For more context and information on how to setup your TPU environment refer to Google’s documentation and to the very detailed pytorch/xla README.

In this repo, we provide a very simple launcher script named xla_spawn.py that lets you run our example scripts on multiple TPU cores without any boilerplate. Just pass a --num_cores flag to this script, then your regular training script with its arguments (this is similar to the torch.distributed.launch helper for torch.distributed):

python xla_spawn.py --num_cores num_tpu_you_have \
    path_to_script.py \

As an example, here is how you would fine-tune the BERT large model (with whole word masking) on the text classification MNLI task using the run_glue script, with 8 TPUs:

python xla_spawn.py --num_cores 8 \
    text-classification/run_glue.py \
    --model_name_or_path bert-large-uncased-whole-word-masking \
    --task_name mnli \
    --do_train \
    --do_eval \
    --max_seq_length 128 \
    --per_device_train_batch_size 8 \
    --learning_rate 2e-5 \
    --num_train_epochs 3.0 \
    --output_dir /tmp/mnli_output/

Logging & Experiment trackingΒΆ

You can easily log and monitor your runs code. The following are currently supported:

Weights & BiasesΒΆ

To use Weights & Biases, install the wandb package with:

pip install wandb

Then log in the command line:

wandb login

If you are in Jupyter or Colab, you should login with:

import wandb

To enable logging to W&B, include "wandb" in the report_to of your TrainingArguments or script. Or just pass along --report_to all if you have wandb installed.

Whenever you use Trainer or TFTrainer classes, your losses, evaluation metrics, model topology and gradients (for Trainer only) will automatically be logged.

Advanced configuration is possible by setting environment variables:

Environment Variables Options
WANDB_LOG_MODEL Log the model as artifact at the end of training (false by default)
  • gradients (default): Log histograms of the gradients
  • all: Log histograms of gradients and parameters
  • false: No gradient or parameter logging
WANDB_PROJECT Organize runs by project

Set run names with run_name argument present in scripts or as part of TrainingArguments.

Additional configuration options are available through generic wandb environment variables.

Refer to related documentation & examples.


To use comet_ml, install the Python package with:

pip install comet_ml

or if in a Conda environment:

conda install -c comet_ml -c anaconda -c conda-forge comet_ml