Transformers
  • 🌍GET STARTED
    • Transformers
    • Quick tour
    • Installation
  • 🌍TUTORIALS
    • Run inference with pipelines
    • Write portable code with AutoClass
    • Preprocess data
    • Fine-tune a pretrained model
    • Train with a script
    • Set up distributed training with BOINC AI Accelerate
    • Load and train adapters with BOINC AI PEFT
    • Share your model
    • Agents
    • Generation with LLMs
  • 🌍TASK GUIDES
    • 🌍NATURAL LANGUAGE PROCESSING
      • Text classification
      • Token classification
      • Question answering
      • Causal language modeling
      • Masked language modeling
      • Translation
      • Summarization
      • Multiple choice
    • 🌍AUDIO
      • Audio classification
      • Automatic speech recognition
    • 🌍COMPUTER VISION
      • Image classification
      • Semantic segmentation
      • Video classification
      • Object detection
      • Zero-shot object detection
      • Zero-shot image classification
      • Depth estimation
    • 🌍MULTIMODAL
      • Image captioning
      • Document Question Answering
      • Visual Question Answering
      • Text to speech
    • 🌍GENERATION
      • Customize the generation strategy
    • 🌍PROMPTING
      • Image tasks with IDEFICS
  • 🌍DEVELOPER GUIDES
    • Use fast tokenizers from BOINC AI Tokenizers
    • Run inference with multilingual models
    • Use model-specific APIs
    • Share a custom model
    • Templates for chat models
    • Run training on Amazon SageMaker
    • Export to ONNX
    • Export to TFLite
    • Export to TorchScript
    • Benchmarks
    • Notebooks with examples
    • Community resources
    • Custom Tools and Prompts
    • Troubleshoot
  • 🌍PERFORMANCE AND SCALABILITY
    • Overview
    • 🌍EFFICIENT TRAINING TECHNIQUES
      • Methods and tools for efficient training on a single GPU
      • Multiple GPUs and parallelism
      • Efficient training on CPU
      • Distributed CPU training
      • Training on TPUs
      • Training on TPU with TensorFlow
      • Training on Specialized Hardware
      • Custom hardware for training
      • Hyperparameter Search using Trainer API
    • 🌍OPTIMIZING INFERENCE
      • Inference on CPU
      • Inference on one GPU
      • Inference on many GPUs
      • Inference on Specialized Hardware
    • Instantiating a big model
    • Troubleshooting
    • XLA Integration for TensorFlow Models
    • Optimize inference using `torch.compile()`
  • 🌍CONTRIBUTE
    • How to contribute to transformers?
    • How to add a model to BOINC AI Transformers?
    • How to convert a BOINC AI Transformers model to TensorFlow?
    • How to add a pipeline to BOINC AI Transformers?
    • Testing
    • Checks on a Pull Request
  • 🌍CONCEPTUAL GUIDES
    • Philosophy
    • Glossary
    • What BOINC AI Transformers can do
    • How BOINC AI Transformers solve tasks
    • The Transformer model family
    • Summary of the tokenizers
    • Attention mechanisms
    • Padding and truncation
    • BERTology
    • Perplexity of fixed-length models
    • Pipelines for webserver inference
    • Model training anatomy
  • 🌍API
    • 🌍MAIN CLASSES
      • Agents and Tools
      • 🌍Auto Classes
        • Extending the Auto Classes
        • AutoConfig
        • AutoTokenizer
        • AutoFeatureExtractor
        • AutoImageProcessor
        • AutoProcessor
        • Generic model classes
          • AutoModel
          • TFAutoModel
          • FlaxAutoModel
        • Generic pretraining classes
          • AutoModelForPreTraining
          • TFAutoModelForPreTraining
          • FlaxAutoModelForPreTraining
        • Natural Language Processing
          • AutoModelForCausalLM
          • TFAutoModelForCausalLM
          • FlaxAutoModelForCausalLM
          • AutoModelForMaskedLM
          • TFAutoModelForMaskedLM
          • FlaxAutoModelForMaskedLM
          • AutoModelForMaskGenerationge
          • TFAutoModelForMaskGeneration
          • AutoModelForSeq2SeqLM
          • TFAutoModelForSeq2SeqLM
          • FlaxAutoModelForSeq2SeqLM
          • AutoModelForSequenceClassification
          • TFAutoModelForSequenceClassification
          • FlaxAutoModelForSequenceClassification
          • AutoModelForMultipleChoice
          • TFAutoModelForMultipleChoice
          • FlaxAutoModelForMultipleChoice
          • AutoModelForNextSentencePrediction
          • TFAutoModelForNextSentencePrediction
          • FlaxAutoModelForNextSentencePrediction
          • AutoModelForTokenClassification
          • TFAutoModelForTokenClassification
          • FlaxAutoModelForTokenClassification
          • AutoModelForQuestionAnswering
          • TFAutoModelForQuestionAnswering
          • FlaxAutoModelForQuestionAnswering
          • AutoModelForTextEncoding
          • TFAutoModelForTextEncoding
        • Computer vision
          • AutoModelForDepthEstimation
          • AutoModelForImageClassification
          • TFAutoModelForImageClassification
          • FlaxAutoModelForImageClassification
          • AutoModelForVideoClassification
          • AutoModelForMaskedImageModeling
          • TFAutoModelForMaskedImageModeling
          • AutoModelForObjectDetection
          • AutoModelForImageSegmentation
          • AutoModelForImageToImage
          • AutoModelForSemanticSegmentation
          • TFAutoModelForSemanticSegmentation
          • AutoModelForInstanceSegmentation
          • AutoModelForUniversalSegmentation
          • AutoModelForZeroShotImageClassification
          • TFAutoModelForZeroShotImageClassification
          • AutoModelForZeroShotObjectDetection
        • Audio
          • AutoModelForAudioClassification
          • AutoModelForAudioFrameClassification
          • TFAutoModelForAudioFrameClassification
          • AutoModelForCTC
          • AutoModelForSpeechSeq2Seq
          • TFAutoModelForSpeechSeq2Seq
          • FlaxAutoModelForSpeechSeq2Seq
          • AutoModelForAudioXVector
          • AutoModelForTextToSpectrogram
          • AutoModelForTextToWaveform
        • Multimodal
          • AutoModelForTableQuestionAnswering
          • TFAutoModelForTableQuestionAnswering
          • AutoModelForDocumentQuestionAnswering
          • TFAutoModelForDocumentQuestionAnswering
          • AutoModelForVisualQuestionAnswering
          • AutoModelForVision2Seq
          • TFAutoModelForVision2Seq
          • FlaxAutoModelForVision2Seq
      • Callbacks
      • Configuration
      • Data Collator
      • Keras callbacks
      • Logging
      • Models
      • Text Generation
      • ONNX
      • Optimization
      • Model outputs
      • Pipelines
      • Processors
      • Quantization
      • Tokenizer
      • Trainer
      • DeepSpeed Integration
      • Feature Extractor
      • Image Processor
    • 🌍MODELS
      • 🌍TEXT MODELS
        • ALBERT
        • BART
        • BARThez
        • BARTpho
        • BERT
        • BertGeneration
        • BertJapanese
        • Bertweet
        • BigBird
        • BigBirdPegasus
        • BioGpt
        • Blenderbot
        • Blenderbot Small
        • BLOOM
        • BORT
        • ByT5
        • CamemBERT
        • CANINE
        • CodeGen
        • CodeLlama
        • ConvBERT
        • CPM
        • CPMANT
        • CTRL
        • DeBERTa
        • DeBERTa-v2
        • DialoGPT
        • DistilBERT
        • DPR
        • ELECTRA
        • Encoder Decoder Models
        • ERNIE
        • ErnieM
        • ESM
        • Falcon
        • FLAN-T5
        • FLAN-UL2
        • FlauBERT
        • FNet
        • FSMT
        • Funnel Transformer
        • GPT
        • GPT Neo
        • GPT NeoX
        • GPT NeoX Japanese
        • GPT-J
        • GPT2
        • GPTBigCode
        • GPTSAN Japanese
        • GPTSw3
        • HerBERT
        • I-BERT
        • Jukebox
        • LED
        • LLaMA
        • LLama2
        • Longformer
        • LongT5
        • LUKE
        • M2M100
        • MarianMT
        • MarkupLM
        • MBart and MBart-50
        • MEGA
        • MegatronBERT
        • MegatronGPT2
        • Mistral
        • mLUKE
        • MobileBERT
        • MPNet
        • MPT
        • MRA
        • MT5
        • MVP
        • NEZHA
        • NLLB
        • NLLB-MoE
        • NystrΓΆmformer
        • Open-Llama
        • OPT
        • Pegasus
        • PEGASUS-X
        • Persimmon
        • PhoBERT
        • PLBart
        • ProphetNet
        • QDQBert
        • RAG
        • REALM
        • Reformer
        • RemBERT
        • RetriBERT
        • RoBERTa
        • RoBERTa-PreLayerNorm
        • RoCBert
        • RoFormer
        • RWKV
        • Splinter
        • SqueezeBERT
        • SwitchTransformers
        • T5
        • T5v1.1
        • TAPEX
        • Transformer XL
        • UL2
        • UMT5
        • X-MOD
        • XGLM
        • XLM
        • XLM-ProphetNet
        • XLM-RoBERTa
        • XLM-RoBERTa-XL
        • XLM-V
        • XLNet
        • YOSO
      • 🌍VISION MODELS
        • BEiT
        • BiT
        • Conditional DETR
        • ConvNeXT
        • ConvNeXTV2
        • CvT
        • Deformable DETR
        • DeiT
        • DETA
        • DETR
        • DiNAT
        • DINO V2
        • DiT
        • DPT
        • EfficientFormer
        • EfficientNet
        • FocalNet
        • GLPN
        • ImageGPT
        • LeViT
        • Mask2Former
        • MaskFormer
        • MobileNetV1
        • MobileNetV2
        • MobileViT
        • MobileViTV2
        • NAT
        • PoolFormer
        • Pyramid Vision Transformer (PVT)
        • RegNet
        • ResNet
        • SegFormer
        • SwiftFormer
        • Swin Transformer
        • Swin Transformer V2
        • Swin2SR
        • Table Transformer
        • TimeSformer
        • UperNet
        • VAN
        • VideoMAE
        • Vision Transformer (ViT)
        • ViT Hybrid
        • ViTDet
        • ViTMAE
        • ViTMatte
        • ViTMSN
        • ViViT
        • YOLOS
      • 🌍AUDIO MODELS
        • Audio Spectrogram Transformer
        • Bark
        • CLAP
        • EnCodec
        • Hubert
        • MCTCT
        • MMS
        • MusicGen
        • Pop2Piano
        • SEW
        • SEW-D
        • Speech2Text
        • Speech2Text2
        • SpeechT5
        • UniSpeech
        • UniSpeech-SAT
        • VITS
        • Wav2Vec2
        • Wav2Vec2-Conformer
        • Wav2Vec2Phoneme
        • WavLM
        • Whisper
        • XLS-R
        • XLSR-Wav2Vec2
      • 🌍MULTIMODAL MODELS
        • ALIGN
        • AltCLIP
        • BLIP
        • BLIP-2
        • BridgeTower
        • BROS
        • Chinese-CLIP
        • CLIP
        • CLIPSeg
        • Data2Vec
        • DePlot
        • Donut
        • FLAVA
        • GIT
        • GroupViT
        • IDEFICS
        • InstructBLIP
        • LayoutLM
        • LayoutLMV2
        • LayoutLMV3
        • LayoutXLM
        • LiLT
        • LXMERT
        • MatCha
        • MGP-STR
        • Nougat
        • OneFormer
        • OWL-ViT
        • Perceiver
        • Pix2Struct
        • Segment Anything
        • Speech Encoder Decoder Models
        • TAPAS
        • TrOCR
        • TVLT
        • ViLT
        • Vision Encoder Decoder Models
        • Vision Text Dual Encoder
        • VisualBERT
        • X-CLIP
      • 🌍REINFORCEMENT LEARNING MODELS
        • Decision Transformer
        • Trajectory Transformer
      • 🌍TIME SERIES MODELS
        • Autoformer
        • Informer
        • Time Series Transformer
      • 🌍GRAPH MODELS
        • Graphormer
  • 🌍INTERNAL HELPERS
    • Custom Layers and Utilities
    • Utilities for pipelines
    • Utilities for Tokenizers
    • Utilities for Trainer
    • Utilities for Generation
    • Utilities for Image Processors
    • Utilities for Audio processing
    • General Utilities
    • Utilities for Time Series
Powered by GitBook
On this page
  • TorchScript flag and tied weights
  • Dummy inputs and standard lengths
  • Using TorchScript in Python
  • Deploy BOINC AI TorchScript models to AWS with the Neuron SDK
  1. DEVELOPER GUIDES

Export to TorchScript

PreviousExport to TFLiteNextBenchmarks

Last updated 1 year ago

This is the very beginning of our experiments with TorchScript and we are still exploring its capabilities with variable-input-size models. It is a focus of interest to us and we will deepen our analysis in upcoming releases, with more code examples, a more flexible implementation, and benchmarks comparing Python-based codes with compiled TorchScript.

According to the :

TorchScript is a way to create serializable and optimizable models from PyTorch code.

There are two PyTorch modules, , that allow developers to export their models to be reused in other programs like efficiency-oriented C++ programs.

We provide an interface that allows you to export 🌍 Transformers models to TorchScript so they can be reused in a different environment than PyTorch-based Python programs. Here, we explain how to export and use our models using TorchScript.

Exporting a model requires two things:

  • model instantiation with the torchscript flag

  • a forward pass with dummy inputs

These necessities imply several things developers should be careful about as detailed below.

TorchScript flag and tied weights

The torchscript flag is necessary because most of the 🌍 Transformers language models have tied weights between their Embedding layer and their Decoding layer. TorchScript does not allow you to export models that have tied weights, so it is necessary to untie and clone the weights beforehand.

Models instantiated with the torchscript flag have their Embedding layer and Decoding layer separated, which means that they should not be trained down the line. Training would desynchronize the two layers, leading to unexpected results.

This is not the case for models that do not have a language model head, as those do not have tied weights. These models can be safely exported without the torchscript flag.

Dummy inputs and standard lengths

The dummy inputs are used for a models forward pass. While the inputs’ values are propagated through the layers, PyTorch keeps track of the different operations executed on each tensor. These recorded operations are then used to create the trace of the model.

The trace is created relative to the inputs’ dimensions. It is therefore constrained by the dimensions of the dummy input, and will not work for any other sequence length or batch size. When trying with a different size, the following error is raised:

Copied

`The expanded size of the tensor (3) must match the existing size (7) at non-singleton dimension 2`

We recommended you trace the model with a dummy input size at least as large as the largest input that will be fed to the model during inference. Padding can help fill the missing values. However, since the model is traced with a larger input size, the dimensions of the matrix will also be large, resulting in more calculations.

Be careful of the total number of operations done on each input and follow the performance closely when exporting varying sequence-length models.

Using TorchScript in Python

This section demonstrates how to save and load models as well as how to use the trace for inference.

Saving a model

To export a BertModel with TorchScript, instantiate BertModel from the BertConfig class and then save it to disk under the filename traced_bert.pt:

Copied

from transformers import BertModel, BertTokenizer, BertConfig
import torch

enc = BertTokenizer.from_pretrained("bert-base-uncased")

# Tokenizing input text
text = "[CLS] Who was Jim Henson ? [SEP] Jim Henson was a puppeteer [SEP]"
tokenized_text = enc.tokenize(text)

# Masking one of the input tokens
masked_index = 8
tokenized_text[masked_index] = "[MASK]"
indexed_tokens = enc.convert_tokens_to_ids(tokenized_text)
segments_ids = [0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1]

# Creating a dummy input
tokens_tensor = torch.tensor([indexed_tokens])
segments_tensors = torch.tensor([segments_ids])
dummy_input = [tokens_tensor, segments_tensors]

# Initializing the model with the torchscript flag
# Flag set to True even though it is not necessary as this model does not have an LM Head.
config = BertConfig(
    vocab_size_or_config_json_file=32000,
    hidden_size=768,
    num_hidden_layers=12,
    num_attention_heads=12,
    intermediate_size=3072,
    torchscript=True,
)

# Instantiating the model
model = BertModel(config)

# The model needs to be in evaluation mode
model.eval()

# If you are instantiating the model with *from_pretrained* you can also easily set the TorchScript flag
model = BertModel.from_pretrained("bert-base-uncased", torchscript=True)

# Creating the trace
traced_model = torch.jit.trace(model, [tokens_tensor, segments_tensors])
torch.jit.save(traced_model, "traced_bert.pt")

Loading a model

Now you can load the previously saved BertModel, traced_bert.pt, from disk and use it on the previously initialised dummy_input:

Copied

loaded_model = torch.jit.load("traced_bert.pt")
loaded_model.eval()

all_encoder_layers, pooled_output = loaded_model(*dummy_input)

Using a traced model for inference

Use the traced model for inference by using its __call__ dunder method:

Copied

traced_model(tokens_tensor, segments_tensors)

Deploy BOINC AI TorchScript models to AWS with the Neuron SDK

  1. Easy-to-use API with one line of code change to trace and optimize a TorchScript model for inference in the cloud.

Implications

Dependencies

Converting a model for AWS Neuron

Copied

from transformers import BertModel, BertTokenizer, BertConfig
import torch
import torch.neuron

You only need to modify the following line:

Copied

- torch.jit.trace(model, [tokens_tensor, segments_tensors])
+ torch.neuron.trace(model, [token_tensor, segments_tensors])

This enables the Neuron SDK to trace the model and optimize it for Inf1 instances.

AWS introduced the instance family for low cost, high performance machine learning inference in the cloud. The Inf1 instances are powered by the AWS Inferentia chip, a custom-built hardware accelerator, specializing in deep learning inferencing workloads. is the SDK for Inferentia that supports tracing and optimizing transformers models for deployment on Inf1. The Neuron SDK provides:

Out of the box performance optimizations for .

Support for BOINC AI transformers models built with either or .

Transformers models based on the architecture, or its variants such as and run best on Inf1 for non-generative tasks such as extractive question answering, sequence classification, and token classification. However, text generation tasks can still be adapted to run on Inf1 according to this . More information about models that can be converted out of the box on Inferentia can be found in the section of the Neuron documentation.

Using AWS Neuron to convert models requires a which comes preconfigured on .

Convert a model for AWS NEURON using the same code from to trace a BertModel. Import the torch.neuron framework extension to access the components of the Neuron SDK through a Python API:

To learn more about AWS Neuron SDK features, tools, example tutorials and latest updates, please see the .

🌍
TorchScript documentation
JIT and TRACE
Amazon EC2 Inf1
AWS Neuron
improved cost-performance
PyTorch
TensorFlow
BERT (Bidirectional Encoder Representations from Transformers)
distilBERT
roBERTa
AWS Neuron MarianMT tutorial
Model Architecture Fit
Neuron SDK environment
AWS Deep Learning AMI
Using TorchScript in Python
AWS NeuronSDK documentation